Restructured endoplasmic reticulum generated by mutant amyotrophic lateral sclerosis-linked VAPB is cleared by the proteasome

Intracellular inclusion bodies (IBs) generated by the accumulation of aggregated malfolded mutant proteins are a common feature of neurodegenerative pathologies such as amyotrophic lateral sclerosis (ALS) and Alzheimer's, Parkinson's and Huntington's diseases (Selkoe, 2003). Protein aggregation occurs when the amounts of malfolded or incompletely folded proteins exceed the capacity of the ubiquitin-proteasome system (UPS), and IB formation occurs when the protein aggregates become large enough to be visible at light microscope resolution (Kopito, 2000).

Although the link between protein aggregation and disease has been recognized for many years, the exact role of IBs in cellular degeneration is still incompletely understood and has turned out to be quite complex. On the one hand, by sequestering functional proteins, IBs can be toxic to cells (Olzscha et al., 2011). In particular, recruitment of UPS components to the ubiquitinated mutant protein aggregates perturbs protein homeostasis (Balch et al., 2008; Bence et al., 2001; Ross and Pickart, 2004). On the other hand, sequestration into IBs of small, submicroscopic, aggregates, which are thought to be highly toxic, can be cytoprotective (Arrasate et al., 2004; Douglas et al., 2008; Tanaka et al., 2004). One potentially important factor contributing to the Jekyll and Hyde characteristics of IBs may be the ease of their elimination, with a higher toxicity expected for IBs that are difficult to clear. Of the two main cellular degradation pathways, the UPS or autophagy, the latter plays the dominant role in IB elimination (Fortun et al., 2003; Olzmann and Chin, 2008; Ravikumar and Rubinsztein, 2004). The bulk sequestration of IBs into autophagosomes avoids the need to extract and unfold each single substrate molecule, a prerequisite for access to the narrow proteasomal channel (Finley, 2009).

An unusual type of inclusion is generated by mutant forms of the protein VAMP-associated protein-B (VAPB). VAPB and its homologue VAPA are tail-anchored proteins of the Endoplasmic Reticulum (ER), consisting in an N-terminal cytosolic domain homologous to the nematode major sperm protein (MSP), a coiled-coil domain, and a C-terminal hydrophobic stretch, which serves as membrane anchor. The VAPs are adaptor proteins involved in a variety of important functions, among which membrane traffic, ER interaction with microtubules, the unfolded protein response, and lipid transport mediated by exchange proteins (reviewed in Lev et al., 2008). Many of these functions are mediated by the ability of the cytosolic domain to recruit ligand proteins bearing the so-called FFAT motif [two phenylalanines in an acidic tract (Kaiser et al., 2005; Loewen and Levine, 2005; Loewen et al., 2003)].

VAPB is the product of one of the 10–13 genes that have been causally linked to familial amyotrophic lateral sclerosis (FALS) (reviewed by Andersen and Al-Chalabi, 2011), a lethal disease, characterized by degeneration of upper and lower motor neurons. A dominant missense mutation, resulting in substitution of proline 56 with serine in the MSP domain (P56S mutation), was identified in patients from eight related Brazilian families of Portuguese origin (Nishimura et al., 2005; Nishimura et al., 2004) and then in patients from an unrelated German family (Funke et al., 2010). Subsequently, another mutation (T46I) in the MSP domain was identified in a patient in the United Kingdom (Chen et al., 2010). Both the P56S and the T46I mutant proteins aggregate and form inclusions in transfected cells (Nishimura et al., 2004; Prosser et al., 2008; Teuling et al., 2007). Structural studies have revealed that P56 of the wild-type (wt) protein adopts the unusual cis-peptide bond conformation and that introduction of a trans bond by substitution with serine results in gross misfolding of the MSP domain (Shi et al., 2010). The misfolded MSP domain induces homo-aggregation, which involves also the coiled-coil and transmembrane domains (Kim et al., 2010; Suzuki et al., 2009).

In a previous study, we characterized the biogenesis and structure of P56S-VAPB inclusions (Fasana et al., 2010). We found that the newly synthesized mutant protein inserts normally into the ER membrane, where it rapidly clusters. These clusters than coalesce into larger paranuclear structures. At variance with all other IBs linked to neurodegenerative disorders, mutant VAPB inclusions correspond to domains of dramatically restructured ER that contain all of the mutant protein and that consist of ribbons of stacked cisternae interleaved by an electron-dense cytosolic layer of uniform width.

Here, we have investigated how cells handle these unusual inclusions. Unexpectedly, we find that the P56S-VAPB inclusions are efficiently cleared by the proteasome in a process involving at least one component of the ER-associated degradation (ERAD) machinery (Bernasconi and Molinari, 2011). We discuss the implications of these results for the pathogenic mechanisms of mutant, ALS-linked, VAPB.

