Response to crowded conditions reveals compact nucleus for amyloid formation of folded protein

Abstract

Although the consequences of the crowded cell environments may affect protein folding, function and misfolding reactions, these processes are often studied in dilute solutions in vitro. We here used biophysical experiments to investigate the amyloid fibril formation process of the fish protein apo-β-parvalbumin in solvent conditions that mimic steric and solvation aspects of the in vivo milieu. Apo-β-parvalbumin is a folded protein that readily adopts an amyloid state via a nucleation–elongation mechanism. Aggregation experiments in the presence of macromolecular crowding agents (probing excluded volume, entropic effects) as well as small molecule osmolytes (probing solvation, enthalpic effects) revealed that both types of agents accelerate overall amyloid formation, but the elongation step was faster with macromolecular crowding agents but slower in the presence of osmolytes. The observations can be explained by the steric effects of excluded volume favoring assembled states and that amyloid nucleation does not involve monomer unfolding. In contrast, the solvation effects due to osmolyte presence promote nucleation but not elongation. Therefore, the amyloid-competent nuclei must be compact with less osmolytes excluded from the surface than either the folded monomers or amyloid fibers. We conclude that, in contrast to other amyloidogenic folded proteins, amyloid formation of apo-β-parvalbumin is accelerated by crowded cell-like conditions due to a nucleation process that does not involve large-scale protein unfolding.

Introduction

The intracellular environment is highly crowded, with an estimate of 80–400 mg ml⁻¹ macromolecules (proteins, lipids, polysaccharides), corresponding to a volume occupancy of 5–40% (Ellis and Minton, 2003; Kuznetsova et al., 2014). Also extracellular space is crowded, where for example blood plasma has a protein concentration of 60–80 mg ml⁻¹ (Leeman et al., 2018). High concentrations of macromolecules restrict the available molecular space by their mutual impenetrability and repulsive interactions, often termed excluded volume effect. The excluded volume effect is accompanied by increased opportunities for non-specific (chemical or ‘soft’) interactions, increased (heterogeneous) viscosity, reduced macromolecule dynamics, as well as changes in solvent polarity and water activity, which in total describes the macromolecular crowding effect (Gao and Winter, 2015; Davis and Gruebele, 2018). The excluded volume effect is an entropic effect in which (due to steric restrictions) compact and assembled protein species are favoured (Ralston, 1990; van den Berg et al., 1999; Minton, 2001, 2005a; Kuznetsova et al., 2014). Many in vitro studies of excluded volume effects (using so-called macromolecular crowding agents, for example, sugar-based polymers such as Ficoll and dextran) on protein folding and stability have shown increased folded-state stability and slower unfolding reactions in the presence of macromolecular crowding agents (van den Berg et al., 2000; Ellis, 2001; Christiansen and Wittung-Stafshede, 2013; Christiansen and Wittung-Stafshede, 2014). Notably, macromolecular crowding agents such as Ficoll and dextran are reported to be inert agents with little or no protein interactions; thus, they are useful to specifically study excluded volume effects (Christiansen and Wittung-Stafshede, 2014; Gao and Winter, 2015).

Cellular environments also contain high concentrations of small organic molecules termed osmolytes; so-called ‘protective’ osmolytes are often upregulated (de novo synthesis or import) to cope with water stress, there aiding in equilibration of cellular osmotic pressure (Yancey, 2005; Ignatova and Gierasch, 2006; Burg and Ferraris, 2008; Sukenik et al., 2013). High amounts of protecting osmolytes increase the viscosity but they also stabilize folded proteins via enthalpic effects as they are specifically excluded from protein surfaces (Timasheff, 1993, 2002). Thereby, protective osmolytes affect protein hydration and this results in stabilization of folded and compact protein structures that minimize interactions with the osmolyte, as compared to the many more interactions that are possible in the expanded unfolded state (Liu and Bolen, 1995; Ignatova and Gierasch, 2006; Auton and Bolen, 2007). Taken together, excluded volume effects are entropic in nature (and can be probed using macromolecular crowding agents), whereas solvation effects (probed via small molecule osmolytes) have enthalpic origin. Nonetheless, these systems are complicated and there is more to be revealed; there are for example studies implying that also osmolytes act via entropic forces (Ishrat et al., 2018).

