An increasing body of evidence suggests that amyloid formation is the fundamental cause of protein deposition diseases1, 2. The nature of the pathogenic species and the mechanism by which the aggregation process results in cell damage are, however, the subject of intense debate3, 4, 5, 6, 7, 8, 9, 10. In systemic non-neurological diseases the accumulation of large quantities (sometimes kilograms) of aggregated species within a variety of organs and tissues may itself be the major cause of clinical symptoms11. In other cases, particularly the degenerative neurological diseases, it appears likely that impairment of cellular function is directly linked to the interaction of protein aggregates with cellular components12, 13. One specific indicator of the significance of aggregate formation in pathological conditions is the evidence for a high variability in the age of disease onset14, an observation that has recently been linked to the evidence that aggregates form by nucleation15. Further clues to the molecular basis of amyloid diseases and the biological significance of protein aggregation have been provided by recent observations that a range of proteins not associated with amyloid diseases are able to aggregate in vitro into fibrils indistinguishable from those found in pathologic conditions16, 17, 18, 19. This finding has led to the proposal that aggregation can be viewed as a general property of polypeptide chains rather than one restricted to a small number of sequences2. In the light of this conclusion we have now examined the effects on cell viability of aggregated species produced in vitro from two such proteins.
The SH3 domain from bovine phosphatidyl-inositol-3'-kinase (PI3-SH3) and the N-terminal ('acylphosphatase-like') domain of the E. coli HypF protein (HypF-N) are two examples of small globular proteins that can form fibrillar aggregates in vitro under appropriate conditions17, 20. Evidence that the aggregates formed from PI3-SH3 and HypF-N can be classified as amyloid fibrils has been obtained from electron microscopy, specific tests such as Congo red and thioflavine T binding and, in the case of PI3-SH3, X-ray diffraction17, 20. The heterogeneous nature of the in vitro aggregation process is a potential difficulty in experiments aimed at probing the nature of aggregate pathogenicity; this problem can hinder the identification of the particular species responsible for any observed toxicity. We have found, however, that highly homogeneous populations of various types of aggregates of PI3-SH3 or HypF-N can be obtained by incubating either protein under well defined solution conditions for specific lengths of time. These sequentially and structurally unrelated proteins are therefore exceptionally favourable systems for investigating any inherent cytotoxicity of specific types of proteinaceous aggregates.
Cytotoxicity of PI3-SH3 aggregates
Incubation of PI3-SH3 in either a 50-mM acetate buffer solution, pH 5.5, containing 25% (v/v) trifluoroethanol (TFE) or a H2O/HCl mixture, pH 2.0, results in the formation of granular or fibrillar aggregates, respectively (Fig. 1). Both types of aggregates enhance the fluorescence of thioflavin T (ThT) by factors of more than 30, indicating the presence of amyloid-like features within these structures (data not shown). Examination by transmission electron microscopy (TEM) reveals that the aggregates formed rapidly at pH 5.5 in solutions containing TFE appear as granules 4–200 nm in diameter (Fig. 1d–f) without any detectable fibrillar species. Prolonged incubation at pH 2.0, however, yields 7–13-nm-wide fibrils in the absence of any detectable granular aggregates (Fig. 1b); the width of the fibrils observed here is characteristic of the ex vivo amyloid fibrils extracted from patients suffering from various forms of amyloid disorders21. Some fibrils of this type formed from PI3-SH3 have been shown to consist of a double helical arrangement of two pairs of ribbon-like protofilaments wound around a hollow core22. In the samples analysed here, such fibrils are occasionally found to split into two 6–7-nm-wide fibrils or to associate further in a parallel fashion to produce 23- or 33-nm-wide sheet-like fibrils, also in agreement with previous findings22.
Figure 1: PI3-SH3 aggregation and cytotoxicity.
a, c, Cytotoxic effect of fibrillar (a) and granular (c) PI3-SH3 aggregates on NIH-3T3 cells. Reported values are those after treatment with 20 
M native protein (Native), with cell medium containing an aliquot of the solution where aggregates were formed (HCl or TFE) or with aggregates at the indicated protein concentrations (PI3-SH3). Values are relative to those of control cells treated with complete medium alone. b, d–f, Electron micrographs showing fibrillar (b) and granular (d–f) aggregates formed from PI3-SH3. Labels 1 and 2 indicate large granules and clusters of small granules, respectively (e, f).
