VPS13A is closely associated with mitochondria and is required for efficient lysosomal degradation

Chorea-acanthocytosis (ChAc) is a rare disease that leads to progressive neurodegeneration. Movement disorders such as chorea, dystonia or parkinsonism, and the presence of acanthocytes (misshapen red blood cells) are characteristic of this fatal disease. To date, therapeutic approaches are based on a few published empirical observations and aim only at mitigating symptoms (Walker, 2015a). Therefore, any finding on the molecular etiology of ChAc could have important implications for the development of effective treatments to cure or slow down the progression of this disease.

Mutations in the vacuolar protein sorting 13 A (VPS13A) gene, which mostly lead to the absence of the protein (also known as chorein), are the cause of ChAc (Dobson-Stone et al., 2002, 2004; Rampoldi et al., 2001; Tomiyasu et al., 2011; Ueno et al., 2001). The other three members of the human VPS13 family are also involved in disease. Mutations in VPS13B lead to Cohen syndrome (Kolehmainen et al., 2003); mutations in VPS13C have been identified as a cause of an autosomal-recessive, early-onset and severe form of Parkinson's disease (Lesage et al., 2016; Schormair et al., 2018); and, most recently, mutations in VPS13D have been linked to other movement disorders (Gauthier et al., 2018; Seong et al., 2018). In addition, genomic data have identified VPS13A-D variants in other neurological disorders (Fromer et al., 2014; McCarthy et al., 2014; Meda et al., 2012), in various types of cancer (An et al., 2012; Furukawa et al., 2011; Morisaki et al., 2014; Park et al., 2016b; Yang et al., 2016b) and in diabetes (Grarup et al., 2011; Saxena et al., 2010; Strawbridge et al., 2011; Windholz et al., 2011).

VPS13 proteins are very large proteins that share conserved domains or structural features. They are widely conserved during eukaryotic evolution, from unicellular organisms to humans (Velayos-Baeza et al., 2004), so their study can be addressed in different models (Rzepnikowska et al., 2017). In Saccharomyces cerevisiae, only one Vps13 protein is present, but it participates in several cellular processes such as protein trafficking, vesicular transport and fusion (Brickner and Fuller, 1997; De et al., 2017; Redding et al., 1996), sporulation (Park and Neiman, 2012), mitochondrial homeostasis (Park et al., 2016a) and functional compensation of the loss of endoplasmic reticulum (ER)-mitochondria encounter structure (ERMES), which are membrane contact sites (MCS) between the mitochondria and the ER (Lang et al., 2015; Park et al., 2016a; Xue et al., 2017). In other model organisms, such as Tetrahymena thermophila, Dictyostelium discoideum and Drosophila melanogaster, several VPS13 proteins are present and have been implicated in different cellular processes such as phagocytosis (Leiba et al., 2017; Muñoz-Braceras et al., 2015; Samaranayake et al., 2011), autophagy (Muñoz-Braceras et al., 2015) and mitochondrial homeostasis (Anding et al., 2018). The functions attributed to the human VPS13 proteins are particularly diverse, especially for VPS13A, which has been reported to be involved in cytoskeleton organization (Alesutan et al., 2013; Föller et al., 2012; Honisch et al., 2015a; Schmidt et al., 2013; Stanslowsky et al., 2016), phosphoinositide regulation (Park et al., 2015), tyrosine phosphorylation dependent on LYN (De Franceschi et al., 2011), secretion (Honisch et al., 2015b; Schmidt et al., 2013), calcium homeostasis (Yu et al., 2016) and autophagy (Lupo et al., 2016; Muñoz-Braceras et al., 2015). VPS13B is required for the organization of the Golgi apparatus and protein glycosylation (Duplomb et al., 2014; Seifert et al., 2011), VPS13C for adipogenesis and lipolysis (Ramseyer et al., 2018; Yang et al., 2016a), and both VPS13C and VPS13D for the maintenance of mitochondrial homeostasis (Anding et al., 2018; Lesage et al., 2016). Despite the growing evidence of the involvement of VPS13 proteins in many cellular processes, the mechanism by which their alteration leads to different human diseases and the relative importance of each process in the pathogenesis of these diseases are still unknown.

Macroautophagy, hereafter referred to as autophagy for simplicity, is a cellular degradation process essential for the maintenance of cellular homeostasis. Its proper functioning is particularly crucial for neuronal survival and has therefore been implicated in several neurodegenerative diseases (Kiriyama and Nochi, 2015). In addition, autophagy is necessary during erythropoiesis (Vessoni et al., 2012). We previously reported a partial defect in autophagy in HeLa cells lacking VPS13A (Muñoz-Braceras et al., 2015) and hypothesized that this defect could play an important role in the etiology of ChAc. A deficiency of autophagy during this disease was supported by the work of Lupo and colleagues, who described a delayed autophagic clearance during erythroid maturation in patients (Lupo et al., 2016). In addition, increased levels of the autophagic proteins ULK1, ATG13, ATG7 and LC3-I were observed in ChAc erythrocyte lysates (Lupo et al., 2016), but the exact mechanism leading to defective autophagy remains unknown.

In this study, we aimed to investigate in more detail the role of VPS13A in autophagy by first using D. discoideum as a model organism and then human cells. Our results suggest that the defects observed in autophagy in the absence of VPS13A are most likely the consequence of a more general impairment of the endolysosomal pathway. In addition, we investigated the subcellular localization of VPS13A and found an unexpected predominant localization to mitochondria, which provides valuable insight into the possible mechanisms by which the absence of VPS13A may lead to lysosomal dysfunction.

