Effect of crowding, compartmentalization and nanodomains on protein modification and redox signaling – current state and future challenges

Biological milieus are highly crowded and heterogeneous systems where organization of macromolecules within nanodomains (e.g. membraneless compartments) is vital to the regulation of metabolic processes. There is an increasing interest in understanding the effects that such packed environments have on different biochemical and biological processes. In this context, the redox biochemistry and redox signaling fields are moving towards investigating oxidative processes under conditions that exhibit these key features of biological systems in order to solve existing paradigms including those related to the generation and transmission of specific redox signals within and between cells in both normal physiology and under conditions of oxidative stress. This review outlines the effects that crowding, nanodomain formation and altered local viscosities can have on biochemical processes involving proteins, and then discusses some of the reactions and pathways involving proteins and oxidants that may, or are known to, be modulated by these factors. We postulate that knowledge of protein modification processes (e.g. kinetics, pathways and product formation) under conditions that mimic biological milieus, will provide a better understanding of the response of cells to endogenous and exogenous stressors, and their role in ageing, signaling, health and disease.


Introduction
Biological systems are highly crowded, dynamic and heterogeneous environments composed of a vast array of different molecules and macromolecules (e.g.proteins, lipids, nucleic acids and polysaccharides).Despite this high complexity and heterogeneity, the majority of the biomolecules that compose these milieus have well-defined functions and react and/or interact with high efficiency and specificity with their corresponding partners/substrates/cofactors.Thus, biochemical reactions take place, for example, in the heavily crowded cytoplasm of eukaryote, or prokaryote cells, where the total concentration of molecules can reach ~400 mg mL − 1 , resulting in up to 45% of the total volume being occupied.This results in altered diffusion, reaction kinetics, and other aspects that can alter cell function and signaling.In particular, there is considerable evidence to support the hypothesis that the formation of membraneless compartments (defined as distinct regions within a cell that are not enclosed by a traditional lipid membrane, that arise from liquid-liquid phase separation), nanodomains (small regions within a cell with a distinct structure and function) and altered local viscosities, may play a significant role (Fig. 1) [1][2][3][4][5][6].
Amongst the macromolecules present in biological systems, proteins stand out not only for their significant role in cell structure, function, signaling and replication, but also for being the most abundant macromolecules, with 2-4 million proteins per cubic micron in both bacteria and mammalian cells [7].High concentrations of proteins are also encountered in the extracellular matrix and in biological fluids such as blood plasma (~80 mg protein mL − 1 ), demonstrating that molecular crowding is a general hallmark of biological systems.Similar crowded conditions are found in foods and some pharmaceutical preparations (where protein concentrations of 100-200 mg protein mL − 1 are common) indicating that this scenario is not unique to cells (Fig. 2) [8,9].
Despite the emerging relevance of crowding, nanodomains and compartmentalization for biology, most of our knowledge about biochemical reactions such as protein modification (in terms of reaction rates, pathways and structural consequences) comes from studies carried out using dilute and homogenous conditions.Under these conditions, molecules move freely and reactions are driven by the chemical properties of the reactants and products (thermodynamic and entropic factors, oxidation/reduction potentials) and their diffusivity in homogenous systems.This is far from biological reality where crowding, increased local viscosity, and confinement are common features.Data have been reported that indicate that these phenomena can alter, amongst others, macromolecular association events, and protein folding, diffusion (rotational and translational), interactions and aggregation [10][11][12][13].In this context, a recent study by Vorontsova and coworkers confirmed that crowding is essential for cellular homeostasis and normal optics in zebrafish lenses [14].Alterations in the development of the crowded environment by mutations on aquaporins revealed key aspects of the regulation of this complex phenomenon.
The hypothesis that tight packinggenerated from having a high concentration of molecules in solutionmodulates the rate, extent and pathways (mechanisms) of protein modification reactions is therefore logical.
In this work, we focus on reviewing the knowledge available to date on the impact of crowding on biochemical processes, with discussion of the consequences that these may have for protein oxidation, glycation and redox signaling.We will firstly address, from a chemical point of view, the characteristics of crowded environments, and then discuss how such packed environments can modulate the diffusion rates of oxidants, targets (e.g.proteins) and co-factors, as a consequence of increased local viscosity, the presence of nanodomains, and altered escape of oxidants/radicals from solvent cages or local environments.This may be of particular significance for species such as H 2 O 2 which can be formed at specific locations (e.g. by enzymes located at specific sites, such as membrane-bound NADPH oxidases, or organelles such as mitochondria), resulting in marked gradients in oxidant concentrations.These phenomena are expected to alter chain reactions, dimerization events, and oxidant transfer processes, either in a positive or negative manner, depending on the site and characteristics of the vicinity where these are formed.Furthermore, altered hydrogen-bonding networks and intermolecular interactions in highly packed systems can result in anomalous rates of hydrogen and electron transfer reactions (e.g.photochemical reactions, radical migration, and transfer of damage to other biomolecules including DNA and RNA) as reported recently [15].Subsequently, we discuss the consequences of crowding on protein conformation and surface dynamics (e.g.altered protein folding) which can have consequences for protein oligomerization, resulting in enhanced protein aggregation and the formation of fibrils that have been associated with some pathologies [10].Finally, we review experimental strategies that can be employed to study protein oxidation and glycation in highly packed and heterogeneous systems, and discuss potential future directions for this developing field.We encourage the reader to consult other excellent reviews that have covered physical and chemical aspects of crowding in depth, to get more insights about this relevant, but neglected field [2,3,16].

Biological interfaces and the excluded volume effect
The high concentrations of proteins, nucleic acids and polysaccharides in the cell cytoplasm, along with the presence of organelles and membranes, results in most interactions and reactions between molecules inside cells occurring near to a surface interface rather than in a bulk homogenous solventthis is also true for other heavily crowded biological milieus (Fig. 2).Considering this, collisions between molecules, equilibria and other properties of the system (e.g.free energy, entropy, enthalpy and volume) are governed by interfacial physical chemistry.Hence, biological systems are considerably more difficult to interrogate than the dilute single-phase (homogenous) systems employed in most in vitro assays.Unsurprisingly, greatly slowed and anomalous diffusion rates have been reported for both small (e.g.dyes and peptides) and large molecules (e.g.proteins), as well as for the solvent (e.g.water), when these have been investigated in denselypopulated systems [6,17].In addition to effects on diffusion, other interactions and steric effects must also be taken into account in packed environments, as collisions between particles are likely to occur more often than predicted.
