Glycation-mediated inter-protein cross-linking is promoted by chaperone–client complexes of α-crystallin: Implications for lens aging and presbyopia

Lens proteins become increasingly cross-linked through nondisulfide linkages during aging and cataract formation. One mechanism that has been implicated in this cross-linking is glycation through formation of advanced glycation end products (AGEs). Here, we found an age-associated increase in stiffness in human lenses that was directly correlated with levels of protein–cross-linking AGEs. α-Crystallin in the lens binds to other proteins and prevents their denaturation and aggregation through its chaperone-like activity. Using a FRET-based assay, we examined the stability of the αA-crystallin–γD-crystallin complex for up to 12 days and observed that this complex is stable in PBS and upon incubation with human lens–epithelial cell lysate or lens homogenate. Addition of 2 mm ATP to the lysate or homogenate did not decrease the stability of the complex. We also generated complexes of human αA-crystallin or αB-crystallin with alcohol dehydrogenase or citrate synthase by applying thermal stress. Upon glycation under physiological conditions, the chaperone–client complexes underwent greater extents of cross-linking than did uncomplexed protein mixtures. LC-MS/MS analyses revealed that the levels of cross-linking AGEs were significantly higher in the glycated chaperone–client complexes than in glycated but uncomplexed protein mixtures. Mouse lenses subjected to thermal stress followed by glycation lost resilience more extensively than lenses subjected to thermal stress or glycation alone, and this loss was accompanied by higher protein cross-linking and higher cross-linking AGE levels. These results uncover a protein cross-linking mechanism in the lens and suggest that AGE-mediated cross-linking of α-crystallin–client complexes could contribute to lens aging and presbyopia.

tion of AGEs, which contribute to lens aging and cataract formation.
␣-Crystallin is a major component of lens proteins, accounting for ϳ40% of total proteins. It consists of ␣A-crystallin (␣AC) and ␣B-crystallin (␣BC) subunits. They are typically present at a 3:1 ratio in the lens (23). They belong to the family of small heat-shock proteins. Similar to other members of this family, both ␣AC and ␣BC exhibit chaperone-like properties (24,25). Through these properties, they prevent aggregation of denaturing proteins in an ATP-independent manner. This function of ␣-crystallin is suggested to be necessary to preserve lens transparency during aging. The noncovalent association between ␣-crystallin and other proteins, possibly through the chaperone-like activity of the former, has been demonstrated in human lenses (26). However, ␣-crystallinclient complexes in the lens, especially in lens fibers, are likely to be permanent. This is because metabolic activity in fiber cells, which make up a large part of the lens, is extremely low, and therefore, there is less likelihood for the existence of mechanisms that can dissociate ␣-crystallinclient complexes. We hypothesized that the proximity of two proteins in such complexes could render them susceptible to AGE-mediated nondisulfide intermolecular cross-linking and aggregation, especially when lens GSH levels are reduced during advanced aging. Such cross-linking and aggregation could explain the precipitous reduction of watersoluble ␣-crystallin in aged lenses. Moreover, cross-linking and aggregation of a major protein in the lens could decrease the accommodative ability of the lens, contributing to age-associated presbyopia. This study investigated AGE-mediated crosslinking of the ␣-crystallinclient complexes and the relationship between cross-linking AGEs and stiffness in aging human lenses.
Next, we determined the effect of age on human lens stiffness. Stiffness increased progressively until ϳ40 years of age and then exponentially until ϳ75 years of age (Fig. 2G). We found a strong correlation between the total cross-linking AGEs (the levels of MODIC, GOLA, glucosepane, and pentosidine together) and lens stiffness (Fig. 2H), suggesting a causal role for cross-linking AGEs in lens stiffness. In these measurements, we used thawed lenses that were stored frozen at Ϫ80°C. To determine whether freezing of lenses affects stiffness measurements in our setting, we measured the axial strain of freshly-obtained human lenses (donor age: 15,22,42, and 53 years) and after their freezing at Ϫ80°C for 2 days and thawing at 4°C for 16 h. Although freezing and thawing slightly altered the axial strain (6.86 Ϯ 6.24%, Fig. 2I), the age-associated increase in lens stiffness was still evident.

␣-Crystallinclient complexes are stable under conditions of the lens
To determine the stability of ␣-crystallinclient complexes, a fluorescence resonance energy transfer (FRET)-based assay was developed. Human recombinant ␣AC was conjugated with Lucifer Yellow iodoacetamide (LYI), and human recombinant ␥D-crystallin (␥DC) was conjugated with 4-acetamido-4Ј-((iodoacetyl)amino)-stilbene-2,2Ј-disulfonic acid (AIAS). We found 1 mol of LYI was bound to 1 mol of ␣AC, and 3 mol of AIAS was bound to 1 mol of ␥DC. The ␣AC-␥DC complex was generated by incubating the two proteins at 70°C for 0.5 h. Tagging AIAS to ␣AC or LYI to ␥DC did not appreciably alter the ␥DC aggregation or the ability of ␣AC to function as a chaperone (Fig. 3, A and B). FRET was observed in the ␣AC-␥DC complex but not in the protein mixture that was not thermally stressed (Fig. 3C). The size-exclusion chromatographic (SEC) profile revealed that more than 95% LYI-␣AC and AIAS-␥DC were present in the complex (Fig. 3D). The ␣AC-␥DC complex did not show a change in FRET intensity when Protein cross-linking in the lens incubated alone, with human lens epithelial cell lysate, or with human lens homogenate, even in the presence of 2 mM ATP, for up to 12 days tested (Fig. 3, E-G). Together, these observations suggest that the ␣-crystallinclient protein complexes are likely to be stable under the conditions of the human lens.

