Genetic expansion of CAG triplets in protein-coding regions of otherwise unrelated genes is the underlying cause of a family of dominantly inherited neurodegenerative diseases including Huntington disease (HD) and spinocerebellar ataxias2
. The tracts of polyglutamine (polyQ) homopolymers (Q ≥ 40) encoded by these expanded CAG triplets cause the normally soluble protein products of these genes, or fragments thereof, to form cytotoxic protein aggregates2
. CAG expansion diseases therefore, belong to a much larger family of protein “conformational diseases,” including systemic and organ-specific amyloidosis, Alzheimer's disease and prion encephalopathy. Pathogenesis in these diseases is tightly linked to the formation of high molecular weight, fibrillar, β-sheet rich, insoluble protein aggregates, termed “amyloid,” that accumulate in characteristic sites either inside or outside of the cell1, 3
In amyloidosis, insoluble protein fibrils derived from normally soluble secreted proteins are deposited in the extracellular
milieu causing damage to surrounding viscera, blood vessel walls and connective tissue4
. Whether organ damage is a consequence of tissue disruption or obstruction due to the sheer mass of deposited protein, as in the case of systemic amyloidosis4
, or to an intrinsic cytotoxicity of amyloids or their oligomeric precursors, as in the case of neuropathic amyloidosis5
, remains a critical but unresolved question. In contrast to amyloidosis, most neurodegenerative diseases are caused by alterations in the conformation and oligomeric state of normally well-behaved intracellular
proteins that, in diseased states, accumulate within cytoplasmic or nuclear inclusion bodies6
. Emerging evidence suggests that oligomeric precursors to these large assemblies are cytotoxic and directly impair crucial cellular functions which cause the neuronal dysfunction and ultimately death associated with these disorders7
Many extracellular amyloids and amyloid precursors, including those associated with systemic amyloidosis, neurodegenerative disease, and even those not associated with disease7
, can be taken-up by a wide variety of cell types including macrophages, neurons, fibroblasts, and epithelial cells7-10
. This uptake is reported to occur via phagocytic or endocytic processes that result in delivery to lysosomes which may suppress their toxicity by degrading them9, 10
. However, all of these mechanisms would deliver aggregates to an endomembrane compartment, and not to the cytosol. Surprisingly, a recent study reported that healthy fetal tissue grafted into the brains of Parkinson's disease patients, acquired cytoplasmic alpha-synuclein- rich Lewy bodies, suggesting a potential “prion-like” transmission of nucleating species from the recipient's diseased brain to the healthy grafted tissue11
The ability of amyloid to cross a membrane barrier and access the nucleocytoplasmic compartment, a necessary step to effect conversion of a cytoplasmic protein like α-synuclein by extracellular aggregates, has never been directly demonstrated. The starting point of the present work was the demonstration by Yang et al that fibrillar, insoluble amyloid formed from synthetic polyglutamine peptides or an amyloidogenic bacterial protein, Csp-B1, are readily taken up by mammalian cells in culture8
. Those studies did not determine whether the “intracellular” amyloids were present within lysosomal or other endomembrane compartments- the demonstrated route for entry of other amyloids into mammalian cells- or the cytosol, which would necessitate the unlikely possibility that these large protein assemblies had crossed a biological membrane. Although they did not directly test this possibility, Yang et al8
reported that exogenously administered amyloids to which a nuclear localization sequence (NLS) had been appended appeared to gain access to the nucleus, raising the possibility that at least some aggregate-associated NLS had become accessible to importins in the cytosol. We therefore sought to directly test whether large polyQ amyloid assemblies can move from outside the cell into the cytosol.
