β-strand interactions at the domain interface critical for the stability of human lens γD-crystallin

Abstract

Human age-onset cataracts are believed to be caused by the aggregation of partially unfolded or covalently damaged lens crystallin proteins; however, the exact molecular mechanism remains largely unknown. We have used microseconds of molecular dynamics simulations with explicit solvent to investigate the unfolding process of human lens γD-crystallin protein and its isolated domains. A partially unfolded folding intermediate of γD-crystallin is detected in simulations with its C-terminal domain (C-td) folded and N-terminal domain (N-td) unstructured, in excellent agreement with biochemical experiments. Our simulations strongly indicate that the stability and the folding mechanism of the N-td are regulated by the interdomain interactions, consistent with experimental observations. A hydrophobic folding core was identified within the C-td that is comprised of a and b strands from the Greek key motif 4, the one near the domain interface. Detailed analyses reveal a surprising non-native surface salt-bridge between Glu135 and Arg142 located at the end of the ab folded hairpin turn playing a critical role in stabilizing the folding core. On the other hand, an in silico single E135A substitution that disrupts this non-native Glu135-Arg142 salt-bridge causes significant destabilization to the folding core of the isolated C-td, which, in turn, induces unfolding of the N-td interface. These findings indicate that certain highly conserved charged residues, that is, Glu135 and Arg142, of γD-crystallin are crucial for stabilizing its hydrophobic domain interface in native conformation, and disruption of charges on the γD-crystallin surface might lead to unfolding and subsequent aggregation.

Introduction

Human age-onset cataract is an exceptionally common protein deposition disease; it is the major cause of blindness affecting 25 million people worldwide.^1,2 The development of novel treatments for cataract is currently hindered by our incomplete understanding of the process(es) of cataract formation in lens cells. Cataracts removed from the aged human eye lens are composed of aggregated states of partially unfolded and/or covalently damaged crystallins.³ These proteins are synthesized in utero, but must remain soluble throughout our lifetime to maintain the transparency of the lens despite the continual UV exposure and potential oxidative stress.^4,5 Crystallins are present in very high protein concentration (200–400 mg/mL) in the lens, with β− and γ-crystallins constituting over 50% of total lens proteins. The monomeric γ-crystallins and oligomeric β-crystallins that are comprised of homologous double Greek key domains are mainly structural proteins. All known γ-crystallins from extant vertebrate species have duplicated Greek key domains that are presumably evolved by gene duplication and gene fusion from an ancestral single domain crystallin.^6,7 In fact, such single domain crystallins have been identified in the Sea squirt Ciona,⁸ descendants of lineages that are candidates for the origin of the vertebrates.

The increased stability conferred by the interfacial interactions between the two domains is likely very important for maintaining the long life time of proteins of the lens nucleus, and thus explains the selection for duplicated forms. The crucial role of the domain interface interactions in the folding, stability and oligomerization of both β- and γ-crystallins is also evident from recent experimental results.^9–14 In the crowded lens nucleus, those interactions at the domain–domain surface are likely crucial in preventing incorrect protein–protein association leading to cataract formation.

Human γD-crystallin (γD-crys), one of the most abundant γ-crystallins in the lens,¹⁵ is comprised of 173 amino acids. As shown in Figure 1(a), the single domains of γD-crystallin, each consisting of two intercalated antiparallel β-sheet Greek key motifs, termed N-td₁, N-td₂, C-td₃, and C-td₄, are connected by a six residue linker peptide.⁵ Each of the Greek key motif consists of four β-strands, namely a, b, c, and d. The ab folded hairpin turn between the first two strands, a and b, is a characteristic feature of the Greek key motif of βγ-crystallins. In γD-crys molecule, the domain interface is composed of (1) a cluster of six hydrophobic residues and (2) two pairs of polar peripheral residues flanking the hydrophobic cluster. The hydrophobic patch is comprised of three amino acids from each domain, which are Met43, Phe56, and Ile81 from the N-terminal domain (N-td) and Val132, Leu145, and Val170 from the C-terminal domain (C-td). Peripheral polar interactions are between Gln54/Gln143 and Arg79/Met147.

Figure 1.

