|Home | About | Journals | Submit | Contact Us | Français|
The pathways leading from aberrant Prion protein (PrP) metabolism to neurodegeneration are poorly understood. Some familial PrP mutants generate increased CtmPrP, a transmembrane isoform associated with disease. In other disease situations, a potentially toxic cytosolic form (termed cyPrP) might be produced. However, the mechanisms by which CtmPrP or cyPrP cause selective neuronal dysfunction are unknown. Here, we show that both CtmPrP and cyPrP can interact with and disrupt the function of Mahogunin (Mgrn), a cytosolic ubiquitin ligase whose loss causes spongiform neurodegeneration. Cultured cells and transgenic mice expressing either CtmPrP-producing mutants or cyPrP partially phenocopy Mgrn depletion, displaying aberrant lysosomal morphology and loss of Mgrn in selected brain regions. These effects were rescued by either Mgrn overexpression, competition for PrP binding sites, or preventing cytosolic exposure of PrP. Thus, transient or partial exposure of PrP to the cytosol leads to inappropriate Mgrn sequestration that contributes to neuronal dysfunction and disease.
Mammalian Prion protein (PrP) is a cell surface GPI-linked glycoprotein implicated in several neurodegenerative diseases including scrapie, bovine spongiform encephalopathy, Creutzfeldt-Jakob disease, and Gerstmann-Straussler-Scheinker disease (Aguzzi et al., 2007; Collinge and Clarke, 2007). The most extensively studied aspect of these diseases is their transmissibility via an unusual agent (termed prion) composed largely, if not exclusively, of a misfolded isoform of PrP termed PrPSc. Prion propogation is thought to occur when PrPSc converts the normal cellular form of PrP (PrPC) into additional copies of PrPSc. Although this explains how altered protein conformation can form the basis of disease transmission, relatively little is known about the pathways of cellular dysfunction that culminate in neurodegeneration in PrP-associated diseases.
Because PrPSc is highly insoluble and aggregation-prone, it was long assumed that its accumulation in the central nervous system would be intrinsically harmful to neurons. However, this view appears to be overly simplistic since several experimental paradigms have partially or fully uncoupled PrP aggregate deposition from downstream neuropathology (Brandner et al., 1996; Mallucci et al., 2003; Chesebro et al., 2005). Conversely, several familial PrP mutations cause neurodegeneration with little or no generation of PrPSc or transmissible agent (Tateishi and Kitamoto, 1995; Tateishi et al., 1996; Chiesa et al., 2003). These and other observations suggest that neurodegeneration might involve different aspects of PrP metabolism beyond just PrPSc accumulation (Hetz and Soto, 2006; Chakrabarti et al., 2009), prompting investigation into other isoforms of PrP that might mediate neurotoxicity.
One minor isoform of PrP, termed CtmPrP, spans the membrane once [at a hydrophobic domain (HD) from residues ~112–135] with the N-terminal domain exposed to the cytosol (Hegde et al., 1998). Remarkably, both a natural and several artificial mutants within the HD that lead to even modestly increased generation of CtmPrP (between 5–20% of total PrP) cause neurodegeneration in transgenic mice (Hegde et al., 1998; Hegde et al., 1999). Furthermore, several familial diseases in humans are associated with hydrophobicity-increasing mutations in the HD (e.g., A117V; Hsiao et al., 1991) that may increase CtmPrP generation (Hegde et al., 1998). Indirect evidence in transgenic mice suggests that CtmPrP levels might also be increased (or perhaps stabilized from degradation) upon PrPSc accumulation (Hegde et al., 1999). Thus, at least a subset of familial neurodegenerative diseases, and perhaps also PrPSc-mediated transmissible disease, are associated with generation of CtmPrP.
In separate studies, a small proportion of PrP was found to be degraded in the cytosol by the proteaseome (Yedidia et al., 2001; Ma and Lindquist, 2001). The observation that improving the efficiency of the PrP signal sequence markedly reduces the proportion of PrP degraded by the proteasome suggested that inefficient forward translocation into the ER is a major source of cyPrP (Rane et al., 2004). Interestingly, enforced cyPrP expression in transgenic mice caused neurodegeneration in a cell-type selective manner (Ma et al., 2002). However, the relevance of this observation to either familial or transmissible disease caused by PrP has been unclear.
More recently, several observations have suggested an indirect means to potentially link cyPrP production to prion disease pathogenesis. First, translocation of PrP into the ER is reduced during ER stress (Kang et al., 2006; Orsi et al., 2006), leading to increased cyPrP production. Second, ER stress appears to be an indirect consequence of prion infection and PrPSc accumulation (Hetz and Soto, 2006; Rane et al., 2008). Third, reduced PrP translocation at levels comparable to that seen during ER stress was sufficient to cause mild age-dependent neurologic dysfunction in transgenic mice despite essentially quantitative degradation of cyPrP (Rane et al., 2008). And finally, proteasome activity may decline with age (Dahlmann, 2007) and upon PrPSc accumulation (Kristiansen et al., 2007). Thus it is plausible that by the combined effects of a weak PrP signal sequence, reduced PrP translocation during ER stress, and reduced proteasome activity upon PrPSc accumulation, cyPrP is generated in sufficient amounts during prion disease to be a contributing factor in neurodegeneration (Rane et al., 2008).