PolyQ peptides (K2
), labeled with fluorescein, rhodamine or biotin were converted to fibrillar aggregates12
that appeared by transmission electron microscopy to be composed of bundles of individual fibrils measuring 3-5 nm in width () These polyQ amyloids have been extensively characterized and exhibit characteristic β-sheet circular dichroism spectra, bind thioflavin T and react with monoclonal antibodies to amyloid13
. Fluorescent K2
aggregates were efficiently internalized by COS7 cells () and by other cell lines including HEK293 and neuro2A (-) as well as CHO and HeLa S3 (not shown). Following overnight incubation with cells, K2
aggregates were enriched in a juxtanuclear, pericentriolar region that was labeled with antibodies to γ-tubulin (). Although this cellular region is enriched in late endosomes, lysosomes () and autophagosomes (staining for LC3; not shown), we failed to detect any significant colocalization of aggregate foci with markers for these organelles. By contrast, strong colocalization was observed between internalized K2
aggregates and cytosolic `quality control' components including ubiquitin, proteasome subunits, and Hsp70 (), suggesting that internalized fibrillar polyQ aggregates did not accumulate within an endosomal compartment.
Synthetic polyglutamine peptides form filamentous aggregates that are internalized by mammalian cells in culture
Internalized polyglutamine peptide aggregates access the cytosolic compartment
Propagation of homotypic polyQ aggregation in cell culture
To investigate the possibility that the aggregates had gained access to the cytoplasmic compartment, we used deep-etch transmission electron microscopy of “unroofed”14
cells that had been briefly exposed to polyQ aggregates (). These images reveal the presence of large fibrillar aggregates, which appear indistinguishable from those used to inoculate the cells, attached to the inner surface of the plasma membrane, resting on a “bed” of cortical actin. There was no evidence in these images that aggregates associated with, or were surrounded by, any endomembranous structures or clathrin.
To further assess whether externally applied polyQ amyloids are in direct contact with cytosolic proteins, we determined whether externally applied K2
aggregates are capable of nucleating the polymerization of a cytoplasmic polyQ rich protein “reporter”, exploiting the well known ability of fibrillar polyQ amyloids to nucleate the polymerization of normally soluble proteins containing polyQ tracts below the threshold (Q ≤ ~37) for spontaneous aggregation. Control experiments confirmed that K2
aggregates are capable of nucleating aggregation of bacterially-expressed glutathione S-transferase (GST) linked to the N-terminal exon of huntingtin (Htt) with either a non-pathogenic (18Q) or a pathogenic (51Q) repeat (Supplementary Information, Fig. S1
, online). We therefore assessed the aggregation state of cyan fluorescent protein (CFP)-tagged Htt exon 1 (CFP-HttQ25
) in HEK293 cells following infection with extracellular aggregates (). CFP fluorescence in untreated cells exhibited a diffuse nucleocytoplasmic distribution, as expected of a soluble HttQ25
fragment (). Strikingly, CFP fluorescence following infection with K2
aggregates redistributed into distinct foci () that colocalized extensively with the internalized K2
puncta. Quantification of the extent of colocalization indicated that nearly 90% of the K2
puncta were coincident with CFP puncta, suggesting that the intracellular and extracellular proteins had indeed coalesced into common intracellular foci. Conversion of the cytoplasmic reporter to a punctate distribution is first detectable within 1hr following exposure to aggregates (data not shown); this phenotype is present in more than 90% of reporter cells by 24 hr (). To determine if the ability to convert diffuse cytoplasmic Q25
to a punctate distribution is a property unique to aggregates generated from synthetic Q44
peptides or a more general property of polyQ amyloids, aggregates were generated in vitro
from purified, bacterially expressed Htt exon 1 containing either 18 or 51 glutamines (). These assemblies were used to infect HEK293 cells stably expressing a Q25
reporter fused to cherry fluorescent protein (chFP-HttQ25
). Reporter cells exposed to non-fibrillar HttQ18
aggregates (, inset) maintained a diffuse pattern of fluorescence (). By contrast, reporter cells exposed to highly fibrillar HttQ51
aggregates (, inset) exhibited extensive conversion to a punctate distribution (). These data indicate that cell penetration and cytosolic nucleation is a property common to fibrillar polyQ amyloids and not simply an artifact of amyloids produced from synthetic peptides.