As previously demonstrated, P56S-VAPB expressed in transiently transfected cells segregates to restructured ER domains in the juxtanuclear region that consist of tightly apposed smooth-surfaced cisternae separated by an electron-dense layer (Fasana et al., 2010) (Fig. 1A,B). To gain insight into the overall organization of these unusual structures, we reconstructed by means of STEM-HAAD tomography a large portion of a transfected HeLa cell (see supplementary material Movie 1 for a visualization of the entire tomogram, collected from three, 300nm-thick, plastic serial sections). As shown in Fig. 1C (in which mainly the cytosolic electron-dense layer is represented) and in supplementary material Movie 2 (in which the limiting cisternal membranes are shown), the structures consist of a continuous ribbon of two to three parallel cisternae that follow a tortuous and sometimes convoluted pathway with frequent U-shaped turns.

In our previous work (Fasana et al., 2010), we generated HeLa-TetOff cell lines inducibly expressing a Myc-tagged version of P56S-VAPB, and showed at the light microscope level that the mutant protein, even when expressed at levels comparable to those of the endogenous protein, generated inclusions. To confirm that the structures visualized by immunofluorescence indeed correspond to the reorganized ER seen in transiently transfected cells, we analyzed the P56S-VAPB HeLa-TetOff cells by TEM, after either full [culture in the absence of doxycycline (Dox)] or intermediate induction (culture in the presence of low – 0.1 ng/ml – concentrations of the antibiotic). Under the latter condition, the levels of P56S-VAPB are similar to those of the endogenous, wild-type (wt) protein (Fasana et al., 2010). Under both conditions, structures identical to those seen in transiently transfected cells were readily detected (Fig. 1D–G), although, as expected, in the cells expressing low levels of the mutant protein, the restructured ER domains were less extended than in the fully induced or transiently transfected cells.

To investigate the fate of P56S-VAPB inclusions, we took advantage of the HeLa-TetOff system. Cells were first allowed to accumulate the mutant protein for 5–6 days in the absence of Dox; then, to stop synthesis of VAPB, the antibiotic was reintroduced into the medium, and the levels of Myc-tagged protein were subsequently analyzed during this chase period (Fig. 2A). To compare the degradation of the mutant with the wt protein, we generated HeLa-TetOff cell lines inducibly expressing wt Myc-tagged VAPB. As shown in supplementary material Fig. S1, the addition of Dox efficiently repressed expression of the transgene (panel A). After removal of the antibiotic, immunofluorescence analysis revealed that the wt protein localized to a normally branching ER, extending from the nuclear envelope to the cell periphery (panel B).

We first investigated the rate of decay of VAPB mRNA by RT-qPCR in the wt- and P56S-VAPB HeLa TetOff cells subjected to the ‘pulse-chase’ protocol illustrated in Fig. 2A. As shown in Fig. 2B, 6 h after re-introduction of Dox into the medium, VAPB mRNA levels were reduced by over 80% in both cell lines, with a further small reduction over the next four hours. We then analyzed the decay of the two proteins by immunoblotting. In these experiments, equal numbers of cells were seeded into wells two days before re-introduction of Dox. At each time point, equal aliquots of the entire cell population of one well were analyzed. By this procedure, dilution of VAPB (cell proliferation in the absence of VAPB synthesis) did not contribute to our analysis, as would have been the case if equal amounts of protein had been loaded in the lanes. These experiments showed an initial lag in the clearance of the two proteins, presumably due to translation of residual mRNA at the early time points of the analysis. At later time points (from 6 h on), the analysis revealed that P56S-VAPB is removed from the cells more rapidly than the wt protein (Fig. 2C,D): the level of mutant protein was reduced to ∼50% at 7 h after Dox addition, compared to the ∼17 h required for wt-VAPB to be similarly reduced.

We then analyzed whether P56S remains in the inclusions during the chase period. As shown in Fig. 3A, P56S-VAPB was visualized within inclusions at times as long as 10 h after Dox re-addition, when most of the mutant protein has already been degraded (Fig. 2C,D). At this time, fewer cells had detectable P56S-VAPB inclusions, as expected (21 vs 26% at T₁₀ and T₀, respectively, Fig. 3B). At later time points, the proportion of inclusion-bearing cells further decreased, until essentially complete depletion at 24 h of chase (Fig. 3B).

The state of aggregation of mutant VAPB in comparison with the wt protein was also assessed with a biochemical assay, determining the solubility of the two proteins in Triton X-100 after cell lysis. Remarkably, whereas wt VAPB was recovered only in the Triton-soluble fraction, P56S-VAPB remained largely Triton-insoluble as long as it could be detected (Fig. 3C). A higher proportion of mutant protein was solubilized in the 10 h sample than at earlier time points, possibly because of the lower concentration of the protein in cell lysates and the smaller size of the inclusions at this time. Taken together, the biochemical and the immunofluorescence data indicate that degradation of P56S-VAPB is not preceded by bulk dissolution of the inclusions.