And finally, PrPSc was shown to directly inhibit the proteasome in vitro (Kristiansen et al., 2007). Because proteasome activity was observed to be decreased with prion infection in cells and mice, it was proposed that cytosolic PrPSc inhibits the proteasome to cause neuronal death during disease pathogenesis. While it is not yet clear how PrPSc (normally formed in extracellular or endo-lysosomal compartments) could access the cytosol, its cytosolic mislocalization was a key point of this model. Thus, one theme that emerges from the above paradigms of neurodegeneration is the exposure of PrP to the cytosolic environment. Although only partially or very transiently exposed, this minor population of PrP could conceivably have adverse consequences for certain cells under certain conditions if it were to make inappropriate interactions with cellular factors whose functions become compromised. However, candidate interacting partners for cytosolic PrP are poorly studied and their roles in disease unknown.
In the context of this hypothesis, the cytosolic protein Mahogunin (Mgrn) is especially intriguing. Cloned as the gene causing the mahoganoid coat color phenotype, mice containing a homozygous loss of Mgrn function were discovered to show late-onset spongiform neurodegeneration in selected brain regions (He et al., 2003). The resemblance of Mgrn and prion disease pathology raised the possibility of a mechanistic relationship. However, a functional connection between Mgrn and PrP was not immediately apparent. Although Mgrn has E3 ubiquitin ligase activity, PrP is not a substrate in vitro and does not accumulate in vivo in the absence of Mgrn (He et al., 2003). Nevertheless, a growing appreciation that minor populations of PrP are either partially (in the case of CtmPrP) or transiently (in the case of cyPrP) exposed to the cytosol during disease led us to consider the alternative hypothesis that cytosolically mislocalized forms of PrP might interact inappropriately with Mgrn to inhibit its function. This would phenocopy Mgrn depletion, leading to region-selective neurodegeneration. Here, we examine this hypothesis using in vitro, cell culture, and mouse models.
Under normal circumstances, the amounts of CtmPrP and cyPrP are minor and often transient. Even mutations that favor production of these isoforms result in modest increases that, while relevant for disease over long time periods in certain cell types, nonetheless make analysis of potential protein-protein interactions daunting. To circumvent this problem, we initially explored the possibility of an interaction between Mgrn and cytosolically exposed PrP using artificial systems that greatly exaggerate their abundance and stability. This strategy allowed the evaluation of potential interactions, mapping of interacting domains, characterization of downstream phenotypes, and detailed functional analysis in a robust experimental system. The physiologic relevance of the results from such exaggerated systems was validated subsequently in cellular and mouse models that more accurately reflect the disease state.
Expression of PrP in the cytosol leads to its rapid degradation by the ubiquitin-proteasome system. Degradation ensues regardless of whether cytosolic PrP is generated by mutation, by deletion of the signal, or by inhibitors of translocation (Ma and Lindquist, 2002; Kang et al., 2006). The very low steady state levels of cytosolic PrP therefore makes it difficult to assess a potential interaction with Mgrn in vivo without proteasome inhibitors that could have many indirect effects. To avoid this, we took advantage of the serendipitous observation that fluorescent protein (FP) tagged PrP lacking the N- and C-terminal signals is poorly degraded and artifactually forms aggregates in nearly all cells (Sup. Fig. S1). Such aggregates remained affixed in the cell upon selective release of freely diffusible cytosolic contents by digitonin-mediated semi-permeabilization of the plasma membrane. We exploited these observations to develop an in vivo interaction assay based on co-association of a FP-tagged test protein with FP-tagged cytosolic PrP aggregates (Fig. 1a).
Co-expression of RFP with CFP-PrP40–231 (CFP fused to residues 40–231 of PrP) followed by digitonin permeabilization, led to a rapid and essentially complete loss of RFP signal (within ~2–5 min) from the nucleo-cytoplasmic compartment (Fig. 1b; Sup. Fig. S1). By contrast, RFP-Mgrn was partially retained in the cell upon permeabilization, co-localizing precisely with aggregates formed by CFP-PrP40–231 (Fig. 1c; Sup. Fig. S1). Co-aggregation was seen with PrP and Mgrn regardless of the FP tags used (we have used CFP, GFP, and RFP in various combinations), in cells with widely varying expression levels of Mgrn (spanning at least 20-fold), and with aggregates of various sizes and morphology (unpublished observations). Evidence for an interaction between RFP-Mgrn and CFP-PrP40–231 could also be observed without permeabilization, especially in cells where the RFP-Mgrn was expressed at lower levels and the non-aggregated population did not confound the imaging (Fig. 1c; Sup. Fig. S1). Furthermore, the observation that Mgrn-RFP was typically retained in co-association with the aggregate over an hour after permeabilization (unpublished observations) suggests that its sequestration was not rapidly reversible. Importantly, Mgrn sequestration was specific to PrP aggregates, since aggregates formed by a GFP-tagged Huntingtin (Htt) fragment containing 103 glutamines failed to co-associate with RFP-Mgrn (Fig. 1d). Thus, mis-localized PrP (artificially immobilized into aggregates in this case) can interact selectively with Mgrn in cultured cells.