As two molecules cannot occupy the same location at the same time, one of the dominant phenomena that occurs under crowding conditions is the volume exclusion effect [18].The excluded volume represents that portion of the total volume available that cannot be occupied at a particular time because of the presence of other molecules.Thus, as the concentration of a specific macromolecule is increased in solution, the fraction of volume available for other species decreases.This is key to understanding biochemical reactions at biological interfaces as it can generate clusters where high concentrations of specific reactants can be encountered, or it can impede translational diffusion (e.g.ingress or escape) of molecules as a result of an increased number of obstacles.Individual proteins have, on average, a diameter between 2 and 7 nm (for proteins with molecular masses ranging from 8.5 to 66.5 kDa [19][20][21]), whereas protein complexes (e.g.ferritins) and peripheral sheets of the endoplasmic reticulum can have diameters of 12 and 50 nm in eukaryotic cells, respectively [22,23].A similar scenario occurs with some reactive species.Some are of a similar size to water molecules (e.g.small oxidants such as •OH, ONOOH, NO•, H 2 O 2 , O 2 • -and 1 O 2 ), but other can be up to ~7 nm in diameter (e.g.phospholipid-or protein-derived peroxyl radicals).These differences in diameter might not seem significant, but if we assume spherical shapes for molecules then the volume that they occupy in solution is given by equation (1) where r is the radius in nm: Fig. 1.Crowding decreases the fraction of volume available for molecules present in biological systems and acts as a driving force for the occurrence of interactions and macromolecular association events.This result in altered local viscosities (e.g.due to hydrogen-bonding networks and electrostatic interactions) and formation of nanodomains (small regions within the cytosol with different composition to the bulk system) which may enhance the local concentration of proteins and oxidants, thereby modulating reaction kinetics and oxidation pathways.(Eq. 1) Considering that most proteins have similar densities (typically 1.22-1.55g cm − 3 , with an average of 1.35 g cm − 3 [24]), a relationship between mass and volume can be easily obtained for different proteins, with these values allowing estimates to be made of the total volume that different concentrations of specific proteins occupy in solution.Taking as an example, hemoglobin (molecular mass ~64,000 g mol − 1 , radius ~3.2 nm [25]), with a concentration of 4.8-5.4mM in erythrocytes from healthy individuals [26], this protein occupies a total volume of 4.0-4.5 × 10 23 nm 3 per L. This equates to 40-45% of the total volume of red blood cells (~89 ± 18 μm 3 [27]), and this is only one of the many proteins present in these cells.Furthermore, it is worth noting that erythrocytes represent a simplified example of the packing within cells, as these lack the nuclei and other internal structures present in most eukaryotic cells.Moreover, the macromolecules (e.g.proteins, lipids, polysaccharides, RNA and DNA) present within cells have multiple heteroatoms and/or conjugated π-systems (e.g.alkenes and aromatic rings) in their structures that can generate specific, or non-specific, non-covalent interactions between them.Therefore crowding can be a driving force for macromolecular association and formation of membraneless nanodomains.This helps explain the slower and anomalous diffusion of solutes and solvents near biological interfaces (or within compartments).Increasing evidence indicates that these effects are not isolated to proteins, with a recent report indicating that mammalian oocytes store mRNA's in a mitochondria-associated membraneless compartment [28].

Viscosity and diffusion are altered near biological interfaces
Waterthe solvent that prevails in living organismsforms hydration layers around proteins (hydrodynamic radii) which result in a decreased rate of diffusion of water molecules (by a factor of 2-4 relative to bulk solution) [6,17,29].This also holds for small hydrophilic oxidants of moderate reactivity and selectivity such as H 2 O 2 .The altered diffusion of species near biological interfaces can be explained by the formation of hydrogen-bonding networks and electrostatic interactions with surface species, as clearly illustrated by heavily-hydrated proteoglycans (e.g.versican, aggrecan and to a lesser extent other species such as perlecan) and glycosaminoglycans (e.g.hyaluronan) present in many extracellular matrices and in synovial fluid [30][31][32].Theoretical and experimental studies have examined the molecular mechanisms and key features that govern this phenomenon.These studies have conceptualized fluids near surfaces as glass-like or supercooled systems where heterogeneous dynamics prevail [6,33,34].These considerations have driven a greater understanding of the effects of increases in local viscosity near biological interfaces, which this being up to four times higher than that expected for a spatially homogeneous application of the Stokes-Einstein relation, and this explains the decoupling of diffusional motion from the effective viscosity [6].This same phenomenon has been described for the diffusion of larger molecules in cells due to increased local viscosities, and is therefore expected to be a universal phenomenon rather than an isolated example [35][36][37].For example, the diffusivity (D) of bovine serum albumin has been calculated to decrease from 6.9 × 10 − 7 cm 2 s − 1 in water to 1.0 × 10 − 8 cm − 2 s − 1 in cells [38].This calculation is supported by neutron backscattering data, which have shown that self-diffusion of bovine serum albumin in crowded aqueous solutions strongly decreases with increasing protein volume fraction [39].A 20% slowdown in the diffusion of bovine serum albumin has been reported by Roosen-Runge et al. with this attributed (solely) to hydrodynamic interactions, as the medium used the same protein as crowding agent [39].However other reports indicate that crowding decreases the cytosolic diffusion of molecules by approximately 4-fold compared to pure water [29].Furthermore, changes in protein mobility would not be expected to be equal between species; studies carried out using polydisperse systems have demonstrated that the motion of large proteins is slowed down to a greater extent than for small proteins [40].
