Osmolytes and crowders regulate aggregation of the cancer-related L106R mutant of the Axin protein

Abstract

Protein aggregation is involved in a variety of diseases, including neurodegenerative diseases and cancer. The cellular environment is crowded by a plethora of cosolutes comprising small molecules and biomacromolecules at high concentrations, which may influence the aggregation of proteins in vivo. To account for the effect of cosolutes on cancer-related protein aggregation, we studied their effect on the aggregation of the cancer-related L106R mutant of the Axin protein. Axin is a key player in the Wnt signaling pathway, and the L106R mutation in its RGS domain results in a native molten globule that tends to form native-like aggregates. This results in uncontrolled activation of the Wnt signaling pathway, leading to cancer. We monitored the aggregation process of Axin RGS L106R in vitro in the presence of a wide ensemble of cosolutes including polyols, amino acids, betaine, and polyethylene glycol crowders. Except myo-inositol, all polyols decreased RGS L106R aggregation, with carbohydrates exerting the strongest inhibition. Conversely, betaine and polyethylene glycols enhanced aggregation. These results are consistent with the reported effects of osmolytes and crowders on the stability of molten globular proteins and with both amorphous and amyloid aggregation mechanisms. We suggest a model of Axin L106R aggregation in vivo, whereby molecularly small osmolytes keep the protein as a free soluble molecule but the increased crowding of the bound state by macromolecules induces its aggregation at the nanoscale. To our knowledge, this is the first systematic study on the effect of osmolytes and crowders on a process of native-like aggregation involved in pathology, as it sheds light on the contribution of cosolutes to the onset of cancer as a protein misfolding disease and on the relevance of aggregation in the molecular etiology of cancer.

Significance

This work demonstrates the effect of osmolytes and macromolecular crowders on the aggregation of a protein involved in cancer. Our findings begin to answer in what way the molecular environment within the cell may determine and regulate the onset of cancers. This has special relevance considering that on the one hand, tumors often correlate with metabolic dysfunctions and cancer cells display accelerated metabolism and biosynthesis, whereas on the other hand, the relation between molecular crowding and cancer is unexplored. The model for Axin L106R aggregation provided here may serve to interpret similar aggregative processes of other cancer-related proteins. More specifically, this work elucidates the molecular mechanism of the hepatocellular cancer correlating with the L106R mutation, potentially paving the way for future tailored therapeutic strategies.

Introduction

Protein aggregation is a widespread phenomenon that results from an unbalance of proteostasis (1). Starvation, thermal or oxidative stresses, and aging-related dysfunctions in the proteosome are known to trigger aggregative events (1). The mechanism of aggregation and the structure of the aggregates define the biological process outcomes. Several forms of aggregates have been described: 1) native aggregates, which include proteins that retain their native fold and are stored within the native aggregates as a protective mechanism. The protein functionality is maintained within these aggregates or regained upon release (2); 2) native-like aggregates, which include proteins undergoing a limited aberration of their folding that can assemble into these native-like aggregates while retaining most of their native structure but with altered function (3); 3) amorphous aggregates, which include proteins that undergo larger modifications to native folding that can induce collapse into amorphous aggregates lacking a well-defined structure, with a complete loss of function (4); and 4) amyloid aggregation, whereby proteins can undergo extensive structural rearrangements to β-sheet-rich structures that give rise to amyloid aggregates characterized by a cross-β structure and an unbranched fibrillar morphology (5). The formation of mature intra- or extracellular amyloid deposits is preceded and accompanied by soluble intermediates such as transient amyloid oligomers, characterized by heterogeneous structure and inherent toxicity (6).

The implications of protein aggregation for health are tremendous. The pathogenesis of a vast ensemble of neurodegenerative and systemic diseases correlates specifically with the appearance of amyloid aggregates (5). Moreover, the etiology of various types of cancer is traced back to either amyloid or native-like aggregation of proteins that regulate key processes for cell homeostasis. For instance, misfolded p53 mutants undergo amyloid fibrillation via prion-like propagation, causing related forms of cancer (7). In another mechanism, destabilized p53 cancer mutants coaggregate with its paralogue proteins p63 and p73 to form soluble amyloid-like assemblies that are unable to exert the tumor suppressor function (8).

A newly reported case of protein aggregation in cancer is represented by Axin, the scaffold protein that orchestrates Wnt signaling in the cytoplasm (9). The L106R cancer mutation destabilizes the RGS domain of Axin and leads to the aberrant assembly of a soluble, multiprotein nanoaggregate that is unable to carry out the Wnt-suppressive function of the wild-type (WT) protein (10). The uncontrolled Wnt activation then promotes cell proliferation and induces the onset of a cancer phenotype (10).

Protein-protein interaction (PPIs), including those involved in aggregation, are widely affected by the complexity of the cellular environment (11). The cytoplasm contains a variety of small molecules such as nutrients and metabolites at concentrations that are on the order of hundreds of millimolar or more. We collectively refer to these molecules here as osmolytes because of their involvement in the cellular response to osmotic stress (12). Proteins and other macromolecules also occupy a large volume fraction and contribute to creating a highly crowded environment in the cytoplasm. We use the term “macromolecular crowders” to designate the excluded-volume effect of this class of molecules (13). Via their preferential exclusion from protein surfaces and other quinary interactions, osmolytes and crowders affect protein structure and dynamics, modulate PPIs, and drive association phenomena, including aggregation (14, 15, 16, 17, 18).

Osmolytes typically exert an enthalpically driven stabilization of compact folded conformations that minimizes the extent of the exposed protein surface and its hydration shell (15). Thus, osmolytes are often able to oppose amyloid aggregation of structured and disordered proteins alike (19). An investigation on the mechanism of inhibition using a model peptide revealed that polyols inhibit all stages of amyloid aggregation by favoring the adoption of a folded β-hairpin state over an ensemble of unfolded states (20). Yet, the extent of inhibition varies with the chemistry of the osmolyte and the protein. For instance, different polyols affect distinct stages of insulin fibrillation, and trehalose inhibits Aβ40 oligomers and fibrils but only suppresses Aβ42 fibrils and expedites the onset of α-synuclein early aggregation while decreasing the total amount of fibrillated protein (21, 22, 23). Polyols exert a similar protective effect on the heat or pH-induced amorphous aggregation of folded proteins like conalbumin and monoclonal antibodies (24, 25, 26). Amino acids display a varied effect. Proline suppresses the fibrillation of insulin, whereas it redirects huntingtin amyloidogenesis to amorphous aggregation (27,28). At low concentrations, it enhances the thermal aggregation of a folded enzyme, whereas at high ones, it has an inhibitory effect (29). Betaine also abolishes insulin fibrillation but, unlike proline, substantially speeds up that of huntingtin (27,28).

