Global structural motions from the strain of a single hydrogen bond

Abstract

The origin and biological role of dynamic motions of folded enzymes is not yet fully understood. In this study, we examine the molecular determinants for the dynamic motions within the β-barrel of superoxide dismutase 1 (SOD1), which previously were implicated in allosteric regulation of protein maturation and also pathological misfolding in the neurodegenerative disease amyotrophic lateral sclerosis. Relaxation-dispersion NMR, hydrogen/deuterium exchange, and crystallographic data show that the dynamic motions are induced by the buried H43 side chain, which connects the backbones of the Cu ligand H120 and T39 by a hydrogen-bond linkage through the hydrophobic core. The functional role of this highly conserved H120–H43–T39 linkage is to strain H120 into the correct geometry for Cu binding. Upon elimination of the strain by mutation H43F, the apo protein relaxes through hydrogen-bond swapping into a more stable structure and the dynamic motions freeze out completely. At the same time, the holo protein becomes energetically penalized because the twisting back of H120 into Cu-bound geometry leads to burial of an unmatched backbone carbonyl group. The question then is whether this coupling between metal binding and global structural motions in the SOD1 molecule is an adverse side effect of evolving viable Cu coordination or plays a key role in allosteric regulation of biological function, or both?

Dynamic motions of folded proteins in several cases are found to be essential for allosteric control of ligand binding, adaptive orientation of active-site moieties, and communication between distant sites in protein structures and complexes (1–4). Despite this fundamental role of dynamic motions in biological function, the molecular factors that control structural fluctuations and conformational heterogeneity within folded proteins remain largely unexplored. Also, it is not yet known if there is any relation, or principal difference, between the functional motions of proteins and the supposedly adverse structural fluctuations implicated in protein misfolding and aggregation (5, 6). An interesting system for shedding more light on these questions is the enzyme superoxide dismutase 1 (SOD1) associated with pathological misfolding and aggregation in the neurodegenerative disease amyotrophic lateral sclerosis (ALS). Native SOD1 is a symmetric homodimer of two Ig-like β-barrels, each of which coordinates one catalytic Cu^+/2+ ion and one structural Zn²⁺ ion. The redox active Cu^+/2+ ion is ligated directly to the side of the monomer β-barrel, whereas the Zn²⁺ ties together the long loops IV and VII to a densely packed dome over the active site. The role of this dome is twofold. It composes a selectivity filter for substrates to the Cu^+/2+ site (7) and also provides a critical part of the dimer interface (8). As a consequence, the dimerization of SOD1 becomes tightly controlled by metal binding via the structure of loops IV and VII, and the conserved C57–C146 disulfide linkage (9): if the metals and disulfide bond are lost, the protein dissociates into free monomers. Metal loss also unleashes characteristic dynamic motions of the central barrel scaffolds (10–13) that are most pronounced in the β-sheets facing the metal-binding sites. Together, these features form an intricate network of structural communication within the SOD1 molecule, spanning from the dynamic motions of the monomeric scaffold, via metal binding and the active-site loops, to the dimer interface. As a clue to the functional role of the communication, the binding of a single Zn²⁺ ion to one subunit of the dynamic apoSOD1 dimer is found to alter the structure and stability of the entire molecule (14). Based on the general view that globular proteins with compromised structural rigidity are at increased risk of misfolding (5, 6), the dynamic apoSOD1 molecule also has drawn attention as a possible precursor for pathological aggregation in ALS (12, 13, 15). In this study, we show by a combination of NMR, X-ray crystallography, and protein engineering that the origin of these dynamic motions is the built-in strain of a conserved hydrogen-bond linkage through the hydrophobic core. Upon truncation of this linkage by the mutation H43F, the SOD1 structure relaxes through H-bond swapping to a more ideal sheet architecture and the dynamics disappear. At a dynamic level, the results unravel the mechanism of an intricate allosteric crosstalk and show how a single side chain in the core can indirectly control a broad repertoire of functional traits. As a side effect of this structural crosstalk, however, the properties of the SOD1 molecule become globally sensitive to mutational perturbation, shedding light on its high sequence conservation and gain of neurotoxic function upon structurally distant mutations.