(a) Cartoon representation of γD-crys monomer is shown by specifying the four Greek key motifs (N-td₁, N-td₂, C-td₃, and C-td₄) in different colors. The heavy sidechain of the residues at the interdomain surface are shown in ball-stick representation. The predicted substitution site E135 is also shown that forms a salt-bridge interaction with R142 in the wild-type protein during unfolding simulations. Yellow color is used for nonpolar residues, while polar residues are shown in orange. Acidic residues are colored in red and basic residues are colored in blue. (b) Time evolution of the fraction of native contacts, Q, for the single domains in isolation at two different temperatures, 380 K (dashed) and 425 K (solid). (c) Time evolution of the fraction of native contacts, Q, for the single domains within the full monomer at two different temperatures, 380 K (dashed) and 425 K (solid). (d) Time evolution of the fraction of native contacts, Q, of the N-td (solid line) in isolation and in the full monomer and the fraction of native contacts, Q, at the domain interface (dashed line).

A number of experimental investigations have provided valuable information on the importance of the interdomain interactions governing the complex folding/unfolding of γD-crys.^9–14 The individual domains of γD-crys exhibit differential stability with the C-td being more stable than the N-td.^16–18 Equilibrium unfolding/refolding experiments of γD-crys at near-physiologic conditions (at neutral pH and 37°C using Guanidinium Chloride [GdmCl] as denaturant) has suggested sequential folding of its domains with the C-td refolding first.¹⁶ Furthermore, these experiments suggest the existence of a partially folded intermediate state with its C-td largely folded and N-td fully or partially unstructured.¹² This folding intermediate undergoes aggregation below 1.0 M GdmCl that competes with productive refolding.¹⁹ Site-specific mutagenesis experiments demonstrated that the domain interface residues nucleate the N-td folding.^12–14 Although both isolated domains could refold efficiently by themselves, the N-td exhibited lower stability than the C-td in isolation.²⁰ Stability comparisons of the isolated domains with the full monomer revealed that the domain interface contributes a ΔG_H2O of ∼4.2 kcal/mol to the stability of the full γD-crys monomer.²⁰ Nevertheless, the detailed effects of the interdomain interactions on the folding mechanism of the individual domains are not evident from the experiments.

β-sheets are found in more than 80% of the reported crystallographic structures, and misfolding and aggregation of β-sheets have been implicated in many neurodegenerative diseases such as Alzheimer's disease and mad cow disease. The difficulty in deciphering the sequence control of β-sheet folding reflects the importance of interactions between side chains that are distant in the primary amino acid sequence. Although, significant progress has been made in understanding the folding pathways of several β-sheets^21–23 and isolated SH3 domain^24–28 using experimental and theoretical/simulation techniques, detailed characterization of the folding/unfolding landscape of a biologically important β-sheet protein still remains a challenge.²⁹ For example, even for a small 16-residue β-hairpin (GB1), there are debates over a hydrogen-bond zipping mechanism versus a hydrophobic core collapse mechanism.^30–34 In a separate example, even the use of high resolution experimental techniques such as coupling of H/D exchange with 2D NMR only allows detection of intermediate states with almost native-like five-strand β-sheet conformation during the earliest stage of Ubiquitin folding.^35,36

Computer simulations of protein models at different resolutions, from simple lattice models (and off-lattice)^37–39 to continuum solvent models^38,40,41 to all-atom explicit solvent models,^42,43 have been used to supplement existing experimental techniques in understanding aspects of protein folding.^44,45 In principle, molecular dynamics (MD) simulations with an atomistic description of the protein and solvent molecules should provide the most realistic description of the protein folding landscapes. However, such folding simulations starting from fully extended states are often limited to small peptides,^42,43 as the sampling of the huge conformational space of a typical protein is currently inaccessible even with the fastest supercomputers. In this endeavor, unfolding simulations, performed in either chemical denaturing or thermal denaturing conditions, have been widely used to shed light onto the folding scenario. According to the principle of microscopic reversibility, the folding pathway should be reverse of the unfolding pathway under same conditions. In addition, previous studies have shown that unfolding simulations, similar to those performed in this study, can accurately capture crucial details of the folding landscapes of a number of proteins in quantitative agreement with experiments (see Ref. 46 and references therein).