In the Drosophila model, mutant VAP inclusions are stained by anti-ubiquitin (Ub) antibodies (Ratnaparkhi et al., 2008), and biochemical assays have demonstrated that P56S-VAPB is polyubiquitinated (Tsuda et al., 2008). In mammalian cells, mutant VAPB was found to be conjugated to overexpressed HA-tagged Ub (Kanekura et al., 2006), however, by immunofluorescence analysis Ub accumulation in P56S-VAPB inclusions was weak or absent (Chen et al., 2010; Teuling et al., 2007; Tudor et al., 2010). To clarify this discrepancy, we carried out immunofluorescence analysis with anti-polyUb antibodies on induced P56S-VAPB HeLa-TetOff cells. Co-localization of polyUb with the inclusions was apparent (Fig. 4A), although the intensity of polyUb staining was quite variable (compare upper and middle rows of Fig. 4A), perhaps explaining the differences in published results. Inclusions remained positive for polyUb also 10 h after Dox removal (Fig. 4A, lower row). Specificity of the labeling was confirmed by the absence of signal on VAPB-positive inclusions when anti-polyUb primary antibody was omitted from the immunostaining reactions (supplementary material Fig. S2).

To confirm that mutant VAPB is linked to endogenous Ub, we analyzed anti-Ub immunoprecipitates. Lysates prepared from induced cells (T₀) or induced cells 10 h after re-addition of Dox (T₁₀) were subjected to immunoprecipitation with anti-Ub antibodies, and the resulting immunoprecipitates were analyzed by immunoblotting with anti-Myc antibodies to detect P56S-VAPB-Ub conjugates (Fig. 4B). Several Myc-positive bands, which were absent in control samples (lanes 4, 6, and 8), were detected in the anti-Ub immunoprecipitates from both T₀ and T₁₀ cells (lanes 5 and 7), confirming that P56S-VAPB becomes ubiquitinated. In both the T₀ and T₁₀ samples, some non-ubiquitinated P56S-VAPB co-precipitated with the higher M_r species. This observation suggests association between the ubiquitinated and non-ubiquitinated forms, although we cannot exclude that some deubiquitination of VAPB occurred during preparation of the immunoprecipitates for SDS-PAGE.

To follow the timing of P56S-VAPB ubiquitination, we adopted a cDNA microinjection protocol previously used to analyze the time course of P56S-VAPB inclusion formation (Fasana et al., 2010). In accordance with our previous observations, as soon as mutant VAPB became detectable (2 h after microinjection of the corresponding cDNA) it appeared in dispersed small clusters, which later congregated in the juxtanuclear region (Fig. 5A). At 2 and 4 h after microinjection, Ub showed a dispersed punctate pattern, however, close inspection of the cells revealed that even the small P56S-VAPB clusters were positive for polyUb (see insets). At 6 h after microinjection, polyUb concentration on the large juxtanuclear inclusions was much easier to detect. However, quantitative analysis showed that the ratio of polyUb to P56S-VAPB staining remained constant over the entire time course (the values of this ratio are indicated below the merged images of Fig. 5A), indicating that ubiquitination of P56S-VAPB occurs rapidly after its synthesis, and does not increase once clusters have formed. To confirm this interpretation, we stopped protein synthesis with cycloheximide at 6 h after microinjection, and fixed the cells 3 h thereafter. The ratio of polyUb to P56S-VAPB staining actually decreased after this treatment (fourth column of Fig. 5A), in agreement with the hypothesis that P56S-VAPB clusters/inclusions are not accessible to ubiquitin ligases.

The congregation of peripheral clusters of ubiquitinated mutant proteins is characteristic of a particular type of IB, known as aggresome (Kopito, 2000). Aggresomes are formed by transport of peripheral clusters of malfolded ubiquitinated proteins along microtubule tracks to a juxtanuclear region in the vicinity of the microtubule organizing centre. To investigate whether microtubules are involved also in the transport of P56S-VAPB clusters from the periphery to the cell centre, we followed the generation of inclusions in microinjected cells after disruption of the microtubule cytoskeleton by nocodazole. Again, dispersed clusters were seen as soon as the protein was detectable (Fig. 5B, left panel), and concentration in a large juxtanuclear structure was complete in most cells at 6 h after the microinjection (Fig. 5B, third panel), in a manner indistinguishable from that of untreated cells (Fig. 5B, right panel). This result reveals an important mechanistic difference between the genesis of classical aggresomes and of P56S-VAPB inclusions.

In recent years it has become apparent that Ub serves as a signal for the delivery of proteins not only to the proteasome but also to the macroautophagosomal (autophagosomal) degradation pathway (Bjorkoy et al., 2005; Kirkin et al., 2009). We therefore investigated the contribution of the proteasome and of autophagocytosis to P56S-VAPB elimination by the use of specific inhibitors of each of these pathways. The proteasomal inhibitor MG-132 nearly completely blocked degradation of the mutant protein (Fig. 6A,B), with the consequent persistence of ubiquitinated inclusions (Fig. 6B,C). In contrast, treatment of the cells with the autophagocytosis inhibitor 3-Methyladenine (3-MA) had no effect on P56S-VAPB degradation (Fig. 6A,B), although it did inhibit starvation-induced autophagocytosis (supplementary material Fig. S3).

