The thermodynamics of neurodegenerative disease

Abstract

The formation of protein aggregates in the brain is a central aspect of the pathology of many neurodegenerative diseases. This self-assembly of specific proteins into filamentous aggregates, or fibrils, is a fundamental biophysical process that can easily be reproduced in the test tube. However, it has been difficult to obtain a clear picture of how the biophysical insights thus obtained can be applied to the complex, multi-factorial diseases and what this means for therapeutic strategies. While new, disease-modifying therapies are now emerging, for the most devastating disorders, such as Alzheimer's and Parkinson's disease, they still fall well short of offering a cure, and few drug design approaches fully exploit the wealth of mechanistic insights that has been obtained in biophysical studies. Here, I attempt to provide a new perspective on the role of protein aggregation in disease, by phrasing the problem in terms of a system that, under constant energy consumption, attempts to maintain a healthy, aggregate-free state against the thermodynamic driving forces that inexorably push it toward pathological aggregation.

INTRODUCTION

Since the identification of prions,¹ pathological protein aggregates have been increasingly identified as key drivers of pathology across numerous neurodegenerative disorders.^2,3 In these aggregation-related diseases, proteins which are normally soluble and functional misfold and/or clump together. The aggregates, a term used here to refer to any structure consisting of two or more protein monomers, can take a range of sizes and structure. Smaller, less ordered aggregates, usually termed oligomers, have been highlighted both as intermediates on the pathway to formation of larger structures and as a potential culprit for the toxic effect of aggregated proteins.^4–6 The aggregates that are characteristic of disease are amyloid; large, ordered, fibrillar structures, consisting of hundreds to thousands of monomeric subunits, here also referred to simply as fibrils.^2,7 Finally, more recently, protein condensates formed during liquid–liquid phase separation have emerged as other potential intermediate structures on the pathway to large ordered aggregates.⁸ While a large variety of proteins and peptides aggregate in different diseases, their amyloid forms share a key similarity in both structure and mechanism of formation. Amyloid fibrils generally form via a nucleated linear polymerization mechanism, in contrast, for example, to the hierarchical assembly of intermediate filaments.^9–11 Inter-molecular β-sheets form the core of the fibrils and are chiefly responsible for their high stability.^12–17 This high stability means that amyloid fibrils can persist under mechanical stresses and temperature variations, and many are resistant to digestion by proteases.^18–21

In disease, normally functional, soluble proteins convert to their amyloid forms, often leading to toxicity, inflammation, and cell death. The triggers and mechanisms of this conversion are not well understood in most diseases, although ageing appears to play an important role in many disorders. (In the section “Transition from Healthy to Diseased States,” I will introduce a way to classify these triggers by considering the underlying energy landscape of the aggregation reaction in vivo.) The most well-known aggregation-related diseases include prion disease, which involves the aggregation of the prion precursor protein and can either be acquired by consumption of food contaminated with prions or can occur spontaneously, generally in individuals with a genetic disposition.²² Many dementias, in particular Alzheimer's disease and Parkinson's disease, are also aggregation-related disorders, but by contrast to prion disease, infection does generally not appear to play a key role. In Alzheimer's, both the Aβ peptide and the tau protein aggregate, whereas Parkinson's disease predominantly involves the aggregation of the α-synuclein protein.²

In all these diseases, a central hallmark of pathology is the increasing amounts of persistent aggregates, in more and more regions of the brain as the disease progresses. This persistence and the extreme stability of amyloid across a diverse range of proteins have led to the suggestion that the amyloid fold constitutes a more stable state than the functional fold, for most proteins at their physiological concentrations.^23–25 Thus, living systems have to employ strategies to prevent or reverse the formation of these structures, in order to maintain proteins in their functional states and prevent the formation of potentially toxic aggregates, see Fig. 1.^26–28 In this review, I frame our current understanding of the biophysics of aggregation in the context of it taking place in such an in vivo environment. In contrast to the test tube, in this in vivo environment many active processes work to remove aggregates or prevent their formation, while at the same time the aggregation-prone monomer never runs out and is constantly resupplied by protein synthesis.

FIG. 1.

Protein aggregation in vivo. In living systems, proteins are constantly being synthesized and then, often with the help of chaperones, fold into their functional conformations (turquoise). However, misfolding and aggregation can occur, sometimes resulting in very stable amyloid forms of the proteins (blue). In a healthy organism, protein quality control mechanisms are able to remove and recycle many such misfolded or aggregated proteins (purple). This process, under the consumption of energy, allows the system to maintain an out-of-equilibrium state, with predominantly folded proteins and few aggregated or misfolded forms (free energy schematic on right; this is vastly simplified for illustration purposes. In reality, the energy surface will be multi-dimensional and rough, with different pathways and many intermediates).