On the basis of these data that indicate that there are altered local viscosities in crowded solutions, it can be hypothesized that the diffusion of oxidants, and their rates of reaction will also be affected.This may be of particular relevance for low reactivity species that induce redox signaling (e.g.H 2 O 2 ), charged (e.g.O 2 • -), and large oxidants (e.g.some carbon-centred and peroxyl radicals), as these are likely to be more markedly affected.Small, highly-reactive oxidants (e.g.•OH and HOCl) are thought to be less affected by crowding and local viscosity as these species are typically restricted to local environments, and have very limited diffusion radii, with reaction occurring at, or near, their site of formation due to their high reactivity (cf.rate constants for reaction of HO• with most biomolecules of 10 9 -10 10 M − 1 s − 1 [41]).Winterbourn has estimated the diffusion distance for different oxidants including •OH, HOCl and H 2 O 2 in the presence of 2 mM GSH [42].These were estimated using the expression: where l is the distance diffused by the oxidant before a drop in its initial concentration from C 0 to C, k is the rate constant for the reaction (in M − 1 s − 1 ), S the concentration of the substrate (in M), and D is the diffusion coefficient of the oxidant in aqueous solution.Although D is not known for these oxidants in cellular environments, recent studies have reported that D for HO• and H 2 O are similar at room temperature and 1 atmosphere pressure, which is as expected for small molecules in aqueous solutions [43].Considering this, we have recalculated the values given by Winterbourn (2008)  for HOCl [45] and 0.9 M − 1 s − 1 for H 2 O 2 [46].The calculated distances for the diffusion of the oxidants before the concentration of these drops to 10% of the original concentration are 0.03, 0.5 and 2920 μm for HO•, HOCl and H 2 O 2 , respectively (Table 1).These values are close to those estimated by Winterbourn with the difference arising from the use of Fig. 2. Concentration of proteins utilized in diluted in vitro assays designed to understand protein modification can differ up to 100-fold from protein concentrations encountered in 'real world' biological systems.
different diffusion coefficients [42].However, considering that viscosity and diffusion are interrelated by: where k B is the Boltzmann constant, T is the temperature, r H is the hydrodynamic radius and η is the viscosity, it is clear that an increase in the viscosity of the medium will be reflected in a decreased diffusion coefficient (D) as these variables are inversely proportional.Thus, a 3-fold increase in the viscosity of the medium near to biological interfaces will be reflected in a 3-fold decrease in the diffusion coefficient of these oxidants with this being particularly relevant for more selective species (e.g.H 2 O 2 ).Hence, a recalculation of the distance diffused by H 2 O 2 before a drop to 10% of its initial concentration in a 3-fold more viscous medium (i.e.D ~965 μm 2 s − 1 ) containing 2 mM GSH would be ~1685 μm (Table 1).It should also be noted that the diffusion of H 2 O 2 is dramatically affected by the substrate utilized to do these calculations.
Whilst most thiols react with H 2 O 2 with rate constants of the order of ~1 M − 1 s − 1 , the active site cysteine (Cys) residues in peroxiredoxins react with rate constants, k 2 , in the range 10 5 -10 7 M − 1 s − 1 [47], thus the diffusion of H 2 O 2 in a compartment containing 20 μM of these enzymes, is much lower (8.7 μm if k 2 is 10 7 M − 1 s − 1 ; cf.2920 μm in the absence of peroxiredoxins) (Table 1).These calculations are however, based on kinetic data determined for dilute and homogeneous environments, thus the rate of diffusion of oxidants such as H 2 O 2 , and also antioxidants (e.g.GSH) or oxidant-removing enzymes (e.g.peroxiredoxins), is also expected to be significantly lower as a result of increased viscosities and the presence of nanodomains (i.e.similar to the evidence obtained for modulated diffusion of water molecules in biological environments).

Effects of crowding on protein folding and surface dynamics
Protein folding and surface dynamics are critical for the regulation of biological activity.As polypeptide chains are synthesized by ribosomes, protein folding is initiated to give three-dimensional assemblies, although these may not reflect the final structures.These initial structures can subsequently undergo re-folding to a final form, as only correctly-folded proteins have significant long-term stability under the conditions within (and external to) cells.Incomplete folding, a failure to fold, or aberrant folding (e.g. as a consequence of structural perturbations induced by oxidation), results in rapid cellular responses, involving multiple pathways, that either increase the capacity to re-fold and/or repair proteins (e.g. via upregulation of chaperone proteins and repair proteins) or remove them (e.g.proteasomes, proteases, lysosomal enzymes).When the accumulation of such species is uncontrolled, various programmed cell death pathways (e.g.apoptosis) are induced [50,51].
As crowding and confinement result in increased local protein concentrations and reduced available volume (excluded volume phenomenon), these are expected to decrease the entropy of the system, thereby resulting in a stabilization of folded structures and enhanced assembly of proteins (and other molecules) into supramolecular structures (e.g.microtubules and nuclear condensates).This has been shown experimentally, in studies where inert crowding agents were used to investigate the effects of crowding on folding thermodynamics and kinetics, with strong correlations reported between the presence of crowding agents and enhanced stability of proteins in their native state [13,16,52,53].This is likely to be of particular relevance for reaction of oxidants with buried amino acid residues in proteins where the partial unfolding of specific domains within a protein structure is essential for reaction to take place.

