Alice I Bartlett & Sheena E Radford, University of Leeds, UK

The wide range of techniques developed over recent years has led to a near-atomistic view of the folding landscape of small, model proteins in vitro. Approaches combining both experiment and simulation have been crucial in achieving this goal, and the synergy of these approaches will become even more important as the field develops in the future. At a fundamental level, much remains to be learned about the biophysics of protein folding: the prediction of folds and folding mechanisms is far from routine, and comparison of results from simulations and experimental approaches remains difficult. The concept of proteins as ensembles and the realization that a rare conformer may possess distinct functional or biological properties provides immense experimental challenges for the detection, structural characterization and biological interrogation of such species. Such phenomena are also important for developing better understandings of concepts such as allostery and cooperativity, which are key to many crucial cellular events. Only by understanding these fundamental concepts can we hope to modulate the properties of proteins at will and understand and exploit the complex networks of interactions that a single genetic product can undertake.

Other future challenges will involve deciphering the folding mechanisms of large proteins and protein complexes, including macromolecular machines with multiple subunits, cofactors and nucleic acids, that perform many of the essential functions of life. Modifications such as phosphorylation, glycosylation, methylation and lipidation that extend the functional repertoire of proteins in the context of their folding landscape must also be tackled. The exploration of how membrane proteins fold is a subject in its infancy biophysically, and there are many challenges ahead in unpicking this complex folding environment. In addition, we need to expand our understanding of folding in the cell to a quantitative level. In the cellular environment, folding commonly involves molecular chaperones and may be challenged by stochastic events such as aggregation. Current research is focusing on developing and applying biophysical techniques to the in vivo environment, with the ultimate goal of a quantitative understanding of the folding landscape in the cell.

Johannes Buchner, Technische Universität München, Germany

In the 1980s, the field of molecular chaperones witnessed a jump-start, with many seminal discoveries of unexpected traits and diversity. This led to the emergence of new concepts that stimulated the entire field of protein folding. An example in case is the finding that in the bacterial cytosol some polypeptide chains fold in splendid isolation one by one within the confined GroEL chamber. The excitement over the uncharted territory opening up and the rapidly growing list of functionally diverse chaperones put the concept of assisted folding in the limelight and, for some, spontaneous folding was 'old school'. In the past years, the pendulum has been swinging back and, by now, may have reached middle ground. The present evidence suggests that, also in the living cell, most proteins fold spontaneously, but there is a certain fraction of proteins that are addicted to chaperone assistance to reach or maintain their functional states. And, as Anfinsen rightly pointed out many years ago, environmental conditions are a key factor in this context: under stress conditions, such as heat shock, chaperone dependence will prevail.

Despite all the progress on chaperone structure and function, we still lack a detailed description of how a chaperone affects client protein conformation during assisted folding. What happens to a protein while it is processed by a chaperone? We got first glimpses for proteins in GroE complexes, but for all the other chaperones information is largely lacking. This is a technically challenging area, but certainly one that is central to propelling the field to the next level. Biophysics and cell biology must join hands to achieve this goal. On the other hand, we need a bird's eye view of chaperone action, that is, to watch the different chaperone proteins present in the same ball park functioning simultaneously. Are they team players with defined functions? Do they talk to each other? Are they organized as relay teams? Is there competition? How flexible is this system? How are viruses and other pathogens exploiting chaperone function?

Given the speed with which this field evolved, it will not be long until we get the first answers—and new unexpected questions.

Fabrizio Chiti, Università di Firenze, Italy

Forty years ago, the Levinthal paradox hypothesized that the process of protein folding would require a time longer than the age of the universe if it occurred via a random search of contacts, postulating the existence of well-defined pathways with intermediates guiding the search to the fully native state.

This reasoning led the field to search for folding intermediates. As more and more details were collected on the structure and dynamics of intermediates, it also became clear that some of them were off-pathway species. Moreover, the new view of protein folding has also clarified that stable intermediates forming on the energy landscape are not necessary for folding to be efficient.