Typically, macromolecular crowding conditions (large molecules), as well as osmolyte presence (small molecules), thermodynamically favour protein–protein interactions as assembled states will reduce the volume occupancy and number of species in solution. However, the effect on reaction kinetics will depend on conformation and surface-area exposure of rate-limiting steps. Here amyloid formation is interesting as this assembly process involves multiple steps and protein-conformational changes (Chiti and Dobson, 2017). Amyloid fibrils are biological polymers composed of monomeric protein units non-covalently assembled through β-strands arranged perpendicularly to the fibril axis, thereby forming a cross-β structure (Chiti and Dobson, 2017). Many proteins in all kingdoms of life (Otzen, 2010) can form amyloid fibrils at certain solvent conditions (Chiti and Dobson, 2017), and this process is directly related to several human diseases, such as Alzheimer’s disease, Parkinson’s disease and type 2 diabetes (Cooper et al., 1987; Jarrett et al., 1993; Wakabayashi et al., 1997). But there are also proteins that form functional amyloids: for example, biofilms are structures used by bacteria to adhere to surfaces which contain amyloids made of curli proteins (Andersson et al., 2013; Evans et al., 2015). Human functional amyloids include proteins in the signaling pathway leading to necrosis (Li et al., 2012) and in melanosome ultra-structures (Dean and Lee, 2019). Amyloid formation reactions of proteins that start from an intrinsically disordered monomeric state will necessarily involve compaction as the extended polypeptide will form β-structure. Thus for disordered/unfolded proteins, it is reasonable that macromolecular crowding via the excluded volume effect accelerates amyloid formation and this has been observed for aggregation of, for example, α-synuclein (Shtilerman et al., 2002; Uversky et al., 2002; White et al., 2010), the amyloid-β peptide (Lee et al., 2012) and a tau fragment (Zhou et al., 2009).

Folded proteins may become amyloidogenic as a result of ligand release, thermal fluctuations, as well as by local or global unfolding (Kuznetsova et al., 2014; Chiti and Dobson, 2017). It is also possible that assembly of folded proteins (native-like aggregates) occurs prior to structural rearrangements leading to amyloid fibers (Bouchard et al., 2000; Plakoutsi et al., 2004; Soldi et al., 2005; Chiti and Dobson, 2017). For folded proteins, it is less straightforward to predict the macromolecular crowding effect on amyloid formation, as the folded states will be stabilized (with respect to unfolding) by macromolecular crowding agents (Jahn and Radford, 2005; Kuznetsova et al., 2014). Indeed, several folded proteins exhibit increased resistance towards amyloid fibril formation in the presence of macromolecular crowding including carbonic anhydrase (Mittal and Singh, 2014), core histones and insulin (Munishkina et al., 2008). However, when placed in denaturing conditions, that is, low pH that partially unfolds the proteins and, for core histones and insulin, dissociates the native oligomeric states, macromolecular crowding was found to accelerate aggregation of the proteins mentioned above (Munishkina et al., 2008; Luo et al., 2016). Acceleration of aggregation by crowding was also shown for partially unfolded apolipoprotein C-II in the absence of lipids (Hatters et al., 2002). Notably, two of these three examples involve proteins with oligomeric native states, and amyloid formation involved oligomer dissociation into monomers as an initial requirement.