Protein self-assembly is at the core of aggregation-related diseases. Therefore, it has attracted much interest from biophysicists as it is a seemingly simple physical process that can both be modeled and readily reproduced using purified protein in the test tube. A large body of test tube aggregation experiments and methods to analyze them has emerged, and work over the last four decades has led to a number of insights into disease.^10,29–31 For example, such studies have contributed to our understanding of the key rate-limiting steps,³² elucidated how protein sequence relates to amyloid formation,³³ and pointed to utilizing targeted mutation or addition of inhibitors and promoters to alter the self-assembly reaction networks.^34–36 More recently, advances in experimental techniques, such as cryo-electron microscopy,^16,17,37 positron emission tomography,^38,39 sensitive biomarker probing strategies,^40,41 and single molecule spectroscopy,^42,43 enable high resolution or real-time measurement of aggregate formation and their biophysical properties in living systems and patient samples.

While we now have detailed mechanistic insights into the aggregation mechanism in the test tube, and understand many of the biological aspects of disease, we still do not have a clear model of how and why aggregates appear, evolve, and persist in patients. Thus, much still remains to be done to translate test tube biophysical measurements into the development of therapeutics for aggregation-related disorders, in particular, the devastating dementias such as Alzheimer's and Parkinson's diseases. The stakes are high, with an ageing population and future dementia numbers projected to put a tremendous strain on healthcare systems.⁴⁴ Although, after decades of research, the first disease modifying treatments for Alzheimer's disease, in the form of the anti-Aβ antibodies aducanumab and lecanemab,^45–47 have recently been approved as human therapies, they still leave much room for improvement. Current therapies can only achieve a slowing, rather than a halting, of disease progression, let alone a reversal to a healthy state. However, many other promising strategies for combating aggregation-related diseases are emerging with a range of therapeutic mechanisms, often significantly different to the approach of aggregate removal by which the currently available clinical antibodies act. Biophysics has the potential to play a transformational role in this space,⁴⁸ as has been demonstrated in the context of two other aggregation-related diseases, sickle cell anemia and familial amyloid polyneuropathy: the discovery of new drugs was driven in both cases by a detailed understanding of the underlying biophysics^49–51

The aim of this review is to provide a clear framework to apply mechanistic insights from the test tube and biophysical intuition to understanding the drivers of neurodegenerative disease. I will first provide a brief overview of the mechanistic in vitro work and the important role that coarse-graining plays in the development of mechanistic models. I will then outline the pathways by which amyloid can emerge in organisms despite the numerous regulatory systems, and why amyloid self-replication is the crucial factor therein. Finally, I will discuss how this framework can provide a biophysical rationale for classifying current treatment approaches and inform the development of future therapeutic strategies.

CHEMICAL KINETICS AND COARSE-GRAINING STRATEGIES TO ESTABLISH REACTION MECHANISMS

Much can be said about chemical kinetics in protein aggregation, but here I only briefly touch on how chemical kinetics is applied, as this topic is covered in great detail in previous reviews, and instead expand more on the ideas of coarse-graining and how the differential rate equations are obtained in the first place, a topic that is also crucial in vivo.^7,30,52 In order to understand how proteins escape from their well-behaved states and form pathological aggregates, one needs to understand the mechanism of aggregation and which steps limit the overall reaction. In this context, chemical kinetic analysis is a powerful tool, allowing one to derive differential rate laws from a model of how proteins interact and assemble. The resulting equations can then be fitted to experimental data to elucidate reaction mechanisms and determine rate-limiting steps. At the core of these kinetic equations is the maintenance of mass balance, by keeping track of all chemical species and the reactions by which they inter-convert. While it is relatively clear what constitutes distinct chemical species in a gas phase reaction, it is less well defined for proteins interacting in solution: proteins are present in a continuous distribution of conformations and will pass many metastable intermediate states when assembling together. This necessitates some level of “coarse-graining,” i.e., grouping of a number of similar states into one effective chemical species, in order to produce a tractable set of distinct species, see Fig. 2. When one is interested in the kinetics of the reaction, a useful strategy is to coarse-grain species that inter-convert quickly on the timescale of the rate-limiting step.^5,53 Moreover, one often includes only a single term for processes that are indistinguishable on the level of the available data, i.e., different pathways which lead to the same products. Using a single term for all multiplication processes (see below) is an example of this approach. Finally, multi-step reactions are often coarse-grained into an effective single step reaction when intermediates are not experimentally visible and the data do not allow the determination of the kinetics of the individual sub-steps. Single step nucleation terms, as illustrated in Fig. 2, are an example of this strategy. While such approaches are extremely effective, care needs to be taken to tailor them to the available data. Too much coarse-graining means one can no longer explain observations, whereas too little coarse-graining leads to the data being over-fitted. Moreover, care needs to be taken when interpreting the results of a coarse-grained analysis. For example, coarse-grained single step nucleation terms contain a reaction order parameter, yet because of the coarse-graining, this reaction order does not necessarily correspond to the nucleus size as one might expect in simpler systems. In the coarse-grained description, it simply reports on the reaction order of the rate-determining step in the nucleation cascade.³¹