In contrast to osmolytes, the exclusion of macromolecular crowders from protein surfaces is typically dominated by entropy and favors the protein compact states so as to maximize the cosolutes free volume (13). Crowders can destabilize folded states and enhance the aggregation of a large number of proteins, diversely modulating the molecular mechanism of the process (30). For instance, they speed up the nucleation rate of the amyloid aggregation of a model peptide and promote the fragmentation of the mature fibrils, which seeds further aggregative events (31). Moreover, polyethylene glycol (PEG) crowders trigger the liquid-liquid separation of α-synuclein, which readily converts into an amyloid hydrogel composed of oligomers and fibrils (32). Yet, although increasing both nucleation and fibrillation rate of monomeric α-synuclein, crowders stabilize the native hexamers and tetramers of human insulin and bovine core histone, which are also compact states, thereby delaying their further aggregation (33). Crowders can assist functional self-assembly processes that contribute to correct proteostasis (30). For instance, they facilitate the complexation of the molecular chaperones GroEL and DnaK/ClpB with unfolding and aggregating clients and induce the association of folded proteins into functional polymers or native aggregates (34,35). Based on these observations, it appears that the cosolutes from these two classes, osmolytes and crowders, tend to exert an effect on protein aggregation that is independent of the mechanism and of the specific type of the aggregate.

The effects of cosolutes on protein aggregation have so far been studied in processes that yield large, insoluble aggregates. How the cellular environment affects the assembly of soluble, aberrant complexes of misfolded proteins (or “nanoaggregates”) is still unknown despite their involvement in disease. To address this question, we assessed the impact of osmolytes and macromolecular crowders on the nanoaggregation of Axin RGS L106R. Some cancers also possess the characteristics of a protein aggregation disease, as cancer cells are hypercrowded environments because of an accelerated metabolism (36). We characterized the RGS L106R aggregation as a native-like process and found that polyols and amino acids generally inhibited it, with sugars and some amino acids acting as suppressors. To contrast, betaine and PEG crowders facilitated RGS L106R aggregation. Based on our results, we propose a mechanism for Axin L106R aggregation in vivo in which osmolytes maintain the protein in the monomeric state as long as it is unbound. Upon recruitment into a multiprotein complex, the increased crowding within the complex triggers the aggregative events that turn it into a dysfunctional yet soluble nanoaggregate.

Materials and methods

Site-directed mutagenesis and cloning

The pGEX-2T vector was modified with a TEV cleaving site by site-directed mutagenesis using the following primers:

• forward—5′-ggcgaccatcctccaaaagagaatctgtatttccagggatccaccgcc-3′ and

• reverse—5′-ggcggtggatccctggaaatacagattctcttttggaggatggtcgcc-3′.

The genes encoding RGS WT and RGS L106R were PCR amplified, purified by clean-up kits (Promega, Madison, WI), and ligated into the TEV-modified pGEX-2T vector.

Protein expression and purification

BL21 codon plus cells were transformed with TEV-modified pGEX-2T plasmids containing RGS WT or L106R inserts. The starter was diluted 1:200 (v/v) in LB broth media and grown under agitation at 37°C in the presence of 100 μM ampicillin and 35 μM chloramphenicol until [optical density]₆₀₀ of 0.6 was reached. The expression was induced overnight with 0.6 mM isopropyl β-D-1-thiogalactopyranoside at 16°C under agitation. The bacteria were harvested at 4500 rotations per minute for 15 min, and the pellet was resuspended in 50 mM TrisHCl (pH 7.4), 300 mM NaCl, 5 mM β-mercaptoethanol, 10 μg/mL DNAase, 10 μg/mL lysozyme, and 1 mM phenylmethanesulfonyl fluoride. The bacteria were lysed with a microfluidizer, and the lysate was separated at 15,000 rotations per minute for 45 min. The lysate was purified using a Merck glutathione S-transferase (GST; Merck, Darmstadt, Germany)-bind 70541 resin and washed with 50 mM TrisHCl (pH 7.4), 150 mM NaCl, and 5 mM β-mercaptoethanol. RGS WT and L106R were cleaved on the resin with 2.2 μM TEV protease in 5 mL buffer upon overnight incubation at 4°C under mild agitation. The cleaved protein was further purified by ÄKTA (GE Healthcare, Chicago, IL)-assisted size exclusion chromatography using two coupled Superdex75 and Superose12 (GE Healthcare) 200 mL columns. The purity was confirmed by SDS-page. The protein was concentrated with Vivaspin 20 5000 molecular weight cut-off centrifuge tubes (GE Healthcare), filtered with FILTSTAR hydrophilic nylon filters 0.22 μm cutoff and stored at −80°C.

UV spectroscopy

UV spectra were recorded with a Shimadzu UV-1650PC spectrophotometer (Kyoto, Japan) using a quartz cuvette of 0.1 cm pathlength for far-UV spectroscopy. The extinction coefficient at 280 nm (ε₂₈₀) of RGS WT and L106R was 22920 M⁻¹ cm⁻¹, as computed by ExPASy Protparam. The extinction coefficient at 360 nm (ε₃₆₀) of 8-anilinonaphthalene-1-sulfonic acid (ANS) was 5700 M⁻¹ cm⁻¹ (37).

Circular dichroism

The circular dichroism (CD) spectra were recorded with a J-810 spectropolarimeter (JASCO, Tokyo, Japan) equipped with a Peltier thermostat (Jasco PTC-423S; JASCO) and Spectra Manager software in a 0.1 cm pathlength quartz cuvette for far-UV CD spectroscopy (in the spectral range between 190 and 260 nm) at 4°C. The aggregation kinetics were monitored by acquiring CD spectra at 0, 15, 30, 45, 60, 90, 120, 180, 240, 360, 720, and 1440 min at 25°C. The melting curves were acquired by measuring the CD at 222 nm with a temperature gradient of 50°C per hour from 4 to 95°C, with readings taken every 0.1°C. The fraction of unfolded protein was determined by fitting the normalized data with the sigmoidal equation

$${\chi }_{unfolded}=1–1/\left(1+\left(\mathrm{e}\mathrm{x}\mathrm{p}\left(\left(\mathrm{T}–{\mathrm{T}}_m\right)/t\right)\right)\right)$$

where T_m was the temperature at which half of the protein is unfolded and t was a constant describing the slope of the curve. In all solutions, the conditions were 8 or 16 μM protein, 50 mM TrisHCl (pH 7.4), 150 mM NaCl, and 5 mM β-mercaptoethanol.