We applied the same semi-permeabilization assay to also illustrate an interaction between Mgrn CtmPrP. We first generated and characterized cell lines expressing FP-tagged wild type PrP or SA-PrP, a construct made exclusively in the CtmPrP form (Sup. Fig. S2). RFP-Mgrn expressed in these cells was then analyzed before and after digitonin semi-permeabilization. While RFP-Mgrn was fully extracted from cells expressing wild type PrP, it was significantly retained in the SA-PrP cells (Fig. 1e). The retained RFP-Mgrn decorated the plasma membrane and intracellular membranous structures containing SA-PrP. Thus, PrP forced into the CtmPrP topology permits an interaction with Mgrn, whereas wild type PrP does not (presumably a consequence of its lack of exposure to the cytosol).
The key interactions observed between Mgrn and PrP could also be demonstrated biochemically in two ways. In the first experiment, PrP from a detergent-solubilized crude brain lysate could be pulled down more efficiently by immobilized recombinant Mgrn than the immobilized BSA control (Fig. 1f). In the second experiment, we co-expressed RFP-PrP aggregates with different GFP-Mgrn constructs lacking either the N-terminus, C-terminus, or RING domain. The cells were then separated into soluble cytosol, wash, and insoluble fractions. Immunoblotting revealed that essentially all of the RFP-PrP was found in the insoluble fraction, consistent with its predominantly aggregated status seen visually (Fig. 1g). Significant GFP-Mgrn, -MgrnΔC, and -MgrnΔR were also recovered with the insoluble fraction, while noticeably less GFP-MgrnΔN was recovered (Fig. 1g, top panel). Importantly, none of the Mgrn constructs were seen in the insoluble fraction in cells lacking PrP aggregates (Fig. 1g, bottom panel). Thus, cytosolically exposed PrP can interact with Mgrn (via its N-terminus; see next section). This interaction is not normally seen with wild type PrP (even though it is capable of interacting; Fig. 1f), presumably because the two proteins are in distinct compartments separated by a membrane barrier.
Serial truncations of the cytosolic GFP-PrP construct (all of which formed cytosolic aggregates; Sup. Fig. S3) combined with the digitonin co-aggregation assay allowed us to map the key region of PrP interacting with Mgrn (Fig. 2a). Interaction was abruptly lost upon deletion from residue 84 to 95, when the last of four identical octapeptide repeats (ORs) is removed from the construct. An immobilized synthetic peptide encoding the OR sequence (PHGGGWGQ) could pull-down Mgrn, but not MgrnΔN or GFP, from the cytosol of cells co-expressing these proteins (Fig. 2b). Neither Mgrn nor FP-Mgrn were captured by control beads lacking peptide (Fig. 2c) or beads conjugated with irrelevant proteins (such as Protein A or Conconavalin A; data not shown). Deletion constructs of Mgrn showed that the N-terminus, in particular the region between residues 199 to 251, was involved in the interaction with cytosolic PrP aggregates (Fig. 2a; Sup. Fig. S4a; Fig. 1g). Importantly, Mgrn need not be functional for this interaction since a construct lacking the RING domain (termed MgrnΔR) still interacts with PrP. Appending only residues 200–250 of Mgrn to a FP was sufficient to allow interaction with cytosolic PrP aggregates (Sup. Fig. S4b). Thus, this 50-residue domain within the N-terminal half of Mgrn interacts with the 8-residue OR sequence, four of which are present in the N-terminal half of PrP.
The observation that overexpressed FP-Mgrn can interact with cytosolically exposed PrP raised the possibility that PrP could similarly sequester endogenous Mgrn to affect its function. Although the functional role or physiologic substrates of Mgrn are not known, its depletion by siRNA was shown to affect lysosomal morphology (Kim et al., 2007; see Fig. 3b). We therefore used altered lysosome morphology as a phenotypic readout of functional Mgrn depletion to ask whether cytosolically exposed PrP would sufficiently influence endogenous Mgrn localization to at least partially phenocopy a Mgrn depletion.
Antibodies selective to Mgrn (Sup. Fig. S5a) revealed that unlike overexpressed Mgrn, endogenous Mgrn is localized in widely distributed puncta (Sup. Fig. S5b) that partially co-localize with markers of the endo-lysosomal system (Kim et al., 2007; data not shown). Upon cytosolic GFP-PrP40–231 expression, Mgrn localization was altered, with clear co-sequestration of at least some Mgrn around the most prominent PrP aggregates (Fig. 3a). It should also be noted that the re-distributed Mgrn that is not with the PrP aggregate also seems to co-localize with PrP that is not visible at this detector gain but is found in either dimmer aggregates or is diffusely cytosolic (unpublished observations). This re-distribution was specific to cytosolic PrP aggregates since neither wild type PrP, Htt aggregates, or GFP-PrP95–231 aggregates (lacking the octapeptide repeat domains) caused any noticeable changes in Mgrn localization (Fig. 3a and data not shown). Because endogenous Mgrn appears to normally be associated tightly with membranes, we could not use selective digitonin extraction to biochemically separate PrP aggregate-associated Mgrn from normal Mgrn. That notwithstanding, the striking correlation between altered endogenous Mgrn localization and PrP constructs that in independent experiments interact with overexpressed FP-Mgrn (e.g., Fig. 1 and and2)2) argues strongly for a physical sequestration and/or redistribution by cytosolically exposed PrP.