In this study, we elucidate the unfolding mechanism for the isolated domains and the full γD-crys monomer by performing chemical denaturing simulations in 8 M urea solution. Further analyses of the atomistic unfolding simulations provide an in-depth understanding of the differential stability of γD-crys domains, which is consistent with previous experiments. Our simulated unfolding data clearly show that the presence of the interdomain interactions affects both the stability and the folding mechanism of the N-td. In addition, a hydrophobic folding core comprised of a and b β-strands from C-td₂ is identified at the domain interface of the C-td. A surprising surface salt-bridge at the end of the ab folded hairpin turn of C-td₂, which is absent in the native conformation of γD-crys, is found to play a crucial role in stabilizing the folding core of the C-td. We further show that the disruption of this non-native salt-bridge by a single alanine substitution (E135A) performed in silico dramatically destabilizes the folding nucleus of the C-td, which, in turn, trigger unfolding of the N-td interface. These results underscore the importance of such non-native interactions in folding of β-sheet domains of γ-crystallins that may offer clues to map pathways of crystallin aggregation leading to cataract formation.

System and Methods

The initial structure of the wild-type γD-crystallin protein [see Fig. 1(a)] has been taken from the crystal structure deposited in the Protein Data Bank (PDB ID code 1hk0). The isolated N-td comprised of 85 residues ends at Gly85, while the isolated C-td comprised of 88 residues begins at Ser87. The chemical denaturing simulations were performed in 8 M urea solution (for details of the preparation of immersed protein system within the 8 M urea solution, see Ref. 47). The final molecular system consisted of a single domain or the full monomer of γD-crys in 8 M urea and contained ∼7650 water molecules and ∼1775 urea molecules. This system consisted of a total of ∼40,000 atoms was minimized for 10,000 steps followed by a 1 ns MD equilibration at 310 K and 1 atm. The particle-mesh Ewald (PME) method⁴⁸ was used for the long-range electrostatic interactions, while the van der Waals interactions were treated with a cutoff distance of 12 Å. The CHARMM32 (c32b1 parameter set) force field⁴⁹ and TIP3P water model⁵⁰ was used to run MD. All simulations wee performed using NAMD2 molecular modeling package^51,52 with a 2 fs time step in NPT ensemble at 1atm.

At least five MD trajectories were run for each system starting from different initial configurations. For each system, simulations were run at two different temperatures, at 380 K and at 425 K. However, all results mentioned in this article are derived from simulations run at 425 K unless otherwise stated, as at 380 K we did not observe complete unfolding of either isolated domains or individual domains within the full monomer even after >0.5 μs MD simulations. This observation is consistent with the experimental data showing that γD-crys is resistant up to 8 M urea denaturant and an acidic pH is required for completely unfolding.¹⁹ However, for a variety of mutant γ-crystallins (such as human V75D γD-crys) that do denature in urea, a partially folded intermediate with the N-td disordered and the C-td folded was populated in urea-induced denaturation/renaturation experiments, similar to the species observed in GdmCl-induced Denaturation (Kate N. Drahos, personal communication). The extent to which the crucial details (such as formation of a non-native salt-bridge) of the unfolding pathway of γD-crys depend on the specific denaturation agent used is currently not clear in experiments. On the other hand, simulations of proteins in a GdmCl solution are particularly challenging due to the less well parameterized GdmCl molecule and the complexity associated with the large number of charges involved.

We therefore performed the denaturing simulations of γD-crys at an elevated temperature (425 K) in 8 M urea solution to study unfolding. The overall agreement between the current simulated results and our recent experimental data indicates that these simulated unfolding pathways of the γD-crys domains capture the essential features observed in the refolding/unfolding experiments, and therefore are useful in providing further insights into the kinetics of unfolding. It should be noted, that these simulations performed under denaturing conditions solely provide kinetic models for γD-crys unfolding and should not be over extended to measure equilibrium quantities.