Crowding alters enzyme kinetics
Excluded volume phenomena have been shown to affect the Michaelis-Menten kinetic parameters of a number of enzymes including adenylate kinase, alcohol dehydrogenase, horseradish peroxidase and lactate dehydrogenase, with the K m and V max values of these enzymes showing significant variations under crowded conditions when compared to dilute [54][55][56][57].The detection of these effects with multiple different enzyme types suggests that this is a general phenomenon.Indeed, in vitro experiments have demonstrated a clear dependence of the rates of catalyzed reactions and tertiary complex formation on the viscosity of the solution [58].Förster resonance energy transfer experiments and solvation studies have revealed that crowding also induces domain displacements and restriction of key sites in adenylate kinase [57].Thus a concentration-and crowding agent-dependent decrease was detected for the K m values of this protein, with kinetic experiments in presence of 200 mg mL − 1 of the crowding agent dextran-70 resulting in a ~2.5-fold lower K m than in a dilute buffer solution.This effect was independent of the crowding agent, with the K m values detected with 200 mg mL − 1 dextran-40 and ficoll-70 being 5-fold and 10-fold lower, respectively [57].Similar findings have been reported for horseradish peroxidase in presence of ficoll and polyethyleneglycol, with a decrease in K m from 5.5 × 10 − 5 M under dilute conditions to 3.1 × 10 − 5 and 2.5 × 10 − 5 M in presence of 30 mg mL − 1 ficoll and polyethyleneglycol, respectively [54].These findings indicate that enzymes show different biochemical/biophysical properties when in crowded environments that more closely mimic likely biological situations, than dilute (typical experimental) conditions.Whether similar effects occur with antioxidant enzymes has yet to be established, but it is likely that their kinetic parameters will also be modulated.Such changes may be particularly marked with enzymes requiring (reducing) co-factors in addition to their substrates, and also for enzymes that react with large targets (e.g.other proteins, phospholipids etc) as diffusion and viscosity effects will diminish the rate of interactions of the enzyme with both co-factors and target substrates.

Crowding and nanodomains as modulators of protein oxidation and glycation
Since crowding can affect protein folding/unfolding and alter protein dynamics it is natural to hypothesize that the sites and extent of modification of proteins exposed to oxidants or glycation agents in biological environments may differ from the data obtained from studies carried out under dilute and homogenous conditions.Thus, under dilute conditions there are more degrees of freedom and protein dynamics are expected to be higher; the opposite is expected under crowded conditions with these likely to be of particular relevance for flexible proteins.
E. Fuentes-Lemus and M.J. Davies modification of plasma proteins exposed to the oxoaldehyde glycation agents methylglyoxal and glyoxal under crowded conditions differ from those detected under dilute conditions [62].These data illustrate that crowding can be a significant factor, and indicate that further studies are needed to elucidate these phenomena.However, as each protein has a unique structure, composition and properties (e.g.flexibility) it is likely that there will be a considerable variation between proteins with regards to the effects of crowding agents and related effects (e.g.altered local viscosity) on protein oxidation and glycation.Therefore studies should be carried out under conditions that mimic as closely as possible in vivo conditions, as crowding and high local viscosities may affect diffusion, folding and surface dynamics and the occurrence of random versus site specific modifications, as these are known to affect the pattern of products and their yields [15,60].

Chain reactions and propagation of oxidative damage in densely populated environments
Proteins are major targets for many one-and two-electron oxidants generated in biological systems [41,63].This is due to their high abundance, and the high reactivity of some amino acid residuesparticularly those containing sulfur atoms (i.e.Cys, cystine and Met) or aromatic sidechains (e.g.Trp, Tyr and His) [41,63].As the general principles, and the structural and functional consequences of protein oxidation and peroxidation on exposure to different biologically-relevant oxidants have been extensively and rigorously reviewed, we refer the interested reader to a number of recent articles on this area [63][64][65][66][67][68].Moreover, as crowding is less likely to affect the rates of reaction and mechanisms of reaction of highly reactive oxidants (e.g.HO•) this topic will not be discussed further here.However, a greater influence of crowding effects would be expected for less reactive initial oxidants, that often diffuse greater distances before reaction, and secondary species (e.g.protein-derived alkyl or alkylperoxyl radicals, sulfenic acids, chloramines) formed on reaction of one-and two-electron oxidants with proteins, that can affect the mechanistic pathway and the yield of final products.The available data for such reactions are therefore discussed in more detail below.
Radical-mediated oxidation of proteins (P-H) can result in the formation of secondary protein-radicals (generally designated as P• from hereon) on the peptide backbone, or on the sidechain of susceptible residues (e.g.Cys•, Trp•, Tyr•).These can be generated by hydrogen atom abstraction, or electron transfer processes followed by deprotonation.However, oxidants with modest redox potentials (e.g.peroxyl radicals) generate P• at relatively slow rates, and secondary radicals are therefore prevalent and important species when considering the effects of crowding on exposure of proteins to oxidants.Once P• are generated, these can recombine, via radical-radical termination reactions (i.e.P• + P• → P-P) that result in protein cross-linking [69], or react with O 2 to give secondary peroxyl radicals (POO•) and subsequently hydroperoxides (POOH), via further hydrogen atom abstraction reactions with other sidechains with available X-H bonds (particularly with X = S, N, or C) [63].These species contribute to the propagation of (modest) radical-chains and damage within protein structures (protein peroxidation) [63].As crowding and formation of membraneless compartments results in high protein concentrations and confinement within local 'structures', the increased proximity of proteins would be expected to modulate these reactions, and particularly the propagation of damage, in a significant manner (Fig. 3).
Whether these are enhanced or diminished is likely to be dependent on the situation.Thus formation of two radicals simultaneously (or near temporally) within a constrained environment may enhance the likelihood of dimerization.In contrast, with low oxidant fluxes, as is usually the case in biological systems, only single oxidized species are likely to be present within a particular constrained domain, and this would be expected to diminish the rates of dimerization (due to decreased diffusion etc., as discussed above).In contrast, these factors would be expected to enhance the rate of damage propagation (chain reactions) or damage transfer processes (Fig. 3).Current data suggest that the latter scenario is the most common.
Pioneering studies by Radi's group demonstrated that exposure of highly concentrated solutions of bovine serum albumin (≤4.5 mM protein, and in the absence or presence of crowding agents) to ONOO − / ONOOH, or metmyoglobin/H 2 O 2 , under aerobic conditions, favors propagation of damage with formation of secondary cysteinyl peroxyl radicals (Cys-OO•) [59].Cys-OO• can oxidize vicinal amino acid residues, with this being favored at high protein concentrations, resulting in propagation of oxidative reactions both within proteins (intramolecularly) or with other proteins in the local vicinity (intermolecularly).The authors also demonstrated that O 2 concentrations play a role in the propagation of radical-dependent protein oxidation [59].This is probably due to the fact that reaction of P• with another X-H (without formation of POO•) merely transfers damage from one site to another (i.e. to give P-H and X•), and therefore the overall extent of damage remains constant.Thus the ratio of products relative to initial oxidant is 1, whilst in the presence of O 2 the ratio is > 1 due to the formation of both POOH and X•.