Our previous study of a member of the VPS13 family, TipC, in D. discoideum provided the first evidence of VPS13 proteins involvement in autophagy. The tipC⁻ mutant presents a multitipped phenotype, which is a characteristic developmental phenotype of autophagy mutants in this social amoeba (Mesquita et al., 2015; Otto et al., 2003, 2004; Tung et al., 2010), and, accordingly, this mutant exhibits impaired autophagy along with additional defects in sporulation and phagocytosis. We found that these phenotypes were largely rescued by the overexpression of the C-terminal region of TipC (amino acids 2725-3848), which contains conserved domains found in virtually all VPS13 proteins, including human VPS13A. In addition, we demonstrated that autophagy is impaired in VPS13A-depleted human HeLa cells (Muñoz-Braceras et al., 2015). Based on these results, we hypothesized that the C-terminal region of TipC in D. discoideum could mediate its interaction with proteins involved in the execution or regulation of autophagy and that this interaction could be conserved for human VPS13A. Therefore, in the present study, we used D. discoideum as a starting point to shed light on the molecular function of VPS13 proteins.

We used liquid chromatography (LC) coupled to tandem mass spectrometry (MS/MS) to identify proteins that co-immunoprecipitate with D. discoideum TipC^2725-3848-GFP and not with a control GFP (Table S1). One of the possible interactors identified was Ras-like in rat brain 7A (Rab7A), a protein involved in autophagy and phagocytosis in D. discoideum and other organisms (Guerra and Bucci, 2016; Rupper et al., 2001). The interaction was confirmed by pulldown experiments using D. discoideum cells expressing hemagglutinin (HA)-tagged Rab7A and TipC^2725-3848-GFP (Fig. 1A). We then analyzed the interaction of the corresponding human proteins in HeLa cells transfected with GFP-tagged wild-type or mutant constitutively active (GTP-bound) or constitutively inactive (GDP-bound) forms of the RAB7A GTPase. We observed that endogenous VPS13A specifically co-immunoprecipitated with GFP-RAB7A, and that VPS13A interacts more with the constitutively active RAB7A mutant than with the constitutively inactive form of the GTPase (Fig. 1B), similarly to Rab-interacting lysosomal protein (RILP), which is a well-known effector of RAB7A (Cantalupo et al., 2001). These results suggest that the ability to interact with RAB7A is conserved among the VPS13 proteins and lead to the hypothesis that VPS13 proteins may participate in autophagy through their interaction with RAB7A.

RAB7A is recruited to endocytic vesicles during their maturation and it is a key participant in the endolysosomal pathway. Its roles in autophagy and the endocytic pathway involve similar processes, as it regulates the movement and maturation of both autophagic and endocytic vesicles and their fusion with lysosomes (Guerra and Bucci, 2016). Moreover, there is a close relationship between the autophagic and endocytic pathways, as autophagosomes can fuse with endosomes to form amphisomes prior to fusion with lysosomes. Therefore, if there is a functional link between VPS13A and RAB7A, VPS13A knockdown may cause alterations of RAB7A itself and the endocytic pathway. Knockdown of VPS13A was performed by RNA interference using two different small interfering RNAs (siRNAs) to rule out possible off-target effects.

First, we evaluated the levels and distribution of RAB7A in the absence of VPS13A. Interestingly, we found a prominent accumulation of endogenous RAB7A in protein lysates of VPS13A-depleted cells (Fig. 2A). This accumulation was also obvious from the confocal microscopy visualization of the protein, which revealed that endogenous RAB7A was notably accumulated in vesicles in a juxtanuclear region of the VPS13A siRNA-treated cells (Fig. 2B). In previous work, we observed a similar pattern of accumulation of microtubule-associated protein 1 light chain 3B (LC3B; also known as MAP1LC3B), a marker of autophagosomes, in a juxtanuclear region of the cell when VPS13A expression was inhibited (Muñoz-Braceras et al., 2015). We detected extensive colocalization of LC3 with RAB7A, and the quantitative assessment of the relative colocalization of both markers revealed a higher Pearson's coefficient of colocalization in VPS13A-depleted cells (Fig. 2B).

Both autophagosomes and endosomes fuse with lysosomes for degradation of their cargo and therefore the accumulation of autophagic and endosomal markers may be due to a reduced number of lysosomes. However, lysosome-associated membrane protein 1 (LAMP1), a lysosomal marker, accumulated similarly in a juxtanuclear area in VPS13A siRNA-treated cells, and clearly colocalized with RAB7A (Fig. 3A) and LC3 (Fig. S1). Lysosomes are normally concentrated in a juxtanuclear region (Pu et al., 2016), to which late endosomes and autophagosomes are transported to promote fusion with lysosomes. The clustering of RAB7A and LC3 together with LAMP1 in the absence of VPS13A shows that endosomes and autophagosomes correctly reach the lysosomes, but their accumulation indicates that there is a defect in the fusion process or in the efficiency of degradation. Although higher Pearson's colocalization coefficients could be indicative of fusion, they may also be the result of pronounced accumulation of markers in the cluster, and therefore we cannot rule out the existence of a slight fusion defect that would not be detectable by conventional confocal microscopy. We then performed transmission electron microscopy (TEM) to better define the identity of the vesicles. Endosomes are single-membrane vesicles with a variable number of intraluminal vesicles and a rather electron-lucent content that distinguishes them from endolysosomes, characterized by the presence of heterogeneous and electron-dense material. Amphisomes and autolysosomes are included in the same category as endolysosomes, as they are also single-membrane vesicles with a heterogeneous content partially degraded and electron dense. Autophagosomes, on the other hand, have an intact double membrane and a non-degraded heterogeneous cytoplasmic content (Huotari and Helenius, 2011; Klumperman and Raposo, 2014). Based on these criteria, few autophagosomes were detected and no significant changes in their number were found upon VPS13A knockdown. Similarly, no differences in the number of endosomes were observed. In contrast, depletion of VPS13A resulted in the accumulation of endolysosomes containing electron-dense and partially degraded material (Fig. 3B). Together, these data indicate that fusion between autophagosomes, endosomes and lysosomes is not severely altered in VPS13A siRNA-treated cells and suggest that, although degradation should occur to some extent given the presence of endolysosomes containing electron-dense material, the blockage of autophagy in the absence of VPS13A occurs at the level of degradation.

