Characterization of a -Crystallin-Plasma Membrane Binding*

a -Crystallin, a large lenticular protein complex made up of two related subunits ( a A- and a B-crystallin), is known to associate increasingly with fiber cell plasma membranes with age and/or the onset of cataract. To understand better the binding mechanism, we developed a sensitive membrane binding assay using lens plasma membranes and recombinant human a A- and a B-crystallins conjugated to a small fluorescent tag (Al-exa350 ® ). Both a A and a B homopolymer complexes, as well as a reconstituted 3:1 heteromeric complex, bind to lens membranes in a specific, saturable, and partially irreversible manner that is sensitive to both time and temperature. The amount of a -crystallin that binds to the membrane increases under acidic pH conditions and upon removal of exposed intrinsic membrane protein domains but is not affected at high ionic strength, suggesting that a -crystallin binds to the fiber cell plasma membranes mainly through hydrophobic interactions. The binding capacity and affinity for the reconstituted 3:1 heteromeric complex were measured to be 3.45 6 0.11 ng/ m g of membrane and 4.57 6 0.50 3 10 2 4 m g 2 1 of membrane, respectively. The present membrane binding data support the hypothesis that the physical properties of a mixed a -crystallin complex may hold particular rel-evance for the function of a -crystallin within the lens.

␣-Crystallin, a large lenticular protein complex made up of two related subunits (␣A-and ␣B-crystallin), is known to associate increasingly with fiber cell plasma membranes with age and/or the onset of cataract. To understand better the binding mechanism, we developed a sensitive membrane binding assay using lens plasma membranes and recombinant human ␣Aand ␣B-crystallins conjugated to a small fluorescent tag (Al-exa350 ® ). Both ␣A and ␣B homopolymer complexes, as well as a reconstituted 3:1 heteromeric complex, bind to lens membranes in a specific, saturable, and partially irreversible manner that is sensitive to both time and temperature. The amount of ␣-crystallin that binds to the membrane increases under acidic pH conditions and upon removal of exposed intrinsic membrane protein domains but is not affected at high ionic strength, suggesting that ␣-crystallin binds to the fiber cell plasma membranes mainly through hydrophobic interactions. The binding capacity and affinity for the reconstituted 3:1 heteromeric complex were measured to be 3.45 ؎ 0.11 ng/g of membrane and 4.57 ؎ 0.50 ؋ 10 ؊4 g ؊1 of membrane, respectively. The present membrane binding data support the hypothesis that the physical properties of a mixed ␣-crystallin complex may hold particular relevance for the function of ␣-crystallin within the lens.
␣-Crystallin is a large protein complex comprised of 30 -40 copies of the A (␣A-crystallin) and B (␣B-crystallin) subunits in roughly a 3 ␣A to 1 ␣B molar ratio and can represent up to 50% of the total protein in vertebrate lenses (1,2). Human ␣Aand ␣B-crystallin polypeptides have molecular masses of about 20 kDa, making the 3:1 ␣A/␣B heteromeric complex found in the lens between 600 and 800 kDa (1). Each subunit shows significant sequence homology to small heat shock proteins such as mouse HSP25 and human HSP27, but they diverge somewhat in their NH 2 and COOH termini (1,3,4). Both protein subunits are also known to exchange readily between soluble ␣-crystallin complexes in a time-(Ͼ6 h) and temperature-(Ͼ25°C) dependent manner, suggesting that the heteromeric complex has a dynamic quaternary structure (1,5).