While these conclusions were reached, investigators started to realize that proteins have an intrinsic potential to form misfolded protein aggregates, such as amyloid-like fibrils or other structured aggregates, and that this propensity is particularly high for fully or partially unfolded states, such as those formed during folding. Protein-folding intermediates can have a role that was unexpected until 10 years ago, that is, to modulate the propensity of a protein to aggregate. As partially folded states can be more or less amyloidogenic than fully unfolded states, depending on conditions and the nature of the structure present inside them, the formation of intermediates along the folding reaction, their stability and on- or off-pathway nature can depend on their ability to counteract aggregation. This is a major challenge that the field has to face in conjunction with that of protein misfolding. A full description of the energy landscape of proteins that takes into account both folding and aggregation can help us to understand the shape of a particular energy landscape.

The intimate link between folding and aggregation is also essential for our ambition to clarify protein folding in vivo. Chaperones are not species devoted only to helping the intramolecular search of native contacts within a protein; they also inhibit intermolecular contacts between distinct folding molecules. An understanding of the fascinating mechanism by which the protein-folding machinery works in vivo cannot prescind from this concept.

Douglas Cyr, University of North Carolina, USA

Protein quality-control factors act in integrated networks to maintain non-native proteins at optimal levels for proper folding and assembly. This is crucial for survival, as escape of misfolded proteins from surveillance leads to the accumulation of toxic protein species and a number of protein conformational diseases. Molecular chaperones protect cells from such pathologies by suppressing protein aggregation, promoting protein refolding and even facilitating protein aggregation. All these cellular efforts aim to limit the accumulation of toxic protein oligomers implicated in neurodegenerative diseases as agents of death. Emerging data indicate that specific Hsp70-Hsp40 chaperone pairs also function as substrate selectors for a network of multisubunit quality-control E3 ubiquitin ligases that target misfolded proteins for degradation. Chaperone-dependent ubiquitin ligases are localized to the cytoplasmic face of the endoplasmic reticulum, the cytosol and the nucleus. In addition, we are just realizing that there are also single-subunit quality-control E3s with built-in chaperone domains that function independently of Hsp70 to directly recognize and target misfolded proteins for degradation.

The CHIP-Hsp70 multisubunit complex was the first chaperone-dependent E3 ubiquitin ligase identified, and elevation of CHIP activity suppresses toxicity of neurodegenerative proteins to different degrees. Molecular chaperones have clear preference for non-native proteins that show different aspects of non-native protein structure. Therefore, it is possible that members of the chaperone-dependent ubiquitin-ligase network preferentially recognize different disease proteins. If true, it is possible that fluctuations in the expression pattern of chaperone-dependent ubiquitin-ligase family members are related to the selective vulnerability of neurons to specific disease proteins. Tests of these concepts will position the field to develop compounds that modulate activity of individual chaperone-dependent ubiquitin ligases to treat protein conformational disease.

Judith Frydman, Stanford University, USA

One of the big intriguing issues in the field is that, even after all these years of intensive research, we still don't know what chaperones really do in the cell. Which proteins need specific chaperones and why? How much overlap is there among different chaperone systems? Experiments from our lab and others clearly show that chaperones are organized in robust networks within eukaryotic cells, and these networks somehow become less functional during age and disease. Understanding such decline will require addressing the poorly understood question of the exact structure and composition of cellular folding networks.

Another challenge ahead is to explore the subcellular organization of folding and quality control networks within the cell. It is increasingly clear that chaperones interact physically with ribosomes for de novo folding; similarly, we and others have shown that chaperones appear to be organized in subcellular locales dedicated to refolding and quality control. Hsp90 localizes to the Z-line in muscle cells, where it helps in myosin folding and assembly: perhaps these are instances of a broader phenomenon, namely that folding and quality control are compartmentalized within the cell.

These are important questions to figure out the underlying logic of the folding network in the cell, as well as issues with practical implications for our understanding of disease. Thus, differences among the folding network composition and/or localization in different tissues and cells may underlie the cell specificity observed for folding and aggregation diseases. The concept of protein homeostasis provides a simple framework to harness the growing links between impairment of cellular folding and a wide variety of gain-of-function and loss-of-function diseases.

