Abstract
Extracellular vesicles (EVs) are a collection of secreted vesicles of diverse size and cargo that are implicated in the physiological removal of nonfunctional proteins as well as the cell-to-cell transmission of disease-causing-proteins in several neurodegenerative diseases. We have shown that the molecular chaperone, cysteine string protein (CSPα; DnaJC5), is responsible for the export of disease-causing-misfolded proteins from neurons via EVs. We show here that CSPα-EVs efficiently deliver GFP-tagged 72Q huntingtinexon1 to naive neurons. When we analyzed the heterogeneous EV pool, we found that the misfolded GFP-tagged 72Q huntingtinexon1 cargo was primarily found in EVs between 180-240nm. We further determined that cargo-loading of GFP-tagged 72Q huntingtinexon1 into EVs was impaired by resveratrol. Importantly, in addition to CSPα, we identified two other J protein co-chaperones, DnaJB2 and DnaJB6, that facilitate EV export of GFP-tagged 72Q huntingtinexon1. While human mutations in CSPα cause the neurodegenerative disorder, adult neuronal ceroid lipofuscinosis, mutations in DnaJB6 cause limb-girdle muscular dystrophy and mutations in DnaJB2 are linked to the neurodegenerative disorders, Charcot Marie Tooth disease, distal hereditary motor neuropathy, spinal muscular atrophy and juvenile Parkinsonism. Our data provides new insights into the parallels between cellular quality control strategies and EV export, as three J proteins linked to disease in humans are the same as those that mediate EV genesis and export of misfolded proteins.
Introduction
The cell-to-cell transfer of extracellular vesicles (EVs) is a conserved process. In vivo, the continuous exchange among different cells generates a dynamic and heterogeneous pool of EVs 46. EVs come in different sizes and carry different cargoes that exert profound effects in recipient cells following uptake. The physiological roles of EVs are still ill-defined but include exchanging physiological information between cells as well as removing unwanted proteins from cells 39,46,72. EVs are also implicated in disease progression. How EVs facilitate the spread of disease in cancer and neurodegenerative disease and what distinguishes physiological from pathological EVs is a current focus of investigation 36,72. While complex cargoes of DNA, RNA, proteins, lipids and metabolites are packaged in EVs for delivery to recipient cells, our understanding of the mechanisms that target proteins to EVs is rudimentary in comparison to the conventional secretory pathway. We have recently identified an EV protein-sorting pathway, the CSPα-EV export pathway, that exports misfolded disease-causing proteins from neurons 15. These findings are consistent with those of Fontaine and colleagues who also report the export of disease-causing proteins via this pathway 22. We found that the molecular co-chaperone, CSPα (cysteine string protein, DnaJC5) promotes EV export of distinct misfolded disease-causing proteins and that this export pathway is ‘druggable’ as demonstrated by the ability of the polyphenol, resveratrol, to block CSPα-mediated export of misfolded proteins 15. Resveratrol was initially linked to CSPα function in a chemical screen as a compound that alters life span of CSPα c. elegans mutants 41.
Cysteine string protein (CSPα, DnaJC5) is a presynaptic co-chaperone that is critical for synaptic proteostasis 6,20. It is a central player in the synapse-specific machinery that protects against local proteostasis imbalances at the synapse which can lead to neurotransmission deficits, accumulation of misfolded, nonfunctional proteins, synaptic loss and neurodegeneration. Human mutations in CSPα, L115R and L116Δ, cause the neurodegenerative disorder adult neuronal ceroid lipofuscinosis (ANCL) 3,54,70. Deletion of CSPα leads to neurodegeneration in multiple experimental models. CSPα knock-out mice exhibit fulminant neurodegeneration and have a reduced lifespan with no mice surviving beyond 3 months 20. Loss-of-function CSPα Drosophila mutants demonstrate uncoordinated movements, temperature-sensitive paralysis and early lethality 78. And, in C elegans, CSPα null mutants display neurodegeneration and reduced lifespan 42. Not surprisingly, CSPα dysfunction has been implicated in several neurodegenerative disorders in addition to ANCL, including Alzheimer’s, Parkinson’s and Huntington’s disease 7,16,34,50,69,77. CSPα is a member of the large J Protein (DnaJ) co-chaperone family 40. J proteins are Hsp70-interacting proteins that have diverse roles in proteostasis. Disruption or absence of a number of different J proteins is known to compromise neuronal function 38,43,76.