Results

Dynamic Motions of the apoSOD1 Barrel.

The dynamic motions of the apoSOD1 barrel are localized mainly to the β-strands facing the active site, and are indicated experimentally by diminished NMR NOE couplings, relaxation-dispersion measurements (11), and backbone hydrogen-exchange [hydrogen/deuterium (H/D)] kinetics (16). To enhance the structural resolution of these dynamic motions as much as possible, we focus here on a SOD1 variant with truncated loops IV and VII (apoSOD^ΔIVΔVII) (16). An advantage of this variant is that it eliminates interference from the active-site loops, which otherwise leads to NMR line broadening at low temperatures, at which the barrel motions are most pronounced. In most other respects, apoSOD^ΔIVΔVII behaves as the wild-type monomer (apoSOD1^pwt) (16). Measurements of ¹⁵N Carr–Purcell–Meiboom–Gill (CPMG) relaxation profiles show here that 34 of the 110 apoSOD^ΔIVΔVII residues are involved in dynamic exchange (Fig. 1 and Table S1). Global data analysis (SI Text) shows further that this exchange is well described by a two-state relaxation (k_ex = k₁ + k₋₁) between a highly populated ground state (GS) and a high free-energy state (HS) with occupancies (p_GS) and (p_HS), respectively.

Fig. 1.

Dynamic motions of apoSOD1^ΔIVΔVII measured by CPMG NMR. (A) The dynamic motions (red) are centered in the active-site sheet (β4 and β5) and the adjacent strand β6. (B) Representative data showing the CPMG amplitude decrease at high temperature. (C) Calculated chemical shift differences between the species undergoing dynamic exchange. (D) The population of the high-energy state decreases by protein destabilization by temperature, pH, and urea. (E) Sequence outline of apoSOD1^ΔIVΔVII showing the positions with dynamic exchange (red), the corresponding dynamics of apoSOD1^pwt (stars), and the GAG replacements of loops IV and VII.

The highest density of dynamic residues is found in β4 to β6 (Fig. 1), in good agreement with previous CPMG measurements of apoSOD1^pwt (11) and H/D exchange analysis (16). Besides some loss of signal at positions that previously interfaced the dynamic loop-IV structure, the main effect of loop removal overall is lower CPMG amplitudes. The reason for this slight amplitude loss seems to be decreased occupancy of the high-energy state, consistent with the slight gain in global protein stability of apoSOD^ΔIVΔVII (Table 1 and SI Text). As a sharp outlier in this trend, the CPMG amplitude of I99 in β6 displays a nearly eightfold increase upon loop removal, rendering this position the strongest signal in apoSOD^ΔIVΔVII. Judging by the unaffected H/D exchange rates of β6 and the neighboring β3, this signal increase likely is not the result of the increased dynamics of β6 itself (16). Rather it seems induced by altered fluctuations of β5 against which the I99 side chain packs (Fig. 1).

Table 1.

Kinetic and thermodynamic parameters determined from CPMG NMR relaxation-dispersion experiments

Parameter	apoSOD1pwt	SOD1ΔIVΔVII	SOD1ΔIVΔVII in 1.5 M urea
(%)*		1.80 (0.07)	0.63 (0.15)
(%)*		0.70 (0.03)	0.14 (0.02)
(%)*	0.80 (0.10)	0.34 (0.02)	0.06 (0.02)
(%)*		0.16 (0.02)	0.03 (0.01)
(kHz)*		5.85 (0.4)	2.80 (0.23)
(kHz)*		3.68 (0.3)	2.39 (0.34)
(kHz)*	2.59 (0.2)	2.95 (0.4)	3.75 (0.6)
(kHz)*		2.00 (0.6)	3.18 (1.0)
ΔGGS-HS (kcal/mol)†	2.79 (1.15)‡	3.72 (2.69)	4.94 (6.25)
ΔHGS-HS (kcal/mol)†	−13.6 (1.2)‡	−19.23 (1.89)	−23.76 (4.33)
ΔSGS-HS (cal/molK)†	−55 (2)‡	−80 (10)	−96 (20)

Experimental and fitting errors are shown in parentheses.