Chaperone-client complex formation increases with increasing temperature
To optimize conditions for the formation of ␣-crystallinclient complexes, ␣AC or ␣BC and citrate synthase (CS) were mixed and incubated at 25, 34, or 43°C for 1 h. ␣AC or ␣BC and alcohol dehydrogenase (ADH) were mixed and incubated at 25, 37, or 49°C for 1 h. The samples were then passed through an FPLC-SEC column, and the elution profiles were monitored. Fig. 4, A-D, clearly shows the formation of complexes of ␣AC or ␣BC with ADH and ␣AC or ␣BC with CS. To determine whether prior thermal stress applied to proteins before they were mixed would lead to complex formation, ␣AC or ␣BC was incubated at 49°C, and ADH was incubated at 43°C for 1 h, cooled to room temperature for 1 h, and then mixed. Similarly, ␣AC or ␣BC was incubated at 43°C, and CS was incubated at 37°C for 1 h, cooled to room temperature for 1 h, and then mixed. Client proteins were thermally stressed at a slightly lower temperature than that used to promote complex formation to avoid their aggregation (occurs at 49°C for ADH and 43°C for CS). Formation of an ␣-crystallinclient complex was not observed in these protein mixtures, suggesting that the complex forms only when ␣-crystallin and the client protein are together and thermally stressed, but not when they were thermally stressed separately and then mixed.

Inter-protein cross-linking by AGEs depends on the extent of formation of ␣-crystallinclient complex
The formation of chaperone-client complexes of ␣AC or ␣BC with CS was determined by incubating the protein mixtures at 25, 34, and 43°C for 1 h. Similarly, formation of complexes of ␣AC or ␣BC with ADH were determined at 25, 37, and 49°C. The idea was that higher temperatures would promote complex formation and predispose them to greater levels of AGE-mediated inter-protein cross-linking. The samples were glycated by incubating with either 200 M MGO or 250 M erythrulose for 7 days under sterile conditions at pH 7.4 and 37°C. The samples were then dialyzed against PBS and analyzed by SDS-PAGE using 4 -20% gradient gels under reducing conditions (Fig. 5).
Densitometric analyses revealed that MGO promoted the formation of nondisulfide protein cross-linking in ␣AC-ADH samples (Fig. 5). The cross-linked protein content was higher by 28 and 29% in samples incubated at 49°C than in samples incubated at 25 and 37°C, respectively. The ␣AC-CS samples incubated at 43°C showed 57 and 56% higher cross-linked protein content than those incubated at 25 and 34°C, respectively. When glycated with erythrulose, the ␣AC-ADH samples incubated at 49°C showed 45 and 42% higher cross-linked protein content than those incubated at 25 and 37°C, and the ␣AC-CS samples incubated 43°C showed 77 and 73% higher cross-linked protein content than those incubated at 25 and 34°C.
With MGO as the glycating agent, ␣BC-ADH samples incubated at 49°C showed 61 and 53% higher cross-linked protein content than samples incubated at 25 and 37°C (Fig. 5). In ␣BC-CS samples incubated at 43°C, the increase was 47 and 45% over samples incubated at 25 and 34°C. With erythrulose as the glycating agent, ␣BC-ADH samples incubated at 49°C showed 53 and 52% higher cross-linked proteins than samples incubated at 25 and 37°C (Fig. 5). The increase was 67 and 81% in ␣BC-CS samples incubated at 43°C compared with those incubated at 25 and 34°C. Thus, inter-protein nondisulfide cross-linking in ␣-crystallinclient complexes by AGEs depended on the extent of the two proteins complexed. In contrast, samples in which ␣-crystallin and client proteins (ADH or CS) were incubated separately at temperatures as described in Fig. 5, cooled, and mixed and then glycated with either MGO or erythrulose as above, the cross-linked protein levels were relatively low (last 3 lanes in Fig. 5). These data reiterate the notion that only the complexes formed between ␣-crystallin and a client protein, but not individual proteins mixed together, become susceptible to AGE-mediated cross-linking.
To further confirm that the cross-linked proteins were formed as a result of inter-protein cross-linking, we performed A FRET-based assay was used to investigate the stability of ␣-crystallinclient complexes. A, conjugation of AIAS to ␥DC did not appreciably change its thermal aggregation. AIAS-␥DC or ␥DC at 0.1 mg/ml in 50 mM phosphate buffer, 1 mM EDTA, pH 7, was incubated at 70°C for a period of 30 min. Conjugation of LYI to ␣AC did not alter chaperone activity toward thermally aggregating ␥DC. For this assay, ␣AC was used at 0.1 mg/ml, and/or LYI-␣AC was used at 0.1 and 0.3 mg/ml in 50 mM phosphate buffer, 1 mM EDTA, pH 7, and incubated over a period of 30 min at 70°C. The percent protection by ␣AC and LYI-␣AC against aggregating ␥D and AIAS-␥D is shown in B. C, FRET (excitation/emission wavelengths of 335/525 nm) was observed only when thermal stress (70°C) was applied for 30 min to a protein mixture containing LYI-␣AC and AIAS-␥DC in 50 mM phosphate buffer, 1 mM EDTA, pH 7. Excitation and emission bandpasses used were 2.5 and 5 nm. D, FPLC elution profile for LYI-␣AC (0.3 mg/ml) and AIAS-␥DC (0.1 mg/ml) with and without thermal stress at 70°C. Stability of the complex of LYI-␣AC and AIAS-␥DC (0.2 mg/ml) was verified by incubating the complex alone (E) or in the presence of lens epithelial cell lysate (ECL, 1 mg protein/ml) (F) and human lens homogenate (HLH, 6 mg protein/ml) (G) in the presence or absence of 2 mM ATP. Incubations were done at 37°C in 50 mM phosphate buffer, pH 7.4. Samples were collected at regular intervals between 0 and 12 days, and FRET was measured as above. The bar graph represents the means Ϯ S.D. of triplicate measurements. NS, not significant.
Western blottings using specific antibodies for ␣AC, ␣BC, ADH, and CS. In all AGE-mediated cross-linked proteins, we observed ␣-crystallin and the client proteins (Fig. 6). The individual protein content in cross-linked proteins was highest in samples incubated at 49°C and glycated (for ADH) and in samples incubated at 43°C and glycated (for CS) (lanes marked with arrows in Fig. 6). This phenomenon was not observed in samples in which ␣-crystallin and client proteins were subjected to thermal stress separately (last three lanes, Fig. 6). Together, these data affirm that ␣-crystallinclient complexes are susceptible to inter-protein cross-linking by AGEs.