In addition to their many similarities, ␣Aand ␣B-crystallin have some notable differences. First, ␣A-crystallin is expressed almost exclusively in lens fiber cells, whereas ␣B-crystallin is expressed strongly in both the lens and non-ocular tissues such as heart, liver, and brain (6). Second, a study utilizing twodimensional 1 H NMR spectroscopy demonstrated that ␣A-crystallin is more stable than ␣B-crystallin with respect to the denaturing effects of extreme temperature, pH, and chaotropic agents such as urea and guanidine (7). Interestingly, studies using knockout mice suggest major differences in the in vivo roles of ␣Aand ␣B-crystallins. Deletion of ␣B-crystallin has little effect on lens morphology; however, the removal of ␣Acrystallin causes early onset cataract associated with accumulation of inclusion bodies containing large amounts of ␣B-crystallin (8,9). From these observations, it seems likely that these two subunits are not functionally equivalent and that there may be some functional significance in the combination of ␣A and ␣B subunits found in the native complex.
Despite these observations, the biological roles of ␣-crystallin are not yet firmly established. It is thought to give vertebrate lenses their refractive and transparency properties through short range ordering, which could result from the repulsive behavior of the native protein complex (10 -13). In addition, ␣-crystallin has been shown to inhibit stress-induced aggregation of proteins in vitro, reflecting a potential in vivo role in maintenance of lens transparency (14 -18). Finally, ␣-crystallin is known to bind subcellular elements such as intermediate filaments and plasma membranes (19 -22), but these interactions are not well characterized, nor is it clear what role they may play in lens biology (23).
It is well known that the amount of soluble ␣-crystallin found in the cytoplasm of lens fiber cells falls steadily with increasing age and that this change is mirrored by an increase in the water-insoluble fraction of ␣-crystallin (24 -28). The prevailing explanation for these observations is that the amount of soluble ␣-crystallin available to bind partially denatured proteins becomes depleted with age, and the resulting high molecular mass aggregates slowly become insoluble. Interestingly, the amount of crystallin protein, especially ␣-crystallin, bound to the membrane also increases dramatically with increasing age and/or cataract formation (29,30). It could be that the progressive insolubilization of ␣-crystallin is due, in part, to increased membrane binding that is also associated with aging and cataract formation. These observations suggest that increased membrane association of ␣-crystallin may be closely correlated with the loss of transparency in the lens, thus underscoring the need to understand this interaction in greater detail.
To date, few studies have examined the in vitro conditions that promote binding of ␣-crystallin to membranes. In a previous study, the interaction of native ␣-crystallin with membranes was shown to be markedly sensitive to ionic strength while reaching an optimum at 37°C and a pH of 7.5 (22). In addition, saturable binding to protein-free phospholipid vesicles has been observed, but the measured capacity is reduced dramatically upon incorporation of cholesterol (31)(32)(33). Finally, it has been postulated that ␣B-crystallin is largely unable to bind membranes in the absence of ␣A-crystallin (20,23).
To understand better the nature of ␣-crystallin membrane association and the relative contribution of each ␣-crystallin subunit to membrane binding, we developed a binding assay using extracted bovine cortical fiber plasma membranes and recombinant human ␣Aand ␣B-crystallins. The effects of time, temperature, ionic strength, and pH on the membrane association of both homopolymeric complexes and the 3:1 heteromeric complex were compared. In addition, the reversibility, binding capacities, and binding affinities for each complex were determined. Our results show that ␣-crystallin homopolymeric complexes composed of ␣Aor ␣B-crystallin interact with lens fiber cell plasma membranes in a similar but distinct manner and that membrane binding of the heteromeric complex containing ␣A and ␣B subunits in a 3:1 molar ratio differs from either of the homopolymeric complexes.

EXPERIMENTAL PROCEDURES
Overexpression and Purification of Recombinant Human ␣Aand ␣B-Crystallin-Cloning of human ␣A-crystallin cDNA was described previously (15). A full-length human ␣B-crystallin cDNA was obtained as an IMAGE clone (GenBank accession number N35834) from Genome Systems, Inc. (St. Louis). Coding regions of each cDNA were cloned into pET23d(ϩ) (Novagen) for overexpression in Escherichia coli strain BL21, and purification was performed essentially as described previously, with two minor changes (15). First, primary separation was performed on a DEAE-Sepharose ion exchange column at pH 8.0. Second, all columns were operated at 4°C. Proteins were estimated to be Ͼ99% homogeneous judging from the appearance of a single band following SDS-PAGE 1 and Coomassie Blue staining. The 3:1 ␣A to ␣B crystallin heteromeric complex was made by combining the two proteins in the appropriate molar ratio and incubating them for 24 h at 37°C in PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 ,1.4 mM K 2 HPO 4 , pH 7.3) (2,5,34,35).