We have recently reported similar data for intrinsically disordered casein proteins.Exposure of dilute and crowded solutions of these proteins (up to 1.2 mM) to AAPH-derived peroxyl radicals, or riboflavin/ light, resulted in enhanced ratios of amino acid loss per radical generated, and evidence for short chain reactions at the highest protein concentrations studied [61].Trp and Cys residues were the most affected residues (as might be expected from the ease of oxidation of these sidechains), but enhanced damage to Tyr and Met residues was also observed.This enhanced oxidation is probably due to radical rearrangement and/or transfer reactions [60,61].Casein proteins were chosen for these studies as they have a high capacity to self-assemble into micelles, a process that is favored at high protein concentrations, with the close proximity between protein chains favoring propagation of oxidative damage [61].The formation of chain-carrying Trp-derived Fig. 3. Overview of the biological fate of secondary radical species generated on sidechains (represented as XH) on exposure to one-electron oxidants.Reactions that participate in a change of the site of modification and/or the propagation of damage are highlighted in blue.Radical-radical termination reactions are highlighted in red.Crowding and nanodomains modulate these reactions with increased propagation of damage and decreased radical-radical termination reactions probably due to altered radical diffusion.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)E. Fuentes-Lemus and M.J. Davies peroxyl radicals (TrpOO•) able to oxidize other targets such as Met and Tyr residues was confirmed, in agreement with previous reports [59,60].Studies with free amino acids and short-peptides, have confirmed that crowding and nanodomains generated by addition of inert crowding agents (dextran or ficoll) to dilute buffer solutions, enhances both the rate, and overall extent of Trp oxidation [60].The product distribution was also different between dilute and crowded conditions, suggesting that crowding enhances the rates and overall extent of termination reactions, in addition to affecting the initial oxidation process (i.e. initial rates).This effect was confirmed by analyzing the extent of formation of casein cross-links by electrophoresis, with significantly less cross-links observed with high concentrations of crowding agent (e.g.3.4 mM dextran) (Fig. 4), which along with the increased loss of amino acids determined under these experimental conditions, confirms that crowding and nanodomains modulate oxidation pathways [60].It should be noted that the kinetics and the overall extent of Trp loss was determined after short periods of incubation (e.g. 30 min), thus studies that examine the effect of crowding and hydrophilic nanodomains over longer time periods would appear to be warranted.
Modulation of the pathways and kinetics of oxidant reactions is also likely to occur when damage is initiated by two-electron oxidants such as H 2 O 2 , ROOH, HOCl and ONOO − /ONOOH, though these reactions have been examined in less detail so far (though some data for ONOO − / ONOOH are provided in Ref. [59]).Thus, the subsequent reactions of species such as sulfenic acids (RS-OH, from reaction of the thiol group of Cys with each of these oxidants) and chloramines (RNHCl, from reaction of HOCl with amine sidechains), that can oxidize other protein sidechains (e.g.other Cys residues in the case of RS-OH, and Cys and Met for RNHCl) are likely to be affected by both viscosity (diffusion effects) and crowding.
Crowding effects have also been observed for glycation reactions, in which albumin and transferrin were exposed to reactive oxoaldehydes, such as methylglyoxal and glyoxal, in the presence of dextran or ficoll of different molecular masses (Fig. 4).These species form adducts with Cys, Lys, Arg and His residues.Although the chemistry of these glycation reactions are completely different, significantly less cross-links were detected for transferrin (but not albumin) exposed to methylglyoxal under crowded conditions when compared to dilute [62].Moreover, the yield of protein carbonyls, detected at specific time points was also affected, consistent with changes to the kinetics, or mechanisms of reaction [62].Together these data indicate that crowding can modulate modification pathways, or the fate of the species formed, likely via modulation of secondary reactions.

Potential effect of crowding on molecule-molecule reactions
In the previous sections, we have focused on discussing the effects of crowding on the reaction between proteins (and derived components) with one-or two-electron oxidants.However, sulfur-containing proteins (i.e. containing Cys, disulfides or Met) are abundant species (c.f.data for intracellular protein thiol concentrations of 10-40 mM [70]) and kinetically important targets of one-and two-electron oxidant species.However, thiols/thiolate anions (and also to a lesser extent nitrogen-based nucleophiles such as amines) can also react with soft electrophiles present in biological systems (e.g.α,β-unsaturated carbonyls and quinones/quinimines via Michael addition reactions, and aldehydes/ketones via Schiff base reactions) [71,72].Some of these reactions proceed with modest rate constants when compared to oxidant-mediated reactions.Thus, the reaction of methylglyoxal, a reactive di-aldehyde that is elevated in the plasma of people with diabetes, with albumin has a rate constant of ~7 × 10 − 3 M − 1 s − 1 [73]).However, some have much higher rate constants.Thus the Cys residue of glutathione (GSH) reacts with unhindered quinones with rate constants in the range 10 5 -10 2 M − 1 s − 1 [74], whereas the Cys34 residue of bovine serum albumin reacts with k 2 10 4 -10 1 M − 1 s − 1 [74], with these value being much higher than for a large number of oxidant reactions (reviewed [75]).These reactions result in the formation of protein-quinone or protein-carbonyl adducts of variable stability, with some of these species detected on proteins in vitro and in vivo [71][72][73][76][77][78].Molecular crowding and hydrophilic nanodomains generated by adding inert crowding agents (e.g.dextran or ficoll) have been shown to affect the mechanisms and extent of modification reactions of albumin and transferrin with methylglyoxal and glyoxal [62].These findings suggest that these reactions are likely to be modulated both in cells and in plasma.In this context, it would therefore appear desirable to obtain kinetic data for these reactions under crowded conditions to gain a better understanding of their relevance in health and disease.This may help elucidate the key drivers of these reactions in vivo, which clearly occur despite the low (apparent) rates of reaction.Thus, an increased molar ratio of carbonyl-or quinone-species to target proteins within domains or compartments may increase the rate, extent and selectivity of modification reactions.E. Fuentes-Lemus and M.J. Davies

Compartmentalization is a crucial regulator of redox signaling in biology
The data discussed above underline the importance of understanding the effects that molecular crowding and microenvironments have on protein modification, and the importance of using appropriate experimental systems to mimic these.This is also true for redox signaling processes.Recent literature have reported that redox signaling and phase separation (formation of membraneless compartments where proteins are confined in a small volume) are strongly associated with signaling processes.