Wild-type VAPB is known to be sequestered within P56S-VAPB IBs, and this sequestration has been suggested to underlie the dominant inheritance of the mutant protein (Kanekura et al., 2006; Suzuki et al., 2009; Teuling et al., 2007). We asked whether the presence of P56S inclusions might accelerate the degradation of the sequestered, endogenous, wt protein. As shown in Fig. 6D, expression of neither wt nor P56S-VAPB had any effect on the steady state levels of the endogenous protein (which has a higher mobility in SDS-PAGE than the Myc-tagged wt and mutant proteins). This result is consistent with the notion that P56S-VAPB degradation proceeds molecule by molecule, and is instead difficult to reconcile with a bulk elimination process mediated by autophagy.

To further investigate the relationship between P56S-VAPB inclusions and autophagocytosis, we compared the distributions of mutant VAPB and the autophagosomal, Ub binding protein p62 (also known as sequestosome) (Bjorkoy et al., 2005; Komatsu et al., 2007; Pankiv et al., 2007). In striking contrast with the observations on IBs that characterize other neurodegenerative diseases (Kuusisto et al., 2001; Nagaoka et al., 2004; Zatloukal et al., 2002), we detected no colocalization between mutant VAPB inclusions and p62 (Fig. 6E, upper row). We also analyzed cells in which autophagy was induced by serum starvation (Fig. 6E, lower row). As expected, starvation caused a reduction in the overall intensity of p62 staining, consistent with its enhanced degradation in this condition (Bjorkoy et al., 2005; Komatsu et al., 2007), while p62-positive puncta, corresponding to autophagosomes/autolysosomes, became more apparent. Also in these serum-starved cells, we found no co-localization between VAPB and p62, although, quite strikingly, p62-positive structures were often closely juxtaposed to the VAPB inclusions. Similar results were obtained with the autophagosomal marker LC3 (Kabeya et al., 2000) (supplementary material Fig. S4).

The 43 kDa TAR DNA-binding protein (TDP-43) has been found as a major component of ubiquitinated, p62-positive, protein aggregates in ALS patients (Ince et al., 2011). We therefore investigated whether such aggregates are present in our induced cells, and, if so, whether they coincide with the P56S-VAPB inclusions. Our analysis failed to reveal any dislocation of TDP-43 from its normal nuclear localization to cytoplasmic inclusions (supplementary material Fig. S5, top row). Instead, similar to our observations on transiently transfected cells (Fasana et al., 2010), we observed accumulation in the inclusions of Bap31, an ER protein that has been implicated in chaperoning ER membrane proteins to ERAD (Wakana et al., 2008) (supplementary material Fig. S5, bottom row).

Since VAPB is an ER transmembrane (tail-anchored) protein, and the inclusions that it generates consist of restructured ER domains, we reasoned that its delivery to the proteasome must involve extraction not only from the aggregates, but also from the ER membrane. We therefore investigated a possible involvement of the AAA chaperone, p97 (also known as valosin-containing protein, VCP), an ATPase that extracts ERAD substrates from the ER to deliver them to the proteasome (Rabinovich et al., 2002; Ye et al., 2001). To analyze the role of p97/VCP in P56S-VAPB degradation, induced P56S-VAPB TetOff cells were transfected with cDNAs coding either for the wt or a mutant (QQ), ATPase-deficient, p97 (Ye et al., 2001). The QQ mutant has been shown to act as a dominant negative (DN), thereby inhibiting substrate extraction from the ER (Ye et al., 2001). As shown in Fig. 7, in cells transfected with wt-p97, P56S-VAPB was degraded with the same time course as in the non-transfected cells. In contrast, transfected DNp97 stabilized the mutant protein, more than doubling its concentration at T₁₀ relative to T₀ in comparison to the control cells.

Since its discovery (Nishimura et al., 2004), FALS-linked P56S-VAPB was observed to form inclusions in cultured cells that appeared similar to those generated by mutant proteins underlying other neurodegenerative diseases such as polyQ expanded Huntingtin in Huntington's disease, mutated synuclein in Parkinson's disease, and neurofibrillary tangles in Alzheimer's disease (Ross and Poirier, 2004). Further investigations, however, revealed important differences between P56S-VAPB inclusions and other IBs. More specifically, mutant VAPB inclusions correspond to a profoundly restructured ER domain and not to a cytosolic protein aggregate, as is the case for most IBs. Our previous work (Fasana et al., 2010) demonstrated that P56S-VAPB, like the wt protein, inserts into the ER membrane via its C-terminal tail, but that after insertion it rapidly clusters, generating paired ER cisternae apparently held together by the interactions between the mutant cytosolic domains (Fasana et al., 2010). We classified these restructured ER domains as a particular form of organized smooth ER (OSER) (Snapp et al., 2003), which consists in different varieties of organized arrays of stacked smooth cisternae. Differently from previously described OSER, however, the stacks of P56S-VAPB inclusions do not form multi-layered arrays; rather, a limited number (two to three) of stacked cisternae form a ribbon that undergoes a convoluted, tortuous pathway, as confirmed here by tomographic analysis.