Results and Discussions

Figures 1b and c show the time dependence of the fraction of native contacts, Q, for the unfolding of the individual domains in isolation and within the full protein, respectively. Here, a native contact between residues i and j is counted if any heavy atom of residue i is within 6.5 Å of any heavy atom of residue j in the crystal structure of the protein. Time evolution of Q clearly indicates that the isolated N-td unfolded ∼2 times faster than isolated C-td at 425 K, in agreement with experiments (16,20). The N-td denatured faster than the C-td in the full monomer as well, which leads to the formation of a partially unfolded intermediate state [RMSD_Cα (full protein) ∼10 Å, Q (full protein) ∼0.4] during the full monomer unfolding [Fig. 1(c)]. As revealed from Figure 1(c), this intermediate state has its C-td mostly structured (Q_C-td >0.5), but the N-td fully or partially unfolded (Q_N-td <0.1), in agreement with recent time-resolved fluorescence experiments.^16–18 The stability of the C-td (both in isolation and in full protein) over the N-td is also manifested by the unfolding simulations at 380 K. Interestingly, a comparison of the time evolution of Q_N-td in isolation with the same in the full monomer shows that the unfolding of the N-td in full monomer started much later (>100 ns) compared to that of isolated N-td [Fig. 1(d)] at 425 K, suggesting the contribution of the C-td in stabilizing the N-td. The contribution of interdomain interactions becomes clearer, as the domain interface native contacts start to weaken before those within the N-td [Fig. 1(d)]. Therefore, our simulations reveal that the interdomain interactions contribute to the stability of the N-td of γD-crys, in line with the experimental observation that the mutations weakening interdomain interactions lower both the stability and the folding rate of the N-td, while the C-td remains unaffected.^12–14

To obtain a more detailed view of the conformational changes during the unfolding processes, we monitored the secondary structure change as a function of time for the isolated domains, as well as of the full monomer (Supporting Information Fig. S1). We find that N-td₂, the one closer to the domain interface, of the isolated N-td unfolded first [Fig. S1(a)]. This might be related to the fact that there is no topological equivalent of Arg142 (from motif 4) in motif 2, but there is one in motif 1 (see more discussion below). Surprisingly, this motif remained more structured during the N-td unfolding in the presence of the C-td interface [Fig. S1(b)]. This finding strongly implies that the absence of interdomain surface interactions catalyzes unfolding of N-td₂ in the isolated N-td, indicating that the domain interface interactions are most likely needed to stabilize this motif in its native conformation. Taken together, these results clearly demonstrate the importance of interdomain native interactions in determining the stability of the N-td.

During the isolated C-td unfolding, Greek motif 3 C-td₃ disrupted much faster than motif 4 C-td₄ [Fig. S1(c)]. Within C-td₄, strands a and b remain fairly stable until the very end of the simulation despite the fact that RMSD_Cα grows up to >14–16 Å and the fraction of native contacts decreases to less than 40%. This observation indicates that these two β-strands from C-td₄ form a stable folding nucleus in the isolated C-td, which functions as a folding scaffold for the N-td. Our identification of this folding nucleus within the C-td confirms the mutational data,^12–14 suggesting population of two major intermediates during γD-crys refolding; the first one most likely has C-td₄ folded, while rest of the protein remains unstructured. The second intermediate has the C-td fully folded and the N-td mainly unfolded. In addition, these results obtained from all-atom simulations explain the three-state mechanism observed during earlier equilibrium/kinetic folding/unfolding experiment of the isolated C-td.²⁰

To investigate the physicochemical factors governing the surprisingly high stability of strands a and b of C-td₄ observed during our chemical denaturing simulations, we calculated the probabilities of native contact formation for the unfolded conformational ensemble (RMSD_Cα >10 Å and Q <0.3) populated during isolated C-td unfolding [Fig. 2(a)]. A cluster of native contacts formed between strands a and b from C-td₄ that encompass residues 131–135 and residues 141–145 remained unperturbed in the unfolded ensemble, as seen in Figure 2(a). An additional cluster of native contacts was also found in the unfolded ensemble of isolated C-td between strands a and d of C-td₄, however, this cluster was much weaker than the one found between a and b strands. Figure 2(b) provides a more precise and dynamic view of unfolding to this point, which shows the time evolution of the probabilities of long-range (|i-j| >6) native contact formation per residue, Q_res, for C-td₄. This calculation allows us to identify residues that maintained more than 30% of the long range native interactions until the very end of the unfolding simulations. Clearly, these residues are 132–135 and 142–145. Taken together, these results suggest the existence of a folding core encompassing strands a and b of C-td₄ that is comprised of mainly hydrophobic residues at the C-td domain interface (V132 and L133 from the a strand, and Y144 and L145 from the b strand).