Kato and coworkers have reported that ataxin-2 senses the activity of mitochondria to regulate the molecular target of rapamycin (mTOR) pathway in yeast [79].Ataxin-2 forms liquid droplets (i.e. a compartmentalized system with high concentrations of the protein) when the Met residues in the low-complexity domain of the protein are in their native (thioether) form, however, under conditions of oxidative stress (e.g. release of mitochondrial H 2 O 2 ) oxidation of these Met residues occurs, with this resulting in disassembly and melting of the liquid droplets [79] (Fig. 5A).This can be reversed by methionine sulfoxide reductase enzymes demonstrating that compartmentalization and oxidative modification of proteins play a significant role in key signaling processes [79].
Huang and collaborators have recently reported an inverse effect, in which protein oxidation triggers the formation of phase-separated domains (i.e.confinement) that modulate the signaling pathways that participate in plant flowering [4] (Fig. 5B).Thus, the authors demonstrated that the terminating flower transcription factor forms a network of intermolecular disulfide bonds between highly conserved Cys residues located in two intrinsically disordered regions of the protein.
Confinement and phase separation via formation of this disulfide network was shown to be mediated by H 2 O 2 produced within the proximal space of the protein condensates [4].Together, these findings illustrate the importance of formation of membraneless subdomains in cells in redox-regulated processes, but indicate that oxidation events can modulate these processes in different manners.
Crowding and nanodomain formation might help rationalize critical unresolved issues in redox biology and signaling, and particularly those related to how oxidant signaling occurs within cells despite the presence of high levels of oxidant removal and scavenging systems, such as GSH (up to 10 mM in hepatocytes [80]) and highly efficient peroxide-removing enzymes such as peroxiredoxins, glutathione peroxidases and catalase.In particular, there is considerable evidence to indicate that oxidation by H 2 O 2 (and others) of specific Cys (or selenocysteine, Sec) residues on 'sensor' proteins can subsequently transfer this data to effector proteins [81][82][83][84][85][86][87][88][89].These events involve the generation of short-lived sulfenic acid (RS-OH), or possibly sulfenylamide (RS-NHR') species, that interact with (oxidize) Cys residues on another protein.
These reactions occur, in the main, via the formation of new disulfide bonds that can be readily repaired, thereby resulting in a transient and controlled transmission of the signal (i.e.rapid switch-on/switch-off events).Similar reactions can occur with Cys residues conjugated to GSH (i.e.glutathionylated proteins) and nitric oxide (RSNO species) [81].The reaction of the sulfenic acid (or related species) with the target protein occurs against a background of high concentrations of other protein thiols (20-40 mM depending on the compartment) and 2-10 mM GSH [90,91].There is therefore considerable potential for aberrant oxidation of non-intended thiols.Despite this, there is abundant evidence for efficient signal transmission between the source and receptor, including between peroxiredoxin 2 (Prx2) and the transcription factor STAT3 [92], peroxiredoxin 1 (Prx1) and thioredoxin/ASK1 [93,94], Prx2 and kinases (MEKK1, MEKK4) in the p38 signaling pathway [95], the yeast glutathione peroxidase 3 analogue Orp1 and the transcription factor Yap1 [96], and the MICAL (molecule interacting with CasL)-Prx1-CRMP2 (collapsing response mediator protein 2) system [84].
The specificity and efficiency of such signaling can be rationalized if the sensor and target proteins form tight interacting complexes (i.e.physically interact, for which there is evidence in some cases), or via these events occurring with membraneless compartments or nanodomains (summarized in Fig. 6).These would minimize unwarranted (aberrant) interactions by enhancing the concentration of the reaction partners and/or minimizing the levels of species that might interfere with the required signaling event.Furthermore, in some of the examples indicated above, the generation of the oxidant (and particularly H 2 O 2 ) is reported to occur at specific and discrete locations, so there are locally enhanced oxidant concentrations when compared to the bulk system (as indicated by the dark grey zones in Fig. 6).Thus in the case of the Prx2-STAT3 system, the formation of H 2 O 2 is believed to occur via interaction of stimulants with the interleukin-6 (IL-6) receptor, which results in NAPDH oxidase (NOX)-mediated formation of H 2 O 2 .This then oxidizes (dimeric or decameric) Prx2, which in turn oxidizes STAT3 (to dimers or tetramers), with this only occurring in the presence of the scaffold protein annexin-2 (AnxA2) [85] (Fig. 6A).Of particular interest is data indicating that this multi-protein system associates to form a (non-covalently bound) trimeric complex before H 2 O 2 is generated, so that this system is poised for rapid signaling and responses.It is possible that further proteins are also involved in this system, as it has (so far) proved impossible to reconstitute this system in vitro [85].As this system Fig. 5. Protein confinement through phase separation and oxidation reactions are strongly connected components that modulate signaling outputs.Studies have shown that oxidation of proteins confined in liquid droplets (A) induced by H 2 O 2 leads to the disintegration of such membraneless compartments modulating the signaling outputs mediated by these proteins.In contrast, for other sensor/target proteins it has been shown that oxidation of highly-conserved Cys residues leads to protein confinement through formation of disulfide networks (B), with this process modulating the final signal outputs.involves (at least) three proteins, there must be a significant energetic costs to cells in maintaining this ternary complex in a pre-assembled state.This may be counteracted by molecular crowding and membraneless nanodomains, which may 'contain' the components at elevated concentrations within particular locations (illustrated schematically in Fig. 6B).STAT3 also interacts with the highly homologous protein Prx1, but Prx2, which is more sensitive to H 2 O 2 (i.e.reacts at more rapid rates with low concentrations of this oxidant) is the favored interaction partner [85].This preference makes logical sense, as it would allow sensing and transcriptional regulation, via STAT3, at lower concentrations of H 2 O 2 than with Prx1.In contrast, Prx1 signaling occurs via a stress-signaling kinase pathway involving thioredoxin/ASK1 [83,93,94], which would be expected to be activated only at higher H 2 O 2 concentrations.ASK1 has also been shown to associate with Prx1 (and Prx2) in the absence of H 2 O 2 with this suggested to be due to unidentified scaffold proteins, and possibly facilitated by molecular crowding, suggesting that this may be a common process involved in the formation of efficient redox relay systems [89].