In the present study, we have uncovered additional interesting characteristics of P56S-VAPB inclusions that set them apart from other IBs. First, although the inclusions contain polyubiquitinated P56S-VAPB and are located in a juxtanuclear position, they do not require the microtubule cytoskeleton to attain this location; and, second, the inclusions are cleared rapidly by the proteasome by a process requiring the AAA ATPase p97. We discuss these two findings in turn.

Microtubules, associated with negative end directed molecular motors, are involved in transport of cellular components from the periphery to the juxtanuclear region. In particular, the aggresome, a particular type of IB, is generated by transport of peripheral clusters of malfolded proteins along microtubule tracks (Kopito, 2000). The proteins in the peripheral clusters are usually polyubiquitinated, and the deacetylase HDAC6 serves as adaptor between ubiquitin and the dynein motor (Kawaguchi et al., 2003). Given the generality of these findings, we expected to find a role for microtubules in the congregation of P56S-VAPB inclusions. In contrast to the previously investigated protein inclusions, however, disruption of the microtubule cytoskeleton was without effect on the localization of the mutant protein. This negative result, while uncovering another difference between mutant VAPB inclusions and well characterized IBs linked to neurodegeneration, leaves the mechanism of congregation of P56S-VAPB inclusions unresolved. We note, however, that in cultured mammalian cells ER cisternae are enriched in the paranuclear region (Puhka et al., 2007), and suggest that simple geometrical constraints could favour entrapment of both normal, ribosome-covered cisternae and P56S-VAPB paired cisternae in the central, thicker and less dynamic portions of the cell.

The observation that P56S-VAPB is cleared from cells more rapidly than its wt counterpart was also unexpected. Indeed, although many mutant proteins are destabilized, large aggregates are thought to be thermodynamically stable (Fink, 1998). Even more surprising was the involvement of the proteasome and the lack of involvement of the autophagy pathway, as indicated by the near complete inhibition of P56S-VAPB degradation by the proteasomal inhibitor MG-132, by the lack of effect of the autophagy inhibitor 3-MA, and by the absence in the inclusions of LC3 and p62, a Ub receptor that is thought to mediate the selective autophagy of ubiquitinated protein aggregates (Kirkin et al., 2009; Komatsu et al., 2007; Pankiv et al., 2007). Our microinjection experiment demonstrated that mutant VAPB becomes polyubiquitinated shortly after its synthesis; nonetheless at least a part of the ubiquitinated protein initially avoids the proteasome, forming the restructured ER inclusions that are only subsequently cleared from the cell. Thus, the rapidity of the aggregation process may initially protect mutant VAPB from proteasomal attack.

A number of IBs are known to be cleared by autophagocytosis, including those formed by polyQ or polyA expanded proteins (Iwata et al., 2005; Ravikumar et al., 2002), and by synuclein (Olzmann and Chin, 2008; Webb et al., 2003). Autophagocytosis also plays a role in the elimination of inclusions generated by mutant cystic fibrosis transmembrane conductance regulator (CFTR) (Fu and Sztul, 2009), and of organized smooth ER (Lingwood et al., 2009). Thus, P56S-VAPB inclusions appear to stand apart from other IBs, both in the mechanism of their genesis and of their clearance from the cell.

Delivery of unfolded/malfolded ER proteins to the proteasome involves a complex process known as ERAD (Bernasconi and Molinari, 2011). A keyplayer in ERAD is the AAA-ATPase p97/VCP, which utilizes ATP hydrolysis to pull substrates out of or across the ER membrane into the cytosol (Rabinovich et al., 2002; Ye et al., 2001). Our results with a dominant negative form of p97/VCP (QQ-p97) indicate that delivery of P56S-VAPB to the proteasome requires the activity of this ATPase; however, at present we do not know its exact role in the process. By fluorescence recovery after photobleaching (FRAP), P56S-VAPB appears to be immobilized within the inclusions that it generates (Teuling et al., 2007). Thus, p97/VCP may be required not only to pull the transmembrane domain out of the lipid bilayer but also to dissolve the protein-protein interactions underlying inclusion formation. We cannot, however, exclude that mutant VAPB molecules slowly exchange with the surrounding ER at a rate too low to be detected by the FRAP experiments of Teuling et al. (Teuling et al., 2007). In this case, p97/VCP would be acting as a trap, extracting single P56S-VAPB molecules from the bilayer as they gradually become available.