To assess whether this re-distribution might affect Mgrn function, the lysosomal morphology and distribution in these cells were visualized with Lysotracker. A change in lysosomal appearance toward larger structures (either larger lysosomes, or possibly clustering) was noted in cells expressing GFP-PrP40–231 aggregates, but not GFP, GFP-Htt aggregates, or GFP-PrP95–231 (Fig. 3c and data not shown). A histogram of the diameters of lysotracker-stained structures (Fig. 3d) showed a clear shift in size: while only ~5–10% of lysosomal structures were 0.8 um or larger in control cells, such enlarged structures represented up to ~50% of staining particles in GFP-PrP40–231 aggregate-containing cells. Importantly, the distribution of lysosomal sizes in both Htt and GFP-PrP95–231 aggregate-containing cells was similar to control cells, with less than 10% of lysosomes greater than 0.8 um (Fig. 3d). Using this diameter as a cutoff for assessment of ‘enlarged’ lysosomes, we could quantitatively compare morphologic effects upon expression of different PrP or Mgrn constructs. Because lysosomes were quantified individually, this assay has the capacity to discern partial effects (e.g., only some of the cell’s lysosomes being enlarged) that might be expected from partial Mgrn depletion.
The specificity of the lysosomal phenotype to Mgrn depletion (and not other effects of cytosolic aggregates) was validated in five ways. First, we could show that the change in lysosomal morphology correlated precisely with constructs that were shown in earlier experiments to interact with Mgrn and cause its re-distribution (Fig. 3e). Second, we observed that in a cell type that does not express any endogenous Mgrn (the commonly used N2a cell line), no changes in lysosomal morphology were seen upon expression of the same PrP constructs that otherwise have dramatic effects on lysosomes in HeLa cells (Fig. 4a). Third, the most severe phenotypes seen in cells containing GFP-PrP40–231 aggregates closely mirrored that seen with siRNA knockdown of Mgrn, and the two treatments were not further additive (Fig. 4b). Fourth, the lysosomal morphology phenotype could be partially reverted by overexpression of functional Mgrn, but not a catalytically inactive Mgrn lacking the RING domain (Fig. 4c,d). And fifth, overexpression of Mgrn200–250 tagged with Cerulean (Cer; a variant of CFP), which interacts with PrP (Sup. Fig. S4b) and can therefore compete for endogenous Mgrn, substantially rescued the lysosomal phenotype (Fig. 4e). Note that this competition also explains the subtle (but reproducible) partial rescue seen with MgrnΔR (Fig. 4d), which typically expresses at more modest levels than Mgrn200–250-Cer. Considered together, these results show that cytosolically exposed PrP, whether presented as aggregates or as a transmembrane protein in the CtmPrP topology, interacts with Mgrn, influences its localization, and at least partially phenocopies Mgrn depletion to cause lysosomal morphology changes.
While SA-PrP and GFP-PrP40–231 are quantitatively exposed to the cytosol (in either the CtmPrP topology or as aggregates), only a small proportion of total PrP is likely to become exposed to the cytosol during either inherited or transmissible diseases caused by PrP. To assess whether situations of only partial PrP exposure would also have similar effects, we analyzed several CtmPrP-favoring mutants previously characterized in transgenic mouse models (Hegde et al., 1998; Hegde et al., 1999). These included the artificial mutants PrP(AV3) and PrP(KH-II), as well as the naturally occurring human disease mutation PrP(A117V). These constructs were co-expressed with either wild type Mgrn or the catalytically inactive Mgrn-ΔR, and the lysosomal morphology was assessed by quantitative microscopy (Fig. 5a). Little or no change in lysosomal morphology was noted in Mgrn- or MgrnΔR-expressing cells with wtPrP, consistent with the fact that wtPrP does not substantially interact with Mgrn. By contrast, each of the CtmPrP favoring mutants showed increased proportions of enlarged lysosomes in the catalytically inert MgrnΔR-expressing cells. Importantly, co-expressing these same constructs with Mgrn reverted the lysosomal morphology close to wild type levels. It should be further noted that these mutants showed a less dramatic effect on lysosomal morphology (as judged by % enlarged lysosomes) compared to SA-PrP, consistent with the fact that they only partially generate CtmPrP.
In another experiment, we asked whether cytosolic PrP generated as a consequence of reduced PrP translocation could affect lysosomal morphology in a Mgrn-dependent manner. For this purpose, we used Ifn-PrP, a construct whose inefficient signal sequence mimics the lower translocation efficiency seen for PrP during ER stress (Rane et al., 2008). To stabilize the non-translocated population of Ifn-PrP, we also briefly inhibited proteasome function (as might also occur during prion infection; Kristiansen et al., 2007). As with the CtmPrP-favoring constructs, Ifn-PrP also caused alterations in lysosomal morphology (in MgrnΔR-expressing cells) that were largely normalized in cells overexpressing Mgrn (Fig. 5b). Interestingly, little or no effect was seen for Ifn-PrP in the absence of proteasome inhibition where it is degraded highly efficiently (data not shown). Thus, multiple situations that result in either partial and/or transient exposure of PrP to the cytosolic environment at elevated levels lead to alterations in lysosomal morphology that can be rescued upon co-expression of Mgrn, but not Mgrn-ΔR. Importantly, the constructs used for this analysis [e.g., PrP(A117V) and Ifn-PrP] lead to more modest phenotypes than the exaggerated situations with GFP-PrP aggregates or SA-PrP, further supporting a direct correlation between the extent of cytosolic PrP exposure and Mgrn dysfunction.