Collectively, this work suggests that CSPα may contribute to proteostasis by exporting misfolded proteins in EVs for off-site disposal in recipient cells. This hypothesis predicts that in the absence of an EV export pathway, misfolded proteins accumulate and neurodegeneration occurs in donor cells. The hypothesis further predicts that dysregulated disposal of misfolded proteins by recipient cells may trigger toxic protein spreading or neurodegeneration. To begin to study these possibilities, we explored two aspects of the CSPα-EV export pathway: (i) whether the molecule resveratrol inhibits secretion of EVs in general, or CSPαEVs selectively and (ii) whether members of the J protein co-chaperone family, other than CSPα, export misfolded proteins such as GFP-72Q huntingtinexon1. Our results confirm and extend our previous findings. We show that within the heterogeneous EV population released from CAD cells, the 180-240nm subpopulation of EVs predominantly contain the GFP-tagged 72Q huntingtinexon1 cargo, and that CSPα increases secretion of this EV subpopulation. Moreover, we find that resveratrol impairs cargo-loading of CSPα-EVs. We further demonstrate that, in addition to the CSPα-EV pathway, misfolded proteins can be secreted via the DnaJB2- and DnaJB6-EV export pathways. CSPα-, DnaJB2-as well as DnaJB6-EVs are delivery competent in that they transfer misfolded GFP-72Q huntingtin cargo to recipient cells. Our data highlight a role of signal sequence lacking-co-chaperones in the sorting of misfolded cargo into specific EVs for export and demonstrate that this EV transit pathway can be targeted pharmacologically.
Results
The CSPα EV export pathway
Trinucleotide repeat expansions of the huntingtin gene cause Huntington’s disease, a progressive neurodegenerative disorder that manifests in midlife 31. Aggregates of polyglutamine-expanded huntingtin are found within genetically normal tissue grafted into patients with progressing Huntington’s Disease, revealing cell-to-cell transit of huntingtin aggregates in vivo 13. We have previously shown that CSPα—but not the loss-of-function CSPαHPD-AAA mutant—promotes cellular export of the polyglutamine expanded protein 72Q huntingtinexon1 as well as mutant superoxide dismutase-1 (SOD-1G93A) suggesting the CSPα-EV export pathway is a general pathway for the secretion of misfolded proteins 15. Extracellular vesicles (EVs) are known to serve as mediators for the intercellular delivery of a wide array of diverse cargo to recipient cells and this heterogeneity of EVs raises the question, which EVs comprise the CSPα-EV export pathway?
In order to address this question, we first characterized total EVs released from CAD neurons. The size distribution of CAD cell EVs is shown by transmission electron microscopy (Figure 1 left panel). To characterize the heterogeneity of EVs by nanoscale flow cytometry and nanoparticle tracking analysis (NTA), we collected media from CAD cells expressing CSPα and GFP-72huntingtinexon1, centrifuged at 300Xg for 5 min to remove cell debris and without further processing, evaluated EVs in the media (Figure 1A). The heterogeneity of particles in this unfractionated media is evident by nanoscale flow cytometry (Figure 1A, middle panel). The most frequent size distribution of EVs was found by NTA to be 144+/-2.7nm, somewhat larger than the 129+/-2.2nm 15 EVs purified by differential centrifugation/exoquick techniques as expected (data not shown).
We next sought to determine which EVs contain misfolded-GFP cargo. Nanoscale flow cytometry analysis reveals that resveratrol increases total EV secretion from control, vector and CSPα-expressing cells (Figure 1B). Secretion of EVs containing GFP-tagged 72Q huntingtinexon1 is dramatically increased when cells express CSPα (p < 0.0001) and this CSPα-EV export of EVs containing GFP-72Q huntingtinexon1 is significantly reduced by resveratrol (p < 0.0001). Comparison of the flow cytometry light scatter is shown in the right hand panel (Figure 1B). The majority of EVs released from CSPα-expressing CAD cells in the absence of resveratrol were the smaller <180nm, presumably exosome EVs (81.6%), while the 180-240nm EVs and the >240nm EVs were 17% and 1.4% respectively (Figure 1C, supplementary Figure 1). The bulk of EVs containing GFP-72Q huntingtinexon1 fall within the 180-240nm range (71% (p < 0.0001)) and CSPα increased export of EVs containing GFP-72Q huntingtinexon1 within this subpopulation by >400% (p < 0.0001). In the presence of CSPα, EVs containing GFP cargo account for 0.06% (<180nm), 1% (180-240nm) and 1.6% (>240nm) within their respective subpopulations.