Overexpression and Purification of Recombinant Human ␥D-Crystallin-Cloning, overexpression, and purification were performed exactly as described previously (15). The protein was stored at Ϫ80°C in PBS until use.
Lens Plasma Membrane Fractionation-Bovine lenses were isolated, decapsulated, and the outer half (i.e. cortical fiber cells) removed using a scalpel. Membrane preparation was then performed using the cortical fiber cells as described by Russell et al. (36). The dry weight was quantified by drying aliquots of the membrane suspension in preweighed centrifuge tubes in a spin-vacuum (Heto VR1, Denmark). Weight measurements were performed on an analytical balance (Mettler-Toledo, Columbus, OH) and the concentration (g/ml) of membrane in the suspension determined by averaging at least four repetitions.
␣-Crystallin Conjugation to Alexa350 ® -Purified recombinant ␣Aand ␣B-crystallin were conjugated to the Alexa350 ® fluorescent tag (molecular weight 410) as described by the manufacturer (Molecular Probes, Eugene, OR). Briefly, ␣-crystallin subunits were mixed with Alexa350 ® powder in PBS supplemented with 100 mM sodium bicarbonate. Conjugation was allowed to proceed for 1 h at room temperature. The reaction was stopped with the addition of hydroxylamine. Conjugated protein was separated from nonreacted Alexa350 ® using a prepacked desalting column according to the manufacturer's protocol (Bio-Rad Econo-Pac 10DG column). The purified Alexa350 ® -conjugated ␣-crystallin subunit was analyzed using A 280 /A 346 readings in a Varian Cary 1E UV-visible spectrophotometer. Protein concentration and degree of conjugation (i.e. conjugation efficiency) were determined with the following equations, where 0.19 is a correction factor for the absorbance of Alexa350 ® at 280 nm, 19,000 is the molar extinction coefficient for Alexa350 ® , A 280 and A 346 are the measured absorbance values at 280 and 346 nm, respectively, and ⑀ protein is the molar extinction coefficient for ␣-crystallin. Covalent attachment of Alexa350 ® to the ␣-crystallin subunits was verified by denaturing the ␣-crystallin conjugates with 6 M urea followed by dialysis for 16 h into 2,000 volumes of PBS using a 14000 MWCO membrane (Spectra/Por ® ). The number of mol of Alexa350 ® detected per mol of ␣-crystallin protein subunit before and after urea treatment and dialysis was not significantly different (data not shown).
The specific activity of the ␣-crystallin conjugates was determined by analyzing known amounts in a Hoefer Dyna-Quant Spectrofluorometer. The average specific activity was then calculated and expressed in F/g of protein (F ϭ fluorescence units). Also, emission spectra were collected using an LS50B luminescence spectrometer (Perkin-Elmer) with an excitation wavelength of 346 nm.
Chaperone-like Activity (CLA) Measurements-CLA measurements were determined on the Alexa350 ® -conjugated and nonconjugated ␣Aand ␣B-crystallin using the previously described insulin reduction chaperone assay (17,(37)(38)(39)(40). The insulin stock solution (1 mg/ml) was made by dissolving 10 mg of insulin (Sigma) in 10 ml of dH 2 O with 10 l of 1 N NaOH added to dissolve the protein. The activity was calculated by the following equation, then scaled such that nonconjugated ␣-crystallin had 100% activity.