FIG. 2.

Coarse-graining. Given the complexity of the energy landscape of a reaction involving protein molecules in solution, some level of coarse-graining is generally necessary to obtain a finite set of chemical species for which rate laws can be formulated. The degree of coarse-graining required is determined both by the complexity of the reaction network and the amount of information available from experimental data. Coarse-graining turns a rough energy landscape with many local minima into a smoother one (top) or summarizes a series of individual steps into a single effective one (bottom).

In protein aggregation, the natural description is to treat every aggregate size as a different chemical species. In this way, due to the large size of aggregates, one quickly obtains a complex system of hundreds of coupled differential equations,^10,54 so a further simplification of this system is often useful: rather than tracking the time-evolution of the entire aggregate size distribution, we consider only the low order moments of the distribution.⁵⁵ Specifically, we use kinetic equations that describe the time evolution of aggregate number, P(t), and total aggregate mass, M(t), yielding a set of just two differential equations, Fig. 3. These include the processes of primary nucleation, the formation of an aggregate directly from monomeric protein, elongation, the growth of an existing aggregate by the addition of further protein monomers to its ends, and multiplication processes or secondary processes, the formation of new aggregates via the involvement of existing aggregates, for example in the form of fibril fragmentation or fibril surface-catalyzed secondary nucleation.^32,56 When they are both present, the elongation and multiplication processes couple together in a positive feedback loop, allowing aggregates to self-replicate and giving rise to exponentially growing aggregate amounts. As the presence of an elongation mechanism is essentially a given when fibrils are formed, the presence of a multiplication process is equivalent to a fibril being able to self-replicate.

FIG. 3.

Chemical kinetics yields molecular mechanisms. Using the coarse-grained differential rate laws (top left, example of equations for a general aggregation mechanism from Meisl et al.,⁶⁰ k_g and k_m are the rate constants of growth and multiplication, respectively, m₀ is the available free monomer concentration) to analyze measurements of aggregate formation over time (top right, simulated data), one can determine the mechanism of aggregation. The minimal set of processes that can describe such measurements across a wide variety of proteins¹¹ consists of primary nucleation, elongation, and multiplication and is shown schematically.

Using in vitro experimental measurements of total aggregate mass over time, M(t), this chemical kinetics framework can then be applied to determine the molecular aggregation mechanism. Again, carefully considered coarse-graining that takes into account the fundamental physical requirements of the underlying reaction network and what level of information is available from experimental data is crucial to achieve reliable mechanistic insights. This approach is particularly robust when reactions can be grouped into effective steps that have orthogonal effects on the measured parameters or that can be independently affected by changing experimental conditions. The following two examples illustrate this point. Nucleation and multiplication processes mainly increase the number of aggregates, P(t), without significantly affecting the total aggregate mass, M(t), whereas elongation does not change the number of aggregates, but does increase aggregate mass [this is reflected in the approximate rate equations shown in Fig. 3 where each of the processes appears only in the equation for $\dot{P}(t)$ or in the equation for $\dot{M}(t)$ but not both].⁵⁷ Thus, measurements of the average size of aggregates, given by $M(t)/P(t)$, can be used to distinguish between the effects of these two groups of processes.⁵⁸ Similarly, primary nucleation is independent of the presence of aggregates, whereas the multiplication rate increases with the concentration of existing aggregates. The two processes can therefore be distinguished by observing the effect of introducing preformed aggregates into a sample of monomeric protein.^11,31,59