To test the effect of macromolecular crowding on an amyloidogenic protein that is folded and monomeric, we turned to G. morhua β-parvalbumin (β-PV). β-PV is an α-helical folded protein capable of binding Mg²⁺ and/or Ca²⁺ through two active EF-hand domains (Rall, 1996). Upon loss of the metal-ion ligands, which may occur due to low pH protonating the coordinating residues, or the chelating agent, ethylenediaminetetraacetic acid (EDTA), apo-β-PV readily forms amyloid fibrils (Martinez et al., 2015; Castellanos et al., 2018). We recently showed that apo-β-PV amyloid formation follows a nucleation–elongation mechanism where the nucleation step involves folded dimers connected by intermolecular disulfide bonds (Werner et al., 2020). The dimers are on path but need to further assemble and/or structurally convert to form competent nuclei that can then be elongated (by monomer and dimer additions) to amyloid fibrils. Thus, the amyloid formation process of apo-β-PV, initiated by assembly of folded proteins, is similar to that reported for only a few other proteins (Bouchard et al., 2000; Plakoutsi et al., 2004; Soldi et al., 2005). Since apo-β-PV is folded, one would predict that macromolecular crowding will slow down amyloid formation by stabilization of the folded monomer. Still, as a folded dimer is on path towards amyloid fibers (Werner et al., 2020), the presence of macromolecular crowding may favor dimers and thereby accelerate nucleation of the amyloid process.

The macromolecular crowding agent Ficoll 70 is a sucrose-based branched polymer often used to mimic excluded volume (Zhou et al., 2008; Wang et al., 2010). Although there may sometimes be exceptions (Nasreen et al., 2020), Ficoll 70 has been shown to only provide steric, excluded volume effects without chemical interactions with proteins (Christiansen et al., 2010; Christiansen and Wittung-Stafshede, 2013, 2014). Using Ficoll 70, we here show that amyloid formation of apo-β-PV, in contrast to data for folded oligomeric proteins, is promoted by macromolecular crowding. From additional protective osmolyte (glycerol and sucrose) kinetic experiments, it appears that the competent nucleus for amyloid formation is compact, and its formation does not involve large-scale conformational changes. However, addition of monomers during amyloid fibril elongation must involve larger structural conversions (resulting in increased exposure of surface area) as this is retarded by protective osmolytes.

Materials and method

Chemicals

Powders of sucrose (BioXtra, Sigma, St Louis, MO, USA), Ficoll PM 70 (Sigma, St Louis, MO, USA) (70 kDa) and poly(ethylene glycol) (PEG) 35,000 (BioUltra, Sigma, St Louis, MO, USA) (35 kDa) were dissolved in 25 mM Tris-HCl buffer, pH adjusted for 7.4 at 37°C and volume adjusted to desired concentration. Molecular biology grade glycerol (Fisher Scientific, Waltham MA, USA) was used.

Protein preparation

Just prior to aggregation assays of G. morhua β-parvalbumin (A51874, Gad m1), expressed and purified as described in Werner et al. (2020), the his-tag was removed by treating β-PV with thrombin (35.5 NIH units μmole⁻¹) for 4–5 h at room temperature and 800 RPM shaking followed by size exclusion chromatography (Superdex 75 10/300, GE healthcare, Uppsala, Sweden) in 25 mM Tris-HCl, supplemented with 1 mM CaCl₂ to stabilize the monomeric state.

Thioflavin T-aggregation assay

The ThT assay was performed in a quiescent condition at 37°C in 25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM CaCl₂, 5 mM EDTA, 7.7 mM NaN₃ and 20 μM recrystallized ThT in 96 well plates. Fluorescence was measured in a plate reader (Fluostar Optima or Fluostar Omega; BMG Labtech, Ortenberg, Germany). Emission was recorded at 480 nm with excitation at 440 nm every 20 min.