Resveratrol significantly reduced secretion of the 180-240nm GFP labeled EV pool in a concentration-dependent manner while increasing export of total EVs (Figures 1B&C). In the presence of CSPα and 50μM resveratrol EVs containing GFP cargo account for 0.03% (<180nm), 0.53% (180-240nm) and 0.78% (>240nm) of the respective subpopulations (data not shown).
To investigate whether neurons secrete free floating GFP-tagged 72Q huntingtinexon1 entities, in addition to GFP-72Q huntingtinexon1 as EV cargo, we labeled EVs with Cell Mask Deep Red plasma membrane stain for 30 min at 37°C. We found that 95% of GFP-72Q huntingtinexon1 particles costain with Deep Red stain, confirming the GFP particles are membrane bound vesicles (Figure 1D and supplementary Figure 2). By comparison, control protein aggregates of anti CD38 do not demonstrate Deep Red staining (0.6%), while 80% of EVs containing the GFP-tagged membrane protein PC3 Palm-GFP costain with Deep Red. The GFP-containing 180-240nm membrane bound vesicles were detergent-sensitive (data not shown), as anticipated. Altogether, these results suggest that the CSPα-EV export pathway involves EVs primarily within the 180-240nm range and that resveratrol selectively inhibits the CSPα-EV export of GFP-tagged 72Q huntingtinexon1 while increasing overall EV export.
We next evaluated the deliverability of CSPα-EVs to naïve cells. When media (total) was applied to naive CAD neurons, GFP-72Q huntingtinexon1 cargo was delivered to recipient cells (Figure 1E). GFP cargo was observed in recipient cells as early as 1 hr after application of media and remained at 48 hours (Supplementary Figure 3). All recipient cells found to contain GFP-72Q huntingtinexon1 were rounded and without processes, however we observed a range of GFP signal intensity among GFP-positive cells suggesting the cells received different amounts of cargo. When media contained less EVs with GFP-72Q huntingtinexon1 cargo, for example media from neurons treated with 50μM resveratrol, GFP cargo was still detected in recipient cells (Figure 1E). These results indicate that the CSPα-EVs containing GFP-tagged 72Q huntingtinexon1 cargo are delivery competent. Considering that GFP positive EVs represent a fraction of total EVs, CSPαEV transfer of misfolded proteins to remote cells is potent.
Next, we subjected EVs to nanoscale flow cytometry and western blot. This parallel analysis shows that increases in cellular CSPα expression correlates with increases in 180-240nm EV export of GFP-72Q huntingtinexon1 without changes in the mean size of released EVs (Figure 2). Cell viability was evaluated after media collection, and was not influenced by CSPα expression (Figure 2D). The observation that transgenic expression of α-synuclein abolishes the lethality and neurodegeneration seen in CSPα knock out mice 7, raised the possibility that α-synuclein might also stimulate EV-export. That is, more than one mechanism may exist to package and export misfolded proteins from neurons. α-synuclein is the main component of the cytoplasmic inclusions formed in Parkinson’s disease patients, and mutations in human α-synuclein or triplication of the α-synuclein gene cause Parkinson’s disease 7,48. We tested the possibility that α-synuclein facilitated EV-release by comparing EV export of GFP-72Q huntingtinexon1 from cells expressing α-synuclein or CSPα. As shown in Figure 2E, western blot analysis reveals that increases in α-synuclein expression did not result in mutant huntingtin export. Our results rule out α-synuclein driven EV export as the mechanism underlying prevention of neurodegeneration in CSPα knock out mice and is consistent with the hypothesis that α-synuclein does not replace CSPα, but instead functions downstream of CSPα 7.