*Determined from fitting Eq. S2 to CPMG relaxation-dispersion data.

^†Determined from the Van’t Hoff equation, Eq. S4.

^‡Data from Teilum et al. (11).

Dynamic Exchange with a High-Energy State That Is More Compact than the Ground State.

To shed further light on the molecular nature of the GS-to-HS transition (Scheme 1), we examined how the relaxation-dispersion behavior of apoSOD^ΔIVΔVII responds to changes in temperature, pH, and [urea]. The results show that the population of the high-energy state decreases with increasing temperature, from p_HS = 1.8 ± 0.1% at 278 K to p_HS = 0.16 ± 0.02% at 298K (Fig. 1 and Table 1), consistent with previous data for apoSOD1^pwt (11). Correspondingly, the exchange rate constant decreases from k_ex = 5,800 s⁻¹ to k_ex = 2,000 s⁻¹ (Table 1). Such reversed temperature dependence of transition from ground state to high-energy state is not expected for a local unfolding event unless the system is in the cold-unfolding regime (17). Rather, elevated temperature is predicted to favor disorder. As a complement to the thermal scans, we monitored how the transition from ground state to high-energy state responds to lowered pH. The CPMG profiles obtained at pH 7.3, 6.3, and 5.3 yield p_HS values of 3.20 ± 0.08%, 1.8 ± 0.1%, and 0.98 ± 0.08%, respectively (Fig. 1, Table 1, SI Text, and Fig. S1A). Accordingly, the occupancy of high-energy state displays a consistent decrease with protein stability, and not the other way around, as expected for high-energy states that are partly ruptured intermediates. As a final test of this peculiarity we destabilized apoSOD^ΔIVΔVII with urea, which is the general method for quantifying changes in solvent-accessible surface area in protein-folding equilibria (18). The results show that the occupancy of the high-energy state decreases from p_HS = 1.8 ± 0.1% to p_HS = 0.6 ± 0.2% upon addition of 1.5 M urea (Table 1, SI Text, and Fig. S1B), that is, the transition between ground state and high-energy state involves burial of solvent-accessible surface area. From the value of , it is further evident that the extent of burial is sizable and corresponds to 29% of that in the global folding transition (SI Text and Fig. S1C). Because the CPMG data show no response to protein concentration (SI Text and Fig. S1D), we conclude that the burial is not an effect of intermolecular association but stems from consolidation of the monomer itself. To allow for such consolidation, the solution state of apoSOD1 needs to be expanded more than indicated from X-ray data, consistent with earlier findings by Banci et al. (13). It also follows that the transition from ground state to high-energy state is not a local unfolding event but, quite the opposite, a folding transition into a more compact state. An intriguing possibility would be that this compact state actually is analogous to the X-ray structure of the apoSOD1 molecule, selected by crystallization because of its high structural order.

Truncation of Conserved Hydrogen-Bond Linkage Eliminates β-Sheet Dynamics.