This overall strategy has been successfully applied to establish the mechanism of aggregation across a wide variety of proteins, providing fundamental insights into the drivers of aggregation, effect of solution conditions, and roles of mutations.^{31,32,36,61–63} Moreover, they have enabled the determination of the mechanism of action of different promoters and inhibitors of aggregation and are central to a rational design of aggregation inhibitors.^34,64,65 However, despite these successes in the analysis of in vitro data, it is not clear a priori that the mechanistic insights gained in vitro would translate to living systems: in vivo, proteins will experience many more interactions, both with their functional binding partners and with the cellular machinery that has evolved to keep them in their functional state, all of which interactions will significantly affect their aggregation behavior. Indeed, a study comparing the aggregation of purified Aβ42 in buffer and in solutions containing increasing concentrations of cerebrospinal fluid (CSF) showed that the speed of the aggregation reaction was altered considerably.⁶⁶ However, the mechanism of aggregation, with a surface catalyzed secondary nucleation step dominating the formation of new aggregates, was still retained in CSF. Similarly, studies of mutant proteins show, in many cases, that mutations associated with more aggressive disease also show increased aggregation propensity, suggesting that mechanisms and relative rates may translate across systems, from in vitro to in vivo.^36,62,67

A recent meta-study more thoroughly investigated this correlation between in vitro aggregation behavior and disease association, across a wide variety of proteins and diseases.¹¹ As expected, absolute reaction rates do not necessarily translate from in vitro to in vivo. However, there was a clear correlation between the in vitro reaction mechanism and the biological role of the proteins. The speed at which aggregates are able to self-replicate in vitro emerged as the key predictive property on in vivo behavior of the aggregating protein. This may be rationalized by the fact that the ability to self-replicate fundamentally alters the aggregation behavior: in the absence of self-replication, each aggregate needs to assemble through a usually slow primary nucleation process. By contrast, when self-replication is possible, a single parent fibril formed by primary nucleation can give rise to thousands of new fibrils. The resulting exponential amplification is a powerful phenomenon, a runaway reaction that quickly consumes all available protein and is very difficult to arrest or counteract. The aforementioned meta-analysis¹¹ revealed two key facts about aggregate self-replication across different proteins. First, self-replication is surprisingly ubiquitous, occurring even for artificial proteins. Second, all disease-associated aggregates show significant self-replication and only aggregates that have evolved to aggregate in a specific functional context do not self-replicate in vitro, see Fig. 4. Remarkably, despite the very different conditions proteins experience in vitro and in the context of disease, the analysis of in vitro measurements provides a clear delineation of disease-associated and non-disease-associated proteins. This work shows that biophysical measurements of the purified protein are indeed able to provide crucial mechanistic insights for the aggregation in living systems, not only in special cases, but across aggregating systems, from the small Alzheimer's disease-associated Aβ peptide, to the proteins that have evolved to form the structural components of bacterial biofilms.

FIG. 4.

In vitro self-replication rate correlates with biological function. The time taken for the aggregate number to double via self-replication in the aggregation of purified protein (horizontal axis) is plotted against the time over which the same aggregates form in living systems (vertical axis), for a wide range of different aggregating proteins. In the top left half, the in vivo aggregation time exceeds the doubling time. If the in vitro measurements are predictive of the in vivo behavior, then in the top left region there is enough time for self-replication to be relevant in the living system. By contrast, in the bottom right half self-replication is too slow to be relevant in vivo. The fact that this plot clearly allows separation of proteins by their biological function and association with disease indicates that in vitro biophysical measurements of aggregation kinetics are indeed predictive and that self-replication is the key process in predicting the biological effect of aggregation. Adapted from Meisl et al., Sci. Adv. 8, eabn6831 (2022). Copyright the Authors, some rights reserved; exclusive licensee AAAS. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).¹¹

In light of these findings, the extension of the kinetic models to living systems is promising. Its utility is clear: aggregation-related diseases generally progress slowly, over the course of decades, and are often linked to ageing; thus, the availability of animal models that faithfully reproduce the disease is limited. Mechanistic models describing aggregation across systems would be invaluable in bridging the gap between our detailed biophysical understanding of aggregation under controlled conditions and the related diseases. The aggregation reaction network is vastly more complex in a living organism, affected by molecular compartmentalization, concentration gradients, and active removal processes specifically evolved to prevent aggregation. Nonetheless, significant insights can be gained from in vivo data through careful model building, which ensures that models capture the key features of behavior while still being simple enough to provide genuine deeper understanding of the determinants of aggregation. This strategy has identified rate-limiting steps and discovered factors that determine the characteristic spatial progressions of disease.^58,68–70 While these results are promising, much still remains to be answered about the mechanisms of in vivo aggregation to apply biophysics to the development of therapies across diseases. The early triggers of disease, the role that aggregate removal processes play in this, and in particular, how in vitro observations of aggregate inhibition can be used to predict in vivo effects, remain key unknowns.^71,72 In the remainder of this work, I thus outline a framework for a more general intuition for translating mechanisms to in vivo, by considering the out-of-equilibrium nature of living systems.