Our results suggest that loss of VPS13A leads to a compromised ability of cells to degrade cargos contained into autophagic and endocytic vesicles. To test this hypothesis, we studied the levels of specific cargos that reach the lysosomes through autophagy or endocytosis. p62 (also known as SQSTM1) binds ubiquitinated cargos and LC3 and allows engulfment of cargo by autophagosomes (Pankiv et al., 2007). We observed that the knockdown of VPS13A led to a juxtanuclear accumulation of p62 (Fig. 4A), thus supporting the hypothesis of a deficient degradation. Ferritin is another autophagic cargo (Ott et al., 2016), and we also observed its accumulation in VPS13A siRNA-treated cells (Fig. 4B). Epidermal growth factor receptor (EGFR) is internalized upon EGF binding, traffics along the endocytic pathway and is degraded in lysosomes (Futter et al., 1996). We visualized the receptor after EGF stimulation and found that, although most EGFR disappeared 1 h after EGF addition in control siRNA-treated cells, internalized EGFR was still present and accumulated in the juxtanuclear region when VPS13A was knocked down (Fig. 4C). Together, these results strongly support our hypothesis that VPS13A is necessary for efficient lysosomal degradation.

Lysosomal degradation is carried out by lysosomal hydrolases, which are synthesized as proenzymes, bound to mannose-6-phosphate receptors, and delivered to the lysosomes. In the acidic environment of lysosomes, proenzymes are released from their receptors and proteolytically processed into active forms (Bonifacino and Rojas, 2006; Ishidoh and Kominami, 2002). Considering the described role of yeast Vps13 in vesicular trafficking (Brickner and Fuller, 1997; De et al., 2017; Redding et al., 1996), it was tempting to speculate that VPS13A was necessary in the transport of proenzymes to the lysosomes. To study this possibility, the localization of the lysosomal cysteine protease cathepsin B (CTSB), an example of lysosomal hydrolase, was analyzed. We found that CTSB accumulated in the juxtanuclear area and colocalized with LAMP1 in VPS13A-depleted cells (Fig. 4D), indicating that there is not a major impairment of CTSB delivery to lysosomes. Lysosomal hydrolases such as CTSB may be properly delivered to lysosomes but may not be fully active, as activation requires the proteolytic processing of the enzyme. In the case of CTSB, the cleavage of its propeptide generates an active single-chain protease and this mature form can be further processed into a double-chain form by an additional cleavage that generates a heavy chain and a light chain (Ishidoh and Kominami, 2002). CTSB processing was analyzed by western blotting and found to be decreased in the absence of VPS13A (Fig. 4E). This further supports the idea that VPS13A is necessary for efficient lysosomal degradation, although further research is needed to determine the mechanism by which loss of VPS13A leads to reduced CTSB processing.

To shed light on how the lack of VPS13A negatively affects cell competence for lysosomal degradation, we investigated the subcellular localization of the protein. Immunofluorescence analysis of endogenous VPS13A revealed a faint signal distributed in a network pattern that disappeared in VPS13A-depleted cells, confirming the specificity of the antibody (Fig. 5A). Because the signal detected by immunofluorescence was low, probably due to the low abundance of the endogenous protein, we set out to overexpress and visualize a GFP-tagged version of VPS13A to continue the analysis of the subcellular localization of the protein. Lang and colleagues had previously reported that yeast Vps13 was not fully functional when tagged at the N or C terminus. However, GFP insertion at a specific internal location did not affect the function of the protein (Lang et al., 2015). The need to place the GFP in an internal position was later confirmed by Park and colleagues, who obtained similar results using an insertion of the GFP at a different position (Park et al., 2016a). We generated, by gap repair, an expression plasmid for VPS13A in which EGFP was inserted in-frame into the human VPS13A sequence after alanine 476 (the equivalent position to that of yeast Vps13 reported by Lang and colleagues). The gap repair strategy to obtain this plasmid is detailed in the Materials and Methods section and depicted in Fig. S2. Overexpressed VPS13A^GFP had the expected molecular weight (Fig. 5B) and presented a network-like pattern similar to that observed for endogenous VPS13A (Fig. 5C). Then, we tested the colocalization of VPS13A^GFP with different organelle markers and found that most VPS13A^GFP clearly decorated the mitochondrial network. We also observed a partial colocalization of VPS13A^GFP with the ER and no colocalization with the Golgi apparatus. To our surprise, we detected only occasional and partial colocalization of VPS13A^GFP with RAB7A and merely sporadic overlap with LC3 and LAMP1 (Fig. 5C,D). The major localization to the mitochondria, partial colocalization with the ER, endosomes and lysosomes, and scarce or no colocalization with the Golgi apparatus were confirmed by detection of endogenous VPS13A by immunofluorescence (Fig. 5E; Fig. S3). Additionally, VPS13A^GFP was observed to colocalize with lipid droplets, according to BODIPY staining (Fig. 5F).