The observation that PrP and Mgrn are expressed in very similar patterns within the central nervous system (Sup. Fig. S6; Lein et al., 2007) raised the possibility that PrP could influence Mgrn metabolism and/or function in mouse models of PrP-mediated disease involving its excessive exposure to the cytosol. To test this idea, we analyzed the status of Mgrn and lysosomes in transgenic mice expressing either Human PrP(A117V) or Ifn-PrP (Hegde et al., 1998; Rane et al., 2008). As controls, we also analyzed transgenic mice expressing Opn-PrP (Rane et al., 2008), a previously characterized version of PrP in which its signal sequence has been replaced with another signal whose efficiency is slightly higher than wtPrP (and hence, does not show increased cytosolic exposure as either CtmPrP or cyPrP).
Mgrn immunostaining revealed widespread expression throughout the CNS (Sup. Fig. S5d), consistent with previous in situ data (Sup. Fig. S6). In Purkinje cells, where expression was especially prominent, reduction in Mgrn staining was observed selectively in HuPrP(A117V) and Ifn-PrP mice (Fig. 6a). Reduced staining was also observed in the piriform area of the cortex for the HuPrP(A117V) mice, but interestingly, not for the Ifn-PrP mice. Conversely, in the subiculum, near the hippocampal region, reduced staining was seen for Ifn-PrP, but not HuPrP(A117V) mice. In the case of Purkinje cells, we could be certain that lack of staining was not due to the loss of cells, since not only were the cells clearly present as judged by their characteristic morphology, but also by co-staining with the Purkinje cell marker Calbindin (Sup. Fig. S7). In the other brain regions, we cannot be certain whether the reduced staining is due to selective cell loss or altered expression. It is noteworthy that in the Opn-PrP brain, no changes in Mgrn staining relative to non-transgenic mice were observed in any brain region in either young or old mice.
At this point, we do not know whether the reduced staining represents reduced protein levels (due perhaps to co-degradation of Mgrn upon interaction with PrP), or to reduced immunoreactivity due to sequestration. While punctate staining is seen in cell areas lacking the expected diffuse Mgrn staining pattern, we cannot be certain that these represent mislocalized or aggregated Mgrn since similar autofluorescent structures (seen with pre-immune samples) confound the interpretation. It should be noted that differences in Mgrn levels were not detected by immunoblotting of brain lysates (data not shown), consistent with the region-selective effects observed by immunohistochemistry.
To examine lysosomal morphology, we analyzed Purkinje cells by staining for the lysosomal enzyme Cathepsin D (CatD). These cells were chosen because they were conclusively identifiable, showed clearly altered Mgrn staining in both HuPrP(A117V) and Ifn-PrP mice, and had not degenerated. Remarkably, a qualitatively obvious increase in CatD staining was seen selectively in the mice that also showed altered Mgrn staining (Fig. 6b). Interestingly, this effect was age-dependent, as no changes in CatD staining was observed in any of the mice at 4 months of age. Thus, mouse models of cytosolic PrP exposure, including a naturally occurring human disease mutation [HuPrP(A117V)], result in altered Mgrn expression (or localization) in an age-dependent and cell type dependent manner. In at least one cell type, altered Mgrn expression is correlated with aberrant lysosomal morphology as was seen in cultured cells.
The data so far indicate that upon cytosolic exposure, PrP can interact with and functionally titrate Mgrn to cause cellular dysfunction. However, the forms of PrP implicated in this mechanism (CtmPrP and cyPrP) are made at very low levels, even for disease-causing mutations that favor their generation. This raised the crucial question of whether such minor populations of PrP are realistically capable of titrating cellular Mgrn. We therefore quantified Mgrn in brain, and found its level to be ~66 pmol per gram total brain protein (Sup. Fig. S8a). By contrast, several studies have carefully determined PrP levels in normal brain to be at least 2 nmol (Pan et al., 1992) and up to ~6 nmol (Bendheim et al., 1988) per gram (i.e., ~70–200 ug PrP per gram total protein). Given that PrP and Mgrn share very similar patterns of expression in brain (Sup. Fig. S6), their molar ratio in most cells will be between ~30:1 to 90:1. This means that as little as between 1–2% of PrP exposed to the cytosol may be sufficient to titrate Mgrn. Importantly, CtmPrP in brain of transgenic mice expressing wild type PrP represents ~1% of total PrP, while CtmPrP in PrP(A117V) mice represents ~6% of total (Sup. Fig. S8b). Thus, CtmPrP exceeds Mgrn on a molar basis for PrP(A117V), but not wtPrP.