AN OUT-OF-EQUILIBRIUM VIEW OF PROTEIN AGGREGATION IN VIVO

Living systems are characterized by maintaining a state far from chemical equilibrium for long periods of time, through the constant consumption of energy.⁷³ From functional proteins to ion concentration gradients, the features that enable complex life are present only in metastable states which are always being pulled toward more stable forms. Living systems have to constantly fight this decay to equilibrium, keeping species kinetically trapped in their functional states and consuming energy to maintain this out-of-equilibrium state.

For proteins in particular, the region of conformational space that corresponds to their correctly folded and functional conformation is minuscule. To maintain function, other misfolded conformations and potentially more stable forms such as amyloid need to be avoided.²⁴ Thus, complex machinery has evolved to fold proteins into their functional forms, and subsequently to maintain that state.^26,28 Chaperones ensure that new proteins fold into the correct structure and can even refold misfolded proteins, while degradation systems such as the proteasome can purge potentially pathological species. Recent work has demonstrated that these systems, under the consumption of ATP, are also capable of dis-aggregating amyloid fibrils.^74,75 This constant turnover of protein and the multitude of mechanisms that remove unfolded or misfolded proteins, under considerable use of energy, serve to maintain protein homeostasis.^27,76

By contrast, in test tube aggregation reactions of purified protein, no protein quality control mechanisms exist, and only the intrinsic kinetic barriers prevent proteins from aggregating. Therefore, given enough time, in vitro samples will reach an equilibrium state with their aggregated forms. This aggregated mixture can contain a wide range of species, being governed by the thermodynamics on a complex potential energy surface.⁷⁷ At the simplest level is the equilibrium between monomer and fibrils, which is determined by the solubility or critical aggregation concentration of monomer.^78,79 Most strategies to determine fibril stability measure this free monomer concentration at equilibrium, with techniques paralleling those used in protein folding.^80,81 Beyond this simple monomer–fibril equilibrium, the fibril length distributions also approach an equilibrium state, albeit over much slower time-scales which reflect the small energy differences associated with such changes in fibril length.⁸² Finally, in recent years the advances in structural techniques, in particular cryo-electron microscopy, have revealed that a multitude of different fibrillar structures form during aggregation and then gradually equilibrate over time-scales much longer than those over which monomeric protein is initially depleted.^83,84 While different fibril structures appear to be associated with different pathologies of the same protein,^17,85 we here coarse-grain this level of granularity and instead focus only on the major aspect, the stability of the aggregated state relative to the monomeric one.

In support of the idea that the aggregated state constitutes a free energy minimum, many proteins whose aggregation is not associated with disease can form stable amyloid fibrils under the right laboratory conditions, suggesting that amyloid formation may in fact be a universal property of most proteins.^{11,25,86–89} Despite this widespread ability of proteins to adopt the amyloid fold, surprisingly few examples exist of systems that have evolved to use amyloid in a functional context, for example as structural elements in bacterial biofilms.^90–92 It may in fact be the intrinsic stability of amyloid that makes it unsuitable for most applications within organisms, with life seemingly preferring more dynamical assemblies, which are less energetically favorable and can thus be readily disassembled.^93,94 By contrast, it is extremely difficult to reverse the formation of highly stable amyloid fibrils.

Given the intrinsic stability of amyloid, the emergence of an aggregation-related disease can thus be seen as a failure of homeostasis: proteins escape the tightly controlled system evolved to keep them in their functional state, making their way down the free energy gradient to misfold and/or to aggregate and accumulate in their more stable aggregated forms. In some cases these aggregated proteins are relatively inert and their presence leads to disease primarily through depletion of the functional protein, inducing loss-of-function toxicity.^2,95 In other cases, such as several dementias, the aggregates themselves are believed to be toxic or to produce toxic species.^96,97 In particular, intermediate species between monomeric and highly structured aggregates, termed oligomers, have been found to exert toxicity via several mechanisms, including disruption of cellular membranes and aberrant interactions with cell receptors.^96,98 In either case, the common theme throughout is that these protein aggregates do not behave like foreign pathogenic agents in infectious disease (although those may be the original triggers of aggregation), or mutated versions of the organism's own cells as in cancer (although mutations can destabilize functional conformations). Rather, they have simply broken free from the kinetic trap of their functional conformations and progressed down the free energy gradient to aggregates. How this may happen is discussed in the following section.