In view of the above results, and based on the studies on yeast Vps13 that demonstrated the localization of the protein at the MCS between the mitochondria and the vacuole or endosomes (Lang et al., 2015; Park et al., 2016a), we wanted to determine whether VPS13A could also be present at the interface between mitochondria and endolysosomes. Triple imaging of VPS13A, mitochondria and RAB7A showed that ∼3% of VPS13A^GFP overlapped simultaneously with mitochondria and RAB7A-positive vesicles, suggesting that the human VPS13A protein might similarly localize at MCS between mitochondria and endolysosomes (Fig. 6A). In addition, we observed that, on average, 12% of VPS13A^GFP overlapped with both the mitochondria and the ER, indicating that VPS13A may also be present in the MCS between the mitochondria and the ER (Fig. 6B). The visualization of VPS13A localization by immunoelectron microscopy would be useful to confirm the association of the protein with mitochondria at contact sites with other organelles. However, to date, the low transfection efficiency of the overexpressed protein and the modest signal obtained with the anti-VPS13A antibody make this task challenging.

Our observations on VPS13A localization at mitochondria-endolysosome and mitochondria-ER MCS (MAM) led us to examine whether the lack of VPS13A had an impact on these MCS. By using TEM, we found endosomal vesicles occasionally adjacent to mitochondria (at a distance of less than 30 nm) in both control and VPS13A siRNA-treated cells. The length of those MCS per mitochondria length was reduced upon VPS13A knockdown (Fig. 6C); however, the formation of the contacts was not hampered, indicating that the loss of VPS13A does not have a major impact on their formation. With respect to the mitochondria-ER MCS or mitochondria-associated ER membranes, we also found a decrease in the relative length of the contacts (Fig. 6D). The study of the mitochondria-endolysosome MCS suggests that the impaired degradative capacity of lysosomes observed in the absence of VPS13A, a protein with a mitochondrial localization, is not caused by the lack of a physical connection between mitochondria and lysosomes, although the absence of VPS13A may lead to a moderate decrease in the formation or stabilization of these MCS and, possibly, to their functional impairment.

In this study, we have shown that the absence of VPS13A, a condition that leads to ChAc in humans, results in defective lysosomal degradation in HeLa cells. This is probably the cause of the partially impaired autophagy reported in our previous work (Muñoz-Braceras et al., 2015) and the accumulation of non-degraded material found in this study. This defect can potentially lead to the presence of protein aggregates in ChAc, as in many other neurodegenerative diseases. Although no protein aggregates have, as yet, been detected in neuropathologic samples (Walker, 2015b), potential protein aggregates have been visualized at the ultrastructural level in the skeletal muscle of patients with ChAc (Melone et al., 2002). Moreover, p62 is accumulated in ChAc red blood cells, where remnants of vesicles, mitochondria and lysosomes are also indicative of delayed autophagic clearance, at least during erythroid maturation (Lupo et al., 2016). VPS13A is also likely to participate in autophagic clearance in brain neurons, because overexpression of human VPS13A was able to reduce the accumulation of Ref(2)P (the D. melanogaster p62 homolog) and ubiquitin-positive aggregates in the central nervous system of a fly mutant for the VPS13A homolog (Vonk et al., 2017).

Although this is the first time that an interaction of RAB7A with human VPS13A and with a D. discoideum VPS13 protein has been demonstrated, the interaction of other members of the VPS13 family with different RAB GTPases has already been reported. The interaction of VPS13C with diverse RAB GTPases, including RAB7A, has been revealed by different proteomic studies (Behrends et al., 2010; Huttlin et al., 2015; McCray et al., 2010), and VPS13B has been described as an effector of RAB6 (also known as RAB6A) (Seifert et al., 2015). In addition, the three D. melanogaster VPS13 proteins were isolated by affinity chromatography using different GTP-RABs (Gillingham et al., 2014). RAB7A is involved in many different stages of endosomal and autophagosomal trafficking, lysosomal fusion and lysosomal maintenance, among other processes (reviewed by Guerra and Bucci, 2016). Defects in any of these processes can lead to defective lysosomal degradation. It is therefore conceivable that VPS13A may play a role in autophagy and in the endolysosomal pathway through its interaction with RAB7A. Alternatively, Lupo and colleagues have reported that VPS13A co-immunoprecipitates with ATG7, which may indicate a more direct role of VPS13A in the autophagic pathway (Lupo et al., 2016). However, we found that the absence of VPS13A leads to the accumulation of vesicles with partially degraded content but not to the accumulation of immature autophagosomes, a phenotype that would be expected for a defect in the early stages of autophagosome formation.