The cytotoxicity of the two types of aggregates formed in such experiments was examined by adding aliquots of the aggregates, at a range of final protein concentrations (see Fig. 1 legend), to cell culture media. Aggregate cytotoxicity was evaluated by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reduction inhibition assay, a standard indicator of cell physiological stress thought to be related to changes in intracellular trafficking, particularly in the pathway of exocytosis23, 24. The experiments reveal that the highly structured PI3-SH3 fibrils formed by prolonged incubation at pH 2 do not significantly modify MTT reduction in either NIH-3T3 or PC12 cells even at the highest concentration tested (Fig. 1a). The presence of the granular aggregates formed after short incubation periods, however, significantly reduces cell viability (Fig. 1c, P < 0.01 at PI3-SH3 concentrations above 1 M). No significant decrease of MTT reduction was detected when the cells were exposed either to 20 M native PI3-SH3 or to the buffer solutions used to form the aggregates in the absence of added protein (Fig. 1).
Cytotoxicity of HypF-N aggregates
In a second series of experiments, the toxicity of aggregates formed by the protein HypF-N was examined. HypF-N was incubated at 0.3 mg ml-1 concentration in 50 mM acetate buffer, pH 5.5, in the presence of 30% (v/v) TFE at room temperature; aliquots of this solution were withdrawn at regular time intervals for examination by TEM and for ThT and cytotoxicity assays on cultured NIH-3T3 and PC12 cells. Under these conditions, aggregates develop within minutes; these initial aggregates exhibit both a CD (circular dichroism) spectrum indicative of 
-sheet structure and an enhancement of the ThT fluorescence by over 50 times relative to that of the soluble native domain. At this stage of incubation, however, the aggregates appear completely non-fibrillar and non-granular in TEM images (Fig. 2b), whereas after 48 h of incubation some fibrillar character is clearly evident (Fig. 2c). The latter aggregates have a width of about 4–8 nm and are extremely short (typically 25–60 nm), resembling the protofibrils observed with other proteins at relatively early stages of fibril formation25. The ends of these aggregates appear disordered, suggesting they are in the process of undergoing a transition from amorphous to fibrillar structures. After about 20 days, aggregates of this type can no longer be detected in the samples; at this time all the aggregates appear as long unbranched fibrils with widths of either 3–5 or 7–9 nm (Fig. 2d, e), characteristic of constituent protofilaments and mature amyloid fibrils, respectively20.
Figure 2: HypF-N aggregation and cytotoxicity.
a, Differential toxicity of amorphous, protofibrillar and fibrillar HypF-N aggregates. Cell viability was checked by the MTT inhibition reduction test, after addition to the cell medium of either 20 M native protein (black circles) or different concentrations of protein incubated in TFE: 20 M (blue), 5 M (red), 1 M (green), 0.2 M (purple) and 0.04 M (cyan). Values are relative to control cells treated with complete medium alone. The electron micrographs show amorphous aggregates of HypF-N, formed after 6 h incubation (b), amorphous aggregates developing into fibrils after 48 h incubation (c) and mature amyloid protofilaments (d) and fibrils (e) after 20 days incubation.
Addition to the cell medium of the aggregates formed after a 6 h incubation period resulted in a marked decrease in the MTT reduction by both NIH-3T3 (Fig. 2a) and PC12 cells (data not shown). This decrease is statistically highly significant at all protein concentrations tested (0.04–20 M; P < 0.01 at 0.04 M) with respect to controls performed by incubating the same cells with 20 M HypF-N in its soluble form or with the buffer solutions in the absence of protein. A series of experiments revealed that aggregate toxicity depends on protein concentration and initially increases with the length of exposure to the conditions that result in aggregation, reaching a maximum after 48 h (Fig. 2a). Incubating cells with protein aggregates formed at this time results in an inhibition of MTT reduction ranging from 20% at the lowest protein concentration used here (0.04 M) to 70% at the highest protein concentration (20 M) with respect to the control experiments. In addition to impairing cellular function, HypF-N aggregates were also found to lead to cell death, as revealed by the trypan blue internalization test (Fig. 3). The results show that the rate of cell mortality increases until it reaches nearly 40% at the highest protein concentrations, indicating that the cellular dysfunction shown by the MTT assay leads to cell death. Addition to the cell medium of protein samples incubated for times longer than 48 h, however, resulted in a progressive decrease of cell impairment that correlates, at least qualitatively, with the increase in the proportion of mature fibrils relative to other aggregates in the protein samples. Even at the highest protein concentrations analysed, MTT reduction approaches the value of the controls when the cells were exposed to protein samples incubated for 21 days (Fig. 2a), conditions where only mature fibrils are evident in the TEM micrographs. A lack of toxicity was also found when cells were exposed to well defined amyloid fibrils formed at low pH (50 mM citric acid, pH 3.0) (data not shown).