Dominant inheritance, as is the case for the P56S-VAPB mutation, can be explained by gain of function (toxicity of the altered gene product) and/or loss of function (negative dominance, and/or haploinsufficiency) or by combinations of these mechanisms. It has recently been reported that overexpressed P56S-VAPB may exert a cytotoxic effect by inhibiting the proteasome (Moumen et al., 2011); however, in that study, the wild-type and mutant protein had similar effects. Thus, any toxic effect specifically attributable to P56S-VAPB inclusions remains to be demonstrated. Instead, the finding that wt VAPB (and to a lesser extent its homologue VAPA) is sequestered within mutant VAPB inclusions (Kanekura et al., 2006; Suzuki et al., 2009; Teuling et al., 2007), has favoured the hypothesis that negative dominance underlies the pathogenicity of P56S-VAPB. This hypothesis predicts that expression of the mutant protein in the background of a normal VAPB genotype should be sufficient to induce ALS pathology. In fly models, this prediction is borne out (Chai et al., 2008; Ratnaparkhi et al., 2008), however, the situation in mammals is less clear. While heavy overexpression of P56S-VAPB, but not of the wt protein, does increase cell death of primary neurons and neuronal cell lines (Suzuki and Matsuoka, 2011; Teuling et al., 2007), P56S-VAPB transgenic mice do not develop any motor deficit (Tudor et al., 2010). Furthermore, and importantly, a recent study failed to detect VAPB inclusions in motor neurons derived from induced pluripotent stem cells of patients carrying the P56S-VAPB mutation (ALS8 patients), but did reveal reduced levels of VAPB within these cells (Mitne-Neto et al., 2011). These findings are fully consistent with our results demonstrating the relatively rapid clearance of mutant P56S-VAPB inclusions.

On the basis of our results, and of the studies on transgenic mice (Tudor et al., 2010) and cells from human patients (Mitne-Neto et al., 2011), we suggest that haploinsufficiency may be the principal mechanism underlying P56S-VAPB pathogenicity. Of note, two studies have reported a reduction in VAP levels in spinal cord motor neurons of sporadic ALS patients and of transgenic mice (G93A superoxide dismutase-1) with ALS pathology (Anagnostou et al., 2010; Teuling et al., 2007), suggesting that the reduction of VAPB levels, caused either by mutation or by other unknown causes, may play a general role in the pathogenesis of both the sporadic and the familial forms of ALS.

VAPB has been implicated in a variety of functions, among which interorganellar contacts and non-vesicular transport of lipids between the ER and other organelles (Kawano et al., 2006; Loewen and Levine, 2005; Peretti et al., 2008; Stefan et al., 2011), organelle-microtubule interactions (Amarilio et al., 2005; Rocha et al., 2009), neurite extension (Saita et al., 2009), vesicular trafficking (Foster et al., 2000; Skehel et al., 1995; Soussan et al., 1999), and the Unfolded Protein Response (Chen et al., 2010; Gkogkas et al., 2008; Kanekura et al., 2006; Suzuki et al., 2009). At present, one can only make educated guesses as to which of these processes is/are impaired in cells with reduced levels of VAPB. Identification of the VAPB functions that require homozygous levels of the protein to maintain motor neurons in good health will be an important milestone towards unravelling the factors underlying sporadic and familial ALS.

The pCAN-myc-P56S-VAPB plasmid, coding for Myc-tagged human P56S-VAPB and used for expression in transiently transfected and microinjected cells, has been described (Fasana et al., 2010). For inducible expression of wt VAPB in HeLa-TetOff cells, a cDNA fragment coding for Myc-VAPB (wt), a gift from Sima Lev (Weizmann Institute of Science, Rehovot, Israel) was subcloned into the pTre tight vector (Clontech). Plasmids pCDNA3-His-p97 (wt and dominant negative QQ mutant) and pGFP-LC3 (Kabeya et al., 2000) were gifts from Yinhong Ye (NIH/NIDDH) and Sharon Tooze (Cancer Research UK), respectively.

To visualize Myc-tagged VAPB by immunofluorescence or immunoblotting, we generally used anti-Myc monoclonals (clone 9E10 from Santa Cruz or Sigma). In some cases it was necessary to visualize VAPB with polyclonal antibodies; in these cases, we used either a polyclonal anti-Myc (from Cell Signaling) or an anti-VAPB antibody donated by Casper C. Hoogenraad (Utrecht University). Anti-polyUb (clone FK1) and anti-mono and polyUb (clone FK2), used for immunofluorescence and for immunoprecipitation, respectively, were from BioMol. Anti-TDP43 and anti-tubulin polyclonals were from Sigma. Anti-HA was from Santa Cruz, and anti-p62 was obtained from AbCam. RSG-His antibody was from Qiagen. An anti-Bap31 polyclonal was a gift from Mitsuo Tagaya (Tokyo University of Pharmacy and Life Sciences).