Influence of CSPα mutants on EV export
We next evaluated the effect of CSPα mutations on EV genesis and export. L115R and L116Δ are human mutations that cause the lysosomal storage disease, ANCL, the CSPαHPD-AAA is a loss-of-function mutation that does not interact with Hsp70 ATPases 3,6,54,70. The domain structure of CSPα and location of the CSPαL115R, CSPαΔ116, and CSPαHPD-AAA mutations is shown in Figure 3A. We have previously shown that CSPαL115R, CSPαΔ116, and CSPαHPD-AAA, like wild type CSPα, are exported from cells, and while CSPαL115R and CSPαΔ116 promote export of GFP-72Httexon1, CSPαHPD-AAA does not 15. We used nanoparticle tracking analysis (NTA) to evaluate the influence of mutant CSPα’s on EV export in the absence of GFP-72Q huntingtinexon1. The NTA profile of CSPαHPD-AAA, CSPαL115R and CSPαΔ116 mutations relative to CSPα (blue line) is shown for comparison (Figure 3B). No difference in mean size (total) of EVs was observed as a result of CSPα mutations (Figure 3C). Furthermore, no difference in relative sizes (<180nm, 180-240nm, >240nm) of EVs was observed as a result of CSPα mutations (data not shown). We then applied conditioned media to naïve cells and monitored EV delivery of GFP-72Q huntingtinexon1. Recipient cells that received control media from vector, CSPαHPD-AAA, and α-synuclein transfected cells showed low levels of GFP signal at 24hrs, likely due to endogenous mechanisms. Like CSPα-EVs, CSPαL115R- and CSPαΔ116-EVs were delivery competent. GFP-72Q huntingtinexon1 aggregates were clearly observed in recipient cells as early as 1 hr and remained visible 24 hrs after media application (Supplementary figure 4). Again, recipient cells contained a range of GFP signal intensity suggesting the receipt of different amounts of cargo and all GFP positive cells were rounded and without processes.
How does resveratrol reduce the export of GFP-72Q huntingtinexon1 by CSPα-EVs? To address this question neurons were treated with 50μM resveratrol 6 hrs following transfection. Figure 4A shows the NTA profile of EVs collected in the presence and absence of 50μM resveratrol (red Line). Resveratrol reduced secretion of GFP-72Q huntingtinexon1 but not CSPα (Figure 4B), consistent with earlier observations 15. Moreover, resveratrol reduced secretion of GFP-72Q huntingtinexon1 mediated by CSPα, CSPαL115R and CSPαΔ116. Cell viability following media removal is shown in the right hand panel. Nanoscale flow cytometry analysis verified that resveratrol increased total EVs exported by CAD cells in all conditions, while decreasing GFP-72Q huntingtinexon1-containing EVs. Resveratrol also significantly, increased the mean size of vesicles in all conditions (Figure 4C). One possibility is that resveratrol may directly influence cargo-loading machinery of the CSPα EVs. It is also conseivable that resveratrol might indirectly influence cargo-loading by modulating other proteostatic mechanisms that target GFP-72Q huntingtinexon1 (eg proteasome, lysosome clearance). Taken together, these results indicate that resveratrol reduces cargo-loading of GFP-tagged 72Q huntingtinexon1 into EVs without reducing export of CSPα or the CSPαL115R and CSPαΔ116.
DnaJB2 and DnaJB6-EV Export Pathways
CSPα (DnaJC5) is a type III J Protein with a well-defined domain architecture that includes an N terminal J domain and a cysteine string region after which the protein is named 5. J proteins are a family of chaperones that contain a wide variety of domains but the defining signature domain is the ~70 amino acid J domain that contains a histidine-proline-aspartic acid (HPD) motif required for activation of Hsp70 40. This led to the speculation that other J proteins may also export misfolded proteins.
In order to test this prediction we screened several J proteins with reported links to neurodegeneration 76 for EV-export activity. DnaJB2 was identified by both western analysis (left panel) and nanoscale flow cytometry (right panel) as a J protein that exports misfolded huntingtin in EVs similar to CSPα (Figure 5A&B). Alternative splicing of the DnaJB2 gene produces two isoforms, a DnaJB2short and a 42kDa DnaJB2long isoform 10 that undergoes C terminal geranylgeranylation 8, we studied DnaJB2long. The related J proteins, DnaJA1, DnaJB1, DnaJB11, DnaJC14, DnaJC13 and DnaJC19 did not export misfolded huntingtin. Extracellular milieu includes free floating co-chaperones in addition to EVs, and chaperones that adhere to EVs were expected to co-isolate with EVs. We found that DnaJA1, DnaJB1, DnaJB2, DnaJB6, DnaJB11 and DnaJC14 were secreted from CAD cells (Figure 5C). In fact, DnaJB11 as well as DnaJA1 were robustly secreted by CAD neurons under all conditions evaluated (Figure 5C). DnaJB11(Erdj3) is known to be secreted via the conventional secretory pathway 24, on the other hand DnaJB1 is secreted in exosomes 68 and the mechanisms underlying DnaJA1 and DnaJC14 secretion remain to be determined. Whether or not secreted J proteins activate Hsp70 ATPase for conformational work would depend on ATP levels and remains to be fully elucidated.