Given the requirement of VPS13A for the degradative function of lysosomes and its interaction with RAB7A, it was surprising to observe that the protein only occasionally colocalizes with endolysosomal compartments. Instead, VPS13A mainly colocalizes with mitochondria. Recently, RAB7A, which has been widely characterized as an endosomal marker, has been shown to be present on other endomembranes, including mitochondria, although at low levels (Jimenez-Orgaz et al., 2018). We have not found an evident colocalization of VPS13A and RAB7A along mitochondrial membranes. However, as it has been reported that the visualization of RAB7A at mitochondria requires a high laser power, we cannot rule out that a small amount of RAB7A, undetectable with our confocal settings, is present at mitochondria together with VPS13A. In contrast, we have clearly observed the occasional colocalization of VPS13A with RAB7A in discrete locations corresponding mainly to the overlap of VPS13A in mitochondria with RAB7A-positive vesicles. This observation raises the possibility that VPS13A interaction with RAB7A takes place locally on mitochondrial membranes in contact with endolysosomal vesicles. The restricted contact between these organelles would explain the low but reproducible proportion of the total VPS13A and RAB7A that colocalize and interact with each other. In line with this, a recent report suggests that RAB7A cycling is regulated specifically at mitochondria-lysosome contacts (Wong et al., 2018); and S. cerevisiae Vps13 can localize at MCS between the mitochondria and endosomes [known as endosome–mitochondria junctions (EMJs)], and between the mitochondria and the vacuole [known as vacuole-mitochondria patches (vCLAMPs)] (Lang et al., 2015; Park et al., 2016a). In addition to VPS13A colocalization with both mitochondria and RAB7A-positive endolysosomes, we detected VPS13A at mitochondria in close proximity to the ER, suggesting that VPS13A can also be present at multiple MCS in mammalian cells. Our results are in agreement with a recent report, published during the revision of this work, which shows VPS13A localization at the junctions between mitochondria and ER tubules (Kumar et al., 2018). Similarly, VPS13A colocalization with lipid droplets was also reported by Kumar et al., but they did not describe the presence of VPS13A at MCS between mitochondria and endosomes. The percentage of lysosomes in contact with mitochondria at any given time is small and the contacts are very dynamic, as further described recently (Wong et al., 2018). Therefore, it is difficult to visualize the contacts between mitochondria and endolysosomes. This might explain the small percentage of VPS13A that overlaps with both mitochondria and endolysosomes that we have described here and the lack of mention of a similar observation by Kumar et al. in their study (Kumar et al., 2018).

Although our analysis by TEM of MCS between mitochondria and endolysosomes, or between mitochondria and the ER, suggests that VPS13A is not essential for their formation, the observed reduction of the length of mitochondria in contact with endosomes or with the ER in VPS13A siRNA-treated cells may indicate that VPS13A is involved in the stabilization or extension of the contacts. Our observation on mitochondria-ER MCS as a consequence of VPS13A knockdown complements that of Kumar and colleagues, consisting of an expansion of those contacts upon VPS13A overexpression (Kumar et al., 2018). Regarding the contacts between mitochondria and endolysosomes, we speculate that a role of VPS13A in these MCS would explain how the loss of mitochondria-localized VPS13A can result in the alteration of the endolysosomal pathways. The function of the MCS between mitochondria and endolysosomes is currently unclear. In yeast, their formation can compensate for the absence of ERMES and, interestingly, Vps13 is required for this effect (Lang et al., 2015; Park et al., 2016a). Studies in yeast suggest that the MCS between the mitochondria and endosomes or the vacuole, like those between the mitochondria and the ER, may be involved in the trafficking of lipids, ions or other metabolites to and from the mitochondria (Elbaz-Alon et al., 2014; Hönscher et al., 2014); and, interestingly, the recent report by Kumar and colleagues showed that the N-terminal region of S. cerevisiae Vps13 can transfer lipids, raising the possibility that VPS13A may be required for lipid trafficking at MCS (Kumar et al., 2018). In mammalian cells, the MCS between mitochondria and endosomes have only recently been described. To date, they have been implicated in iron transfer (Das et al., 2016; Hamdi et al., 2016), mitochondrial fission (Hsu et al., 2018; Wong et al., 2018), and lysosomal dynamics and RAB7A hydrolysis (Wong et al., 2018). Therefore, a hypothetical role of VPS13A in the MCS between mitochondria and endolysosomes may be required for a correct local regulation of RAB7A, leading to the observed changes in RAB7A levels and localization, and could potentially explain the alteration of lysosomal function when VPS13A is absent.

Although loss of VPS13A function at MCS between mitochondria and endolysosomes is the most direct explanation of the defects observed upon depletion of the mitochondria-localized VPS13A protein, we cannot rule out an alteration of a more indirect crosstalk between organelles as the cause of the endolysosomal dysfunction. Many cases have been reported in which the dysregulation of one organelle affects the activity of the other (Raimundo et al., 2016; Soto-Heredero et al., 2017). As a consequence of this functional connection, both mitochondrial and lysosomal homeostasis are often compromised in neurodegenerative disorders, including Parkinson's disease, Alzheimer's disease and lysosomal storage disorders (Audano et al., 2018; Plotegher and Duchen, 2017). Interestingly, defects in the morphology and trafficking of mitochondria and lysosomes have recently been described in medium spiny neurons derived from ChAc patient induced pluripotent stem cells (Glaß et al., 2018). These findings suggest that the mechanism of crosstalk between mitochondria and lysosomes that we describe in HeLa cells may also be relevant in the cells affected in the disease, a possibility that warrants further investigation.

The described endolysosomal dysfunction leading to the accumulation of vesicles with non-degraded material in VPS13A knockdown cells may be clinically relevant, as the modulation of lysosomal-related processes (such as autophagy, lysosome biogenesis or lysosomal degradation) is considered a promising potential therapeutic approach for many neurodegenerative diseases with compromised endolysosomal function (Morel et al., 2017; Sardiello, 2016). In conclusion, our findings provide the basis for the study of lysosomes and their communication with mitochondria in the context of ChAc, but further research is necessary to shed more light on how the absence of VPS13A, a protein with a predominant mitochondrial localization, has a negative impact on lysosomal degradation.