Protein cross-linking in the lens
samples at 37°C (Fig. 7). In ␣BC-CS samples, the GOLA and GOLD levels were 6.9 Ϯ 1.5 and 2.9 Ϯ 0.13 pmol/mol leucine equivalent at 43°C compared with samples at 25°C (GOLA: 3.65 Ϯ 0.55, p Ͻ 0.01; GOLD: 0.16 Ϯ 0.04, p Ͻ 0.0001) and 34°C (GOLA: 3.7 Ϯ 0.55, p Ͻ 0.01; GOLD: 1.97 Ϯ 0.04, p Ͻ 0.0001) (Fig. 8). The levels of pentosidine were 0.34 Ϯ 0.008 pmol/mol leucine equivalent in ␣BC-ADH samples at 49°C, which were higher than those incubated at 25°C (0.184 Ϯ 0.04, p Ͻ 0.0001) and 37°C (0.238 Ϯ 0.022, p Ͻ 0.01) (Fig. 7). The levels were 0.293 Ϯ 0.026 for the ␣BC-CS complex formed at 43°C, and this was higher than samples incubated at 25°C (0.152 Ϯ 0.03, p Ͻ 0.0001) and 34°C (0.162 Ϯ 0.008, p Ͻ 0.0001) (Fig. 8). The GODIC levels were higher in ␣BC-ADH mixtures that were incubated at 49°C (390.5 Ϯ 56.1 pmol/mol leucine equivalent) than in samples incubated at 25°C (115.1 Ϯ 22.5, p Ͻ 0.0001) and 37°C (110.9 Ϯ 29.4, p Ͻ 0.0001) (Fig. 7). No such change in GODIC levels was noticed in the ␣BC-CS protein mixtures Figure 5. Thermal stress and glycation lead to nondisulfide cross-linking of ␣-crystallin and client proteins. ␣AC or ␣BC was mixed with ADH or CS and then incubated at temperatures ranging from 25 to 49°C for 1 h. The protein mixture was then glycated using 200 M MGO or 250 M ERT for 7 days at 37°C. The protein samples (15 g) were analyzed by SDS-PAGE using 4 -20% gradient gels under reducing conditions. Arrows point to lanes that have the highest level of cross-linked proteins. Densitometric analysis of the region (black brackets) was performed to quantify cross-linked proteins. The samples in lanes that have a red bracket had ␣AC or ␣BC or client proteins that were incubated separately at respective temperatures for 1 h as shown in the figure, cooled at room temperature for 1 h, mixed, and glycated. The bars represent means Ϯ S.D. for three independent experiments. ***, p Ͻ 0.001; ****, p Ͻ 0.0001.

Protein cross-linking in the lens
incubated at 43°C compared with those incubated at 25 and 34°C (Fig. 8).
When ␣BC (49 or 43°C), ADH (43°C), or CS (37°C) was subjected to thermal stress individually, mixed, and then glycated with either MGO or erythrulose, the levels of the cross-linking AGEs were similar to those observed for the protein mixtures (␣BC ϩ ADH or ␣BC ϩ CS) incubated and glycated at 25 and 34 or 37°C. Together, these results suggest that ␣-crystallinclient protein complexes, but not the noncomplexed proteins, are susceptible to AGE-mediated cross-linking. . ␣-Crystallinclient complexes undergo higher cross-linking when subjected to combined thermal stress and glycation than noncomplexed proteins. ␣AC or ␣BC and client proteins were mixed before subjecting them to thermal stress and glycation using ERT (250 M) or MGO (200 M) for 7 days at 37°C. Subsequently, proteins (15 g) were run on SDS-PAGE using 4 -20% gradient gels under reducing conditions and subjected to Western blotting (WB) against ␣AC, ␣BC, CS, and ADH. Arrows above lanes indicate the highest levels of covalently cross-linked proteins. In lanes marked with brackets, proteins were first incubated at the indicated temperature for 1 h separately, cooled for 1 h at room temperature, mixed, and then glycated.