Characteristic of Ig-like structures is the conserved “tyrosine corner motif,” comprising a buried tyrosyl moiety that hydrogen bonds to a backbone carbonyl group at position n – 4 or 5 (19). In SOD1, this feature is replaced with a conserved “histidine corner motif”: the buried imidazole of H43 hydrogen bonds to the backbone of T39 in loop III (Fig. 2). Of particular interest is that H43 also binds to the backbone of the Cu^+/2+ ligand H120, thereby forming the hub for a buried hydrogen-bond chain that extends through the SOD1 core (Fig. 2). Via the backbone of L38, the chain connects further to the backbone of G93 in loop V, a hotspot for ALS-provoking mutations (15). To examine the impact of this conserved linkage on the observed barrel dynamics, we replaced H43 with a phenylalanine. The apo mutant maintains a well-dispersed ¹H-¹⁵N-HSQC (heternonuclear single quantum coherence) spectrum characteristic of a structurally ordered protein (Fig. 2, SI Text, and Fig. S2A). Even so, the chemical shift differences induced by the mutation are significant and reveal long-range effects (Fig. 2 and Table S2), which matches well with the regions found to undergo dynamic exchange in apoSOD1^pwt and apoSOD^ΔIVΔVII (SI Text). The H43F mutation also seems to increase the protein’s content of β structure as measured by secondary structure propensity (SSP) analysis, particularly in the regions of β2, β4−β6, and β7 (Fig. S2A and SI Text). This indicates that the mutation H43F imposes order on the SOD1 structure. Consistently, the mutation H43F also abolishes the relaxation-dispersion dynamics of the SOD1 barrel: the CPMG decays flatten completely, save for a single residual signal from I99 in β6 at low temperatures (Fig. 2 and SI Text). The same loss of barrel dynamics follows mutation H43F in apoSOD1^pwt (SI Text and Fig. S3). On this basis, we conclude that the conserved hydrogen-bond linkage mediated by the side chain of H43 induces the dynamic motions of the SOD1 barrel.

Fig. 2.

Mutation H43F freezes out the dynamics. (A) Conserved H120–H43–T39 linkage. (B) The mutant H43F has an altered but highly dispersed HSQC spectrum. (C) CPMG profiles showing diminished dynamics for apo.

Mutation H43F Leads to Consolidation of the Active-Site Sheet.

The impact of the H43F mutation on the barrel breathing motions was examined by H/D exchange analysis, which measures the time it takes for each individual amide proton to exchange with solvent deuterons through local or global unfolding. In essence, apoSOD1^pwt and apoSOD^ΔIVΔVII display three levels of H/D exchange rates (16): (i) fast exchange, in which the amide protons are replaced within the experimental dead time (∼5 min); (ii) intermediate exchange in the EX2 regime, in which the exchange rate constant is proportional to the population of locally “open” species; and (iii) slow exchange in the EX1 regime, in which follows the rate constant for global unfolding. For apoSOD^ΔIVΔVII, the fast and intermediate exchanges are localized mainly to the dynamic sheet below the active site, whereas the back sheet exchanges slowly through global unfolding, save for the first few residues of β2 (SI Text and Table S3). The pattern, which fits well with CPMG data (Fig. 1), indicates that the motions of the active-site sheet are complex and energetically disperse, describing openings of less than 2.2 kcal/mol for dead-time exchange and excitations of >3.5 kcal/mol for intermediate exchange (SI Text), as well as transitions into the compact HS (Scheme 1). As observed with CPMG analysis, the dynamic motions undergo radical changes upon truncation of the H120–H43–T39 linkage. The H/D exchange rates of the active-site sheet show a marked decrease where, most notably, the amide of G44 in β4 changes from dead-time exchange in apoSOD^ΔIVΔVII to full protection in apo (Table S3). At the same time, H43F increases the exchange rates in β2 and β3 at the opposite side of the protein, and in the beta cap connecting the two sheets (Fig. 3). Accordingly, the ordering of the active-site sheet occurs at the expense of the structural rigidity of β2, β3, and the beta cap: one frustration seems to replace another.

Fig. 3.

Mutation H43F decreases the H/D exchange rates () in the active-site sheet β4, β5, and β7 and increases in the opposing part of the structure.

X-Ray Crystallography.