Alexa Fluor 488 and 568 secondary antibodies were from Invitrogen. Rhodamine-conjugated anti-mouse IgM antibodies were from Santa Cruz. Peroxidase-conjugated anti-rabbit and anti-mouse IgG were from Sigma, and light chain-specific anti-rabbit IgG was from Jackson ImmunoResearch.

HeLa-TetOff clones expressing Myc-tagged wt VAPB were selected, maintained in culture and induced to express the transgene by growth in the absence of Dox for 5–6 days as previously described for the Myc-tagged P56S-VAPB clones (Fasana et al., 2010). For both the wt and the mutant VAPB two separate clones were used, with similar results. Transfection of P56S-VAPB TetOff cells with the indicated plasmids was carried out by the CaPO₄ method 24 h before harvesting or fixing the induced or non-induced cells.

Non-transfected HeLa cells were maintained in culture and microinjected or transfected with pCAN-myc-P56S-VAPB, as previously described (Fasana et al., 2010).

Equal numbers (∼400,000) of Myc-tagged-wt or P56S-VAPB (referred to simply as wt or P56S-VAPB) HeLa TetOff cells, grown for 4 days in the absence of Dox, were seeded in 60 mm Petri dishes, each containing a coverslip. After 48 h in the absence of the Dox, the coverslips were removed and stained with DAPI; the DAPI-stained nuclei were counted in order to check that each dish contained the same number of cells. Cells from the dish corresponding to the first time point (T₀) were collected and lysed with 0.5 ml pre-warmed (80°C) lysis buffer [2% SDS, 50 mM Tris-HCl, pH 8, plus Complete (Roche) protease inhibitors]. The other dishes were supplemented with Dox (0.5 µg/ml) and returned to the incubator; cells were harvested at the indicated time points with the same volume of lysis buffer as the T₀ cells. Equal volumes of the lysates were analyzed by SDS-PAGE-immunoblotting.

SDS-PAGE and blotting were carried out by standard procedures. Blots were incubated with primary antibodies diluted in Tris-saline, 5% milk, 0.1% Tween, unless alternative conditions were recommended by the supplier. After incubation with peroxidase-conjugated secondary antibodies, the blots were developed with ECL-Plus or ECL-Ultra (GE Healthcare) reagents. The films were digitized, and band intensities were determined with ImageJ software (National Institutes of Health) after calibration with the optical density calibration step table (Stouffer Graphics Arts).

Cells grown to 80% confluence in two 10 mm Petri dish were harvested and resuspended in 0.6 ml of a hypotonic solution, containing 1 mM Tris-HCl, pH 7.4, 15 mM KCl, 0.1 mM EDTA, Complete (Roche) protease inhibitors, and incubated on ice for 5 min. After homogenizing the samples by repeated passages through an insulin syringe, an equal volume of a solution containing 2 mM Tris-HCl, pH 7.4, 0.5 M sucrose, 0.2 mM EDTA, and protease inhibitors was added to restore isotonicity. The samples were centrifuged at 200,000 g at 4°C for 1 h, and the supernatant, containing cytosolic proteins, was discarded. The pellets were lysed with 1 ml of a pre-warmed buffer (80°C) containing 1% SDS, 50 mM Tris-HCl, pH 8, plus protease inhibitors. After shearing the released DNA by repeated freeze-thawing and passage through an insulin syringe, nine volumes of a buffer containing 0.140 M NaCl, 50 mM Tris-HCl, pH 7.5, 5 mM EDTA, pH 8, 2.2% Triton X-100 was added. After removing any insoluble material by centrifugation at 12,000 g for 10 min, a small aliquot of the sample was kept to analyze the input, while the rest was subjected to immunoprecipitation with anti-Ub monoclonals or unrelated mouse IgGs. Immunocomplexes were collected with agarose-Protein G beads and analyzed by immunoblotting with anti-Myc polyclonal antibodies.

All solutions contained Complete (Roche) protease inhibitors. Cells grown to 80% confluence on 60 mm plates were harvested and resuspended with 0.25 ml of a solution containing 10 mM Tris-HCl, pH 7.4, 1 mM EDTA. The samples were then lysed with an equal volume of 2% Triton X-100 in the same Tris-EDTA buffer. After 30 min on ice, 100 µl were removed for the analysis of the total sample, and the remaining lysates were centrifuged at 12,000 g for 10 min. After withdrawal of the supernatants, the pellets were resuspended in 200 µl of 10 mM Tris-HCl, pH 7.4, and treated with DNase (40 µg/ml final concentration) to digest any released DNA. The resuspended pellets were then brought to 400 µl, and equal aliquots of the total samples, pellets and supernatants were analyzed by SDS-PAGE/immunoblotting.