ANS fluorescence

Fluorescence spectra were recorded with a Perkin Elmer LS 55 fluorimeter (Waltham, MA) using a quartz cuvette of 1 cm pathlength for far-UV fluorescence spectroscopy, with excitation at 380 nm and emission between 400 and 600 nm, at 4°C. 300 μL of 20 μM protein solution were diluted with additions of an ANS solution to reach protein/ANS molar ratios between 1:1 and 1:12. The aggregation kinetics were acquired with a BioTek Synergy H1 Hybrid Reader (Thermo Fisher Scientific, Waltham, MA) plate reader using 96-well half-area plates (Costar 3696; Sigmal-Aldrich, St. Louis, MO) for 100 μL volume solutions, with excitation at 380 nm and emission at 470 nm, at 25°C. 25 μL of 32 μM protein solution were added to 75 μL aggregating solution to reach 8 μM protein, 24 μM ANS, and 750 mM or no cosolutes, pipetting 10 times to homogenize. In all solutions, the buffer was 50 mM TrisHCl (pH 7.4), 150 mM NaCl, and 5 mM β-mercaptoethanol.

Dynamic light scattering

The dynamic light scattering (DLS) was measured with a Malvern Zetasizer Nano ZS (Malvern, UK) using disposable plastics cuvettes of 1 cm pathlength. The aggregation kinetics were monitored by acquiring the DLS at 0, 15, 30, 45 60, 90, 120, 180, 240, 360, 720, and 1440 min at 25°C. In all solutions, the conditions were 24 μM protein, 50 mM TrisHCl (pH 7.4), 150 mM NaCl, and 5 mM β-mercaptoethanol.

Transmission electron microscopy

8 μM protein were incubated with or without 750 mM cosolute for 12 h at 25°C. The aggregation reaction solution was diluted 10× before sample preparation. 3–5 μL sample was applied to a glow-discharged transmission electron microscopy (TEM) grid (carbon supported film on 300 mesh Cu grids; Ted Pella, Redding, CA). After 30 s, the excess liquid was blotted, and the grids were stained with 2% uranyl acetate for 30–60 s, blotted, and allowed to dry in air. The samples were examined using an FEI Tecnai 12 G2 TWIN TEM (Hillsboro, OR) operated at 120 kV. The images were recorded by a 4000 × 4000 FEI Eagle CCD camera. In all solutions the buffer was 50 mM TrisHCl (pH 7.4), 150 mM NaCl, and 5 mM β-mercaptoethanol.

Results

The L106R mutation induces partial unfolding of the RGS domain

Axin RGS domain WT and bearing the L106R cancer-related mutation were expressed as fusion proteins with GST. After cleavage of the GST-RGS constructs with TEV protease, RGS WT and L106R were purified as described in the Materials and methods. We performed a comprehensive biophysical characterization of RGS L106R and its aggregation in comparison to RGS WT. CD spectra of RGS L106R and RGS WT displayed minima at 208 and 222 nm that are typical of an α-helical secondary structure. The weaker intensity of the RGS L106R CD spectrum as compared with that of the WT protein correlated with a partial unfolding induced by the cancer-related L106R mutation (Fig. 1 A). The melting temperature of the RGS domain dropped from 56.1 ± 0.1°C for the WT to 28.3 ± 0.1°C for the L106R mutant, indicating a loss of thermodynamic stability upon mutation (Fig. 1 B; Table 1). The RGS L106R melting curve showed a longer denaturation phase, starting at a temperature as low as 5°C (Fig. 1 B), which also reveals that the L106R mutation significantly destabilizes the protein, reflected in the more gradual unfolding compared with RGS WT already at low temperatures. A weakening of RGS folding is consistent with L106 being the pivot of a cluster that holds together the α1, α2, and α3 strands by interacting hydrophobically with F102, C111, L115, and T197 (38). The Leu-to-Arg mutation is hence likely to disrupt the hydrophobic core, resulting in partial unfolding.

Figure 1.

Biophysical characterization of RGS WT and RGS L106R. (A) CD spectra of RGS WT (blue) and RGS L106R (red) at 4°C. (B) Melting curves of RGS WT (blue) and RGS L106R (red) recorded by CD at 222 nm. ANS fluorescence spectra in the presence of (C) RGS WT and (D) RGS L106R at 4°C are shown, with RGS/ANS molar ratio varying from 1:1 to 1:12. To see this figure in color, go online.

Table 1.

Effect of osmolytes and crowders on the T_m of RGS L106R and WT

Protein variant	No cosolute	Glucose	Sucrose	Trehalose	PEG 3350	PEG 6000
RGS WT	56.1 ± 0.1°C	–	–	–	–	–
RGS L106R	28.3 ± 0.1°C	33.0 ± 0.1°C	32.9 ± 0.1°C	31.2 ± 0.1°C	27.9 ± 0.1°C	26.7 ± 0.1°C

We then used the ANS fluorescent probe to follow the aggregation of the RGS WT and L106R. ANS is used to assess the tertiary structure and determine the compactness of folding by detecting the exposure of hydrophobic patches on protein surfaces (39). ANS emission in the presence of RGS WT was indistinguishable from that of the free molecule, with a wavelength of maximal emission of 495 nm and no dose dependence (Fig. 1 C). In the presence of RGS L106R, ANS displayed a steep, dose-dependent increase of the emission intensity and a blue shift of the maximal emission wavelength from 495 to 470 nm (Fig. 1 D). Such a change in the spectral properties is consistent with ANS becoming embedded into a hydrophobic chemical environment or bound to a protein hydrophobic pocket (39). Hence, we conclude that the L106R mutation induced a loosening of the folding of RGS and the formation of solvent-accessible hydrophobic patches capable of ANS binding.

DLS at time 0 of RGS L106R displayed a sharp peak at a hydrodynamic diameter of ∼3.6 nm, compatible with the size of the monomeric protein, as previously reported (10). A smaller, broad peak at ∼11.7 nm (6.7% of the intensity) was indicative of the co-presence of association products of higher molecular weights (Fig. 2 A). The DLS at time 0 of RGS WT displayed a single, sharp peak at ∼5.6 nm, accounting for a dimeric or trimeric complex (Fig. 3 A).

Figure 2.

Biophysical characterization of RGS L106R aggregation. (A) DLS and (B) CD of RGS L106R acquired at fixed time points over 24 h of incubation at 25°C: time 0 (red), 30 min (blue), 1 h (light blue), 4 h (orange), 12 h (yellow), 24 h (black). (C) Time-course kinetics of ANS fluorescence at 470 nm in the presence (red) and in the absence (black) of RGS L106R at 25°C. (D) TEM image of RGS L106R after 12 h of incubation at 25°C. The scale bar represents 100 nm. To see this figure in color, go online.

Figure 3.