Trypsin Treatment of Lens Plasma Membranes-Lens plasma membranes were pelleted by centrifugation at 14,000 ϫ g for 30 min at 4°C and decanted. The membranes were then resuspended in buffer T (10 mM bis-Tris, pH 6.5, and 0.1 mM MgCl 2 ). Trypsin (Sigma) was added to a final concentration of 10 g/ml, and the mix was allowed to incubate at 37°C for 3 h. The reaction was stopped by the addition of antipain serine protease inhibitor (Sigma) to a final concentration of 10 M. The membranes were centrifuged at 14,000 ϫ g for 30 min and decanted. The pellet was washed twice with PBS. The final pellet was resuspended in PBS containing 10 M antipain inhibitor.
Membrane Binding Measurements-Alexa350 ® -conjugated ␣Aor ␣B-crystallin was incubated with bovine cortical fiber cell plasma membranes in binding buffer (PBS supplemented with 5 mM MgCl 2 , except in the pH dependence measurements; see below). In control assays, bovine serum albumin (BSA) or recombinant human ␥D-crystallin was added to the reactions in a 1:1 ratio with ␣-crystallin. After the incubation, the sample was centrifuged at 14,000 ϫ g for 30 min at 4°C and decanted. The pellet, containing plasma membrane and bound Al-exa350 ® -conjugated ␣-crystallin, and the supernatant were then analyzed for fluorescence. The degree of binding (X) was calculated by the following formula.
The binding capacity was calculated from experiments in which 500 g of membranes were used with a varied amount of conjugated ␣-crystallin protein. The horizontal asymptote of the saturation curve represents the maximum amount of ␣-crystallin able to bind a fixed amount of membrane.
Other binding experiments were performed using this general approach, only with varied time (0 -16 h), temperature (4 -45°C), or NaCl concentration (0 -1 M). The effect of pH was determined using a previously described Tri-buffer system (Tris, MES, and acetate) that was designed to give high buffering capacity at a wide pH range (5 to 9 pH) while keeping the ionic strength constant (41).
Reversibility was assessed by first binding Alexa350 ® -conjugated ␣Aor ␣B-crystallin to a membrane sample and then decanting the unbound ␣-crystallin protein following the standard centrifugation. Next, the membrane-␣-crystallin complex was resuspended in either binding buffer only or binding buffer containing a molar excess of nonconjugated ␣-crystallin and incubated at 37°C for 24 h. As controls, the second incubations were also performed in the presence of ␣-crys-tallin and either human ␥D-crystallin or BSA. The degree of removal was determined through fluorescence analysis of the pellet and supernatant immediately after the second incubation.
The binding constants were determined by performing a series of binding assays with varied membrane (i.e. substrate) concentration. The apparent binding constant (K app ) was calculated for a wide range of membrane concentrations, according to the equation below, and then graphed in a log 10 K app versus log 10 [membranes] plot.
The overall binding constant (K) for noncooperative binding events was then obtained by the inverse logarithm of the y axis of the linear fits. For the cooperative binding event, the binding constants (K high and K low ) were determined by the inverse logarithms of the y axes of the horizontal asymptotes from nonlinear regression fitting.

␣-Crystallin
Overexpression and Purification-␣Aand ␣Bcrystallin were both overexpressed in E. coli as described previously. Purification led to homogeneous proteins of greater than 99% purity, as judged by SDS-PAGE and Coomassie Blue staining ( Fig. 1, lanes 2 and 3).
Membrane Fractionation-Plasma membranes were extracted from 20 bovine lenses as described under "Experimental Procedures." The final product was analyzed with SDS-PAGE, which showed no detectable ␣-crystallin upon zinc negative staining and Coomassie Blue staining. The only visible band was at a position corresponding to the major intrinsic membrane protein MP26 (Fig. 1, lane 1) (36).
The membrane sample was also analyzed by transmission electron microscopy. As expected, the trilayered gap-junctionassociated membrane structures were clearly visible at a magnification of ϫ 90,000. The membranes were smooth on both sides, indicating that ␣-crystallin complexes normally associated with lens membranes were removed by the extraction procedure (data not shown).