A similar involvement of (one or more) scaffolding proteins in redox signaling has been shown for the Orp1-Yap1 system, where there is a requirement for the scaffold protein Ybp1 (Fig. 6A).Reaction of Orp1 with H 2 O 2 results in subsequent oxidation of the transcription factor Yap1, and upregulation of antioxidant-, stress-, detoxifying-and thiolreductase proteins, with this only occurring in the presence of Ybp1, and a 1:1:1 ternary complex [88].A similar requirement may hold for the Prx2-CRMP2 system, where reconstitution in vitro has also proved impossible [83].As with the Prx systems, the formation of this ternary system may be aided by crowding and nanodomain formation.
A further mechanism for the generation of selective signals has been reported for the Prx1-CRMP2 system, where the provision of H 2 O 2 for the oxidation of Prx1 comes from the FAD-dependent monoxygenase MICAL which oxidizes NADPH in the presence of O 2 .MICAL interacts with Prx1 via highly specific interactions which results in the formation of a protected 'channel' for H 2 O 2 transfer between the proteins, which is catalase insensitive, thereby ensuring specific oxidation (illustrated schematically in Fig. 6C).These types of events may also be enhanced by crowding (Fig. 6D) as this would be expected to drive and/or help maintain the protein complexes and limit disassembly (e.g. by preventing dissociation and limiting diffusion).Other MICAL proteins (MICAL2 and 3) can also oxidize both Prx1 and Prx2, but MICAL1 appears to be the preferred partner [84].
'Intended' reactions within membraneless nanodomains may also occur (in addition to the 'channeling' process discussed above) with different kinetics (rate constants) to those determined in bulk dilute solution (which are generally low for species such as H 2 O 2 ) due to a decreased requirement for diffusion-limited encounters and entropic limitations.Such domains may also play a major role in generating and maintaining gradients of signaling molecules, especially when diffusion of the initiating species or the downstream target is limited.Such diffusion can be modulated by protein size or other characteristics (e.g.membrane-binding domains), hydrophobicity/hydrophilicity effects (as may also occur with lipid-derived species), or charge effects (for example, with the superoxide radical anion, O 2 • -).It is therefore possible to envisage, fast and efficient signal transduction of signals between initial sensors and downstream effector proteins within domains or compartments where these molecules are localized, concentrated, and pre-assembled into multimeric complexes (or close associations) via crowding and nanodomain effects (Fig. 6).These complexes may only require a small proportion of the total complement of the proteins involved.Thus only 1-5% of the total cell STAT3 is reported to be involved in pre-formed Prx2-STAT3-AnxA2 complexes, with STAT3 believed to be the limiting factor [85], as Prx2 and AnxA2 are both highly expressed proteins in most cells.
These data have been used to generate computational models that account for such localized redox relay systems.These involve specific localized upregulation of oxidant production (with H 2 O 2 being primarily considered, but similar effects should hold for other oxidants), and formation of pre-assembled (or closely associated groupings) of sensor proteins and their signal receptors, in specific locations [86,87].These models may be improved by the inclusion of viscosity effects in these models, as this modulates molecular diffusion in a potentially significant manner [29].Fig. 6.Schematic illustration of potential processes involved in redox signaling under both physiological, and oxidative stress conditions, and the potential effects of molecular crowding on these.(A) Oxidantmediated physiological redox signaling arising from localized enzymatic generation of oxidants (or other controlled or regulated biological processes) at a specific site on a scaffold (protein, membrane or organelle).This would be expected to result in a gradient in the oxidant concentration (indicated by the grey circles of decreasing color intensity, see also text).The oxidizing equivalents may then be rapidly and efficiently transferred (green arrows) to closely associated effector molecules (red), which may be similarly localized to the scaffold.The generator and effector proteins may be pre-localized to the scaffold in covalent complexes, or assemble after oxidant generation.The former is likely to give rise to more efficient signal transfer.The efficiency and likelihood of these events may be enhanced (panel B) by molecular crowding (blue structures, which may be other macromolecules or cellular structures), increased solution viscosity, surface effects and decreased molecular diffusion.(C) Localized redox signaling arising from channeling of a generated oxidant from an enzyme (or other regulated source) to an effector protein within a defined complex.Such complexes may be pre-assembled, or generated ad hoc, and may exist as either bound entities (e.g.attached to a scaffold) or exist freely within a cell or compartment.These interactions, and efficient transfer of oxidation equivalents, may be enhanced by the generation, or maintenance of protein complexes via molecular crowding (panel D). (E) Non-compartmentalized stress signaling arising from unintended oxidant generation in solution (either enzymatically, or from other processes such as bolus oxidant exposure, exposure to radiation, etc) with subsequent diffusion of the oxidant to effector molecules.These processes may also be enhanced (panel F) by molecular crowding, enhanced solution viscosity, and diminished molecular diffusion.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) In the case of other redox signaling systems such as GAPDH and OxyR there is little (current) evidence for scaffold-induced facilitation of oxidation, complex formation with oxidant generators, or compartmentalization, though this may change with further research.In these cases the sensing of, or response to, H 2 O 2 may only occur with higher levels of oxidant, and hence be a response to stress, rather than a transcriptional regulator under normal physiological conditions.In the case of GADPH, the environment of the active state Cys (Cys152 in humans, Uniprot accession code: P04406) is such that this has a lower pK a value (i.e. the Cys is predominantly present in its thiolate, RS − form) and hence has a higher reactivity than other Cys residues.The rate of reaction at this residue may also be accelerated via the formation of peroxycarbonate (HCO  6E and F).This remains to be explored, as do the effects of crowding on these reactions.It is predicted that these may be significant due to decreased diffusion of both the oxidant and target protein.