RGS WT does not display aggregation properties. (A) DLS and (B) CD of RGS WT acquired at fixed time points over 24 h of incubation at 25°C: time 0 (red), 30 min (blue), 1 h (light blue), 4 h (orange), 12 h (yellow), 24 h (black). (C) ANS time-course kinetics in the presence (blue) and in the absence (black) of RGS WT at 25°C. (D) TEM image of RGS WT after 12 h of incubation at 25°C. The scale bar represents 100 nm. To see this figure in color, go online.

RGS L106R aggregates are native like

DLS measurements were used to monitor aggregation of RGS L106R over time. The DLS of RGS L106R recorded at different time points over 24 h revealed that RGS L106R formed aggregates with increasing molecular weights. Within the first hour of incubation, the protein population shifted completely from the monomer to a 13.5- to 18.2-nm-wide association product. Further aggregation occurred, as the hydrodynamic diameter increased up to ∼38 nm in the following 23 h of incubation. The peak broadening accounted for a wide distribution of aggregates of different sizes, with polydispersity increasing over time (Fig. 2 A). TEM images acquired after 12 h of incubation confirmed that RGS L106R forms aggregates characterized by a distinctive elongated, unbranched morphology. The aggregates were monodisperse in width (∼10 nm) but polydisperse in length (between 30 and 150 nm) (Fig. 2 D). Conversely, RGS WT did not undergo aggregation to any extent over time. The DLS displayed a single, sharp peak at ∼5.5–6.5 nm hydrodynamic diameter over the whole incubation time, and the TEM images revealed spheroidal oligomers with a narrow height distribution of around 10 nm (Fig. 3, A and D).

CD spectra recorded over the same time window showed that RGS L106R does not further unfold before aggregation and that it undergoes only a partial unfolding after aggregation has started, with 75% of the native α-helical structure retained after 24 h of incubation (Fig. 2 B). This suggests that both the precursor and the aggregated states are native like in nature. RGS WT, on the other hand, did not show any change in its CD spectrum throughout the incubation time (Fig. 3 B).

Being dependent on hydrophobicity rather than protein structure, the ANS spectral properties proved ideal for characterizing the kinetics of a process of native-like aggregation like that of RGS L106R. The experiments were performed at an RGS/ANS molar ratio of 1:3, corresponding to midsaturation. The ANS time-course kinetics in the presence of RGS L106R displayed a steep, time-dependent increase of the emission at 470 nm over time (Fig. 2 C). The kinetic trace was fitted to the following function:

$$\mathrm{y}={\mathrm{m}}_{\infty }+{\mathrm{m}}_{\mathrm{1}}\times t+{\mathrm{m}}_{\mathrm{2}}\times \mathrm{e}\mathrm{x}\mathrm{p}\left(-k_{\mathrm{ANS}}\times t\right),$$ (1)

where m_∞ is the ANS emission at the end of the kinetics at long times, m₁ is the slope of the linear term that corrects for the absence of a plateau at long times, m₂ is the difference between ANS emission at the end and at the beginning of the kinetics, k_ANS is the rate constant of the exponential growth term, and t is time. Physically, k_ANS indicates the rate at which the number and/or the extent (area) of the hydrophobic patches increase on the RGS L106R surface (39). The exposure of hydrophobic patches correlates with the progression of protein aggregation, so k_ANS quantitatively characterized the kinetics of the process (40). RGS WT maintained its structure and ANS-binding properties were unaltered over the whole incubation time (Fig. 3 C).

Selection of osmolytes and crowders to be tested for their effect on RGS L106R aggregation

The effect of the cellular environment on RGS L106R native-like aggregation was mimicked with an array of osmolytes known to populate the cytoplasm at high concentrations. These osmolytes differ in their molecular weight, structure, and other properties that affect their solvation and hence the dynamics of the aggregation process (Table 2). Glycerol, erythritol, xylitol, and sorbitol display increasing size and number of hydroxyl groups. Sorbitol, myo-inositol, and glucose, as well as sucrose, maltose, and trehalose, have similar molecular weight but different structure. Glycine, proline, and glycylglycine have different constraints on the torsion angle, whereas alanine, serine, proline, and hydroxyproline differ by the addition of a single hydroxyl group on a hydrophobic side chain. We also mimicked large, multidomain hub proteins rich in intrinsically disordered regions such as Axin and its major protein partners, using flexible macromolecules lacking a fixed shape or fold, such as PEG polymers of increasing size and molecular weight: PEG 400, PEG 3350, and PEG 6000.

Table 2.

Effect of osmolytes and crowders on RGS L106R aggregation

Osmolyte	Class	Molecular weight (gmol−1)	kANS (s−1)	ANSend (a.u.)
No osmolyte	–	–	0.61 ± 0.05	44,426.6 ± 4539.5
Glycerol (Gyl)	polyol	92.10	0.48 ± 0.04	36,345.0 ± 1003.3
			p = 0.0094	p = 0.0107
Erithrytol (Eth)	polyol	122.12	0.40 ± 0.05	41,076.8 ± 2900.7
			p = 0.0004	p = 0.2436
Xylitol (Xyl)	polyol	152.15	0.37 ± 0.05	39,585.5 ± 3386.3
			p = 0.0002	p = 0.1207
Sorbitol (Srb)	polyol	182.17	0.48 ± 0.03	43,499.7 ± 3175.9
			p = 0.0060	p = 0.7693
myo-inositol (Ins)	polyol	180.16	0.60 ± 0.05	51,367.5 ± 1440.8
			p = 0.7562	p = 0.0227
Glucose (Glc)	polyol	180.16	0.27 ± 0.04	40,082.7 ± 3681.1
			p = 0.0001	p = 0.2133
Sucrose (Suc)	polyol	342.30	0.13 ± 0.03	33,347.7 ± 2790.1
			p < 0.0001	p = 0.0095
Maltose (Mal)	polyol	342.30	0.28 ± 0.04	29,766.7 ± 966.2
			p = 0.0001	p = 0.0017
Trehalose (Thl)	polyol	342.30	0.22 ± 0.05	39,642.0 ± 1690.4
			p < 0.0001	p = 0.1382
Glycine (Gly)	amino acid	75.07	0.54 ± 0.04	50,120.0 ± 5326.2
			p = 0.049	p = 0.1264
Alanine (Ala)	amino acid	89.09	0.54 ± 0.05	55,462.0 ± 618.5
			p = 0.0945	p = 0.0067
Serine (Ser)	amino acid	105.09	0.29 ± 0.05	51,859.8 ± 5620.6
			p < 0.0001	p = 0.0635
Proline (Pro)	amino acid	115.13	0.32 ± 0.04	28,279.8 ± 3931.4
			p < 0.0001	p = 0.0008
Hydroxyproline (HyPro)	amino acid	131.12	0.27 ± 0.03	32,123.8 ± 3457.2
			p < 0.0001	p = 0.0029
Glycylglycine (GGl)	amino acid	132.12	0.23 ± 0.02	44,462.3 ± 2036.0
			p < 0.0001	p = 0.9904
Betaine (Bet)	trimethylamine	117.15	0.92 ± 0.03	21,307.0 ± 1768.1
			p = 0.0001	p = 0.0002
PEG 400	crowder	400.0 (average)	0.64 ± 0.05	30,375.5 ± 1923.3
			p = 0.4353	p = 0.0007
PEG 3350	crowder	3350.0 (average)	0.99 ± 0.05	27,055.7 ± 1373.6
			p = 0.0001	p = 0.0008
PEG 6000	crowder	6000.0 (average)	1.14 ± 0.07	27,715.3 ± 2011.4
			p < 0.0001	p = 0.0011