While the quantification indicates that CtmPrP levels in disease-causing mutants are sufficient to titrate Mgrn, we sought to test this directly. For this, we took advantage of the observation that CtmPrP generation by these mutants depends critically on a slight but detectable inefficiency of the PrP signal sequence. Thus, replacing the PrP signal sequence with a more efficient signal [from either Prolactin (Prl) or Osteopontin (Opn)] reduces CtmPrP levels to near wild type for mutants such as AV3 and A117V (Kim and Hegde, 2002). Remarkably, Prl-AV3 (the Prl signal fused to the AV3 mutant of PrP) when expressed in cultured cells does not cause the Mgrn-dependent enlarged lysosomal phenotype seen with AV3 (Fig. 7a). A similar rescue of the lysosomal phenotype was seen with Opn-HuPrP(A117V) compared to HuPrP(A117V) (Fig. 7b).
Analysis of transgenic mice overexpressing (at ~4X normal; Sup. Table 1) Prl-AV3 and Opn-PrP(A117V) showed that Mgrn levels remain detectable throughout the life of the animals (Fig. 7c and 7d). Because these mice still contain the pathogenic mutation, differing only in the levels of CtmPrP, this form indeed appears to be responsible for Mgrn titration and a substantial part of the neurodegenerative phenotype. Thus, while CtmPrP is only a minor isoform of PrP, its selective elimination alleviates the Mgrn-dependent phenotype in cell culture (Fig. 7a and 7b) and Mgrn depletion in mice (Fig. 7c and 7d). We therefore conclude that very small amounts of cytosolically exposed PrP are sufficient to influence Mgrn function and contribute to disease.
This study elucidates a novel interaction between two disease causing isoforms of PrP (CtmPrP and cyPrP) and the putative ubiquitin ligase Mgrn, a protein whose absence leads to spongiform neurodegeneration. In cultured cell systems, the interaction between cytosolically exposed PrP and Mgrn leads to a lysosomal morphology phenotype comparable to that seen upon siRNA-mediated depletion of Mgrn. The re-localization of Mgrn in these cells and the ability to rescue the altered lysosomal phenotype with functional Mgrn (but not a catalytically inactive mutant) argues strongly for functional depletion of Mgrn activity upon its interaction with PrP. Accordingly, these same cytosolically exposed PrP constructs had no effect on lysosomal morphology in a cell type lacking Mgrn. The interaction between PrP and Mgrn was specific, since aggregates formed by another neurodegeneration-causing protein (Htt) or PrP aggregates lacking the octapeptide repeats neither interacted with Mgrn nor led to the lysosomal phenotype. Analogous effects on Mgrn immunoreactivity and lysosomal morphology were seen in selected cell types of transgenic mouse models of cytosolically exposed PrP. One of these mouse models corresponds to a naturally occurring mutation [PrP(A117V)] associated with Gerstmann-Straussler-Shienker disease (Hsiao et al., 1991). Remarkably, the Mgrn-depletion caused by this mutant could be rescued by a more efficient signal sequence that acts to selectively minimize PrP exposure to the cytosol. We therefore conclude that inappropriate interaction between cytosolically exposed PrP and Mgrn contribute to the neurodegenerative phenotype in at least a subset of diseases associated with aberrant PrP metabolism. These findings provide a qualitatively new direction for understanding neurodegeneration caused by PrP, and raise a wide range of questions for future studies.
Among the various naturally occurring diseases caused by PrP, our findings most directly relate to two subsets of familial cases. One class of mutations within the central hydrophobic domain (P105L, G114V, A117V, G131V, S132I, and A133V) increase the hydrophobicity of this region and likely leads to increased generation of CtmPrP (as judged by in vitro assays; Hegde et al., 1998; Kim and Hegde, 2002). The other class includes two premature stop codon mutants (at residues 145 and 160) that seem to display reduced translocation into the ER, thereby generating increased cyPrP (Zanusso et al., 1999; Heske et al., 2004). These diseases may not be transmissible (Tateishi and Kitamoto, 1995; Tateishi et al., 1996; Hegde, 1999), and are not ‘prion’ diseases in the true sense; rather, they are better viewed as protein folding diseases caused by aberrant PrP. Thus, an important question is how our findings might relate to either other familial PrP-mediated diseases or the transmissible prion diseases. The answer to these questions awaits further studies, but depends on the extent to which PrP (in particular the N-terminus) is ever exposed to the cytosol during the course of disease pathogenesis.
Due to a slightly inefficient signal sequence, even wild type PrP transits through the cytosol to a small (~10% of total synthesized PrP) but detectable extent en route to its proteasomal degradation (Rane et al., 2004; Levine et al., 2005; Ma and Lindquist, 2001; Yedidia et al., 2001). Importantly, the molar ratio of PrP to Mgrn in brain (~30:1 to 90:1) means that as little as 2% of total PrP is equimolar to cellular Mgrn levels. Furthermore, routing of PrP through the cytosol is increased during ER stress due to its reduced translocation into the ER (Kang et al., 2006; Orsi et al., 2006). One implication of these observations is that there is always a potential opportunity for Mgrn to interact with PrP, and conditions which enhance this potential might contribute to neurodegeneration via Mgrn sequestration. This could happen in any of several ways that might be relevant to both genetic and transmissible prion diseases.