Dictyostelium discoideum original tipC⁻ mutant was obtained from Dicty Stock Center, kindly deposited to the repository by William F. Loomis (Stege et al., 1999). The identity of the strain was confirmed by PCR using diagnostic amplifications of the tipC disrupted gene. The tipC⁻ strains overexpressing GFP alone or the C-terminal region of TipC fused to GFP were generated by electroporation of the pDV-CGFP-CTAP vector itself (GenBank Accession Number EF028672) or a construct containing the DNA genomic region encoding amino acids 2725-3848 of TipC cloned into the pDV-CGFP-CTAP vector, and subsequent selection of the transformed cells as previously described (Muñoz-Braceras et al., 2015). HeLa cells were a gift from Alberto Muñoz from the Biomedical Research Institute, Madrid, Spain, and they were authenticated and tested routinely for contamination.

HA-Rab7A plasmid was generated by PCR amplification of D. discoideum complementary DNA (cDNA), using the oligonucleotides ddRAB7-Fw and ddRAB7-Rv (Table S2), and cloned into the pDM358 extrachromosomal vector (GenBank Accession Number EU912542) (Veltman et al., 2009). The VPS13A^EGFP plasmid was generated by a complex strategy including conventional cloning and a gap repair approach and hence it is described in detail later in the Materials and Methods section. GFP-RAB7A plasmids were a gift from Patricia Boya from the Biological Research Center of Madrid. The rest of the plasmids were purchased from Addgene: mCherry-RAB7A (plasmid 61804), mito-BFP (plasmid 49151) and BFP-KDEL (plasmid 49150). The sequences of all plasmids were confirmed by Sanger sequencing.

Primary antibodies used in this study were as follows: anti-calnexin [BD Transduction Laboratories, 610523; used at 1:100 for immunofluorescence (IF)], anti-CTSB [RyD Systems, AF953; 1:1000 for western blotting (WB) and 1:150 for IF], anti-EGFR (Cell Signaling Technology, 4267; 1:100 for IF), anti-GAPDH (Enzo LifeSciences, ADI-CSA-335; 1:2000 for WB), anti-ferritin (RyD Systems, MAB93541; 1:100 for IF), anti-GFP (Sigma-Aldrich, G1544; 1:4000 for WB), anti-GM130 (BD Transduction Laboratories, 610823; 1:100 for IF), anti-HA (Sigma-Aldrich, 12CA5; 1:1000 for WB), anti-LAMP1 (Cell Signaling Technology, 15665; 1:100 for IF), anti-LC3B (Cell Signaling Technology, 2775; 1:200 for IF; Nanotools, 0260-100/LC3-2G6; 1:100 for IF;), anti-RAB7A (Cell Signaling Technology, 9367; 1:100 for IF and 1:1000 for WB; Abcam, ab137029; 1:100 for IF), anti-RILP (Novus, NBP1-98278; 1:600 for WB), anti-p62 (Epitomics, 3340-1; 1:150 for IF), anti-TOM20 (Santa Cruz Biotechnology, sc-11415; 1:100 for IF) and anti-VPS13A (Sigma-Aldrich, HPA021662; 1:75 for IF and 1:500 for WB). Alexa-Fluor-dye-conjugated secondary antibodies were purchased from Invitrogen and used at 1:500 for IF and horseradish-peroxidase-conjugated secondary antibodies were purchased from Santa Cruz Biotechnology and used at 1:5000 for WB.

The plasmid for the expression of VPS13A with EGFP inserted in-frame after alanine 476 was constructed by gap repair, which takes advantage of the efficient homologous recombination in yeast. A scheme of the strategy to generate the plasmid is shown in Fig. S2 and described in detail below. Overlapping fragments of VPS13A transcript variant A (NM_033305) were obtained by PCR amplification of cDNA from HeLa cells, EGFP was amplified from the pEGFP-C3 plasmid, and the ori (2 μm) and URA3 elements, necessary for plasmid selection and amplification in yeast, were amplified from the pRS426 plasmid. The corresponding oligonucleotide sequences are shown in Table S3. A pEGFP-C3 plasmid without the EGFP sequence (between the AgeI and ScaI restriction sites) was generated and used as a backbone for gap repair after digestion with the AseI and XhoI restriction enzymes. The PCR products and the vector fragments were transformed in the S. cerevisiae strain FY250 for their assembly by gap repair as previously described (Joska et al., 2014). A plasmid of 17 kb was generated, isolated and transformed in DH5α Escherichia coli by a conventional heat shock procedure to be amplified. To reduce the size of the plasmid, we cloned the 11 kb fragment between the AclI and MluI restriction sites into the ClaI and BtgI sites of the pBluescript SK vector and a final VPS13A^EGFP expression plasmid of 14 kb was obtained. The whole VPS13A^EGFP sequence was verified by Sanger sequencing.

Dictyostelium discoideum cells were grown axenically in liquid non-defined HL-5 complete medium (Formedium, HLB0102) supplemented with 0.014 g/ml glucose (Pronadisa, 2150) and 25 U/ml penicillin and 0.025 mg/ml streptomycin (Sigma-Aldrich, P4458) in an incubator at 22°C. HeLa cells were grown in complete Dulbecco's modified Eagle medium (DMEM; Sigma-Aldrich, D5648), supplemented with 10% fetal bovine serum (FBS; Gibco, 10500064) and 50 U/ml penicillin and 0.05 mg/ml streptomycin, in standard culture conditions (37°C, 5% CO₂).

For EGFR degradation assay, cells were incubated overnight in serum-free DMEM and then treated with 100 ng/ml EGF (Sigma-Aldrich, E9644) for the indicated times. Staining of mitochondria was performed by incubating the cells with 100 nM MitoTracker Red CMXRos (Molecular Probes, M7512) for 15 min and lipid droplets were stained using 1 μg/ml BODIPY 558/568-C₁₂ (Molecular Probes, D3835) for 30 min. In all cases, cells were incubated in standard culture conditions during treatment with these compounds.