Effect of osmolytes and crowders on RGS L106R aggregation

To characterize the effect of cell-compatible cosolutes on RGS L106R aggregation we first determined the rate of the ANS time-course kinetics in the presence of 750 mM of osmolyte or ethylene glycol (the PEG building block). The inhibition would result in a decrease of k_ANS but an enhancement in its augmentation. The concentration of choice is comparable to the actual osmolytes volume fraction in cellular environments and is often used as a proxy of the in vivo conditions for aggregation (12). With the exception of myo-inositol, all polyols slowed the rate of RGS L106R aggregation (Figs. 4 A and 7 A; Table 2). The k_ANS varied from 0.61 ± 0.05 h⁻¹ in the absence of cosolutes to 0.13 ± 0.03 h⁻¹ in the presence of sucrose, decreasing linearly as a function of the polyol molecular weight in the order k_ANS,ref ≈ k_ANS,Ins > k_ANS,Gyl > k_ANS,Eth > k_ANS,Xyl > k_ANS,Glc > k_ANS,Suc (Fig. 4 B; Table 2). All the k_ANS values in the presence of the polyols apart from myo-inositol were significantly different from that of the reference, with a p-value below 0.05 (Table 2).

Figure 4.

Effect of polyols on RGS L106R aggregation. (A) Time-course ANS kinetics of RGS L106R aggregation at 25°C in the absence (black) and in the presence of 750 mM glycerol (purple), erythritol (pink), xylitol (light green), sorbitol (light blue), myo-inositol (brown), glucose (red), sucrose (blue), maltose (orange), and trehalose (green). Each trace represents the average of at least three independent measurements. (B) Dependence of k_ANS of RGS L106R aggregation in the presence of 750 mM glycerol, erythritol, xylitol, glucose, and sucrose (blue line) and of sorbitol, myo-inositol, maltose, and trehalose (red) on the molecular weight of the polyol osmolytes. The error bars represent the standard deviation (SD) of the k_ANS obtained from the fitting of at least three independent measurements. To see this figure in color, go online.

Figure 7.

Apparent rate of RGS L106R aggregation. k_ANS (A) and ANS_end (B) of RGS L106R aggregation in the absence of cosolutes (pink) and in the presence of polyols (green), amino acids derivatives (red), and crowders (blue) at a monomer concentration of 750 mM are shown. The error bars represent the SD on the k_ANS obtained from the fitting (A) or on the ANS_end (B) of at least three independent measurements. The pale pink stripe represents the variation of k_ANS or ANS_end in the absence of cosolutes as a reference of significance for the effect of the cosolutes. To see this figure in color, go online.

A strong divergence from the linear trend seen in Fig. 4 B (blue line) was observed for polyols having the same molecular weight but different structure (Fig. 4 B). Sorbitol, glucose, and myo-inositol affected the k_ANS of RGS L106R aggregation differently, with k_ANS,Srb = 0.48 ± 0.03 h⁻¹, k_ANS,Glc = 0.27 ± 0.04 h⁻¹, and k_ANS,Ins = 0.60 ± 0.05 h⁻¹ (Fig. 7 A; Table 2). The disaccharides sucrose, maltose, and trehalose also exerted distinct effects, as k_ANS,Suc = 0.13 ± 0.03 h⁻¹, k_ANS,Mal = 0.28 ± 0.04 h⁻¹, and k_ANS,Thl = 0.22 ± 0.05 h⁻¹ (Fig. 7 A; Table 2). In particular, only glucose and sucrose substantially decreased the k_ANS of RGS L106R aggregation, whereas the impact of sorbitol, maltose, and trehalose was comparable to that of other polyols of lower molecular weight.

Except for glycine and alanine, amino acids also generally decreased the k_ANS of RGS L106R aggregation with p-values much below 0.05, but differently from polyols, we did not observe a linear size dependence on the amino acid molecular weights (Fig. 5, A and B; Table 2). Indeed, the presence of hydrophobic side chains of different sizes resulted in opposing outcomes; the effect of alanine was indistinguishable from that of glycine, with k_ANS,Ala = 0.54 ± 0.05 h⁻¹, whereas proline decreased it substantially to k_ANS,Pro = 0.32 ± 0.04 h⁻¹ (Fig. 7 A; Table 2). Turning a side chain hydrophilic with the addition of a hydroxyl group resulted in varied effects; compared to alanine, serine induced a sizeable decrease to k_ANS,Ser = 0.29 ± 0.05 h⁻¹, whereas the effect of hydroxyproline was comparable to that of proline (Fig. 7 A; Table 2). Similarly to polyols, though, the strongest decrease was observed with a dipeptide, as the presence of glycylglycine lowered the rate to k_ANS,GGl = 0.23 ± 0.02 h⁻¹ (Fig. 7 A; Table 2). Betaine showed a completely different impact to amino acids, as it strongly enhanced the rate RGS L106R aggregation, yielding k_ANS,Bet = 0.92 ± 0.03 h⁻¹ (Fig. 7 A; Table 2).

Figure 5.

Effect of amino acids on RGS L106R aggregation. (A) Time-course ANS kinetics of RGS L106R aggregation at 25°C in the absence (black) and in the presence of 750 mM glycine (pink), alanine (purple), serine (light green), proline (blue), glycylglycine (light blue), hydroxyproline (red), and betaine (green). Each trace represents the average of at least three independent measurements. (B) Dependence of the k_ANS of RGS L106R aggregation in the presence of 750 mM betaine (orange), glycine, alanine, serine, proline, glycylglycine, and hydroxyproline (green) on the molecular weight of the amino acid osmolytes. The error bars represent the SD on the k_ANS obtained from the fitting of at least three independent measurements. The dashed green line is a guide for the eye. To see this figure in color, go online.