Figure 2.

(a) Probabilities of native contact formation for the unfolded conformational ensemble (RMSD_Cα >10 Å and Q <0.3) of the isolated C-td. Darker shades of blue indicate stronger native contacts. Secondary elements are shown for each residue, β-sheets in green and helices in red. (b) The probabilities of native contact formation per residue, Q_res, of C-td₄ as a function of simulation time during isolated C-td unfolding. The quantity, Q_res is colored according to the colorbar.

Visual inspection and further analyses (see later) of the conformational behavior of the folding core of C-td during unfolding strikingly revealed a salt-bridge interaction between residues Glu135 and Arg142 at the γD-crys surface formed that is maintained until the protein completely unfolds. Surprisingly, this salt-bridge interaction, in which the Oɛ2 atom of Glu135 stays at a distance ∼3.2Å from the Nη2 atom of Arg142 during entire unfolding simulation, is not seen in the wild-type crystal structure. In fact, the sidechain carboxylate oxygens of Glu135 are ∼7 Å away from the sidechain nitrogens of Arg142 in the crystal structure. These results strongly imply that, this surprising non-native salt-bridge interaction near the end of the ab folded hairpin turn of C-td₄ protects the two β-strands of the Greek key motif 4 from unfolding, thus promoting the formation of a hydrophobic folding nucleus.¹

Such a favorable role of non-native interaction in the formation of protein folding nucleus is in line with previous experimental and theoretical studies, suggesting that specific non-native interactions may accelerate folding by reducing conformational search to the native state.^53–63 The identification of a critical salt-bridge in a crystallin folding/unfolding intermediate is another example of non-native interactions crucial for protein folding. Similarly, Presta and Rose⁵⁹ described helix stop signals that were not necessarily retained in the native state of the folded proteins. Baldwin and coworkers^64,65 described ion pairs and helix dipole interactions in the S-peptide of RNase, which were not expected to be maintained in the native state of RNase. Dobson, Schwalbe and coworkers⁶⁶ observed long-range non-native interactions in lysozyme unfolding in 8 M urea, which were related to a non-native Arg-Trp-Arg sandwich structure, as found by Zhou and coworkers⁴⁷ with microseconds of MD simulations.

We then performed in silico a single alanine replacement on residue Glu135, a highly conserved residue within γ-crystallin family, to explore the role of this non-native salt-bridge in the unfolding of the isolated C-td and of the N-td in full γD-crys. We expect such a replacement to disrupt the critical Glu135-Arg142 salt-bridge, thereby favoring unfolding of interdomain surface. Indeed, the isolated C-td, as well as the N-td in full γD-crys, of the E135A mutant, unfolded much rapidly compared to the wild-type, as revealed from the time dependence of fraction of native contacts, Q [Fig. 3(a)]. The faster denaturation of both domains in the E135A mutant was a direct consequence of the disruption of Glu135-Arg142 salt-bridge, which is evident from the comparison of the C_α-C_α distance between residue Glu/Ala-135 and Arg142 in the wild-type and mutant protein during unfolding [Fig. 3(b)]. During the wild-type isolated C-td unfolding, those two residues formed the non-native salt-bridge within the first 5 ns of the simulation and maintained as a salt-bridge for the rest of the simulation [Fig. 3(b)]. This persistence of the salt-bridge between Glu135 and Arg142 seems to suggest that they play a key role in the “nucleation process” of the hydrophobic core near the domain interface. In contrast, the distance between Ala135 and Arg142 became larger and fluctuated more in the E135A mutant compared to the wild-type at different stages of unfolding [Fig. 3(b)]. Consequently, the secondary structure of Greek key motif 4 C-td₄ of the mutant disrupted more quickly and more drastically, particularly the β-strand near Arg142 is completely lost around 70 ns [Fig. S2(a)]. Other trajectories also indicate early destruction of this β-strand between 45 and 70 ns. In addition, the β-sheets comprising the Greek key motif 2 N-td₂ also denature much earlier in the E135A mutant (∼50 to 70 ns) than in the wild-type protein (∼160 ns) [Supporting Information Fig. S2(b)], suggesting the critical role of the Glu135-Arg142 salt-bridge in stabilizing the interdomain surface.