To pinpoint the structural rearrangement underlying the diminished dynamics upon mutation H43F, we solved the X-ray structures of apoSOD1^ΔIVΔVII, apo, and holo to resolutions of 1.9 Å, 2.8 Å, and 1.3 Å, respectively (SI Text and Table S4). The most striking feature of the apo structure is that the twisted backbone carbonyl group of H120 (β7), after being released from H43, relaxes into the plane of the sheet and H-bonds to the amide of G44 (β4) (Fig. 4). Following this relaxation, there is a tightening of the H-bond network between β7 and β4, where the average H-bond length decreases from 2.9 ± 0.13 Å to 2.8 ± 0.21 Å. The tightening also extends to β5 and is accompanied by a new surface interaction between the side chains of H120 and D53. The loss of dynamic motions upon mutation H43F thus seem to be the result of ordering of the active-site sheet, fully consistent with the HSQC and solution-state SSP analysis in Fig. S2A. Because of this extensive structural rearrangement, apoSOD1^ΔIVΔVII and apo cannot strictly be seen as the same state. Equally revealing, we find in holo that metal coordination to the H120 side chain wrenches back its backbone carbonyl group into the core (Fig. 4). This shows that the strained geometry of H120 in the wild-type protein is indeed required for metal binding and even establishes, at the cost of desolvating an unmatched polar moiety in the H43F mutant, that there seem to be no alternative metal-binding geometries available.

Fig. 4.

Removal of the conserved H120–H43–T39 linkage by mutation H43F. (A) X-ray structures of apoSOD1^ΔIVΔVII (PDB entry 4BCZ) and apo (PDB entry 4BD4) show that H43F leads to relaxation of the twisted backbone carbonyl of H120 into the sheet where it H-bonds to the amide of G44. (B) Cu²⁺ coordination twists back the carbonyl of H120 into the core, forcing the protein into a native-like metal-binding geometry. Holo (PDB entry 4BCZ).

Effects on Folding, Metallation, and Enzymatic Activity.

The global folding reaction of the apoSOD1^pwt and apoSOD1^ΔIVΔVII monomers is described to a good approximation by a cooperative two-state transition (9, 16):

where U is the globally unfolded species, GS is the folded ground state in Scheme 1, and k_f and k_u are the folding and unfolding rate constants, respectively. The resulting chevron plots of logk_f and logk_u vs. [GdmCl] are shown in Fig. 5, and the kinetic parameters derived from these plots are listed in Table 1. The curvatures of the unfolding limbs are linked to changes of rate-limiting step at high [GdmCl] and are discussed in SI Text. From the very small effects of the H43F mutation on the global stabilities of apoSOD1^pwt and apoSOD1^ΔIVΔVII, it is apparent that the penalty of removing the H120–H43–T39 linkage is compensated for by new favorable interactions in the folded ground state, in full accordance with the structural rearrangement of the apo structure in Fig. 4. In contrast, the H43F mutation yields a pronounced destabilization of the fully metallated species holo, reflected by a selective increase of k_u (Fig. 5 and Table 2). Again, the result is in full accordance with the X-ray data, in which the maintained geometry of the metal-bound species leaves the wrenched H120 backbone amide unsatisfied upon H43 removal (Fig. 4). Removal of the H120–H43 bond also is expected to decrease metal affinity. To examine this possibility, we measured the effect of EDTA on the unfolding kinetics of the holo proteins, which specifically targets the metal-off rates (20) (Fig. 5, Folding Analysis, and Table S5). Consistent with previous studies (20), holoSOD1^pwt displays elevated k_u values in the presence of EDTA, exposing the apparent rate constant of metal loss, k_off (Fig. 5). This rate constant of the fully metallated protein most likely describes the simultaneous loss of both metal ions, because metal loss from the Zn site in the absence of a loaded Cu site is considerably faster (20), i.e., k_off is strongly controlled by the Cu site. Upon H43F mutation, however, k_u of holo converges with that of the apo protein. The mutation thus seems to increase k_off to a point at which metal dissociation no longer limits the unfolding kinetics. Taken together, this shows that the H120–H43–T39 linkage not only induces structural malleability of the apo protein, but also maintains high metal affinity. The obvious question, then, is how does this conserved feature affect enzyme activity? To find out, we compared the activity of holoSOD1^pwt and holo at two different pHs: at pH 9.5 with the direct KO₂ assay and at pH 7.4 with a modified pyrogallol assay (21) (SI Text). The results show that the impact of the H43F mutation on the enzymatic turnover is modest: the activity per Cu^+/2+-loaded monomer goes down by a factor of ∼5 (SI Text). As a control, we also determined the effect of H43F mutation on the activity of wild-type holo dimer and yielded a correspondingly modest reduction factor of ∼3 (SI Text). Notably, this reduction in activity upon H43F mutation is smaller than that observed for splitting the dimer interface (SI Text).