When testing the effect of the PEG crowders, we found that the smallest of the three, PEG 400, did not affect the rate of RGS L106R aggregation (Figs. 6 A and 7 A; Table 2). Conversely, PEG 3350 and PEG 6000 crowders increased it to k_ANS,PEG3350 = 0.99 ± 0.05 h⁻¹ and k_ANS,PEG6000 = 1.14 ± 0.07 h⁻¹, with a p-value much lower than 0.05 (Figs. 6 A and 7 A; Table 2). Such an opposite behavior compared to that of polyols and amino acids is well pointed out by the positive, near-linear dependence of k_ANS on the molecular weight of the crowder (Fig. 6 B).

Figure 6.

Effect of crowders on RGS L106R aggregation. (A) Time-course ANS kinetics of RGS L106R aggregation at 25°C in the absence (black) and in the presence of PEG 400 (green), PEG 3350 (blue), and PEG 6000 (red) at an ethylene glycol monomer concentration of 750 mM. Each trace represents the average of at least three independent measurements. (B) Dependence of the k_ANS of RGS L106R aggregation in the presence of PEG 400, PEG 3350, and PEG 6000 on the molecular weight of the crowders at an ethylene glycol monomer concentration of 750 mM. The error bars represent the SD on the k_ANS obtained from the fitting of at least three independent measurements. The purple line is a guide for the eye. To see this figure in color, go online.

The ANS emission at the end of the kinetics (ANS_end) provided an indication on the extent of the aggregation process, i.e., on the amount, size, and compactness of the aggregate. High emission levels correlate with aggregates larger in number, size, or extension of the hydrophobic surface exposed. None of the cosolutes was able to suppress RGS L106R aggregation by decreasing ANS fluorescence to that of the initial state (Fig. 7 B). Nonetheless, in some cases we observed a substantial decrease in fluorescence with respect to the reference, with p-values below 0.05 (Fig. 7 B; Table 2). In particular, glycerol, sucrose, maltose, proline, and hydroxyproline were able to lower both k_ANS and ANS_end, whereas betaine and PEG crowders induced at the same time an enhancement of k_ANS with a strong inhibition of ANS_end (Figs. 4 A, 5 A, 6 A, and 7 B; Table 2). myo-inositol and alanine instead increased the ANS_end in the absence of any effect on k_ANS (Figs. 4 A, 5 A, and 7 B; Table 2).

The cosolutes determine the size and morphology of RGS L106R aggregates

To get insight on the number, size, and morphology of RGS L106R aggregates, we performed TEM imaging of the protein incubated 12 h in the presence of a subset of the cosolutes. The TEM results showed that none of the cosolutes blocked aggregation, though significant variations in the effect were observed (Fig. 8). Sorbitol induced the formation of RGS L106R aggregates with a similar morphology to those incubated in the absence of osmolyte but shorter in length (Fig. 8 A). Further decrease in the average length of the aggregates was observed in the presence of glucose, though these retained the worm-like structure (Fig. 8, A and B). Sucrose promoted the emergence of a spheroidal morphology and a narrow range of sizes, as all aggregates were round and displayed a diameter of 10–20 nm (Fig. 8 C). Coincubation with maltose resulted in fewer aggregates, and these were also shorter in length. The effect of trehalose was similar as that of glucose (Fig. 8, D and E). Serine favored the formation of a lower number of bulkier, branched assemblies, that seem to stem from the association or entanglement of smaller aggregates (Fig. 8 F). In the presence of proline and hydroxyproline, a relatively low number of mostly round-shaped aggregates were formed (Fig. 8, G and H). Glycylglycine also allowed only few aggregates to form, but these were larger in size and showed a filamentous appearance (Fig. 8 I). Betaine-, PEG 3350-, and PEG 6000-treated samples instead displayed a much higher density of worm-like aggregates of comparable or smaller size (Fig. 8, J–L). Taken together, the TEM studies were in significant agreement with the observed effect in the ANS time-course kinetics. Specifically, we found that the cosolutes having an inhibiting effect on the k_ANS displayed a decrease in the density and/or the size of the aggregates on the gird, whereas the k_ANS enhancers seem to increase the amount of aggregated protein.

Figure 8.

Effect of cosolutes on RGS L106R aggregation upon coincubation. Representative TEM images of RGS L106R after 12 h of incubation at 25°C in the presence of (A) sorbitol, (B) glucose, (C) sucrose, (D) maltose, (E) trehalose, (F) serine, (G) proline, (H) hydroxyproline, (I) glycylglycine, (J) betaine, (K) PEG 3350, and (L) PEG 6000 at a monomer concentration of 750 mM are given. The scale bar represents 100 nm.

Osmolytes and crowders have opposing effects on RGS L106R structure and thermodynamic stability

We then investigated the effect of cosolutes on the structure of RGS L106R aggregates. The CD after 12 h of incubation in the presence of glucose, sucrose, and trehalose revealed that RGS L106R aggregates retained a higher amount of the native structure than in the absence of osmolyte (Fig. 9 A). Such a protective effect toward the folding of the mutated protein correlates with the evidence that such polyols acted as chemical chaperones by assisting the partial refolding of RGS L106R into a more structured conformation immediately after the addition to a solution of the cosolute (Fig. 9 B). By contrast, in the presence of PEG 3350 and PEG 6000, RGS L106R displayed a slightly lower content of secondary structure both immediately after mixing and after 12 h of incubation (Fig. 9, A and B). These findings were consistent with the observation that glucose, sucrose, and trehalose increase the thermodynamic stability of RGS L106R, whereas PEG 3350 and PEG 6000 decrease it (Fig. 9, C and D). The T_m of RGS L106R was indeed found to rise from 28.3 ± 0.1°C in the absence of cosolute to 33.0 ± 0.1, 32.9 ± 0.1, and 31.2 ± 0.1°C in the presence of glucose, sucrose, and trehalose, respectively (Table 1). A decrease was instead registered in the presence of PEGs: 27.9 ± 0.1°C with PEG 3350 and 26.7°C with PEG 6000 (Table 1).

Figure 9.

Effect of cosolutes on RGS L106R stability. CD of RGS L106R incubated at 25°C in the absence (black) or in the presence glucose (pale blue), sucrose (light blue), trehalose (blue), PEG 3350 (orange), and PEG 6000 (red) at a monomer concentration of 750 mM after 12 h of incubation (A) or immediately after mixing (B) is given. CD melting curves of RGS L106R in the absence (black) and in the presence of (C) glucose (pale blue), sucrose (light blue), and trehalose (blue) or (D) PEG 3350 (orange) and PEG 6000 (red) at a monomer concentration of 750 mM are shown. To see this figure in color, go online.