A confounding observation from the parallel analysis was that DnaJB6short was observed to export GFP-72Q huntingtinexon1 by nanoscale flow cytometry but not western analysis (Figure 5A). This suggested to us that cargo in DnaJB6 EVs may undergo degradation during EV isolation. To investigate further we analyzed GFP-72Q huntingtinexon1 export by spotting media on nitrocellulose prior to EV purification. Dot blot analysis of unfractionated media (data not shown) and western blot of EVs (Figure 6A), reveal that at high expression levels DnaJB6short did indeed mediate export of GFP-72Q huntingtinexon1. Alternative splicing of the DnaJB6 gene produces two isoforms, a DnaJBshort and a 36kDa DnaJB6long isoform that contains a C terminal nuclear localization signal 51. We compared DnaJB6short and DnaJB6long for GFP-72Q huntingtinexon1 export activity. To our surprise, EV export of GFP-72Q huntingtinexon1 by DnaJB6long was more robust than DnaJB6short by western analysis. Domain structures of CSPα, DnaJB2long and DnaJB6long are shown in Figure 6B. Each of these co-chaperones contains unique domains in addition to the common J domain. Figure 6C shows the percentage of GFP positive EVs within the total pool determined by nanoscale flow cytometry (left panel) and quantification of GFP-72Q huntingtinexon1 by western analysis (right panel). The percent of EVs containing GFP-72Q huntingtinexon1 is dramatically increased when cells express CSPα (p < 0.0001), DnaJB2long (p < 0.0001) or DnaJB6long (p < 0.0001). EVs containing GFP-72Q huntingtinexon1 are primarily within the 180-240nm range from DnaJB2long and DnaJB6long expressing cells, similar to that found for cells expressing CSPα. We found that exported DnaJB2long- and DnaJB6long-EVs containing GFP-72Q huntingtinexon1 deliver cargo to recipient cells (supplementary figure 5). In contrast to CSPα and CSPα mutants, media containing DnaJB2long and DnaJB6long EVs did not deliver GFP-huntingtinexon1 to recipient cells at 1 hr but were observed to deliver cargo 24 hrs following media application. We then labeled the RNA cargo of EVs (total) with Exo-Red (SBI) and applied labeled EVs to target cells in order to compare GFP-72Q huntingtinexon1 with Exo Red transfer (supplementary figure 5; right panel). Nearly all recipient cells were Exo Red positive, demonstrating the effective delivery of EVs. These results indicate that while DnaJB2long and DnaJB6long EVs transfer GFP-72Q huntingtinexon1 between cells there are differences in recipient cell uptake or proteostasis between DnaJB2long/DnaJB6long and CSPα EVs.
Collectively our results show that three J proteins that are key to neuronal protein quality control, CSPα, DnaJB2long, DnaJB6long shuttle misfolded GFP-72Q huntingtinexon1 from parent to recipient cells in EVs and that the molecule resveratrol modulates this pathway. These removal-based mechanisms represent >1% of the total EVs produced, and yet, efficiently transfer neurotoxic components to naïve cells.
Discussion
Here we report that the J proteins, CSPα (DnaJC5), DnaJB2long and DnaJB6long promote export of misfolded 72Q huntingtinexon1 in extracellular vesicles (EVs). Successful intercellular transmission is achieved via these EVs. We found that GFP-72Q huntingtinexon1 is primarily located in EVs within the 180-240nm size range. Intervention in these J protein removal based mechanisms is promising, given that resveratrol, a common dietary component, strongly inhibits 72Q huntingtinexon1 transport through the CSPα-EV export pathway. J Protein cochaperones orchestrate a number of diverse intracellular proteostasis functions, our data provides a clear role for CSPα (DnaJC5), DnaJB2long and DnaJB6long in facilitating misfolded protein export. Earlier work demonstrated that CSPα increases secretion of TDP-43, α-synuclein and tau from HEK293 cells 22 and exports SOD-1G93A and 72Q huntingtinexon1 in EVs 15 suggesting that several disease-causing proteins transit through this pathway. USP19 has recently been reported to be involved in the CSPα export pathway75.