Fig. 5.

Folding chevron plots (logk_f and logk_u) and the apparent metal-off rate constants (logk_off) for the proteins analyzed in this study. Parameters from fits are listed in Table 2. logk_u in the presence of EDTA, corresponding to logk_off, is shown as red ● and red ▲ for and SOD1^pwt, respectively. logk_off at 0 M GdmCl was determined from fluorescence of the Zn²⁺-specific chelator Zinpyr.

Table 2.

Kinetic parameters determined from folding kinetics

Parameter	apoSOD1pwt	holoSOD1pwt			apoSOD1ΔIVΔVII
log*	−1.75 (0.20)	−0.25 (0.05)	−1.42 (0.09)	−0.20 (0.05)	−1.06 (0.04)	−0.74 (0.07)
mf (M−1)*	−2.85 (0.78)	−2.21 (0.04)	−3.49 (0.39)	−1.98 (0.08)	−2.29 (0.05)	−2.21 (0.10)
log*	−3.36 (0.09)	−8.29 (0.22)	−3.29 (0.04)	−5.45 (0.14)	−4.35 (0.03)	−4.77 (0.10)
mu (M−1)*	1.16 (0.06)	1.64 (0.08)	1.14 (0.02)	1.06 (0.05)	0.77 (0.01)	0.85 (0.03)
log Kpart*	−3.14 (0.15)	−3.87 (0.24)	−3.56 (0.10)	−3.64 (0.29)	−4.09 (0.23)	−3.93 (0.91)
m’’ (M−1)*	1.01 (0.05)	0.96 (0.07)	0.93 (0.02)	0.76 (0.04)	0.74 (0.03)	0.66 (0.11)
ΔGGS-U (kcal/mol)†	2.19 (0.39)	10.93 (0.37)	2.54 (0.18)	7.14 (0.26)	4.47 (0.10)	5.48 (0.23)
MP (M)‡	0.40 (0.10)	2.09 (0.07)	0.40 (0.04)	1.73 (0.01)	1.07 (0.02)	1.32 (0.02)

Experimental and fitting errors are shown in parenthesis. Rate constants are in seconds⁻¹.

*Derived from fitting data in Fig. 5 to a two-state transition model (Eq. S6).

^†ΔG_GS-F = −2.3RT(log− log).

^‡The transition midpoint calculated as the intersection log k_u and log k_f, where [U] = [F].

Discussion

The integrity and structural stability of natural proteins rely generally on funnel-like folding energy landscapes leading to unique final states (22). To comply with biological function, however, this basic funnel sometimes must be perturbed to allow several competing low-energy states and, thus, local motions of the native structure (3, 23–26). In this study, we demonstrate that the molecular basis for such functional complexity may be surprisingly simple: the global dynamic motions of apoSOD1 are induced by the strain of a single conserved hydrogen-bond linkage and may be switched on and off by a single point mutation (Fig. 6).

Fig. 6.

Switching off the dynamic motions by mutation H43F. (A) Dynamic positions of apoSOD1^ΔIVΔVII (red) with intact H120–H43–T39 linkage. (B) H-bond relaxation and loss of dynamics in the mutant apo. (C) Metallation of apo drives the protein back to wild-type–like distortion of the β sheet.

Dynamic Motions from Evolutionary Engineered Frustration.