Glycation-mediated protein cross-linking is higher in thermally-stressed and glycated mouse lenses
Mouse lenses were either exposed to high temperature (50°C for 1.5 or 3 h) or to control conditions (37°C) and then incubated with or without a mixture of glycating agents (25 mM glucose, 500 M MGO, and 500 M erythrulose) for 3 days at 37°C in serum-free minimal essential medium (MEM) in a CO 2 incubator. The lenses were then homogenized in PBS, and the water-soluble proteins were separated by centrifugation and subjected to SDS-PAGE on 4 -20% gradient gels under reducing conditions to assess nondisulfide protein cross-linking. In lenses that were incubated at 50°C for 1.5 and 3 h, the crosslinked protein levels were ϳ30 and 28% higher compared with lenses incubated at 37°C (Fig. 9, A and B). Compared with controls, lenses that were glycated alone showed ϳ18% increase in cross-linked proteins. However, when lenses were first incubated at 50°C and then glycated, the levels of cross-linked proteins were ϳ41 and 44% higher that those in lenses that were glycated alone. The results clearly showed that the combination of thermal stress and glycation leads to more protein crosslinking than either thermal stress or glycation alone.
Western blottings showed that ␣AC, ␣BC, ␤-crystallin (␤C), and ␥-crystallin (␥C) become cross-linked by the combination of thermal stress and glycation in mouse lenses (Fig. 9, C-F). One observation to note is that in Western blottings the thermal stress alone induced nondisulfide intermolecular cross-

Protein cross-linking in the lens
linking of proteins. This could be due to glycation-mediated cross-linking by the pre-existing precursors in the lens, such as MGO and ascorbate oxidation products.

Thermally stressed and glycated lenses exhibit greater levels of cross-linking AGEs than glycated mouse lenses
Four cross-linking AGEs (MOLD, MODIC, GOLA, and pentosidine) were quantified in lens proteins. The levels of these AGEs were higher in lenses that were thermally stressed (1.5 h) and glycated than in lenses that were only glycated; the levels of MOLD, MODIC, GOLA, and pentosidine were 0.079 Ϯ 0.016, 3.91 Ϯ 0.5, 7.8 Ϯ 2, and 0.33 Ϯ 0.052 pmol/mol leucine equivalent in thermally stressed and glycated lens proteins compared with 0.05 Ϯ 0.03, 2.1 Ϯ 0.8, 2.4 Ϯ 1.1, and 0.175 Ϯ 0.048 pmol/ mol leucine equivalent in just glycated lens proteins (Fig. 10). Furthermore, higher levels of cross-linking AGEs were observed in lenses subjected to 3 h of thermal stress and glycation (Fig. 10). Together, the results in Figs. 9 and 10 strongly suggest that the chaperone-client complexes that formed as a result of thermal stress underwent intermolecular protein cross-linking by AGEs.

Thermal stress and glycation increase mouse lens stiffness
The axial and equatorial strain in mouse lenses was unaltered upon glycation alone (Fig. 11, A and B). However, thermal stress at 50°C (1.5 or 3 h) significantly reduced the axial strain (by 46

Protein cross-linking in the lens
and 49%) and equatorial strain (by 61 and 58%). The combination of thermal stress and glycation further reduced the axial strain (by 57 and 58%) and equatorial strain (by 78 and 76%). These results suggest that AGE-mediated cross-linking of the ␣-crystallinclient complexes could increase lens stiffness and contribute to presbyopia.

Both cortical and nuclear proteins of human lens are AGEmodified upon glycation
We tested whether glycation of organ-cultured lenses leads to AGE formation in just cortical proteins or both cortical and nuclear proteins. We used human lenses in these experiments, as mouse lenses are difficult to separate into cortical and nuclear fractions. Freshly-obtained human lenses (donor age: 43, 44, and 48 years) were incubated for 3 h with serum-free MEM and then incubated with or without a glycation mixture (25 mM glucose, 500 M MGO, and 500 M erythrulose) for 3 days in a CO 2 incubator. Lenses were then weighed, and the capsule was removed and stirred for 2 h in 10 ml of PBS at 4°C in a 50-ml flask using a magnetic stirrer. The undissolved nucleus was weighed and homogenized in 3 ml of PBS using a hand homogenizer and centrifuged at 21,000 ϫ g for 20 min at 4°C. The soluble material from the lens after stirring (from lens cortex) was also centrifuged at 21,000 ϫ g for 20 min at 4°C. The supernatant protein from the nucleus and cortex (50 g each) was subjected to Western blotting for CML and methylglyoxal-derived hydroimidazolone-1 (MGH-1). The dilutions were 1:5,000 for the mouse monoclonal anti-CML antibody (from Jes Thorn Clausen, Novo Nordisk, Germany) and 1:1,000 for the mouse monoclonal anti-MGH-1 antibody (generated in our laboratory). Secondary antibody used was HRP-conjugated anti-mouse IgG (diluted 1:5,000). Fig. S1 shows that lenses that were treated with the glycation mixture had 22.3 and 13.5% higher levels of MGH-1 and CML in the cortical proteins, and 22.7 and 24.5% higher in the nuclear proteins when compared with those in untreated lenses. These results suggest that both cortical and nuclear proteins can be AGE-modified by glycating agents and that the increased stiffness observed in mouse lenses upon glycation could be due to AGE modification of both cortical and nuclear proteins.