Discussion

ANS as a probe for studying protein aggregation

We followed the kinetics of RGS L106R native-like aggregation using ANS, a fluorescent probe used for detecting exposed hydrophobic patches on protein surfaces and accompanies aggregation with an increase of the emission over time (39). ANS has proved instrumental in monitoring aggregation processes of all kinds, irrespective of the mechanism and of the internal structure of the aggregate (40). For instance, ANS was employed to follow the amorphous aggregation of carbonic anhydrase; the amyloid aggregation of transthyretin, Aβ, IAPP, and HypF-N; and the heterogeneous amyloid and amorphous aggregation of concanavalin A, immunoglobulin light chain, and β₂-microglobulin (4,41, 42, 43, 44, 45, 46, 47). ANS was also used to characterize the native-like aggregation of β-lactoglobulin and interleukin-1β in the absence of major structural rearrangements within the protein building blocks (48,49). Even though ANS cannot provide precise mechanistic or structural information about an aggregation process, the increase of its emission is treated as proportional to the formation of new ANS-binding sites on protein surface and to the progression of aggregation (39,40). The rate constant of the exponential growth of ANS emission k_ANS hence fully characterizes the native-like aggregation process of RGS L106R. We note that ANS binding to proteins is inhibited only by organic acids, anionic surfactants, and tryptophan, which mimic ANS by having a large hydrophobic moiety with a negatively charged head (40). None of the cosolutes tested in this study share such amphipathic molecular properties, so any alteration in the observed k_ANS is unlikely to be due to ANS displacement during the kinetics. The reliability of ANS as a probe is reinforced by its affinity to the native protein. ANS binds partially folded states like transient intermediates of protein folding and stabilized molten globules, whereas fully folded and unfolded states generally do not bind or only weakly bind ANS (50,51). The decrease in secondary structure and the intense ANS binding to monomeric RGS L106R are hence indicative of its molten globular nature. This suggests that RGS L106R is already in the molten globule form in its native state and that the increase in ANS emission is solely due to aggregation. Consistently, R103, R106, and K107 at the center of the hydrophobic core disrupted upon mutation are potential binding sites for the ANS sulfonate group via ion pairing (40).

The aggregation mechanism of RGS L106R

The DLS and CD time-course kinetics indicate that a limited loss of secondary structure accompanies the aggregation without preceding it. This evidence suggests that RGS L106R aggregation is initiated by a native-like state in the absence of an unfolding event as a trigger. When mixing with the aggregating solution at 25°C, RGS L106R may accesses a further destabilized intermediate conformation having a similar content of secondary structure but a lower kinetic barrier toward the aggregated state. A subset of thermally activated protein molecules may then enable aggregation to start from such a near-native conformation, similarly to the mechanism reported for the TFE (2,2,2-Trifluoroethanol)-activated aggregation of HypF-N (52).

The ANS time-course kinetics of RGS L106R aggregation is fit to a single exponential curve. The absence of a lag phase is consistent with RGS L106R being a molten globule able to initiate aggregation even without first accessing an aggregation-competent monomeric state. This apparent simplicity can be deceiving, as even complex, multistep kinetic processes may display a similar trend (53). RGS L106R forms assemblies having a worm-like, unbranched morphology. The width of the aggregate, as observed by TEM imaging, corresponds to the hydrodynamic diameter of the association products measured by DLS in the first hour of aggregation. This suggests that RGS L106R mature aggregates are formed by smaller aggregates that associate via geometrically constrained aggregation interfaces allowing for an ordered elongation (Fig. 10). This mechanism is reminiscent of the process of Barstar oligomerization followed by lateral association to form amyloid protofibrils, which is also characterized by a single exponential kinetics (53).

Figure 10.

Suggested model of RGS L106R aggregation in vitro. RGS WT has a compact native fold (green) with a cluster of hydrophobic residues buried in the core of the protein (yellow). (A) The L106R cancer-related mutation induces partial unfolding and exposure of the core residues in extended hydrophobic patches on the protein surface. (B) During the first hour of incubation, tetrameric association products of RGS L106R are formed via hydrophobic interactions between the exposed patches. (C) Further aggregation proceeds by piling up of the tetrameric oligomers to form a worm-like, unbranched aggregate. To see this figure in color, go online.

The distinctive worm-like appearance and the absence of structural conversion within the assembly make RGS L106R aggregates similar to those formed by γD-crystallin P23T, which are also native like (3). Conversely, other known native-like aggregates display very different characteristics. The early native-like aggregates of Sso AcP, human/murine prion, and β2-microglobulin readily convert into proper amyloid fibrils, whereas RNase A aggregates combine native-like and amyloid nature, with folded domains decorating an inner amyloid fibril (54, 55, 56, 57, 58). The native-like aggregates formed by human pancreatitis-associated protein and yeast prion Ure2p still display a mature fibrillar morphology even in the absence of amyloid structure (59,60). RGS L106R native-like aggregates are distinctly different from all others also from a biological point of view, as they are the only ones to correlate with cancer (10).

The effect of cosolutes on protein aggregation is independent of both the kinetic mechanism of aggregation and the structure of the aggregate

The effects of osmolytes and crowders on the native-like aggregation of RGS L106R are similar to those observed for other types of aggregation. Polyols such as alcohols and sugars can indeed oppose the amyloid and amorphous aggregation of a variety of proteins to different extents, with the inhibitory effect also being proportional to the size of the molecule (19, 20, 21, 22, 23, 24, 25, 26). Proline and hydroxyproline also inhibit the fibrillation and thermally induced aggregation of a number of proteins (19,27, 28, 29,61). Betaine and PEG crowders are able to trigger some aggregative processes of either amyloid or amorphous nature (19,27,30, 31, 32, 33). Less expectedly, myo-inositol, glycine, and alanine do not display any effect, despite being reported to affect other processes of aberrant protein assembly (15,19). Overall, these findings are consistent with our observations on RGS L106R aggregation, as monitored through ANS fluorescence.