Mechanistically, the H120–H43–T39 linkage across the SOD1 core seems to promote dynamic motions by frustration (24), i.e., the protein is strained into a geometry in which it cracks up and starts to rattle between multiple low-energy minima. Judging from CPMG data (Fig. 1), the epicenter of this cracking is in the active-site sheet (Fig. 6). From the lack of exchange at H120 and G41, it also is evident that the straining bond between H120 and H43 does not transiently break during the dynamic motions but remains intact (Fig. 6). The frustration nevertheless is relieved upon H43F mutation, which relaxes the protein into a fixed ground state with continuous hydrogen bonding between β4 and β7 (Fig. 6). Even so, the influence of the H43F mutation is not limited to the folded ground state but also is seen by H/D exchange at higher energy levels, at which the diminished breathing in the β4–β5 region is accompanied by loosening of the beta cap and strands β2 and β3. This shows that the structural interplay indeed is long range and extends across the entire SOD1 monomer (Fig. 3). Despite the ordered appearance of the crystal structures, the apoSOD1 scaffold thus is equipped with a rich repertoire of correlated movements, consistent with earlier predictions from simulations (27, 28) and NMR studies (13). The apoSOD1 barrel also controls the motions of the functional loops: the CPMG profiles of the barrel are correlated precisely with those of loops IV and VII, and mutation H43F freezes the motions in both regions (SI Text); vice versa, metallation of the protein’s Zn²⁺ site freezes the motions of the barrel via consolidation of loops IV and VII (11). Taken together, this points at a two-way coupling between the barrel and loop motions, with interesting implications for the function of the SOD1 dynamics.

Functional Role of the SOD1 Dynamics.

The coupled structural and dynamic effects of the H43F mutations raise the question, do all these features relate to biological function, or are some merely side effects of conflicting evolutionary design? From X-ray data, it is clear that the H120–H43–T39 linkage is critical for distorting the Cu^+/2+ ligand H120 into native coordination geometry, thus priming the apo structure for metal binding (Fig. 4). Consistently, the H43F mutant selectively destabilizes the holo state and shows radically increased metal-off rates (Fig. 5): elimination of the apoSOD1 distortion weakens the protein’s ability to hang on to its metals. It then is possible that the plasticity arising from the apo-state distortion presents an additional benefit by allowing adaptive orientation of metal-binding ligands. During catalytic turnover, the Cu^+/2+ ion is redox active, which might induce cyclic structural changes between different coordination geometries. Such changes conceivably may lead to adverse stability differences between oxidized and reduced species (29) unless accommodated for by the protein structure. A related system might be the electron shuttle azurin, in which CPMG analysis identifies local structural motions around the coordinated Cu⁺ ion (30). The suggested role of this azurin flexibility, or lack of rigid protein matrix, is to facilitate electron transfer (30). From this perspective, and from the growing evidence that active-site dynamics are critical for enzyme catalysis (31, 32), it is notable that the H43F mutation has only a modest impact on SOD1 activity. One explanation might be that the activity of SOD1 is near diffusion control (7) and not limited by dynamics at the observed timescales. The task of the protein matrix then would be to select the substrate electrostatically and to protect the Cu^+/2+ ion from reacting with other molecules (7). Instead, the most apparent role of the SOD1 dynamics is in structural crosstalk: the coupling between barrel and loop motions explains mechanistically how fractional metallation of one monomer influences not only the binding of subsequent metal ions, but also the structure and stability of the entire homodimer (14, 27, 33, 34). Where there are tuneable dynamics there also is allostery (4), even if it happens to pass undetected in conformationally biased X-ray structures (13). As an indication that the SOD1 crosstalk indeed is functional, it comprises 37% of the protein’s residues, complying well with the benchmark for other allosteric proteins (26). The precise biological function of this structural interplay, however, is not yet clear. One possibility is that it regulates molecular assembly, perhaps by contributing to more concerted enzymatic activation in the maturation pathway (35).

Aggregation and Disease.