Discussion
The objectives of this study were as follows: 1) to investigate whether ␣-crystallinclient protein complexes are stable in human lenses; 2) to test whether ␣-crystallinclient protein complexes are more prone to intermolecular nondisulfide cross-linking by AGEs; and 3) to determine whether the accu- Figure 9. AGE-mediated protein cross-linking is higher in mouse lenses that were thermally stressed (50°C) and glycated than in glycated or thermally stressed lenses. Lenses were glycated using a mixture of 25 mM glucose, 500 M erythrulose, and 500 M MGO in serum-free MEM for 3 days at 37°C in a CO 2 incubator. Water-soluble protein from mouse lenses (50 g) was analyzed by SDS-PAGE using 4 -20% gradient gels (A). The bar graph represents means Ϯ S.D. of densitometric analyses of the intermolecular cross-linked proteins (in brackets) from triplicate measurements (B). Western blots (WB) for ␣AC (C), ␣BC (D), ␤C (E), and ␥C (F) reveal formation of highmolecular-weight covalently-cross-linked proteins, indicated by the brackets. *, p Ͻ 0.05. Figure 10. Cross-linking AGE levels are higher in mouse lenses that were thermally stressed (50°C) and glycated. Lenses were glycated using a mixture of 25 mM glucose, 500 M erythrulose, and 500 M MGO in serum-free MEM for 3 days at 37°C in a CO 2 incubator. LC-MS/MS quantification of AGEs in WS protein from mouse lenses that were subjected to thermal stress and/or glycation. The bars represent means Ϯ S.D. for three independent experiments. NS, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001.

Protein cross-linking in the lens
mulation of cross-linking AGEs is related to increase in lens stiffness.
␣AC and ␣BC constitute ϳ40% of total proteins in human lenses (27). They prevent aggregation of client proteins in the lens, thereby keeping the lens proteins soluble and maintaining lens transparency. ␣-Crystallin binds to aggregation-prone proteins, possibly through a "holdase" mechanism that does not require ATP hydrolysis and prevents further denaturation of the protein (28). Large heat-shock proteins bind to denaturing client proteins in an ATP-dependent manner and use ATP to dissociate chaperone-client complexes, resulting in the release of repaired client proteins (29,30). Whether human lenses have such a mechanism to dissociate ␣-crystallinclient complexes is not known. Our FRET assay showed that under conditions of the human lens, the ␣AC-␥DC complex was stable up to 12 days, the time period we tested. Addition of lens epithelial cell lysate, human lens homogenate or ATP did not cause a dissociation of that complex. This supports the idea that there is no mechanism to dissociate ␣-crystallinclient complexes in human lenses. This idea is further bolstered by the fact that ␣-crystallin cannot itself renature denatured client proteins. A recent study showed that BAG-3 mediates the association of client protein-bound small heat-shock proteins with a large heat-shock protein and such association helps in renaturation of the client protein (31). However, in that study, the release of the bound client protein was not demonstrated. Furthermore, this mechanism has not been demonstrated in the lens. Mogk et al. (32) demonstrated the role of ClpB/DnaK in the release of denatured client proteins bound to small heat-shock proteins, after which they can be renatured by the large heat-shock protein machinery. In the lens, large heat-shock proteins such as HSP70 are mostly present in the metabolically-active anterior epithelium, and possibly the outermost cortical fiber cells. The bulk of the lens comprises metabolically-inactive fiber cells. Therefore, HSP70-mediated dissociation of ␣-crystallinclient complexes in the inner regions of the lens is unlikely. Wang and Spector (33) demonstrated that ATP could alter conformation of ␣-crystallinclient complexes and trigger the release of denatured proteins, subsequently allowing the large heat-shock proteins to renature client proteins. In contrast, Biswas and Das (28) demonstrated that ATP improves chaperone activity of ␣-crystallin. Thus, it remains debatable whether ATP alone can dissociate ␣-crystallinclient protein complexes. Wang and Spector (34) showed that ␣-crystallinbound enzyme can be activated and released in the presence of rabbit reticulocyte lysate and ATP, but such a mechanism is unlikely to occur in the lens. Moreover, there is no experimental evidence to date for the release of client proteins from the ␣-crystallinclient protein complex in the human lens. Thus, it is reasonable to assume that ␣-crystallinclient complexes formed especially in the inner fiber cells of human lenses are unlikely to dissociate and remain permanently tethered.
Human lenses accumulate several chemical modifications with aging. AGE modification is one of the dominant ones, and it can cross-link proteins and form HMW protein aggregates (35). AGE-mediated cross-linking is largely dependent on the proximity and concentrations of cross-linking components (amino groups) in a protein. We hypothesized that formation of ␣-crystallinclient complexes might increase the effective concentrations of critical components for glycation-mediated cross-linking of proteins. Our results clearly showed that ␣-crystallinclient complexes are susceptible to intermolecular cross-linking by glycation. We used thermal stress in this study to enhance formation of ␣-crystallinclient complexes. In human lenses, such complexes occur naturally during aging (26,36).
Mass spectrometric analyses showed greater levels of cross-linking AGEs in glycated ␣-crystallinclient complexes than in glycated ␣-crystallin and client proteins that were not present in complex. Together, these data provide a strong basis for the notion that in the absence of any dissociation mechanism the proximity of ␣-crystallin and client Figure 11. Stiffness increases in mouse lenses subjected to thermal stress and glycation. Mouse lenses (3-4 months, C57BL/6J mouse) in serumfree MEM were subjected to thermal stress (⌬) at 50°C for 1.5 or 3 h. Subsequently, they were glycated using a mixture of 25 mM glucose, 500 M erythrulose, and 500 M MGO for 3 days at 37°C in a CO 2 incubator. Control experiments with lenses that were either thermally stressed or glycated alone were included. Axial (A) and equatorial (B) compressive strains are plotted against a fixed applied load. The bar graphs represent the means Ϯ S.D. of triplicate measurements. NS, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001.