The action of cosolutes hence seems to be independent of both the kinetic mechanism of aggregation and the structure of the aggregate, suggesting that they may impact a kinetic step that lies upstream of the aggregation itself. This led us to hypothesize that cosolutes exerted their effect on RGS L106R native-like aggregation by increasing or decreasing the thermodynamic stability of the molten globular state and modulating its structural properties. A stabilization of the molten globule decreases the number of activated precursors, disfavoring aggregation. A destabilization enlarges the population of states capable of aggregative events and enhances the statistical probability for an aggregation process to start, resulting in acceleration. Accordingly, the k_ANS-decreasing osmolytes glucose, sucrose, and trehalose had a stabilizing effect and increased the amount of secondary structure of molten globular RGS L106R, whereas PEG 3350 and PEG 6000 destabilized its folding and considerably increased its aggregation rate. In support of this, several cosolutes found here to oppose or favor aggregation also either stabilize or destabilize other molten globular proteins. For instance, sucrose and proline induced a contraction of molten globular ribonuclease A, thereby stabilizing a compact state (62). Carbohydrates and alcohols stabilized the molten globule state of equine ferricytochrome c (63,64). Glycerol stabilized an alkaline molten globular state of 5-aminolevulinate synthase, allowing for the retention of ∼1/5 of the k_cat of the native enzyme (65). At the peptide level, computer simulations show that sorbitol shifts the conformational equilibrium of met16 peptide from an unfolded to a folded state with the aggregation-prone residues unexposed, decreasing the rate of nucleation of an amyloid aggregate (66). Conversely, PEG and ficoll crowders destabilize native myoglobulin into a molten globule at physiological conditions (67,68), whereas ficoll and dextran destabilize the molten globular 5-aminolevutinate synthase and induce its association into amorphous aggregates (65). At the atomistic level, PEGs interact more favorably with aromatic C, amide N, and cationic N (69). Such preferential interactions with portions of side chain or backbone are likely to lead to attractive interactions for some exposed residues, inducing the destabilization of the native conformation and unfolding (69).

The effects of the osmolytes on RGS L106R aggregation are variable. The decrease in k_ANS and the TEM observation of a similar density of aggregates as for the protein alone but smaller in size, suggest that sorbitol, glucose, and trehalose slow down the growth phase of the process without hampering the nucleation. ANS_end is unaffected, suggesting that the aggregates are highly hydrophobic and likely to keep growing at longer times. Maltose, proline, and hydroxyproline instead decrease both k_ANS and ANS_end, showing far fewer aggregates of further smaller size. Such a strong inhibitory effect may emerge not only from delaying the growth but possibly also from inhibiting nucleation. Sucrose does not seem to decrease the amount of protein that aggregates, but it rather imposes a tight size and shape control over the aggregates, constraining them to small spheroids. The strong inhibition in terms of k_ANS and ANS_end can result from the aggregates’ compact shape and lack of exposed hydrophobic surface. Glycylglycine and serine show the opposite behavior, as they seem to decrease the amount of aggregated protein but favor clustering into larger assemblies with no size control, which explains the high ANS_end value measured. The diverse impact of polyols with similar MW but different structure (i.e., sorbitol, myo-inositol, and glucose and sucrose, maltose, and trehalose), could relate to the different thermodynamics of their solvation. For instance, alcohols and sugars are known to impact the structure of water differently, and this has a crucial effect on protein folding (70,71). The great difference observed with serine and alanine can depend on similar factors, as hydrophobic side chains build rigid hydration shells, whereas hydrophilic ones keep them less structured (72). Conversely, proline and hydroxyproline show similar effects, probably because of proline’s high solubility.

In contrast, the samples coincubated with betaine, PEG 3350, and PEG 6000 display a much higher density of aggregates that are shorter in length. This can be indicative of a fast aggregation process, in which the cosolutes selectively enhance the nucleation; more nuclei form in the initial stages of the process with an increasing effect on k_ANS, whereas the growth of such aggregates would then be hindered, resulting in the low ANS_end observed. This is consistent with the reported effect of PEG crowders on met16 amyloid aggregation, as a shorter lag phase and lower Thioflavin T emission at the plateau were indicative of the enhancement of secondary nucleation processes and of the stabilization of smaller aggregates (20,31). Our study thus reveals that the process of RGS L106R native-like aggregation can be easily tuned to result in different outcomes via different mechanisms.

A model for how the cellular environment affects Axin L106R aggregation

Based on the effects of osmolytes and crowders on L106R aggregation, we propose a general model for how the cellular environment affects Axin L106R aggregation in vivo. The osmolytes stabilize the molten globular state of RGS L106R and thereby oppose its structural collapse into a nanoaggregate, allowing Axin to reside in the cytoplasm as a monomer. The inhibition of aggregation may thus be a result of suppressed exposure of aggregation-prone regions of the protein in the compact state that could promote protein-protein associations. Upon recruitment into large complexes, such as the destruction complex or the signalosome, RGS L106R is exposed to higher local protein concentrations without directly participating in PPIs. The crowding within the complexes outweighs the stabilizing effect of the osmolytes and triggers aggregation, potentially because of increased probabilities for protein-protein encounters (Fig. 11).

Figure 11.

Model of Axin L106R aggregation in vivo. (A) In the cytoplasm, Axin L106R fluctuates as a monomer, with the DIX (blue) and mutated RGS (green) domains free to move. In the presence of osmolytes (red), the hydrophobic patches (yellow) formed on RGS surface upon L106R-induced partial unfolding remain partially hidden from the solvent, thereby inhibiting aggregation. (B) Axin forms the destruction complex with GSK-3β (red), β-catenin (violet), CK1 (purple), and the large, mostly intrinsically disordered scaffold protein APC (light blue) through the RGS domain. The high local concentration of disordered strands within the multiprotein complex triggers the destabilization of mutated RGS and the exposure of the hydrophobic patches. (C) Axin L106R aggregates via mutated RGS within the protein complex, turning it nonfunctional. To see this figure in color, go online.

This suggested mechanism may explain why cancer-related mutations are so abundant in hubs and scaffold proteins involved in wide networks of interactions, continuously switching from one complex to the other. Such proteins indeed experience elevated local concentration of crowders more than any other protein in the cytoplasm. Under acute crowding stress, even mildly destabilizing mutations may trigger the nanoaggregation of structured domains and induce the cancer phenotype as a cascade event. The mechanism proposed here is potentially applicable to the cancer-related mutations of Axin, p53, and other protein hubs, with the crowding-assisted nanoaggregation of multiprotein complexes being a general mechanism for carcinogenesis in the absence of insoluble depositions.

Conclusions

The dissection of the dynamics of RGS L106R native-like aggregation reveals that macromolecular crowding is a driving force for aggregation in vivo. The involvement of destabilized domains or proteins within large complexes formed onto scaffold or hub proteins such as Axin hence increases the risk of triggering aggregation. Such a mechanism offers new tools for understanding the role of protein aggregation in the molecular etiology of cancer and for designing effective therapeutic strategies.

Author contributions

T.G. designed and performed the experiments and analyzed the data. Y.L.-K. acquired and analyzed TEM images. D.H. and A.F. supervised research. T.G., D.H., and A.F. designed the research. T.G., Y.L.-K., D.H., and A.F. wrote the manuscript.

Editor: Samrat Mukhopadhyay.