Atomic force microscopy

Samples of apo-β-PV fibrils were diluted with Milli-Q water (10 times) and incubated on a freshly cleaved mica for 15 min. The mica was subsequently rinsed five times and dried under a nitrogen gas flow. Intermittent contact mode atomic force microscopy (AFM) in air was used to record the images (NTEGRA Prima setup, NT-MDT) with a single crystal silicon cantilever (TipsNano, NSG01, force constant of ~5.1 N m⁻¹) with a resonance frequency of ~180 kHz. 512-pixel images were acquired with a 0.5-Hz scan rate. Images were analyzed using the WSxM 5.0 software (Horcas et al., 2007).

Circular dichroism spectroscopy

Far-UV circular dichroism (CD) spectra (190–260 nm) of PV monomers (apo- and holo-β-PV) were collected at room temperature (21 ± 1°C) using Chirascan CD spectropolarimeter (Applied Photophysics) in a 1 mm pathlength quartz cuvette. Three scans were collected and averaged with a time-per-point of 2 s, 1 nm bandwith and step size of 1 nm. The buffer signal was subtracted, and intensities converted from millidegrees to mean residue molar ellipticity (degrees M⁻¹ m⁻¹). CD spectra of amyloid fibrils were collected on post-aggregation samples that had been centrifuged at 100 000 g for 5 h, resuspended in 10 mM Tris-HCl, followed by 3 h centrifugation again and final resuspension into 10 mM Tris-HCl (pH 7.8).

Results and discussion

Amyloid formation of apo-β-PV (EDTA added at time zero to remove metal ions) in the presence of Ficoll 70 at 37°C was monitored by the Thioflavin T (ThT) fluorescence assay (Naiki et al., 1989; Xue et al., 2017). A significant increase in the speed of aggregation of apo-β-PV was observed when incubated in the presence of 100 and 200 mg ml⁻¹ Ficoll 70 (Fig. 1a), with a reduction of the half-time (time point when ThT fluorescence has reached 50% of its final value) from 32 ± 3 h (buffer), to 11 ± 1 h (100 mg ml⁻¹ Ficoll 70) and 4.5 ± 0.2 h (200 mg ml⁻¹ Ficoll 70). The resulting β-PV amyloid fibrils formed in the presence of Ficoll 70 were found to share the curvilinear appearance of apo-β-PV amyloid fibrils formed in buffer (Fig. 1b,c) and the far-UV CD spectra of amyloid fibers formed with and without Ficoll 70 were identical (Fig. 1d). From inspection of the data in Fig. 1a, it is evident that the lag time is drastically reduced and the slope of the fluorescence rise (i.e. tangent at half-time point) is larger for the two Ficoll 70 conditions. This implies that both nucleation and elongation processes are accelerated in macromolecular crowded solutions. This is in excellent agreement with excluded volume theory that predicts that assembled species are favored when the volume is reduced (Minton, 2005b; Ellis and Minton, 2006). Also more elaborate modelling (including both scaled-particle and transition-state theories) predicts this for systems where the size of the monomer is smaller when in the amyloid fiber than when in the free form (Schreck et al., 2020). Notably, 200 mg ml⁻¹ Ficoll 70 will occupy 13% of the volume based on its partial specific volume value (Christiansen et al., 2010; Aguilar et al., 2011) and thus the effective concentration of apo-β-PV is not 30 μM but 34 μM in the presence of 200 mg ml⁻¹ Ficoll 70. However, based on the earlier-determined concentration dependence of apo-β-PV aggregation (Werner et al., 2018), this small difference in protein concentration does not increase the aggregation speed significantly.

Fig. 1. Macromolecular crowding agents accelerate amyloid formation but do not affect fiber morphology. ThT fluorescence curves of 30 μM apo-β-PV at 37°C under quiescent conditions in 150 mM NaCl, 1 mM CaCl₂, 5 mM EDTA, 25 mM Tris-HCl, pH 7.4 and 0–200 mg ml⁻¹ Ficoll 70 (a) or 0–40 mg ml⁻¹ PEG 35,000 (e). AFM images of fibrils obtained from apo-β-PV alone (b) or in the presence of 100 mg ml⁻¹ Ficoll 70 (c). (d) CD spectra of resulting apo-β-PV amyloid fibers formed with (red) or without (blue) 100 mg ml⁻¹ Ficoll 70 (curves normalized to the same signal at the minimum). (f) CD spectra of 15 μM apo-β-PV at 21°C in 10 mM Tris-HCl, pH 7.8, 1 mM EDTA and 96 μM CaCl₂, with (red) 50 mg ml⁻¹ Ficoll 70, or (blue) 20 mg ml⁻¹ PEG, or (grey) without addition (i.e. all prior to aggregation).