Protein cross-linking in the lens
proteins provides an ideal situation for AGE-mediated cross-linking. Furthermore, it is possible that the progressive accumulation of such cross-linked proteins leads to protein insolubilization. This could be the reason why AGE levels are higher in water-insoluble proteins than in water-soluble proteins of human lenses (6,18,19). Unlike disulfide bonds, inter-protein cross-linking by AGEs is permanent and could explain why addition of reducing agents cannot completely solubilize the highly-cross-linked water-insoluble proteins of aged and cataractous lenses (37).
The age-associated increase in lens stiffness has been documented in many studies and has been implicated in the onset of presbyopia (38,39). It has been established that the concentration of free ␣-crystallin, especially in the human lens nucleus, decreases with age as it becomes insoluble and is incorporated into HMW protein aggregates (26,40). Whether loss of soluble chaperone protein is the primary trigger for presbyopia needs to be investigated. Many studies have implicated disulfide bond formation due to increased oxidative stress in the genesis of such protein aggregates (41,42). Although there is compelling evidence for such a mechanism, our findings in this study offer an additional mechanism through AGE formation. A recent study showed that thermal stress promotes lens stiffness by enhancing aggregation of proteins (38). Whether promotion of AGE formation from in situ precursors is the reason for the increase in stiffness in that study has to be determined. Our human lens data in this study clearly establishes a positive correlation between cross-linking AGEs and age-associated lens stiffness. Furthermore, we demonstrated that mouse lenses subjected to combined thermal stress and glycation accumulate more nondisulfide cross-linked proteins than lenses subjected to either glycation or thermal stress alone. We also showed that the increase in cross-linked proteins by combined thermal stress and glycation is accompanied by an increase in crosslinking AGEs both in ␣-crystallinclient complexes and in mouse lenses. More importantly, we showed that mouse lenses subjected to thermal stress and glycation were less resilient than those subjected to thermal stress or glycation alone. Thus, our study provides a biochemical mechanism for the onset of presbyopia in humans (Fig. 12).
Given our findings in this study, an important question arises: is the chaperone activity of ␣-crystallin beneficial or detri-mental to the lens, especially in aged lenses? It is likely that chaperone activity is essential to prevent protein aggregation and maintain transparency in aging lenses. This beneficial activity may have a negative consequence after a certain age, possibly after the 4th or 5th decade in humans, when lens defenses such as GSH and antioxidant enzymes become deficient or weak (12,43). This could lead to reduced MGO metabolism by glyoxalase and enhanced oxidation of ascorbate to highly-reactive glycation precursors. In fact, these changes have been documented in aged human lenses (44,45). The lack of dissociation mechanisms for the ␣-crystallinclient protein complexes could thus create a milieu in aged lenses favoring AGE-mediated cross-linking of ␣-crystallinclient protein complexes. Therefore, it is tempting to speculate that chaperone activity is beneficial in early life but could have a negative connotation later in life. The loss and gain of chaperone activity as a result of chemical modifications (46 -50) could also dictate the kinetics of AGE-mediated cross-linking of ␣-crystallinclient protein complexes. The inter-protein association through disulfide bonds, which has been well-documented in human lenses, could also predispose them to AGE-mediated cross-linking, similar to ␣-crystallinclient complexes. This possibility needs to be investigated in future studies. It should be noted that we quantified only four protein cross-linking AGEs. We did not measure other known cross-linking AGEs in the lens such as vesperlysine (19) and K2P (16). Moreover, the human lens is expected to have many additional uncharacterized AGEs (51). Therefore, AGE-mediated cross-linking of ␣-crystallinclient complexes could be even more important than what is appreciated in this study. Finally, the proposed mechanism in this study might also explain the early onset of presbyopia in diabetics (52) because of higher levels of crosslinking AGEs than in nondiabetics (19).   L5006), and carboxypeptidase Y (catalog no. C3888) were obtained from Sigma. LYI (catalog no. L1338) and AIAS (catalog no. A484) were obtained from Invitrogen. The antibody for ␣AC (catalog no. ADI-SPA-221-F) was obtained from Enzo Life Sciences (Farmingdale, NY). The antibody for ␣BC (catalog no. ABN185) was from Millipore-Sigma (Burlington, MA). Antibodies for ␤C (catalog no. SC22745) and ␥C (catalog no. SC22746) were from Santa Cruz Biotechnology (Dallas, TX). The CS antibody (catalog no. 14309S) and anti-rabbit IgG (catalog no. 7074S) were from Cell Signaling Technology (Danvers, MA), and the ADH antibody (catalog no. 200-4143S) was from Rockland Antibodies (Limerick, PA). All other chemicals were of analytical grade.