Figure 3: Percentage of cell deaths induced by 48-h-aged HypF-N aggregates at different protein concentrations.
Solid bars refer to aggregated protein, dotted bars to control experiments performed in the presence of soluble protein (see Methods for details).
High resolution image and legend (24K)The origins of aggregate toxicity
The results of the present study indicate that aggregates formed from PI-SH3 and HypF-N can be substantially cytotoxic. In both cases, the cytotoxicity was found to depend on the supramolecular organization of the amyloid aggregates and is much more pronounced for the rapidly formed non-fibrillar aggregates than for the highly organized fibrillar structures. The results for these two proteins, neither of which is disease-related, are very similar to those reported for the disease-associated fusogenic prion protein fragment, for 
-synuclein, for A(1–42) and for transthyretin3, 4, 5, 6, 7, 8, 9, 10, 26. Remarkably, the levels of cell impairment induced by the most toxic species found in the present study are comparable to those of the highly toxic aggregates formed from A(1–42) (ref. 27). The data, therefore, suggest that impairment of cell viability by protein aggregates of the type that can subsequently form amyloid fibrils could be a general phenomenon and not simply a specific property of the small number of polypeptides associated with clinically recognized protein deposition diseases. This result is of particular significance in the light of the recent conclusion that the ability to form highly ordered amyloid fibrils is itself a generic property of proteins17, 18, 19. In addition, the mature fibrils of the proteins we tested appear to be essentially harmless to cells, providing an explanation for previous data indicating that well defined amyloid fibrils formed by a synthetic peptide are not cytotoxic27 and reinforcing the findings on the cytotoxicity of A, -synuclein and transthyretin pre-fibrillar assemblies both in cultured cells and in transgenic mice3, 4, 5, 9, 10, 26, 28.
The results reported in this paper, therefore, provide evidence that protein aggregates not associated with disease can be inherently cytotoxic and result in substantial impairment of cellular function or even cell death. There is a considerable debate as to whether fully formed mature amyloid fibrils or rapidly formed aggregates that precede their formation are the primary pathogenic species responsible for the onset of disease3, 4, 5, 6, 7, 8, 9, 10. Although there is evidence for the toxicity of mature fibrils in some amyloid diseases, the data reported here support previous suggestions that, at least in some cases, the non-fibrillar aggregates that precede formation of mature amyloid fibrils may be the primary toxic species3, 4, 5, 9, 10, 26, 28. This toxicity is likely to arise because in these early aggregates hydrophobic side-chains and other regions of the polypeptide chain will be much more accessible than in the fully formed mature fibrils. Indeed, the latter are often found to be remarkably inert, for example in their resistance to proteolysis and degradation29.
The inherent cytotoxicity of the aggregates formed by the two non-disease-related proteins studied here suggests not only that the toxicity of aggregated species could be a general phenomenon, but also that the pathogenicity of protein-deposition diseases could be primarily related to the structural nature of the aggregates rather than to the specific sequences of the proteins from which they arise. We suggest that toxicity could primarily arise because on the surface of disordered aggregates there is likely to be a combinatorial display of amino acids enabling these species to interact inappropriately with a wide range of cellular components. Such a conclusion further suggests that the differing clinical manifestations of amyloid formation, ranging from neuronal cell death to the accumulation of large quantities of proteinaceous material, could arise in large part because of variations in the nature and morphologies of the specific aggregates as they form in the different diseases, as well as on the susceptibility of different cell types to the various aggregates. Such variations will occur as a consequence of the specific character of the different proteins involved and their location, and also on the different conditions under which aggregation takes place.
Implications for misfolding diseases and biological evolution
The present findings, that early aggregates formed by a wider range of proteins than those known to be associated with neurological diseases can be cytotoxic, provide new opportunities to define the nature of amyloid diseases and the mechanism of amyloid toxicity at the molecular level. They also raise the possibility that trace amounts of aggregates of a variety of proteins might occur spontaneously, particularly during ageing, and that such aggregates could account for subtle impairments of cellular function in the absence of an evident amyloid phenotype. It would thus be interesting to search for early protein aggregates in systemic and neurologic disorders not presently associated with amyloid formation. More generally, knowledge of the origin and nature of aggregate pathogenicity is of crucial importance in efforts to identify the correct targets for drug design in the search for effective therapeutic protocols.