The development of genetically encoded, and localized, H 2 O 2 generating system (e.g.D-amino acid oxidase/D-amino acid couples) should aid examination of these facets.Data from such systems localized to the nucleus, and other specific sites, suggests that the diffusion distance of H 2 O 2 inside cells is only 100-200 nm (cf.data in section 3) [99,100].
Nanodomains have also been proposed to be induced by, and to control, Ca 2+ signaling at the interface between the endoplasmic reticulum (ER) and mitochondria [101].Evidence has been presented that elevated Ca 2+ enhances oxidant formation and release from the mitochondria, at very specific and localized interaction sites between the organelles, with the increased oxidant formation subsequently enhancing additional Ca 2+ release from the ER, resulting in the formation of Ca 2+ oscillations.Nanodomain formation and oxidant generation, may therefore be critical to both redox signaling, and that induced by Ca 2+ [101].Both processes may be affected, at least to some extent, by the same factors, including reduced diffusion, increased steric barriers and interfacial effects induced by the crowded environment within cells.

Experimental strategies to overcome the complexity of studying the effect of crowding on protein oxidation and glycation in vitro
Multiple factors need to be considered when trying to mimic the complexity and heterogeneity of biological milieus.The tight packaging within cells involves multiple different types of molecules, and also species of markedly different size and characteristics.Moreover, the presence of heteroatoms and partial charges in these species, promotes specific and non-specific non-covalent interactions.Therefore, to successfully address the different features found inside cells and biological milieus, it is necessary to employ chemical approaches that are more complex than those that have typically been used to date, but which still permit control of key parameters.Without this, it would not be possible to have an appropriate understanding of the processes and mechanisms occurring during protein glycation, and protein oxidation and peroxidation.
In this context, inert crowding agents such as polyethylene glycol (PEG), ficoll and dextran have been employed to simulate cellular crowding effects in in vitro assays [59,60,102].The inherent advantage of using these polymers is that these have known chemical and physical properties [103,104].These non-adsorbing polymers generate two phenomena in solution that are present in biological environments: (1) exclusion-volume effect and (2) depletion interactions.The first one is a consequence of the mutual impermeability between the crowding agent and solutes, whereas the second corresponds to the induced attractive interactions between molecules in solution due to the presence of the crowding agent [18,105].Modified PEG, dextran and ficoll polymers with other functional moieties (e.g.charged groups) can be used to mimic reactions occurring near intracellular surfaces that are negatively-or positively-charged.However, not all polymers are suitable for spectroscopic studies and a high purity is a requisite for reproducibility.Moreover, it is likely to be important to examine different polymers as these generate different 'networks' in solution.Thus, whilst PEG and dextran are linear polymers, ficoll is a branched species resulting in different environments in solution.The latter may be a more relevant model in some cases (e.g. to mimic the complex 3-dimensional environment of the extracellular matrix).For example, and as discussed above, the effect of ficoll on the K m value of adenylate kinase (10-fold decrease) was greater than that detected with two different (linear) dextran polymers [57].We have recently reported that dextran and ficoll can be used to obtain kinetic data on protein oxidation and to investigate the modification pathways induced by oxidants of modest oxidation potential, and also for glycation agents [60,62].However, as the link between crowding and compartmentalization, and redox chemistry and biochemistry is a relatively new field, more experimental data is clearly needed to determine the best models to mimic biological environments.
A number of new technologies such as interferometric light scattering (also called mass photometry) have been successfully employed to shed light on the importance of molecular crowding in increasing the rate of generation of protein condensates (liquid droplets) of α-synuclein [106].This appears to be a key process in the formation of the amyloid fibrils of α-synuclein that are associated with the development of Parkinson's disease [107].Additionally, the use of nanoenvironmental sensor molecules (e.g.6-acetyl-2-dimethylaminonaphthalene) as utilized in in vivo experimental approaches, is likely to be an important tool to elucidate the role of crowding on different aspects of redox chemistry and redox biology [14].

Conclusions and future directions
The redox chemistry and biochemistry fields, like the redox signaling field, are gradually moving towards the consideration of crowding, subcellular domains and compartmentalization, rather than studying global bulk processes in order to have a better understanding of biochemical processes in living systems.Accumulating evidence supports the hypothesis that crowding, nanodomain formation, and altered local viscosities regulate the rates, extent, modification sites and products of protein modification induced by oxidants and glycation agents.Moreover, compartmentalization can modulate not only protein modification, but also signaling processes and cross-talk between oxidationand Ca 2+ -mediated signaling processes.As there are inherent complications in monitoring the pathways and kinetics of protein oxidation and glycation in intact systems, in vitro experiments continue to be a fundamental scientific tool, though it is clearly of major importance to verify in vitro findings in vivo.Such model systems need to be carefully designed in order to resemble the high concentration of macromolecules and formation of nanodomains that occur in biological environments.This is particularly potential relevance in understanding the processes that occur in aged cells, as increasing experimental evidence indicates that cell senescence is accompanied by a gradual loss of intracellular water, which would be expected to exacerbate crowding effects in the cell cytoplasm, and possibly exacerbate aberrant and damaging biochemical processes, resulting in a downward spiral [108].

Declaration of competing interest
E.F-L.declares no conflicts of interest.M.J.D. declares commercial consultancy contracts with Novo Nordisk A/S.This funder had no role in the conceptualization and writing of the manuscript, or in the decision to publish this.