Importantly, the Ficoll 70 result implies that the expected excluded-volume effect on the folded-to-unfolded transition of apo-protein monomers is not affecting amyloid formation. Unfortunately, we could not test the effect of Ficoll 70 on apo-β-PV monomer thermal or chemical equilibrium stability, as the apo-form is prone to aggregation during the required experimental time scales. Nonetheless, one expects macromolecular crowding to stabilize the compact folded state relative to the more expanded unfolded state (Christiansen et al., 2010; Stagg et al., 2011; Christiansen and Wittung-Stafshede, 2013; Aden and Wittung-Stafshede, 2014). Because some studies have shown macromolecular crowding to also modulate folded-state structure (Perham et al., 2007; Stagg et al., 2007; Homouz et al., 2008), we tested the influence of Ficoll 70 on the secondary structure of folded apo-β-PV using far-UV CD within time scales without aggregation. However, we found that the presence of Ficoll 70 did not affect the secondary structure of apo-β-PV (Fig. 1f). Taken together, the effect of Ficoll 70 on apo-β-PV aggregation agrees with folded dimers, stabilized by disulfide bonds (Werner et al., 2020), being key players in a nucleation process that occurs without monomer unfolding. The excluded volume effect due to Ficoll 70 presence favors dimer assembly and, thus, overall aggregation of apo-β-PV into amyloids is accelerated. Also, another macromolecular crowding agent, PEG, accelerated apo-β-PV amyloid formation (Fig. 1e) without affecting apo-β-PV folded structure (Fig. 1f). Although PEG may interact differently with proteins, and not only provide steric effects (Sukenik et al., 2011; Gao and Winter, 2015), the similar results found for PEG and Ficoll 70 (two agents with different chemistry) support that the observed kinetic acceleration is due to excluded volume. Notably, if attractive chemical (soft) interactions between protein and crowding agent were occurring, as noted in a few cases (Mikaelsson et al., 2014), such interactions would stabilize the folded monomers and thereby reduce aggregation (which is opposite to our findings).

To specifically test the effect of macromolecular crowding on the amyloid elongation step of apo-β-PV, we performed seeded reactions in which various amounts of preformed sonicated apo-β-PV amyloid fibril seeds were added to monomers. When high amounts of fibril seeds are added, only elongation takes place at early time points and therefore elongation can be exclusively probed. In Fig. 2, we show that seeded reactions (7.5, 15 and 24 μM monomer equivalents of fiber seeds added to 30 μM monomers) is accelerated in the presence of Ficoll 70, as compared to the corresponding seeded reactions in buffer. Again, this can be explained by the excluded volume effect (due to the presence of Ficoll 70) favoring assembled protein states. To elongate, apo-β-PV monomers must structurally convert from the native α-helical structure to the β-strand arrangement in the amyloid fibrils. The observed kinetic acceleration thus show that the macromolecular crowding effect on apo-β-PV monomer stability (folded versus unfolded states) does not kinetically limit the conformational transitions needed for the amyloid fiber elongation process. Analysis of the initial rates in the seeded apo-β-PV experiments reveals that the elongation rate constant is increased by a factor of 1.4–1.5 in the presence of 50 mg ml⁻¹ Ficoll 70 as compared to in buffer (Fig. 2d). This magnitude is in accord with scaled particle theory predictions (thus, based on entropic effects/excluded volume) of amyloid fibril growth: in (White et al., 2010) it was reported that experimental and predicted elongation rates for some different proteins were between 1.2 and 1.8 times faster in 50 mg ml⁻¹ of a macromolecular crowding agents as compared to in buffer.