We used a filter retardation assay to confirm that the observed coalescence of CFP fluorescence into puncta reflects a change in the aggregation state of CFP-HttQ25
(). In control samples not exposed to added aggregates, filter-bound CFP-HttQ25
was undetectable, while GFP-HttQ71
could be weakly detected, reflecting spontaneous aggregation owing to its longer polyQ tract. However, in cells exposed to K2
and, to a greater extent, GFP-HttQ71
, were readily trapped by the filter, confirming that the observed changes in reporter distribution reflect a change in the overall aggregation state of the intracellular protein. Addition of an equivalent amount of Q44
fibrils to cells during lysis was largely ineffective at inducing aggregation of the cellular reporters, indicating that the observed conversion of the endogenous intracellular reporter is not an artifact of sample preparation. K2
aggregates were also observed to induce focal redistribution (Supplementary Information, Fig. S2a
, online) and aggregation (Supplementary Information, Fig. S2b
, online) of cytoplasmic GFP-HttQ60
expressed in differentiated, postmitotic neuro2A neuroblastoma cells15
. Together, these observations strongly suggest that fibrillar polyQ aggregates can enter the cytosolic compartment of mammalian cells where they nucleate the aggregation of otherwise soluble proteins containing polyQ tracts.
Sequence-specificity of amyloid nucleation is a fundamental property that underlies the species barrier and strain selectivity properties of prions in both mammals16
. To test whether the proposed intracellular nucleation of cytoplasmic Htt is selective for proteins with polyQ tracts, we generated fibrillar aggregates from two well-characterized amyloids, the prion domain of the yeast Sup35 prion protein (Sup35-NM) and the Alzheimer's disease amyloid peptide Aβ(1-40) (Supplementary Information,Fig. S3
, online). These aggregates were efficiently internalized by cells and accumulated in a juxtanuclear distribution (), which, like the K2
aggregates, colocalized with ubiquitin and proteasome epitopes (not shown). To assess the selectivity of nucleation of CFP-HttQ25
aggregation we repeated the infection experiment using either `homologous' (K2
) or `heterologous' (Sup35-NM or Aβ(1-40)) aggregates. When cells expressing CFP-HttQ25
were incubated with either Sup35-NM or Aβ(1-40) aggregates, CFP-HttQ25
fluorescence remained diffuse and homogeneous, indistinguishable from untreated cells, while cells infected with K2
aggregates exhibited a punctate redistribution of CFP-HttQ25
(). The heterologous aggregates were unable to induce aggregation of GFP-HttQ25
in the filter retardation assay (), strongly supporting the conclusion that the ability of exogenously added aggregates to nucleate cytoplasmic polyQ aggregation is sequence-specific.
Internalized fibrillar aggregates induce homotypic but not heterotypic aggregation of cytoplasmic reporters
To determine whether selective nucleation of homologous cytoplasmic proteins by internalized fibrillar aggregates is an exclusive characteristic of polyQ-containing proteins we tested the ability of Sup35-NM aggregates to induce the specific aggregation of a cytoplasmically expressed GFP-Sup35-NM reporter (). In the absence of added aggregates, the vast majority of GFP-Sup35-NM remained soluble. By contrast, the protein shifted mostly to the insoluble pellet fraction in cells treated with Sup35-NM, but not K2Q44K2 or Aβ(1-40) aggregates. Therefore, the ability of extracellular fibrillar amyloids to nucleate the aggregation of cytoplasmic proteins in a specific, template-dependent manner in mammalian cells is not an exclusive property of polyQ aggregates and of proteins containing homopolymeric polyQ tracts.
To determine whether the novel phenotype of CFP-HttQ25 expressing cells infected with K2Q44K2 aggregates can be inherited during cell division, we transiently exposed a population of HEK293 cells stably expressing CFP-HttQ25 to extracellular tetramethylrhodamine (TMR)-labeled K2Q44K2 aggregates and quantified the fraction of cells with the novel, punctate CFP fluorescence phenotype in subsequent generations (). The fraction of K2Q44K2 aggregate-infected cells exhibiting this phenotype declined exponentially over the first 15-20 generations after infection, in parallel with the measured loss of fluorescent K2Q44K2 aggregates, suggesting that the initial inoculum was diluted during cell division. The frequency of cells beyond generation 27 with CFP-HttQ25 puncta was consistently and significantly (P = 0.004) higher at each generation following infection with K2Q44K2 aggregates than after control treatments with dextran or Aβ amyloid (). Moreover, cells initially exposed to a ten-fold lower K2Q44K2 aggregate inoculum (0.1μM) also sustained a persistent punctate phenotype that was quantitatively indistinguishable from that induced by the higher concentration of K2Q44K2 aggregates. Thus, persistent aggregation of a strictly nucleocytoplasmic polyQ reporter is a novel and heritable phenotype of cells transiently exposed to extracellularly administered homotypic, but not heterotypic, protein aggregates.