Measurement of lens stiffness
Human lenses (donor age: 13-75 years, n ϭ 17) were obtained from Saving Sight, Kansas City, MO, and were stored at Ϫ80°C until use. Stiffness was measured in thawed lenses using a setup described by Cheng et al. (53) with slight modifications. Briefly, the lens chamber was filled with 65-70 ml of PBS. A right-angled mirror was positioned for the side view of the lenses. Side-view images were captured using a dissecting microscope attached to a camera and computer. Lenses were pictured before placing a chamber slide on them. To increase the load on the lens, a premeasured volume of water was added using a pipette. Total load added was calculated by multiplying the volume of water added with the density of water. Lens stiffness was measured using a load of 2,030 mg. The axial diameter before and after loading a fixed volume of water was measured using ZEN 2.3 lite software (Carl Zeiss Microscopy, LLC, White Plains, NY). Axial strain was calculated as described previously (53).

LC-MS/MS measurement of AGEs
Samples were hydrolyzed with acid or enzymes, as described previously (14), with some minor modifications. For acid hydrolysis, samples were suspended in 6 N HCl and incubated at 110°C for 16 h under argon in sealed glass ampules. The hydrolyzed samples were dried in a vacuum concentrator and suspended in 50 l of water. For enzyme hydrolysis, protein solutions in PBS were incubated with 0.1 unit of Pronase E for 8 h followed by another addition of Pronase E for 16 h, 0.3 unit of leucine aminopeptidase for 8 h, and 0.3 unit of carboxypepti-dase Y for 16 h. The enzyme-digested samples were passed through a 3-kDa molecular mass cutoff filter. The samples (acid-and enzyme-hydrolyzed) were analyzed by LC-MS/MS using the standard addition method, and the concentrations of AGEs were expressed relative to amino acid content (leucine equivalent, determined by the ninhydrin assay), as described previously (14). The LC-MS system consisted of a Waters Acquity UPLC system (Milford, MA) connected to a Sciex 4500 QTrap mass spectrometer (Redwood City, CA). Chromatographic separation was performed at 40°C on a Waters Acquity HSS T3 1.8 M 2.1 ϫ 100-mm column using a gradient program at 0.6 ml/min consisting of solvent A ϭ water and solvent B ϭ 80% (v/v) acetonitrile, each containing 0.12% heptafluorobutyric acid. The gradient program was as follows: 2% B (0 -2.2 min) to 8% B (3.3 min) to 34% B (7.6 min) to 100% B (7.8 -9.5 min), after which the column was re-equilibrated with 2% B prior to the next analysis. For mass spectrometric detection, the scheduled multiple-reaction monitoring mode was used: utilizing collision-induced dissociation of the protonated molecules with compound-specific orifice potentials and fragment-specific collision energies ( Table 1). Ion source parameters were as follows: temperature, 650°C; ion spray voltage, 2.5 kV; curtain gas, 35 ml/min; nebulizer gas, 65 ml/min; heating gas, 70 ml/min. The MRM parameters for AGEs were measured in a positive-ion mode.

Measurement of pentosidine
Proteins were acid-hydrolyzed and suspended in water as above. The acid-hydrolyzed material was analyzed for pentosidine using UPLC and a fluorescence detector. The column was a Waters Acquity HSS T3 1.8 M 2.1 ϫ 100 mm. A linear gradient program was used with solvent A: 100% water with 0.12% heptafluorobutyric acid, and solvent B: 60% acetonitrile with 0.12% heptafluorobutyric acid. The program was as follows: 0 -5 min: 10% B; 5-7 min: 30% B; 7-7.5 min: 70% B; 7.5-10 min: 100% B; and 10 -12.9 min: 10% B, at a flow rate of 0.5 ml/min. The column eluate was monitored in a fluorescence detector (excitation/emission wavelengths ϭ 335/385 nm). Pentosidine content in samples was calculated based on the standard curve generated using a synthetic pentosidine and expressed as picomoles/mol of leucine equivalent.

Protein cross-linking in the lens
Labeling of recombinant ␣Aand ␥D-crystallin using fluorophores Recombinant human ␣Aand ␥D-crystallin were purified as described previously (54,55). The cysteine residues in ␣AC and ␥DC were tagged with LYI and AIAS, respectively, as described previously (56). Labeled LYI-␣AC and AIAS-␥DC were purified by passing through a Sephadex G-25 column equilibrated with 50 mM phosphate buffer, 100 mM NaCl, pH 7.4. The fluorescent peak containing the labeled protein was collected and dialyzed against 50 mM phosphate buffer, pH 7.4, for 16 h at 4°C. The extent of labeling of the fluorophores was calculated as described previously (56).

Measurement of chaperone activity
␥DC or AIAS-␥DC (0.1 mg/ml) was incubated at 70°C in the presence or absence of ␣AC (0.1 mg/ml) or LYI-␣AC (0.1 and 0.3 mg/ml) in 50 mM phosphate buffer containing 1 mM EDTA, pH 7.4. Light scattering at 400 nm was monitored in the kinetic mode for 0.5 h in a UV spectrophotometer.