DNA transfection was performed using Lipofectamine 2000 (Invitrogen, 11668019), according to the manufacturer's protocol. Briefly, Lipofectamine 2000 and DNA were mixed in Opti-MEM reduced serum medium (Life Technologies, 31985062) and DNA-Lipofectamine 2000 complexes were added to the cells. Cells were visualized 24 h after the DNA transfection. VPS13A expression was inhibited by reverse transfection of validated siRNAs (Ambion, siRNA ID s23340, named as VPS13A siRNA1; or siRNA ID s23342, named as VPS13A siRNA2). Silencer select negative control siRNA (Ambion, ID No. 2 4390846) was used to transfect cells as a control of the knockdown experiments. An additional control siRNA (Ambion, ID No. 1 4390843) was used to confirm that the phenotypes observed with several markers (p62, LAMP1, RAB7A and LC3) were not affected by the used siRNA control (data not shown). Transfection of siRNA was performed using Lipofectamine RNAiMAX (Invitrogen, 13778030) according to the manufacturer's instructions. Briefly, siRNA and Lipofectamine RNAiMAX were mixed in Opti-MEM and cells were added to the mixture. Knockdown cells were transfected again 48 h after the first transfection with the same siRNAs and the experiments were carried out 72 h after the second transfection. The level of VPS13A mRNA was measured by TaqMan PCR (Applied Biosystems; VPS13A, Hs00362891_m1; endogenous control gene used for normalization, Hs00183533_m1). Both VPS13A siRNAs significantly decreased VPS13A mRNA levels, although the inhibition of the expression was stronger using the VPS13A siRNA1 (80.0±16.5% for siRNA1 and 53.0±8.0% for siRNA2). The levels of VPS13C mRNA were not altered by any of the VPS13A siRNAs as determined by TaqMan PCR (Applied Biosystems; VPS13C, Hs00419559_m1).

GFP-tagged proteins were immunoprecipitated with GFP-Trap A beads (Chromoteck, gtak-20) according to the manufacturer's protocol. Briefly, D. discoideum or HeLa cells expressing the GFP-tagged protein were washed with PDF buffer (20 mM KCl, 9 mM K₂HPO₄, 13 mM KH₂PO₄, 1 mM CaCl₂, 1 mM MgSO₄, pH 6.4) or phosphate-buffered saline (PBS) 1× (133 mM NaCl, 8 mM Na₂HPO₄, 2 mM KH₂PO₄, pH 7.4), respectively, harvested and lysed using protein lysis buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% NP-40), supplemented with phosphatase inhibitors (2.5 mM NaF, 0.2 mM Na₃VO₄) and 1:100 protease inhibitor cocktail (Sigma-Aldrich, P8340). Cell lysates were centrifuged (12,000 g for 15 min at 4°C) and the supernatant was transferred to a microcentrifuge tube. Pre-washed GFP-Trap A beads were added, and the mixture was diluted in washing buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 2.5 mM NaF, 0.2 mM Na₃VO₄ and protease inhibitors) and incubated for 4 h at 4°C with continuous rotation. Then, the mixture was centrifuged (2000 g for 2 min at 4°C) and the beads were washed five times with washing buffer and prepared for mass spectrometry or western blot analyses.

The protein extract samples were resuspended in a volume up to 100 μl of Laemmli sample buffer and applied onto a conventional 4% stacking and 12% resolving SDS-PAGE gel. Electrophoresis was stopped as soon as the front entered 1 cm into the resolving gel in order to concentrate the whole proteome in the stacking/resolving gel interface. The unresolved protein bands were visualized by Coomassie Blue staining and each gel lane was excised, cut into cubes (1 mm²), deposited into 96-well plates and automatically processed in a Proteineer DP digestor (Bruker Daltonics). Digestion was performed as previously described (Shevchenko et al., 1996), with the following minor modifications: gel plugs were washed first with 50 mM NH₄HCO₃ and then with acetonitrile (ACN) prior to reduction with 10 mM dithiothreitol in 25 mM NH₄HCO₃ solution, and alkylation was carried out with 55 mM iodoacetamide in 50 mM NH₄HCO₃ solution. Gel pieces were rinsed with 50 mM NH₄HCO₃, then with ACN, and dried under a stream of nitrogen. Proteomics Grade Trypsin (Sigma-Aldrich, T6567) at a final concentration of 16 ng/μl in 25% ACN/50 mM NH₄HCO₃ solution was added and digestion took place at 37°C for 4 h. The reaction was stopped by adding 50% ACN/0.5% trifluoroacetic acid (TFA) for peptide extraction and the tryptic eluted peptides were dried by speed-vacuum centrifugation. The peptide samples were analyzed on a nano LC system (Eksigent Technologies nanoLC Ultra 1D plus, AB SCIEX) coupled to a 5600 Triple TOF mass spectrometer (AB SCIEX) with a nanoelectrospray ion source. Samples were injected on a C18 PepMap trap column (5 µm, 100 µm ID ×2 cm, Thermo Scientific) at 2 µl/min, in 0.1% formic acid in water, and the trap column was switched on-line to a C18 nanoAcquity BEH analytical column (1.7 µm, 100 Å, 75 µm ID ×15 cm, Waters). Equilibration was performed in mobile phase A (0.1% formic acid in water) and peptide elution was achieved in a 60 min gradient from 5% to 40% B (0.1% formic acid in acetonitrile) at 250 nl/min. The mass spectrometer operated in data-dependent acquisition mode. For TOF scans, the accumulation time was set to 250 ms and, per cycle, up to 15 precursor ions were monitored. MS and MS/MS data obtained for each sample were processed using Analyst TF 1.5.1 Software (AB SCIEX, Foster City, CA, USA). Raw data were translated to mascot general file (mgf) format and searched against a composite target/decoy database built from the 8418 sequences in the D. discoideum reference proteome at Uniprot Knowledgebase (as of July 2015), using an in-house Mascot Server v. 2.4 (Matrix Science, London, UK). Search parameters were set as follows: carbamidomethylcysteine as fixed modification and methionine oxidation as variable one. Peptide mass tolerance was set to 30 ppm and 0.05 Da, in MS and MS/MS mode, respectively, and two missed cleavage sites were allowed. False discovery rates (FDR) for peptide identification were calculated and only the peptides passing a cutoff of FDR<1% were taken into account. In addition, absence of identification in the control GFP pulldown sample and a coverage above 10% of the protein were set as the parameters to consider the identified proteins as strong candidates for the interaction. However, 60S and 40S ribosomal proteins were ignored, even if meeting the cited criteria.