We propose that the persistence of the CFP-HttQ25
aggregate phenotype at low but significant levels through multiple generations in K2
aggregate-treated mammalian cells reflects the existence of an efficient mechanism to suppress the distribution of cytoplasmic aggregates during cell division or the dissemination of aggregates to neighboring cells following cell lysis. Internally formed polyQ aggregates are efficiently sequestered into pericentriolar aggresomes that are only rarely (~1% of the time) segregated equally among mitotic progeny during cell division18
. Perhaps aggresome formation is a mechanism that has evolved to minimize prion propagation among dividing cells.
The observation of a persistent aggregation phenotype in dividing cells implies that aggregates formed within mammalian cells are themselves capable of nucleating aggregation when distributed through cytoplasmic transfer during cytokinesis. However, in order to be potentially relevant to the pathogenesis of neurological diseases such as HD, in which the target cells are post-mitotic neurons, aggregates formed within a mammalian cell must be able to translocate to neighboring uninfected cells. When HEK reporter cells stably expressing a cherry fluorescent protein-tagged Htt fragment (chFP-HttQ25) were co-cultured together with HEK cells stably expressing GFP-HttQ71, the vast majority of the HttQ25 reporter retained a diffuse cytoplasmic pattern of fluorescence, despite the presence of highly aggregated HttQ71 foci in neighboring cells (). Quantification of the fraction of HttQ25 cells with chFP puncta revealed that cells co-cultured with GFP-HttQ71 exhibited a slightly higher number of chFP-Q25 puncta compared with reporter cells grown alone or co-cultured with GFP-HttQ25 (), possibly due to inefficient transfer of aggregates through cell-cell contact or to uptake of aggregates released by GFP-HttQ71 cells. By contrast, we found that selective lysis of GFP-HttQ71 cells with puromycin (the reporter cells are genetically resistant to the drug) gave rise to a striking increase in the number of cells with chFP-HttQ25 puncta (). No such increase was observed when co-cultured HEK cells expressing GFP-HttQ25 were killed with puromycin (). The precise colocalization of chFP cytoplasmic puncta with GFP fluorescence strongly argues that HttQ71 aggregates from the lysed cells were internalized by the reporter cells and that the internalized aggregates are capable of nucleating homotypic aggregation of the GFP-HttQ25 reporter.
The data reported here establish that cell membranes are far more permeable to large fibrillar aggregates than hitherto suspected. Although other investigators have reported that fibrillar aggregates can be internalized by mammalian cells and sequestered within the endosomal/lysosomal compartment7-10
, our study directly demonstrates penetration of the cytoplasmic compartment by fibrillar aggregates. Perhaps large, rigid polyQ amyloid fibrils are able to physically breach biological membranes, as has been reported for spherical prefibrillar oligomers of amyloidogenic proteins19, 20
. It is unlikely in our experiments that fibrillization of polyQ aggregates occurs directly at the cell surface, as has been proposed for insulin associated polypeptide, IAPP21
, as our preparations are largely devoid of monomer. Further research is clearly necessary to understand the mechanism by which these large proteinaceous assemblies are able to gain access to the cytoplasm.
These findings thus have broad implications for understanding the pathogenesis of the large number of human disorders —including Alzheimer's disease and systemic amyloidosis— that are associated with the deposition of cytotoxic amyloids in the extracellular space. The recent discovery that alpha-synuclein-positive Lewy bodies can be propagated from diseased tissue to healthy transplanted fetal mesencephalic dopaminergic neurons in human Parkinson's disease brains11
strongly suggests that passage of aggregated pathogenic proteins between cells within an individual brain may contribute to the pathogenesis of Parkinson's disease. Our data suggest the possibility that this sort of mechanism may also contribute to polyglutamine diseases- and perhaps other neurodegenerative diseases as well.