Fig. 2. Ficoll 70 accelerates, but sucrose slows down, amyloid fiber elongation. ThT aggregation curves of 30 μM apo-β-PV at 37°C under quiescent conditions in 150 mM NaCl, 1 mM CaCl₂, 5 mM EDTA, 25 mM Tris-HCl, pH 7.4, with or without 50 mg ml⁻¹ Ficoll 70 or 380 mg ml⁻¹ sucrose, in the presence of 7.5 μM sonicated pre-formed fibrils (PFF) (a), 15 μM PFF (b) or 24 μM PFF (c). (d) Initial slope analysis of the curves (a–c) during the first 2.3 h.

To complement the Ficoll 70 results, which focus on excluded volume, we tested the effect of small molecule protective osmolytes (Kim et al., 2006) to assess solvent effects on apo-β-PV amyloid formation. Somewhat surprising, we found that the presence of both sucrose and glycerol (two protective small molecule osmolytes with different chemistry) resulted in faster apo-β-PV amyloid formation (Fig. 3a). The effect was concentration dependent, with higher sucrose or glycerol levels accelerating amyloid formation more. Importantly, the resulting apo-β-PV amyloid fibrils formed in presence of osmolytes appear like those formed in buffer and Ficoll 70 conditions according to far-UV CD (Fig. 3b) and AFM analysis (Fig. 3c,d) Notably, comparing the curves in Fig. 3a and Fig. 1a reveals that the aggregation lag times are not reduced as much, and the slopes (tangents) at the half time points are not as steep, as when apo-β-PV is placed in Ficoll 70, despite the fact that the concentrations of sucrose and glycerol used (in mg ml⁻¹ scale) are higher than the concentrations of Ficoll 70. For example, 380 mg ml⁻¹ sucrose reduced the half time point by a factor two, whereas 200 mg ml⁻¹ Ficoll 70 reduced the half time point by a factor of seven. Thus, excluded volume appears to have a larger effect than osmolyte-mediated hydration changes on the amyloid formation reaction. To better understand this difference, we assessed the effect of one of the osmolytes, sucrose, on the amyloid fibril elongation process of apo-β-PV by performing seeded amyloid formation reactions. We found that in contrast to Ficoll 70, the presence of sucrose reduces the elongation rate as compared to the corresponding reaction in buffer at all three amyloid fiber seed concentrations tested (Fig. 2).

Fig. 3. Osmolytes accelerate the overall amyloid reaction but do not affect amyloid fiber morphology. ThT fluorescence curves of 30 μM apo-β-PV at 37°C under quiescent conditions in 150 mM NaCl, 1 mM CaCl₂, 5 mM EDTA, 25 mM Tris-HCl, pH 7.4 and 0–390 mg ml⁻¹ sucrose (blue shades) or 0–566 mg ml⁻¹ glycerol (red shades) (a). CD spectra of resulting apo-β-PV amyloid fibers formed with (red) or without (blue) 190 mg ml⁻¹ sucrose (curves normalized to the same signal at the minimum) (b). AFM images of fibrils obtained from apo-β-PV in the presence of 380 mg ml⁻¹ sucrose (c) or 566 mg ml⁻¹ glycerol (d).

To explain the protective osmolyte results, that is, that non-seeded apo-β-PV amyloid formation is accelerated but apo-β-PV amyloid elongation is retarded by osmolytes, the nucleation step in the amyloid formation process must be promoted by the osmolytes. Protecting osmolytes such as glycerol and sucrose stabilize folded proteins due to their preferential exclusion from the protein surface (Timasheff, 2002; Ignatova and Gierasch, 2006; Kim et al., 2006). This effect will destabilize expanded unfolded states (that expose more surface area and thus require more osmolyte exclusion) but stabilize compact states of proteins. With this in mind, nucleation of apo-β-PV amyloid formation appears to involve compact structures that are stabilized by the osmolytes. The rate-limiting step for nuclei formation (i.e. the transition state) must also be stabilized by osmolyte presence to result in faster kinetics; thus, nucleation cannot involve large-scale protein unfolding. Again, this result is in accord with folded dimers being key species in apo-β-PV nuclei formation (Werner et al., 2020).