The proteostasis network consists of elaborate and incredibly adaptive protein machinery that protects against age-associated neuronal decline and neurodegenerative disorders 52. The role of CSPα, DnaJB2 and DnaJB6 in neural quality control mechanisms are supported by genetic as well as experimental evidence. Human mutations in CSPα (DnaJC5) and DnaJB6 lead to adult neuronal ceroid lipofuscinosis (ANCL) 3,54,70, and limb-girdle muscular dystrophy 33,57,62,63,67 respectively. Mutations in DnaJB2, depending on their location, cause Charcot Marie Tooth disease type 2 25, distal hereditary motor neuropathy 4, spinal muscular atrophy or juvenile Parkinsonism 61. The association of DnaJB2 with neurofibrillary tangles in Alzheimer’s disease patients 10, the presence of DnaJB6 in lewy bodies in Parkinson’s disease patients 17 and the intralysosomal accumulation of lipofuscin caused by CSPα mutations 3,54,70 link these J proteins to neural quality control mechanisms. In addition, DnaJB2 overexpression reduces huntingtin and SOD-1 aggregation in mouse models of Huntington’s disease and amyotrophic lateral sclerosis 35,45,55. In drosophila and xenopus, overexpression of DnaJB6 suppresses polyglutamine toxicity 19,32 and, in mice, brain-specific overexpression of DnaJB6 delays polyglutamine aggregation and onset of disease 37. In the absence of DnaJB6, mice die mid-gestation due to a failure of chorioallantoic attachment during placental development and keratin aggregation 73. Moreover, the absence of CSPα leads to neurodegeneration in drosophila, mice and c. elegans 20,41,78. In vitro, DnaJB6 inhibits Aβ amyloid aggregation 47. In cellular models, DnaJB6 suppresses huntingtin, ataxin-3 and androgen receptor aggregation 12,26,32, while the absence of DnaJB6 leads to α-synuclein aggregation2 and DnaJB2 suppresses rhodopsin, huntingtin, ataxin-3, parkin and TDP-43 aggregation 8,11,23,60,74. Considering the conserved nature of EV shuttles, alongside the CSPα, DnaJB2 and DnaJB6 links to quality control, the data in this report expands our understanding of proteostasis to include J protein-driven EV transfer of disease-causing misfolded proteins. The newly-identified J protein-EV pathways by which neurons transfer neurotoxic proteins to cells have similarities to tunneling nanotube and exofer (4μm) systems that also transport polyglutamine expanded huntingtin 14,49.
Perhaps the most unexpected observation in this report is that DnaJB6long, which contains a nuclear localization signal (NLS) 51, exports misfolded GFP-72Q huntingtinexon1 in EVs. The cellular machinery that targets proteins, or more specifically misfolded/aggregated proteins, into EVs is not fully elucidated. Plasma membrane anchors are proposed to be involved in EV protein packaging pathways 65. CSPα is one of the most heavily palmitoylated proteins known 56 and DnaJB2long undergoes geranylgeranylation 8. Interestingly, DnaJB2long contains two unique ubiquitin interacting motifs that bind ubiquitylated proteins and target them for clearance 9. We do not know why CSPα, DnaJB2long and DnaJB6long sort GFP-72Q huntingtinexon1 to EVs and the other J proteins failed to do so. Clearly, the common J domain is not the only molecular determinant for EV-export of misfolded 72Q huntingtinexon1 and the specific domains that mediate cargo selectivity of these pathways requires future studies. It is conceivable that the distinct domains of these three J proteins allows for greater level of adaptability in the recruitment of misfolded clients for export. Such molecular characterization of chaperone activity in EVs will undoubtedly be the focus of future experimentation.