For example, transmissible prion diseases are accompanied by both ER stress (Hetz and Soto, 2006; Rane et al., 2008) and reduced proteasome activity (Kristiansen et al., 2007), possibly allowing cyPrP to be both elevated and stabilized. Consistent with such a model, Mgrn was seen to be affected in the Ifn-PrP transgenic mice designed to mimic the reduced translocation that might occur during transmissible prion disease. Familial mutants that act via generation of transmissible prions (e.g., E200K or D178N) could act by a similar indirect manner. In addition, if PrPSc were to ever access the cytosol as has been proposed (Kristiansen et al., 2007), it too could sequester Mgrn in much the same way as our artificial cytosolic PrP aggregates. In other PrP-mediated diseases, the mutations may directly enhance the interaction with Mgrn, as might be the case with octapeptide repeat insertions. This could allow even the normally small amount of cytosolic PrP to contribute to neurodegeneration. Conversely, deletion of repeats seem to attenuate transmissible prion disease severity (Flechsig et al., 2000), perhaps because one adverse downstream event (Mgrn interaction) is minimized. Intriguingly, mice expressing PrP lacking all ORs do not show typical spongiform pathology in the CNS upon prion infection (Flechsig et al., 2000). Thus, via a combination of different mechanisms, it is plausible that an interaction between Mgrn and cytosolically exposed PrP may be a contributing factor in many or all PrP-mediated neurodegenerative diseases, and not just those involving CtmPrP. Each of these hypotheses merit further examination to see if cytosolically exposed PrP is indeed generated and/or stabilized in sufficient amounts to influence Mgrn localization and function. Because PrP is a very abundant protein (and often accumulates to many fold higher levels during disease), even relatively small proportions of it (a few percent) in the cytosol would be sufficient to affect the comparatively low abundance Mgrn.
Depletion of Mgrn by cytoslically exposed PrP is likely to be a contributing factor, and not the sole downstream event leading to neurodegeneration. This supposition is based on the fact that PrP-mediated neurodegeneration in mice can be significantly more severe than simply knocking out Mgrn (where pathology is observed at 6–12 months; He et al., 2003). However, there are several possible ways in which an inappropriate interaction with PrP is actually more detrimental than a knockout. One way is if acute or adult-onset depletion of Mgrn precludes compensatory mechanisms that are otherwise initiated in a germline knockout. Another is if cytosolic PrP partially co-depletes factors that associate with Mgrn. Although Mgrn is largely dispensible, it may associate with other factors whose loss (even partially) is far more detrimental. One candidate is Tsg101 (Kim et al., 2007), a key component of the ESCRT machinery involved in endo-lysosomal trafficking (Hurley, 2008). By depleting this and/or other ESCRT factors, the PrP-Mgrn interaction could more severely influence lysosomal trafficking and cellular function than simply deleting Mgrn. Thus, while it is likely that most instances of PrP-mediated neurodegeneration will involve multiple downstream pathways leading to cellular dysfunction, it is nonetheless plausible that the Mgrn interaction could play a much more central role than might initially appear based on the relatively mild phenotype of Mgrn-null mice. If this is the case, one might predict that prion infection of Mgrn-null mice would lead to a much milder phenotype than otherwise expected upon PrPSc accumulation.
Further insight into the mechanism of neuronal dysfunction may come from a better understanding of Mgrn function. At present, the substrates or site(s) of action for this putative ubiquitin ligase are unknown. It is been suggested on the basis of genetic evidence that Mgrn functions in the same pathway as Attractin, a cell surface receptor implicated in melanocortin signaling (He et al., 2003). In another study, Mgrn was shown to interact with and ubiquitinate Tsg101 to influence endosomal trafficking (Kim et al., 2007). This latter result could mean that Mgrn influences receptor recycling and/or downregulation via receptor mono-ubiquitination, an increasingly common trafficking signal in the endo-lysosomal system (Piper and Luzio, 2007). This would place Mgrn in the ubiquitous and essential pathway of ubiquitin-dependent trafficking of membrane proteins, consistent with the observed localization pattern on intracellular vesicles. However, its role in endo-lysosomal pathways would presumably be non-essential or functionally redundant since the phenotype of Mgrn-null mice is restricted to a small subset of cells despite rather widespread expression (He et al., 2003). Such functional redundancy could explain why despite widespread expression of both cyPrP and CtmPrP (both within and outside the nervous system), the phenotype appears to be relatively focal (Hegde et al., 1998; Ma et al., 2002; Rane et al., 2008). Indeed, in cultured cells that lack Mgrn expression, neither cyPrP aggregates nor CtmPrP lead to alterations in lysosomal morphology. Thus, one explanation for the selectivity of cell death in prion diseases may involve interacting partners, such as Mgrn, whose expression or functional importance is restricted. This would mean that cyPrP and CtmPrP are not intrinsically cytotoxic, but depend critically on their cellular context. It will therefore be important not only to identify other potential interacting partners of cytosolically exposed PrP, but to clearly delineate their expression and function to elucidate how they might contribute to neurodegeneration.