TRANSITION FROM HEALTHY TO DISEASED STATES

In a healthy system, strategies to maintain the out-of-equilibrium functional monomer state can be classified into two scenarios, with a continuous spectrum connecting them. (1) Very high kinetic barriers make the formation of an aggregated species very unlikely during the evolutionarily important part of the organism's life (see Fig. 5, top right). (2) Somewhat lower kinetic barriers are present together with energy-consuming protein quality control mechanisms that remove aggregates when they are formed (see Fig. 5, top left). While the active removal of aggregates as in (2) has been shown to be important in many systems, prion diseases, where introduction of a very small number of aggregates leads to system-wide runaway aggregation, provide evidence for (1) also being a relevant scenario.

FIG. 5.

Different strategies for maintaining the out-of-equilibrium state and their breakdown in disease. Left column: A kinetic barrier keeps proteins in their functional states most of the time, but there is still a significant flux to aggregated species. This is compensated for by energy-consuming protein quality control mechanisms that drive the system into the functional state. In disease, these protein quality control mechanisms decline, and can no longer prevent the accumulation of aggregates. Right column: The system is maintained in its functional state simply by a high barrier to initial aggregate formation. In disease, the introduction of preformed aggregates, such as an infectious prion, circumvents the high-energy barrier of initial aggregate formation by providing the low-energy barrier pathway of seeded self-replication.

In both cases, aggregates may occasionally appear, either through random fluctuations (more likely in case 2) or through introduction of a preformed aggregate (for example through consumption of a diseased organism as in prion disease). When those aggregates are unable to self-replicate, they will likely do little damage: In case 1, it is a rare event, so aggregate concentrations will be negligible as each individual aggregate introduced can only grow itself, not produce any new aggregates. In case 2, removal mechanisms will have plenty of time to remove the individual aggregated species that cannot self-replicate. By contrast, when the aggregates are able to self-replicate, the situation becomes much more problematic: even the introduction of a small number of aggregates, if they are not cleared quickly and effectively, can lead to a runaway reaction, much like in the case of a viral infection. Consequently, those species which are able to self-replicate are by far the most likely to break out of the enforced out-of-equilibrium state. Evidence for this scenario being a realistic representation of the in vivo situation comes from the observed correlation of self-replication propensity and disease across different proteins in the previously discussed meta-study.¹¹

Analogous with how protein aggregation is prevented in the healthy system, the triggers of disease can be grouped into two classes. In the first, disease is triggered by the introduction of a sufficient amount of preformed self-replicating aggregate, from external sources. Infection with prion disease is the prime example of such a scenario. Sporadic cases are very rare, but the disease can be passed from one individual to another by the consumption of infected material. In the biophysical picture, the barrier to spontaneous primary nucleation is very high, but when preformed seeds are introduced, their ability to self-replicate allows them to bypass this barrier and trigger runaway aggregation (see Fig. 5, bottom right).

In contrast to prion disease, in most aggregation-related disorders, pathological aggregates arise sporadically. The precise etiology is often unknown, but it is clear that mutations and ageing can both play a key role. For example, several mutations within the genes encoding the predominantly aggregating proteins, result in early-onset disease.⁹⁹ However, even when mutations are present, ageing is known to be the major contributor to disease onset, with decline in protein quality control processes constituting an important factor.^76,100–102 This gives rise to the second class of triggers, in which the decrease in the rate of misfolded-to-functional state transitions enables the aggregate concentration to multiply and overwhelm the protective systems⁷¹ (see Fig. 5, bottom left).

Often reality will lie somewhere between these extreme cases, and a combination of both decreased protein quality control and increased introduction of aggregated species may combine to give rise to disease. Furthermore, feedback mechanisms that exacerbate the decline of protein quality control mechanisms as aggregates accumulate may also be at play. The sequestration of proteasomes into aggregated structures,¹⁰³ the apparent triggering of tau aggregation by Aβ aggregates in Alzheimer's disease,¹⁰⁴ and the fact that often several different proteins, rather than just a single one, are found to form aggregates in disease¹⁰⁵ are all in support of a feedback mechanism by which existing aggregates trigger the formation of new aggregates from different proteins, for example, by overwhelming the shared protein quality control mechanisms. Such a process will have a very similar effect to the simpler self-replication mechanisms that proceed via secondary nucleation or fragmentation of aggregates, leading to runaway aggregate accumulation.^60,70

IMPLICATIONS FOR THERAPEUTICS AND DRUG DISCOVERY

Having discussed the scenarios which lead to disease, therapeutic strategies can now similarly be classified into a number of different approaches, briefly summarized as (1) support of active removal processes, (2) decrease in aggregation driving forces, or (3) increase in kinetic trapping, see Fig. 6.