Neural export of misfolded proteins offers many advantages. The capacity of EVs to transfer misfolded/aggregated protein cargo like GFP-72Q huntingtin will influence proteostasis of both recipient and parent cells. The export of misfolded proteins could relieve pressure on parent cell proteostasis machinery and deliver misfolded proteins to cells with greater folding/clearance capacity. This would be particularly relevant in neurons where, for example, synapse-specific proteostasis is almost certainly limited, and yet, proteins susceptible to misfolding and aggregation are abundant 28. EV export would greatly improve clearance of toxic proteins that synapses are unable to digest. The strategy to save resources by exporting unfolded and aggregated proteins in EVs for offsite degradation pathways predicts that EVs containing misfolded cargo would be specifically targeted to cells equipped to process toxic protein aggregates. Little is known about the signals that target EVs to recipient cells and the mechanisms underlying uptake differences in EVs that remove unwanted proteins vs EVs that pathologically spread disease-causing proteins. Although, it is widely agreed that in neurodegenerative disorders, there is a relationship between EVs and the spread of misfolded-disease-causing proteins including prions, amyloid peptides, α-synuclein, SOD-1, TDP-43, and tau 1,15,18,21,27,29,30,53,58,59,64,66,71. Several of these toxic proteins have been reported as both free floating aggregates and EV cargo, for example SOD-115, suggesting there are different routes of export. The ultimate question of how CSPα (DnaJC5), DnaJB2long and DnaJB6long EVs function in neurodegenerative disease progression remains to be determined.
To our knowledge, our results illustrate the first examples of J protein co-chaperones that transfer misfolded proteins between cells, enabling clearance of neurotoxic proteins from parent cells and delivery to recipient cells. Our study suggests a mechanism for how neurons restore proteostasis. Dysregulation of this pathway may contribute to pathogenic disease progression in neurodegenerative disorders, although the significance of J protein mediated EV export in vivo is far from clear.
Materials and Methods
EV collection from CAD cells
Maintenance of CAD (catecholaminergic derived CNS cells) was described before 15,44. For expression in CAD cells, cDNAs encoding for Myc-tagged J proteins and GFP-72Q huntingtinexon1 were transiently transfected with Lipofectamine 3000 (Invitrogen) in Opti-MEM™ medium. Media was changed to serum-free media 6 hrs post-transfection, and, resveratrol (Lalilab Inc.) treatment began at the media change. Media was collected 48 hrs after transfection, spun at 300Xg for 5 min to remove cell debris and evaluated without further processing by nanocyte tracking analysis, nanoscale flow cytometry analysis and dot blot analysis. For western analysis EVs were isolated from the unfractionated media by exoquick precipitation solution (SBI) and solubilized in sample buffer. Following media collection, cell viability was determined utilizing an XTT assay (New England Biolabs).
Immunoblotting
Proteins were separated by SDS-PAGE and electrotransferred from polyacrylamide gels to nitrocellulose membrane (0.2μm pore size). Membranes were blocked in tris-buffered saline (TBS) containing 0.1 % Tween 20, 1 % BSA and then incubated with primary antibody overnight at 4°C. The membranes were washed and incubated with horseradish peroxidase-coupled secondary antibody for ~2 hrs at room temperature. Bound antibodies on the membranes were detected by incubation with Pierce chemiluminescent reagent and exposure to Cdigit, LiCor (Mandel). The chemiluminescent signals were quantified using image studio digits software (Mandel). For the dot blots, 10μl of unfractionated media was spotted on nitrocellulose membrane, dried and membrane processed as described above.
Plasmids
cDNAs encoding for CSPα, CSPα mutants, and J proteins were expressed in the plasmid myc-pCMV. GFP-72Q huntingtinexon1 was expressed in the plasmid pcDNA3.1. All amplified regions of all plasmids were sequenced to ensure the absence of any undesired mutations.
Transmission Electron Microscopy
Media was centrifuged at 2,000g for 10 min, and at 10,000Xg for 30 min to discard membrane and debris. The supernatant was then centrifuged at 100,000g for 2 hrs and EVs collected. EVs were suspended in PBS, fixed sequentially on a 150 mesh grid in 4% paraformaldehyde followed by 2% glutaraldehyde and stained with 2% uranyl acetate for 10 min. EVs were imaged with a Tecnai F20 TEM and a Gaten CCD camera.