For protein extraction, cells were washed with PBS, harvested and lysed with lysis buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% Triton X-100, supplemented with phosphatase and protease inhibitors) unless otherwise indicated. Cell lysates were centrifuged (12,000 g for 15 min at 4°C) and the protein concentration of the supernatant was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23225). Equal amounts of protein samples were prepared in 1× SDS-PAGE sample loading buffer, heated, loaded and run for separation in conventional handcast SDS-polyacrilamide gels. For VPS13A detection, precast NuPage Tris-acetate 3-8% polyacrylamide gradient gels (Invitrogen, EA0378) were used. Electrotransference on Immobilon-P polyvinylidene difluoride (PVDF) membranes (Millipore, IPVH00010) was performed by wet transfer. For immunoblot analysis, membranes were blocked in 5% skimmed milk in TBS-T buffer for at least 1 h at room temperature (RT), incubated with primary antibodies in blocking solution overnight at 4°C, washed with TBS-T, incubated with the appropriate horseradish-peroxidase-conjugated secondary antibodies for 1 h at RT, and washed before incubation with Amersham ECL Western Blotting Detection Reagents (GE Healthcare Life Sciences, RPN2134) or Clarity Western ECL Substrate (Bio-Rad, 1705060) for antibody detection. The chemiluminescent signal was captured using CURIX RP2 Plus films (Agfa), which were introduced in an X-ray film-processing machine and scanned. Intensity of bands was performed by densitometry using ImageJ software (NIH).

Cells were grown on glass sterilized coverslips in 24-well plates (Corning, 353047) or in μ-Slide 8-well chambered coverslips (Ibidi, 80826). For immunofluorescence, fixation was performed with 3% paraformaldehyde in PBS, pH 7.4, for 15 min at RT and then washed with PBS and incubated with 100 mM glycine in PBS for 30 min at RT. Permeabilization with cold methanol was performed at −20°C for 10 min. Coverslips were washed with PBS, incubated with blocking solution (3% bovine serum albumin, 0.2% Triton X-100 in PBS) for 1 h at RT, incubated with primary antibodies in blocking solution for 3 h at RT, washed and incubated with the appropriate Alexa-Fluor-dye-conjugated secondary antibodies 1 h at RT. Coverslips were then washed and mounted on slides using Prolong Diamond Antifade Mountant (Molecular Probes, P36970). Images were acquired using an inverted laser confocal microscope (Zeiss, LSM710) with a 63×/1.40 Plan-Apochromatic oil immersion objective and analyzed using ImageJ software.

Fixation of cells was performed with 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.4) for 2 h at RT, followed by a post-fixation with 1% OsO₄ and 1.5% K₃Fe(CN)₆ in water for 1 h at 4°C and dehydration with acetone. In situ flat embedding of samples was performed in Epoxy, TAAB 812 Resin (TAAB Laboratories) according to standard protocols. After polymerization, resin sheets were detached from the substrate and mounted onto resin blocks. Ultrathin sections (80 nm) were obtained, deposited onto slot grids and stained with saturated uranyl acetate/lead citrate. Visualization was performed at 80 kV using an electron microscope (Jeol, JEM-1010) and images were recorded with a digital camera (TVIPS, TemCam-F416).

Analysis of images was performed using ImageJ, either manually using tools to select, quantify and measure elements from micrographs or confocal images, or semi-automatically using available plugins such as JACoP for colocalization analyses. Mander's overlap coefficients were calculated using thresholds to define labeled structures. To quantify the simultaneous colocalization of VPS13A with two different organelle markers, a mask with the pixels that were above the threshold in the images of VPS13A and the two organelle markers was created and used to calculate the percentage of VPS13A overlapping with the two organelles simultaneously. Confocal microscopy images were false colored to make them as colorblind friendly as possible.

Data from independent experiments were analyzed using GraphPad Prism (GraphPad Software). Means±s.d. were calculated and represented in plots with error bars. Statistical differences between groups, determined by two-tailed Student's t-test, are indicated with asterisks (significant at *P<0.05, **P<0.01, ***P<0.001).

We thank the Optical and Confocal Microscopy facility at Biomedical Research Institute Alberto Sols, the Proteomic facility at the National Center of Biotechnology, and the Electron Microscopy facility at Molecular Biology Center Severo Ochoa for technical support.