All of the PrP-derived constructs have been described before (Hegde et al., 1998; Rane et al., 2008; Kim and Hegde, 2002; see Supplementary Methods). The FP-PrPx-231 constructs and Mgrn deletion constructs were generated by standard cloning techniques (see Supplementary Methods). SA-PrP is characterized in Sup. Fig. S2. GFP-tagged Htt exon 1 containing 103 glutamines was a gift of L. Greene (NIH). Antibodies were from the following sources: 3F4 and 6D11 mouse monoclonal against PrP (Signet); Calbindin D28k (Sigma); Cathepsin D (Santa Cruz Biotechnology). The GFP and RFP antibodies were raised against the full length recombinant proteins. Anti-GFP reacts to all GFP-derived FPs (e.g., CFP, Cerulean, YFP, etc.), but not RFP (data not shown). Rabbit anti-Mgrn was raised against purified His-tagged full length Mgrn.
Culture of HeLa and N2a cells, transient transfections, preparation of stable cell lines, immunofluorescent staining, and fluorescence microscopy of fixed and live cells was as before (Rane et al., 2004; Rane et al., 2008). For quantitative analyses and comparisons between multiple samples, images were collected using identical excitation and detection settings within the linear range of the photomultiplier tube without saturating pixels. Immunohistochemistry was with minor modifications of earlier methods (Rane et al., 2008; see Supplementary Methods).
Transfected cells were stained with 500 nM LysoTracker Red DND-99 (Molecular Probes) for 30 min at 37°C, rinsed with cold 1× PBS (4°C), and fixed with 3.7% formaldehyde in PBS for 10 min at room temperature before imaging. Random fields of transfected cells (identified by GFP co-expression) were chosen blindly (without visualization of lysosomal staining), and images were collected in both the GFP and LysoTracker channels. Five or six fields, each containing at least 4 transfected cells were imaged. Using Image J, the lysosome images were converted to black and white images using the threshold function, and the lysosome diameter for each lysosome was manually measured. The data were tabulated in Microsoft Excel, which was used to generate the histograms and perform statistical analyses by the Student’s two-tailed t-test.
Semi-permeabilization and imaging to detect interactions between proteins (e.g., Fig. 1) was as before (Lorenz et al., 2006; characterized in Sup. Fig. S1). Biochemical fractionation by selective detergent extraction has been described (Levine et al., 2005; see Supplementary Materials).
ON-TARGETplus SMARTpool siRNAs against Mgrn and GFP (catalog# L-022620-00-0005 and D-001300-01-20; Themo Scientific Dharmacon products) were transfected using Lipofectamine 2000 (Invitrogen) following the manufacturer’s instructions. Constructs to be analyzed were transfected 48 h after siRNA treatment, and the cells visualized 24 h later.
A synthetic peptide containing the OR sequence, followed by three glycines and a cysteine (PHGGGWGQGGGC) was coupled to Sulfo-Link beads (Pierce). Cytosol for pull-down experiments was generated from 10 cm dishes of transfected N2a cells.
Cell lysate was prepared in 1.5 ml of KHM (110 mM KAc, 20 mM Hepes, pH 7.2, 2 mM MgAc2) containing 200 μg/ml digitonin. Debris was removed by centrifugation, and 1250 μl was incubated with 50 μl of beads (either sepharose or peptide-conjugated sepharose) for 2 hours at 4 °C. The beads were washed several times with KHM containing 100 μg/ml digitonin prior to elution with SDS.
Approximately 3 mg purified recombinant Mgrn and BSA fraction V (Sigma) were immobilized on ~ 1 ml CnBr activated sepharose (Amersham Pharmacia). 200 μl total hamster brain homogenate (10% w/v) prepared in PBS containing 0.5% Triton X-100, 0.5% sodium deoxycholate was clarified by centrifugation and diluted with 1620 μl of KHM containing 100 μg/ml digitonin. This lysate was equally divided and incubated with 50 μl of each of the immobilized-protein beads for 2 hours at 4°C. The beads were then washed with KHM containing 100 μg/ml digitonin, after which they were eluted in SDS.
In vitro translocation assays, pulse-chase analyses, glycosidase sensitivity, and immunofluorescence of PrP (and related constructs) employed previously described methods (Hegde et al., 1998; Rane et al., 2004; Kang et al., 2006).
Ifn-PrP and Opn-PrP transgenic mice have been described (Rane et al., 2008). Transgenic mice expressing HuPrP(A117V), Opn-HuPrP(A117V), and Prl-PrP(AV3) mice were generated as described before (Hegde et al., 1999), and will be characterized in greater detail elsewhere. Transgenic lines 6, 36, and 33 of Prl-PrP(AV3), HuPrP(A117V), and Opn-HuPrP(A117V), respectively, were analyzed (see Supplementary Table I).
We are grateful to N. Rane for providing transgenic mouse tissues and constructs, E. Whiteman for making the initial Mgrn constructs, Y. Abebe for expert animal care, L. Greene, H. Lorenz for constructs and sharing unpublished results, G. Patterson for help with microscopy, and L. DaSilva, G. Mardones, and P. Burgos for advice and reagents. This work was supported by the Intramural Research Program of the NICHD at the National Institutes of Health.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.