Finally, we ask whether the partly ruptured apoSOD monomer is also the mysterious precursor for pathologic aggregation in ALS. After all, its structural plasticity need not be solely a functional advantage but also might carry aberrant side effects arising from conflicting evolutionary optimization (36). Of particular interest here is the high-energy state HS (Scheme 1), as seemingly analogous species are implicated as precursors in β2-microglobulin and lysozyme amyloidosis (5, 6). In contrast to the high-energy intermediates of β2-microglobulin and lysozyme, however, we observe here that the high-energy state HS is not on the pathway to the unfolded state, but is more compact than the solution ground state (Table 1). Together with the finding that destabilization of the apo monomer with urea accelerates the fibrillation kinetics in vitro (37), this puts precise constraints on the aggregation mechanism: the fibrillation precursor needs to be more expanded than the folded ground state, which excludes the involvement of the high-energy state HS. Instead, the precursor for apoSOD fibrillation appears to be the globally denatured state (37), consistent with an observed maximum of the fibrillation kinetics above 5 M urea, at which the occupancy of D is nearly 1 and the occupancy of the high-energy state has decreased from 0.01 to less than 10^−4.7 (Table 1, SI Text, and Fig. S4). On this basis, we conclude that the dynamic motions of the apoSOD1 barrel do not compromise the protein’s solubility. The most imminent explanation as to why the SOD1 motions in this respect are different from other amyloidogenic proteins is that they are functionally evolved. A predicted consequence of the SOD1 crosstalk, nevertheless, is that the protein becomes globally sensitive to mutations. Peripheral mutations need not only perturb locally but also may modulate the structure allosterically over long distances. An example of such long-range modulation is the observation that the common ALS-associated SOD1 mutation G93A affects the metal-binding ligands more than 18 Å away (38). For an enzyme whose stability, maturation, and cellular lifetimes are intimately coupled to the metal-binding equilibria (20), such global effects might be critical, shedding light not only on why the ALS-provoking mutations are so broadly distributed in the SOD1 structure, but also on why the SOD1 sequence is so highly conserved across divergent species (39). To this end, our data show that the evolutionary pathway to such complex functional dynamics may be very short: in this case, the change of a ubiquitous tyrosine corner into a cross-core histidine link is just one mutation away.

Materials

Protein Preparation.

Mutagenesis, expression, and purification were performed as in refs. 9 and 16, and experiments were in Bis-Tris HCl buffer at pH 6.3 unless otherwise stated.

NMR.

CPMG relaxation rates were measured on a Bruker Avance 700-MHz spectrometer. Assignment of apo was by a standard set of experiments (SI Text) on a Bruker 900-MHz spectrometer. Relaxation rates R₁, R₂, and NOE were determined as described in ref. 16 on a Bruker 600-MHz spectrometer. H/D exchange analysis was by 1:1 sample dilution in D₂O, and detection was on a Bruker 700-MHz spectrometer, as described in ref. 16. Detection dead time was 3 min, and HSQC spectra were recorded every 10 min.

X-ray Crystallography.

Crystallization was by hanging drop as described in (SI Text) and data were collected at 100 K at the MAX Laboratory synchrotron in Lund, Sweden. Data collection statistics are listed in (SI Text). All structures were solved by molecular replacement using Phaser with monomeric 1MFM as a starting model. Final refinement, validation, and geometrical statistics are described in (SI Text). Structures are deposited in www.pdb.org with accession numbers 4BCY for holo, 4BCZ for apoSOD1^ΔIVΔVII, and 4BD4 for apo.

Folding Kinetics.

Protein concentration was 4 μM, and mixing was by a PiStar-180 stopped-flow apparatus (Applied Photophysics) or done manually on a Varian Cary Eclipse, with excitation at 280 nm and emission above 305 nm or at 360 nm. Data were analyzed with Kaleidagraph (Synergy Software) (SI Text).

Metal-Off Rates.

Metal-off rates (k_off) were determined by GdmCl-induced unfolding in 2 mM EDTA. At 0 M GdmCl, k_off was determined by mixing holo protein with the Zn²⁺-specific fluorescent chelator Zinpyr (Sigma-Aldrich).