FIG. 6.

Schematic of therapeutic strategies. Left: Increasing aggregate removal, e.g., by anti-aggregate antibodies, increases the driving forces into the out-of-equilibrium functional state, away from the aggregated state. Middle: decreasing the free energy of the functional state relative to the aggregated state decreases the driving force for aggregation. Right: increasing the barrier to aggregate formation keeps proteins in their functional states. While, for simplicity, aggregate removal mechanisms have not been shown in the middle and right cases, they are still present and potentially crucial for maintaining the functional state. The reduced flux of new aggregates in those two cases will allow these quality control mechanisms to clear remaining aggregates much like in the left case.

The first strategy, enhancing active removal processes, essentially compensates for the decline in protein quality control, by increasing the rate at which aggregates are removed, for example by the cellular protein degradation systems.⁷⁶ As this strategy requires energy consuming processes to push the system back into its healthy out-of-equilibrium state, it needs to utilize the organism's intrinsic aggregate removal mechanisms. One way to achieve this is by the use of anti-aggregate antibodies, which recognize the aggregated structures and then trigger the bodies' immune response to remove them.¹⁰⁶ This has led to the first disease-modifying treatments against Alzheimer's disease: aducanumab and lecanemab, both of which have been recently approved for human use, and donanemab, which has shown promising results in phase III clinical trials.^46,47,107 While these are the first class of disease modifying drugs for Alzheimer's disease, several drawbacks will likely mean that these antibody-based treatments will not solve the problem alone: they are expensive, they can trigger severe side effects in the form of cerebral edema and micro-hemorrhages, likely resulting from the aggressive removal of aggregates,^108,109 and the lack of orally available treatments puts a significant burden on patients. Moreover, while Aβ aggregates are present predominantly in extracellular spaces, other aggregated proteins that are important in Alzheimer's disease (the tau protein) or Parkinson's disease (α-synuclein) accumulate inside cells, making them much less accessible to antibody-based treatments. However, novel technologies are emerging to enable the removal of aggregates by intra-cellular antibodies, thus, promoting intra-cellular aggregate removal processes may be a viable strategy for future therapies.¹¹⁰ While small molecules are able to address some of the issues of antibody therapies, designing such molecules to increase the removal of aggregated species is likely to be very challenging. The active removal of species requires an energy input; a small molecule which binds significantly to aggregates must decrease its free energy and thus cannot subsequently induce their dissociation. Instead, in order to destroy aggregates, a small molecule must not only bind to the aggregates to be removed, but also recruit existing cellular protein removal systems, requiring a feat of dual pharmacology.

A more easily accessible avenue for small molecule-based approaches is the second strategy, the decrease in driving forces that favor aggregation. The fundamental idea here is to decrease the free energy difference between the soluble/functional and aggregated states, thus reducing the need for protein quality control mechanisms to keep the system in its out-of-equilibrium state. An example of a successful application of this approach is the small molecule Tafamidis, which is used to treat familial amyloid polyneuropathy. Tafamidis binds to the correctly folded form of transthyretin, thus stabilizing it and preventing its pathological aggregation, functioning as an artificial chaperone.⁴⁹ Another example of such an inhibitor that was developed based on a fundamental understanding of the biophysics of the disease is Voxelotor. It is used for the treatment of sickle cell disease, which is caused by aggregation of hemoglobin within red blood cells. The seminal work by Eaton and Ferrone in the 1980s not only determined the aggregation mechanism of hemoglobin, but also first discovered surface-catalyzed secondary nucleation as an important mechanism in the formation of fibrillar protein aggregates.⁵⁶ Four decades later, the biophysical insights from this work, and in particular, the understanding that the exponential amplification properties of the aggregating system meant that small changes to the concentration of the aggregating proteins could have large effects on the aggregation behavior, have led to the development of a new drug for sickle cell disease.^50,51 This drug acts by stabilizing a functional form of hemoglobin, thus preventing its aggregation and the deleterious effects of this process. It is important to note that both in these examples and more generally, the stability of this drug-protein complex does not need to exceed that of the fibrillar form, it simply needs to be stabilized enough that the rate of conversion from functional to aggregated forms no longer exceeds that of the innate aggregate removal mechanisms. Another way to achieve stabilization of the monomeric state relative to the aggregated state is to lower the monomer concentration. Promising results have been achieved using anti-sense oligonucleotides (ASO) to lower expression levels of tau in both mouse models of the disease and more recently in human phase I clinical trials.^111–113 In these systems, a lowering of the monomer concentration gradually leads to the disappearance of already-formed aggregates, suggesting that the thermodynamic driving force toward aggregates has been sufficiently lowered that removal mechanisms can restore the healthy state. The main potential drawback of such monomer-targeted approaches is that the state or concentration of the monomeric precursor has to be altered, so if this precursor is functional, the therapeutic intervention may negatively affect its ability to function normally.