FRET assay
FRET was employed to determine the complex formation between LYI-␣AC and AIAS-␥DC. FRET was initiated by incubating LYI-␣AC and AIAS-␥DC as described in the chaperone activity assay above. The emission spectrum of the incubation mixture before and after thermal stress was measured in a spectrofluorometer (Spectramax-4, Horiba Jobin Mayer, Edison, NJ). An excitation wavelength of 335 nm was used to measure the emission spectra from 350 to 600 nm. The excitation and emission bandpasses were 2.5 and 5 nm, respectively.

Testing the stability of LYI-␣AC and AIAS-␥DC complex
Stability of the LYI-␣AC-AIAS-␥DC complex was tested (0.2 mg/ml) over a period of 12 days in the presence or absence of ATP (2 mM) or human lens homogenate (300 l, 6 mg of protein/ml) from a young (17 years) or an aged donor (75 years) or human lens epithelial cell (FHL124) lysate (300 l, 1 mg of protein/ml). FHL124 cells were from Dr. Michael Wormstone, University of East Anglia, UK, and cultured as described previously (57). Incubations were performed at 37°C under sterile conditions in 50 mM phosphate buffer, pH 7.4. Samples were collected from the incubation mixtures at 0, 4, 8, and 12 days, and FRET was measured at excitation/emission wavelengths of 335/525 nm.

Generation and purification of ␣-crystallinclient protein complexes
Human ␣AC and ␣BC were cloned, bacterially expressed, and purified as described previously (58). Mixtures containing ␣AC or ␣BC and ADH were incubated at 25, 37, or 49°C in 50 mM phosphate buffer, 100 mM NaCl, pH 7.4. Mixtures containing ␣AC or ␣BC and CS in 10 mM HEPES, pH 7.4, were incubated at 25, 34, or 43°C. The ratios of ␣AC/ADH and ␣BC/ ADH were 1:2 and 1:3 (w/w), respectively. These ratios were chosen for total inhibition of client protein aggregation by ␣-crystallin. The ratios of ␣AC/CS and ␣BC/CS were both 1:0.35 (w/w). The protein mixtures were dialyzed against 50 mM phosphate buffer, pH 7.4, for 16 h at 4°C.
Purification of the ␣-crystallinclient protein complex was performed on a FPLC instrument (NGC Chromatography System, Bio-Rad) equipped with an analytical gel-filtration column (Enrich SEC-650, 10 ϫ 300 mm, 24 ml). The column was equilibrated with 50 mM phosphate buffer, pH 7.4. Five hundred microliters of ␣-crystallinclient protein mixture was then loaded into the column and eluted using 50 mM phosphate buffer, pH 7.4, at a flow rate of 0.7 ml/min, and the eluent was monitored at 280 nm. LYI-␣AC, AIAS-␥DC, and the complex of LYI-␣AC and AIAS-␥DC (formed by incubating at 70°C for 1 h) were similarly purified on FPLC.

Glycation of ␣-crystallinclient protein complexes
The ␣-crystallinclient protein complexes purified as described above were glycated with either 200 M MGO or 250 M erythrulose in 50 mM phosphate buffer, pH 7.4, at 37°C for 7 days. Following incubation, the samples were dialyzed against PBS for 16 h at 4°C.

Thermal stress and glycation in mouse lenses
All animal experiments were reviewed and approved by the University of Colorado, Aurora's Institutional Animal Care and Use Committee (IACUC), and performed under adherence to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Lenses from C57BL/6J mice (3-4 months old) were isolated from eyes and incubated at 50°C for 1.5 or 3 h with serum and phenol-red-free MEM in a CO 2 incubator. Control lenses were incubated at 37°C. Lenses were then glycated with a mixture of 25 mM glucose, 500 M MGO, and 500 M erythrulose for 3 days at 37°C in serum-free MEM in a CO 2 incubator. Lenses incubated simultaneously without glycating mixture served as controls. Lens stiffness was measured as above. Lenses were then homogenized in PBS, and the watersoluble fraction was isolated by centrifugation at 20,000 ϫ g for 30 min at 4°C. Water-soluble protein was subjected to SDS-PAGE on 4 -20% gradient gels under reducing conditions. To measure AGEs in proteins, the samples were processed and analyzed by LC-MS/MS as described above.

Western blotting detection of crystallin cross-linking
For Western blotting, water-soluble protein (50 g) from mouse lenses was separated by SDS-PAGE on 4 -20% gradient gels under reducing conditions and then electrophoretically transferred to a nitrocellulose membrane. The membrane was incubated overnight at 4°C with one of the following antibodies: ␣AC antibody (diluted 1:5,000); ␣BC antibody (diluted 1:5,000); ␤C antibody (diluted 1:2,000), or ␥C antibody (diluted 1:2,000). The membrane was then incubated with HRP-conjugated anti-rabbit IgG (diluted 1:5,000) for 1 h at room temperature. For Western blotting of thermally stressed and glycated ␣-crystallinclient protein complexes (15 g), the primary antibody (incubated overnight at 4°C) dilutions for ␣AC and ␣BC are the same as those used above. The dilution for the CS antibody was 1:5,000 and that for the ADH antibody was 1:4,000.

Statistical analysis
All data are means Ϯ S.D. and are representative of at least three independent experiments. Statistical significance of the Protein cross-linking in the lens data was analyzed using one-way ANOVA with Dunnett's multiple comparison tests using GraphPad Prism 7 software. A p value Յ0.05 was considered statistically significant.