The inherent toxicity in protein aggregates could also help us to understand fundamental aspects of cell biology. It suggests, for example, that avoidance of aggregation could be more important for the proper functioning of biological organisms than was previously suspected if aggregates of proteins are often toxic rather than simply non-functional. In this case, in addition to increasing the efficiency of folding and rescuing misfolded proteins after biosynthesis30, evolutionary developments to prevent aggregate formation, notably molecular chaperones, ubiquitination enzymes and proteasomes, are needed for the preservation of the long-term viability of living organisms. This latter idea is reinforced by recent findings concerning the relationships between neurodegenerative diseases and failure of cellular defence mechanisms targetted towards misfolded proteins and the existence, in both prokaryotic and eukaryotic cells, of a complex regulatory system of intracellular protein degradation (see ref. 31 and references therein). Such results, together with findings that aggregate formation is linked to the inheritance of specific traits in organisms such as yeasts32, provides increasing evidence that the control of protein misfolding and aggregation in addition to being of fundamental importance for cell viability, has been a major driving force in biological evolution.
Methods
Protein aggregate production
PI3-SH3 was expressed and purified as previously reported33. Granular aggregates were formed incubating the protein for 1 h at 20 °C at a concentration of 10 mg ml-1 in 50 mM acetate buffer, pH 5.5, containing 25% (v/v) TFE. Fibrillar aggregates were grown for 1 month at a protein concentration of 10 mg ml-1 in a water/HCl mixture, pH 2.0, at 37 °C. Conditions for HypF-N purification and aggregation have been described previously20.
Cell culture
NIH-3T3 cells (mouse fibroblasts, American Type Culture Collection) were routinely cultured in Dulbecco's modified Eagle's medium (Gibco BRL) containing 10.0% bovine calf serum and 3.0 mM glutamine in a 5.0% CO2 humidified environment, at 37 °C. PC12 cells (rat pheochromocytoma, American Type Culture Collection) were cultured in RPMI medium (Gibco BRL) supplemented with 10.0% horse serum, 5.0% fetal bovine serum and 3.0 mM glutamine in a 5.0% CO2 humidified atmosphere at 37 °C. 100 U ml-1 penicillin and 100 g ml-1 streptomycin were added to both media. Cells were used for a maximum of 20 passages. NIH-3T3 or PC12 cells were plated at a density of 10,000 cells per well on 96-well plates in 100 l of fresh medium. After 24 h, the NIH-3T3 medium was exchanged with 100 l of DMEM, without phenol red, containing 10.0% bovine calf serum, and the PC12 medium was changed with 100 l of RPMI, without phenol red, supplemented with 10.0% horse serum and 5.0% fetal bovine serum.
Incubation of cells in the presence of protein aggregates
Aliquots of solutions containing native or aggregated proteins were centrifuged, dried under N2 to remove TFE when necessary, dissolved in RPMI without phenol red and immediately added to the cell media at 0.1–20 M (PI3-SH3) or 0.04–20.0 M (HypF-N) final concentrations. After 24 h incubation, 10 l of a stock MTT solution in PBS was added to give a final concentration of 0.5 mg ml-1 and incubated for a further 4 h. 100 l of cell lysis buffer (20.0% SDS, 50.0% N,N-dimethylformamide, pH 4.7) was added to each well and the samples were incubated overnight at 37 °C in an humidified incubator. Absorbance values of blue formazan were determined at 590 nm with an automatic plate reader. Cell death was assessed by the trypan blue internalization test34. After a 24 h incubation with 48 h aged HypF-N in TFE or with the soluble domain, NIH-3T3 cells were treated with trypan blue and survival was quantified by counting (three fields per well, two wells per condition, an average of 50 cells per field).
Electron microscopy
TEM images were acquired using a JEM 1010 transmission electron microscope at 80 kV excitation voltage. In each case, 3.0 l of protein solution was placed on a formvar and carbon-coated grid and blotted off after 3 min. The sample was then stained with 3 l of 2.0% uranyl acetate, dried and observed at a magnification of 12–30,000.
ThT staining
60 l aliquots of protein solution were mixed with 0.44 ml of 25 M ThT in 25 mM phosphate buffer, pH 6.0, and the resulting fluorescence measured immediately after mixing using a Shimadzu RF-5000 spectrofluorimeter at excitation and emission wavelengths of 440 and 485 nm, respectively.
-42 are potent central nervous system neurotoxins. Proc. Natl Acad. Sci. USA 95, 6448-6453 (1998). | 