Figure 3.

(a) Time evolution of the fraction of native contacts, Q, for the isolated C-td (solid line) and for the N-td in full protein (dashed line) of the wild-type protein (in black) and the E135A mutant (in red) at 425 K. (b) C_α-C_α Distance between E135 and residue 142 as a function of simulation time in the isolated C-td of the wild-type protein (in black) and the E135A mutant (in red). The time dependence of the distance between the E135-R142 ion-pair in the wild-type isolated C-td is also shown in dashed line. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

The effect of the E135A replacement on the tertiary structure loss of the isolated C-td was even more dramatic, as revealed from the detailed comparison of the wild-type unfolding trajectories with those of the mutant (Fig. 4). The ab folded hairpin turn of C-td₄ became more flexible at a very early stage of unfolding (∼40 to 80 ns) in the mutant than the wild-type. By ∼80 to 120 ns, the wild-type isolated C-td mostly preserves its overall tertiary structure despite the loss of secondary structures in the Greek Key motif 3 C-td₃. On the other hand, by this time the mutant has lost most of its tertiary structure as well as the secondary structures within C-td₄, mainly near the mutation site. Finally, at ∼140 ns, both the wild-type and the mutant fully unfold (Q ∼0.1). By this time, strands a and b of C-td₄ still maintained their secondary structures in the wild-type C-td. However, all of the tertiary structures, as well as the secondary structures, of C-td₄ are completely lost in the mutant at this time, demonstrating the importance of the Glu135-Arg142 ion-pair interaction in preserving the stable folding nucleus of the isolated C-td. This result deserves further validation through direct biochemical experiments.

Figure 4.

Representative conformations of the isolated C-tds of the wild-type and the E135A mutant during 160 ns unfolding simulations. The cartoon representation of the protein is colored in gray to black from N-terminal to C-terminal. Residues 135 and 142 are shown in ball-stick representation. The colors used to represent different types of residues are similar to that used in Figure 1(a). For the wild-type, the non-native salt-bridge between Glu135 and Arg142 was formed within the first 5 ns of the unfolding simulation and maintained as a salt-bridge for the rest of the simulation. On the other hand, for the E135A mutant, there is no such salt-bridge, and the two residues are separated apart in far distance after about 100 ns simulation. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Based on these observations during kinetic unfolding simulations, we propose a detailed folding mechanism of the β-sheet domains of γD-crys. During the folding of the full γD-crys protein, a partially unfolded intermediate state is populated with its C-td mainly folded and N-td mostly unstructured. We also identify potential intermediates in the formation of double Greek key domains, in which one of the Greek key motifs is folded first, while the other motif remains mainly unstructured. In the full γD-crys monomer, the Greek key motifs that fold first within the individual domains are those near the interdomain surface. Thus, C-td₄ that encompasses the C-td interface of γD-crys folds before C-td₃. The C-td interface then serves as the folding template for the N-td, thus facilitating the folding of N-td₂. Within C-td₄, a hydrophobic folding core is formed in the very initial stage of γD-crys folding. This important folding core mainly consists of strands a and b. Surprisingly, a non-native salt-bridge formed between two conserved residues, that is, Glu135 and Arg142, at the end of the ab folded hairpin turn is found to be crucial for the formation of the folding nucleus of the C-td. This non-native salt-bridge interaction allows the early formation of ab folded hairpin turn, thus assisting the folding of strands a and b of C-td₄.