Total RNA was extracted using the RNeasy Mini Kit and accompanying QIAshredder (Qiagen) according to the manufacturer's instructions. Contaminating DNA in the sample was degraded by on-column incubation with DNaseI (Qiagen) for 15 min. 1 µg per sample was reverse-transcribed using the SuperScript III First- Strand Synthesis System for RT-PCR (Invitrogen) in accordance with the manufacturer's instructions. P56S-VAPB mRNA levels were quantitatively analyzed by the SYBR green method, with forward primer (specific for the Myc-tagged sequence) 5′-AGCTGATCTCCGAGGAGGACCTG-3′, and reverse primer 5′-TTTCGGTCTGTCGGGTTGCC-3′. The size of the PCR products was confirmed by gel electrophoresis. Amplification was carried out with the ABI Prism 7000 Sequence Detection System and SDS software version 1.2.3 (Applied Biosystems). The cycling conditions were one cycle of denaturation at 95°C for 10 min, followed by 40 two-segment cycles of amplification (95°C for 15 s and 60°C for 1 min). At the end of the cycles, a dissociation protocol (95°C for 15 s, 60°C for 20 s, and 95°C for 15 s) was run to verify the presence of one gene-specific peak and the absence of primer-dimers. All samples were amplified in triplicate. The 2^−ΔΔCT method was used to calculate the results, which were normalized to GAPDH mRNA, used as internal control (forward primer, 5′-TCGGAGTCAACGGATTTGG-3′, reverse primer, 5′-TGGCAACAATATCCACTTTACCA-3′; Gen-Bank accession no. NM_002046.3).

3-MA, dissolved in water, was used at a final concentration of 10 mM. Cycloheximide was dissolved in PBS and used at a final concentration of 50 µg/ml. Nocodazole and MG-132 were dissolved in DMSO and used at final concentrations of 3 and 10 µg/ml respectively. Control cells received equal volumes of the vector.

Cells grown on coverslips were fixed with 4% paraformaldehyde and processed for immunofluorescence, as described previously, with minor modifications (Fasana et al., 2010), using Alexa- or Rhodamine-conjugated secondary antibodies (Invitrogen). To visualize GFP-LC3 together with P56S-VAPB, paraformaldehyde-fixed cells were permeabilized with methanol instead of Triton, and antibodies were diluted in PBS + 0.1% bovine serum albumin. Cells were imaged with the Zeiss LSM 510 Meta confocal system (Carl Zeiss, Oberkochen, Germany) using a 63× PlanApo lens. Alexa Fluor 488 and GFP were imaged with the use of the 488 line of the Argon/2 laser, a 405/488/543/633 dichroic, and a 505–550 band pass emission filter. For Alexa 568 and Rhodamine, the 544 line of the He/Ne laser was used, with the same dichroic as above, and a 560 longpass emission filter. Wide-field imaging was performed with an Axioplan microscope (Carl Zeiss, Oberkochen, Germany), using the 40× PlanNeofluar lens equipped with a phase contrast ring. Illustrations were prepared with Adobe Photoshop software.

To determine the ratio of polyUb to P56S-VAPB staining, series of images (10 for each condition) were acquired all with identical parameters, taking care to remain below saturation in both the Ub (Alexa 568) and the VAPB (Alexa 488) channels. Each image was segmented in the VAPB channel, and a mask corresponding to VAPB staining was used to determine the mean grey value in both channels with the use of ImageJ software.

Cells were fixed as a monolayer in 2% glutaraldehyde, 0.1 M cacodylate buffer for 30 min at room temperature and further processed for transmission EM as previously described (Fasana et al., 2010; Snapp et al., 2003). Epon thin sections were observed with a Philips CM10 transmission electron microscope (Philips, Amsterdam, The Netherlands) equipped with a Morada EM CCD-camera (from Olympus SIS). Digital images were acquired using the iTEM software (from Olympus-SIS).

STEM-HAAD (High-Angular Annular Dark-Field) tomography was performed with a JEOL JEM 2200 FS EM operating at 200 kV, using a spot size of 1 nm and the smallest condenser aperture resulting in a collection angle of 3 mrad. The sample (300 nm thick serial sections) was mounted on a Fischione Ultra-Narrow Gap Tomography Holder and tilted over a ± 60 range every 2 degrees or according to a ‘Saxton scheme’. Images were acquired using a HAADF detector at a magnification of 40,000×, resulting in a pixel resolution of 7.93 nm.

Serial section tomograms were reconstructed and joined using the IMOD software package (Kremer et al., 1996; Mastronarde, 1997). A tomogram sub-volume was automatically segmented using a K-means clustering algorithm implemented with MATLAB. The binary output was Gaussian filtered to reduce noise and rendered using UCSF Chimera package. Alternatively, selected structures of the tomogram were rendered by manual segmentation using IMOD.

In addition to the people who kindly supplied reagents (listed in Materials and Methods), we acknowledge the Monzino Foundation (Milan, Italy), for its generous gift of the Zeiss LSM 510 Meta confocal microscope. We thank Paola Genevini for her valuable help in the experiments illustrated in Figs 5-7.