Fibril elongation involves the addition of monomers to the ends of growing amyloids with β-structure. Here structural conversions are needed as apo-β-PV monomers, to be added to growing fibers, are folded with α-helical structure. Conversion from α-helical to β-sheet structure will require transient exposure of an increased protein surface area, which will be disfavored in the presence of the osmolytes due to preferential exclusion effects. The retarding osmolyte effect on apo-β-PV amyloid fiber elongation kinetics agrees with findings on insulin and lysozyme at conditions where the proteins retained some folded structure. For both proteins, slower amyloid fibril elongation was detected in the presence of the protective osmolyte glycine (White et al., 2010).

Conclusions and outlook

Folded proteins, in addition to intrinsically disordered proteins, may form amyloid fibrils and this sometimes results in human diseases. In most cases, folded multimeric proteins must first dissociate to monomers, and the monomers must become structurally perturbed, for amyloid formation to occur. In such cases, macromolecular crowding, as found in living systems, will reduce the aggregation reaction as it stabilizes both folded proteins and multimeric states. Here we analyzed a folded monomeric protein that can assemble to amyloids. We show that the macromolecular crowding agent Ficoll 70 speed up amyloid formation although we start from the folded protein. The results are explained by the fact that apo-β-PV amyloid formation is not initiated by monomer unfolding but is instead promoted by folded dimers (Werner et al., 2020). The complementary experiments with small molecule osmolytes support the conclusion that no unfolding is involved in amyloid nucleation and further demonstrates that the nuclei are compact species stabilized in the osmolyte conditions. The fact that amyloid fiber elongation is accelerated by macromolecular crowding agents but retarded by osmolytes is explained by favorable excluded volume effects and unfavorable solvation effects, respectively, on monomer additions to growing amyloid fibers. Taken together, our study shows the first example of a folded monomeric protein that is augmented to adopt amyloid structures at conditions mimicking those found in living systems. The consequences of cell-like conditions on amyloid formation will depend crucially on mechanisms and structures of involved transient species: in the case of apo-β-PV, aggregation into amyloids is favored by the excluded volume effect as the process involves a compact nucleus that is reached without large-scale polypeptide unfolding.

Excluded volume, here mimicked by macromolecular crowding agents, is definitely an important factor of in vivo conditions. However, other aspects of the cell environment are also at play and may in some cases dominate. In addition to excluded volume and solvation effects (studied here), also nonspecific chemical interactions may take place between macromolecules and hydrophobic forces within macromolecules, for example, DNA, were shown to be affected by cell-like conditions (Feng et al., 2019). Many more systematic studies of these effects on the molecular-mechanistic level are desired, including studies using mixtures of macromolecular crowding agents, protein crowders, as well as assessing combinations of macromolecular crowding agents and osmolytes; all in combination with new theoretical approaches (Schreck et al., 2020). So far, known amyloidogenic proteins involved in human diseases are intrinsically disordered proteins or folded (often oligomeric) proteins that require partial unfolding/dissociation (i.e. some destabilization) to initiate amyloid formation. Perhaps we may eventually discover that folded proteins, without the need for destabilization, also participate in amyloid diseases, with amyloid initiation triggered by, for example, oxidative stress that cause covalent bonds between folded proteins. In fact, it may emerge that what we today consider ‘well-characterized’ human diseases may involve yet undiscovered amyloid formation that, in turn, is modulated by the cell environment (e.g. crowding, osmolytes).