It is interesting to note that even though the hydrophobic core comprising strands a and b of C-td₄ plays an important role in the nucleation process as explained above, our simulations also showed that the unfolding of the wild-type isolated C-td is not affected much with a few substitutions to alanine for the hydrophobic residues (V132A, Y134A, Q143A, and L145A) of this core. This result is consistent with the experimental mutational data showing that the stability of the C-td was unaffected in the mutants V132A and L145A (12). Thus, our results indicate the importance of the Glu135-Arg142 salt-bridge for the correct formation of the hydrophobic interdomain surface of γD-crys. Interestingly, our recent in vitro thermodynamic and kinetic folding experiments of single mutants of γD-crys suggest that partial unfolding of the C-td is required to induce aggregation.⁶⁷ Taken together, both experimental and simulation results seem to imply that a partially denatured folding intermediate ensemble of γD-crys with C-td₃ mostly unstructured is likely one of the candidates for the aggregation-prone species leading to cataract formation. The importance of salt-bridge interactions in stabilizing γ-crystallins is also evident from the presence of highly polar residues on their surface, as demonstrated by crystallographic data.⁶⁸ Most of these residues are highly conserved, suggesting their important role in crystallin structure and/or function.⁶⁹ These polar residues form charge networks on the molecular surface, in which more than 50% of the ionic side chains form ion pairs.⁶⁸ Such organized charges strongly contribute to the high thermal stability, as well as to the high water solubility, of γ-crystallins within the lens, as the disruption of the charge network by the binding of small molecules to key charged residues leads to the post-translational adduct-mediated unfolding.⁷⁰ In addition, point mutations associated with congenital cataracts also involve surface polar residues of γ-crystallins.^71,72

Taken together, aforementioned experimental observations suggested that the surface charge network, together with the hydrophobic interactions, plays a major role in maintaining the stability of native γD-crys. The greater abundance of acidic and basic residues within the C-td most likely adds to its higher stability over the N-td in the native state. Our current simulations provided more atomic details to these observations and further identified one critical non-native salt-bridge interaction at the protein surface, in addition to the hydrophobic interactions between strands a and b of C-td₄. This Glu135-Arg142 salt-bridge helps initiating the folding of the ab folded hairpin turn, resulting in the formation of the C-td surface that is essentially used by the N-td as a folding template. Since a correctly formed interdomain surface of γD-crystallin is likely required to prevent unfolding into aggregation-prone species, disruption of charges on the γD-crystallin surface might lead to aggregation and development of opacity.

Conclusions

The γD-crystallins are synthesized in lens fiber cells in utero and infancy, and then must stay folded and stable within the terminally differentiated and anucleated fiber cells for a lifetime. For these proteins, the stability of the native state is a fundamental biological property. Though it seems reasonable that this stability resides in features of the Greek Key β-sheet fold and its tightly packed hydrophobic core, the molecular basis of the stability has been difficult to unravel. The detailed folding mechanism of the individual double Greek key domains of γD-crys has also resisted elucidation. In this article, we have used large-scale explicit solvent atomistic simulations to investigate the unfolding processes of the isolated domains, as well as of the full monomer, of γD-crys. Our simulations strongly suggest that the stability and the folding mechanism of the N-td are regulated by the interdomain interactions. As hypothesized based on fluorescence spectroscopic data,²⁰ a hydrophobic folding core is identified within the C-td interface of γD-crys that is comprised of strands a and b from Greek motif 4 C-td₄. Detailed analyses strikingly reveal that a surface salt-bridge formed between Glu135 and Arg142 initiates folding of the ab folded hairpin turn of C-td₄, allowing formation the stable folding core. However, this salt-bridge is not seen in the γD-crys crystal structure. We show the critical role of this non-native salt-bridge in preventing C-td₄ from complete unfolding by in silico introduction of a single mutation E135A. This substitution that disrupts the Glu135-Arg142 salt-bridge causes significant destabilization to the stable folding core of the isolated C-td, which, in turn, triggers unfolding of the N-td. Our results thus show that certain non-native interactions formed in the highly conserved polar exterior of γD-crys are crucial for correct folding of its hydrophobic domain interface. Therefore, our simulations in combination to our recent experiments provide a complete molecular picture of the complex unfolding landscape of γD-crystallin that might offer new insights into the development of novel treatments for cataracts.