Lila M. Gierasch, University of Massachussetts, USA

While considerable progress has been made in our fundamental understanding of the mechanism of protein folding and aggregation in vitro, we have only scratched the surface in our knowledge of folding and aggregation in vivo, which is absolutely required to develop effective therapeutic interventions for aggregation pathologies, such as neurodegenerative diseases like Alzheimer's and Parkinson's, and other amyloid diseases. Tackling the challenges to elucidate mechanisms of in vivo protein folding and aggregation will require creative and integrative biology as well as bold, powerful, and novel physical chemical tools and principles.

Folding in vivo occurs in a crowded, spatially organized, heterogeneous cellular environment, with a cast of folding assistants; this is drastically different from the highly dilute solutions and optimized buffers used for in vitro folding studies. Yet for the most part, evolutionary pressures have selected for success in protein folding in vivo. Proteins sample an energy landscape inside cells that is shaped not only by the information in their sequences, but also by the very nature of the cellular microenvironment and the many cellular components such as chaperones that interact with incompletely folded proteins. In the cell, proteins encounter this landscape for the first time upon their 'birth' from the ribosomal exit tunnel, and then subsequently fold, unfold, partially unfold, and refold throughout their lives under the energetic constraints of the in vivo energy landscape. The protein folding community must respond to the urgent need to link our fundamental grasp of the behaviors of polypeptide chains in defined media to the successful maintenance of functional, folded proteins in vivo and the consequences when proteins misbehave.

Arthur Horwich, Yale University, USA

There is a set of standing questions regarding protein misfolding, neurodegeneration and cytosolic chaperones.

1. Why do molecular chaperones fail to forestall misfolding when present in the very compartment where misfolding occurs, for example, in the cytosol by α-synuclein or by SOD1?
2. Why is there no heat-shock response mounted by the neuronal cells subject to misfolding by, for example, cytosolic misfolded proteins such as α-synuclein and SOD1?
3. What is the basis to cell-specific toxicity? For example, why are motor neurons affected by SOD1 misfolding and striatal neurons affected by α-synuclein?

In terms of methodological development, it would be very attractive to have a means to directly visualize the polypeptide backbone at the single-molecule level, such that one could look at ensembles of non-native molecules to 'see' their topologies. Molecules would have to be 'frozen' (perhaps with low temperature), their backbones 'imaged' (perhaps via metal chemistry), then 'read out' (perhaps by atomic force microscopy or EM techniques).

Ronald Wetzel, University of Pittsburgh, USA

With respect to protein misfolding, one important area where I believe nature is really poised to give ground is the issue of nucleation of disease-associated amyloid fibril formation. How spontaneous amyloid formation is initiated at the molecular level is of basic interest and also important for therapeutic intervention. In addition, it seems likely that polymorphic structures responsible for, for example, prion strains have their origins in nucleation. In a few exceptional cases, the nucleus seems to consist of energetically unfavorable conformations of the monomer. In the vast majority of cases, however, amyloid formation is preceded by the appearance of spherical oligomers and protofibrils, and it seems that, at least in some cases, nucleation occurs within these prefibrillar condensed phases. Some hints about the mysterious nucleation event are now emerging, but a rigorous kinetic and mechanistic accounting has yet to be made. I think we are poised to make progress on this issue because of a combination of new techniques and the emergence of protein systems allowing exquisite control over aggregation kinetics under physiologically relevant conditions.

Although most of the intense effort in understanding amyloid formation has focused on in vitro studies, the biologically relevant process occurs in vivo. In fact, nucleation of amyloid in vitro is the easy part; we will probably have to wait longer for a detailed description of how amyloid formation is nucleated in vivo. The two processes might be substantially different: except for in some cases where protein concentrations are relatively high, such as amyloid diseases, whose onset is triggered by abnormally elevated protein levels, the concentrations required for spontaneous amyloid formation in vitro tend to be orders of magnitude higher than what is typical for precursor pools in vivo. Moreover, in at least some diseases, precursors are made and/or circulate throughout the body, but amyloid deposits form at only one or a few locations. The hidden hand of some heterogeneous nucleus for amyloid formation seems implicated, but what is it? An extracellular matrix component or surface? A local concentration of a small helper molecule? Even if, as held by the current wisdom, amyloid fibrils are not toxic or even protective, nucleation of amyloid formation would remain relevant to disease as a major route by which cellular pools of more toxic species are diminished. Needless to say, the long-term big questions continue to be just this last issue: the identities of toxic species, and the means by which they cause disease.