Fluorescence Imaging
Images of live cells were taken with the EVOS FL Auto Imaging System. Images of donor cells were taken at the time of media collection. Images of recipient cells were taken between 1-48 hours. The intensity of the GFP-72Q Httexon1 varied among cells and we used the single slider until the on-screen brightness of the lowest intensity aggregate was satisfactory and used this setting to capture images in high-quality mode. Where indicated, after the 300Xg for 5 min conditioned media spin to remove cell debris, 60μl of Exo-Red (SBI) was incubated with 1 ml media containing EVs for 10 min at 37°C to label the RNA EV cargo. 150μl of exo-quick was added, the labeled EVs were placed on ice for 30min and then centrifuged for 3 min at 20,000Xg. The labeled EV pellet was resuspended in 1ml of DMEM-F12 media and applied to target cells.
Nanoparticle Tracking Analysis
For NTA and nanoscale flow cytometry analysis, all samples were diluted 25 fold using 0.2x phosphate buffered saline that had been filtered twice through 0.2micron filter. For NTA analysis, all samples were analyzed using the Nanosight LM10 (405nm laser, 60mW, software version 3.00064). Samples were analyzed in triplicate for 60 seconds per replicate with a count range of 20-100 particles per frame. A variety of NIST (Thermo Scientific 3000 Series) standards (Supplementary Materials) were analyzed each today prior to sample analysis. The system was cleaned between each sample and checked for any sample carryover using the PBS diluent.
Nanoscale Flow Cytometry
Samples were similarly prepared for nanoscale flow cytometry as for NTA. All samples were analyzed using the Apogee A50 flow cytometry platform. Light scatter was provided using the 405nm laser (75mW); GFP signal was generated using the 488nm laser (50mW, 535/35) and far red signal was generated using the 630nm laser (75mW, 680/35). All samples were analyzed for 60 seconds. Optimization was performed to insure single EVs were being analyzed and single events were triggered by light scatter only. The system was cleaned each day prior to sample analysis and a variety of silica and polystyrene standards (Apogee 1493 standards) processed for instrument set up and QC. The silica standards were used to assess the relative size range of EVs (see supplementary Figure 1).
To demonstrate that the GFP-72Q Httexon1 signal was associated with bona fide vesicles and not simply protein aggregates, samples were first mixed with Cell Mask Deep Red plasma membrane stain (Thermo Fisher Scientific (C10046), final concentration of 0.1X), incubated for 30 minutes at 37°C and then diluted in PBS. Palmitoylated-GFP positive EVs were obtained from the conditioned media of a PC3 prostrate cancer cell line; the membrane of these EVs contains the FP proteins. Anti-CD-FITC antibody was used to generate protein aggregates. Antibody (10μl) aliquots were incubated at 50°C for increasing periods of time to generate nonmembrane, protein aggregates. These aggregates as well as the PC PALM-GFP controls were similarly stained with Deep Red cell mask. Aggregate concentration was analyzed by NTA and uptake of the membrane dye was measured using nanoscale flow cytometry. For these experiments, single particle detection was triggered using positive GFP/FITC fluorescence and the associated far red signal analyzed.
Statistical Analysis
All data were graphed and statistically analyzed using GraphPad Prism version 6.01 for Windows, GraphPad Software, La Jolla California USA, www.graphpad.com. Statistics included One-Way and Two Way ANOVA with either Tukey’s or Dunnett’s post test analysis if initial ANOVA was statistically significant (p<0.05; stars indicate significance ⋆ p<0.05, ⋆⋆ p<0.01, ⋆ ⋆⋆ p<0.001, ⋆⋆⋆⋆ p<0.0001). All values are presented as the mean ±SEM where appropriate, otherwise the SD is presented as indicated.
Author Contributions
J.E.A.B. conceived the project. D.P. and J.L. designed and interpreted all NTA and NFC experiments. J.E.A.B designed and interpreted all WB and FI experiments. J.D. provided technical assistance. J.E.A.B. D.P. and J.L. wrote the manuscript.
The authors declare they have no competing financial interests.
Acknowledgements
The work was supported by a grant from the Alzheimer Society of Alberta and Northwest Territories, the Alberta Prion Research Institute and National Research Council of Canada. The authors would like to express their gratitude to Dr. Frank Visser for technical support and Ali Jalloul for technical advice.