This leaves us with the final and least invasive therapeutic strategy: the increase in the effective kinetic barrier to aggregation by use of an aggregation inhibitor that targets aggregated states. (Note: more accurately, an inhibitor that binds an intermediate species on the aggregation pathway also simply stabilizes it and thus hinders its further conversion, rather than actually increasing the barrier height as in the simplified diagram in Fig. 6. However, the presented classification is useful despite this somewhat fuzzy definition because it distinguishes strategies that target the monomeric state and inhibitory strategies that target aggregated states. This is useful because affecting monomeric states has very different implications in the context of designing therapies than targeting aggregated states, as discussed.) In practice, this is achieved by binding to specific species on the aggregation pathway to significantly slow the aggregation reaction, which avoids the major drawbacks of both previously discussed approaches. Because it targets aggregated species, it does not suffer from the same “on-target toxicity” risks as the approaches that aim to interact with the monomeric/functional form of the protein. As this strategy does not require the utilization of the intrinsic aggregate removal mechanism but instead simply relies on binding interactions, it can be more easily achieved with small molecules, allowing for easy intra-cellular action and oral administration. Moreover, designing a molecule to bind a particular molecular species to prevent its conversion or interaction with another binding partner is a tried and tested drug development strategy. Finally, targeting rare aggregated species may be easier to achieve with low concentrations of tight-binding inhibitors than targeting the likely much larger monomer population. However, to realize this approach, a detailed understanding of the aggregation reaction mechanism is required in order to correctly identify the most promising targets for inhibition. This needs to be combined with a high accuracy kinetic assay that allows the mode of action of an inhibitor to be determined and optimized.^34,65 The chemical kinetic framework outlined in the earlier section is crucial here, both in interpreting the kinetic data mechanistically and in predicting which mode of action will be the most effective in vivo. Promising results have already been achieved with this strategy^114,115 and before too long drugs against neurodegenerative diseases from this class may join the first drugs in the other two classes to expand the collection of disease-modifying treatments.

This picture also suggests there is cause for optimism: for dementias in particular, the slow progression of disease and gradual accumulation of aggregates suggest that the systems are not far out of balance, and that removal processes may be just on the verge of keeping aggregation at bay.²⁴ Therapies may not have to drastically alter the system, and even slight decreases of the aggregation rates or increases of the removal rate, in particular in early disease, may be sufficient to shift the balance to net aggregate removal.

CONCLUSIONS

In recent decades, the advances in both biophysical measurements and their analysis have allowed us to gain detailed insights into the molecular processes central to neurodegenerative disease. A key remaining difficulty is linking the biophysical insights into the molecular self-assembly with the processes that control the emergence of disease-causing aggregates in humans. Here, I provide a framework to help address this challenge of unifying the picture across the scales of complexity, from the test tube to human pathology, by recasting the problem as a competition between innate drive to aggregate and active removal of aggregates. Under controlled in vitro conditions, an aggregation-prone system will tend to its free energy minimum, the aggregated state. By contrast, in vivo aggregates are removed and proteins are kept in their out-of-equilibrium functional states, by active, energy-consuming processes. Disease occurs when the system manages to break free from this metastable functional state, either by bypassing the kinetic barriers or because the active processes meant to keep the system from aggregating can no longer function effectively enough. A key characteristic to enable this escape is the ability of protein aggregates to self-replicate, that is existing aggregates can catalyze the formation of new ones, leading to exponential amplification. This view of in vivo protein aggregation then naturally provides a classification for potential therapeutic strategies into those that increase aggregate removal, those that decrease thermodynamic driving forces, and those that increase kinetic trapping. While quantitative comparison and prediction of in vivo effects will require the development of detailed mathematical models, this thermodynamic view of protein aggregation can provide an intuitive way to link changes in molecular processes to effects in disease.

AUTHOR DECLARATIONS

Conflict of Interest

Georg Meisl is an employee of WaveBreak Therapeutics.

Author Contributions

Georg Meisl: Conceptualization (equal); Investigation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

DATA AVAILABILITY

Data sharing is not applicable to this article as no new data were created or analyzed in this study.