Analysis of ␣-Crystallin Alexa350 ® Conjugates-Conjugation of Alexa350 ® to either ␣Aor ␣B-crystallin was performed numerous times throughout the course of the present study. The conjugation reaction proved to be very consistent for both ␣-crystallin subunits (Table I). On average, 0.73 Ϯ 0.12 mol of Alexa350 ® was attached per ␣A-crystallin subunit, whereas 1.0 Ϯ 0.12 mol was conjugated to each ␣B-crystallin subunit. In addition, the absorbance and emission spectra, characterized by an absorbance peak at about 346 nm and an emission peak at about 440 nm, matched those published previously (42) (data not shown).
To determine the specific activity, known amounts of each Alexa350 ® conjugate were analyzed for fluorescence. The specific activity was determined by taking an average of measurements at a wide range of protein concentrations (Table I). These results demonstrated the high sensitivity and reproducibility of detection. Linearity of the fluorescence extended to at least 100 g of protein conjugate, with low error (data not shown).
The average molecular masses for Alexa350 ® -conjugated ␣Aand ␣B-crystallin homopolymeric complexes were measured on a Pharmacia FPLC Superose 6 column and then compared with nonconjugated ␣Aand ␣B-crystallins (summarized in Table I). ␣A-Crystallin without the Alexa350 ® moiety had an average molecular mass of 540 Ϯ 5 kDa, whereas the Alexa350 ® conjugate was 526 Ϯ 4 kDa. Similar results were obtained for ␣Bcrystallin. No significant changes in molecular mass of the complexes were seen upon conjugation with Alexa350 ® . The elution profile of conjugated protein complexes was indistinguishable from that obtained with nonconjugated controls.
CLA of the Alexa350 ® conjugates was tested using the insulin reduction assay. The activity of ␣A-Alexa350 ® was calculated as the percentage of activity seen with an equimolar quantity of nonconjugated ␣A-crystallin. ␣B-crystallin was treated in the same manner. The summarized data indicate no measurable change in CLA for either ␣-crystallin subunit after modification with Alexa350 ® (Table I).
Verification That Alexa350 ® Does Not Alter ␣-Crystallin's Membrane Binding Affinity-To confirm that the Alexa350 ® moiety does not affect or participate in the membrane binding event, we performed control binding experiments in which equal amounts of either conjugated or nonconjugated ␣-crystallin homopolymeric complexes were allowed to bind an equal amount of membrane sample. After the incubation was complete, one-third of each pellet, containing membranes and any bound ␣-crystallin, was run on an SDS-polyacrylamide gel, stained with Coomassie Blue, then analyzed using digital image analysis (ImageQuant version 5.0) to compare the amounts of ␣-crystallin bound with or without the Alexa350 ® attached. Fig. 1 shows a representative gel that demonstrates no measurable difference in binding between the conjugated and non-  2 and 3) were resolved by SDS-PAGE. The membrane sample mainly shows MP26, whereas the amount of endogenous ␣-crystallin is essentially undetectable (labeled arrows indicated along gel). Both ␣Aand ␣B-crystallin subunits were essentially homogeneous. Lanes 4 and 6 show 33% of a pellet resulting from binding assays in which 500 g of membrane was incubated with 16 g of either nonconjugated homopolymeric complex. Lanes 5 and 7 show an equivalent fraction of the pellet resulting from similar binding assays containing Alexa350 ® -conjugated homopolymers. Based on digital image analysis of multiple samples, the amount of ␣-crystallin bound was not affected significantly by Alexa350 ® conjugation, although the protein bands for each conjugate are somewhat more diffuse. Values are shown for conjugation efficiency, fluorescence specific activities, molecular masses, and CLA of Alexa350 conjugates and nonconjugated controls. Conjugation efficiencies and specific activities are average values obtained from six separate preparations of Al-exa350-conjugated proteins. The molecular mass of each complex was estimated from the elution volume of each protein using Superose 6 gel permeation chromatography. CLA measurements were obtained using the insulin reduction assay and then normalized using the nonconjugated ␣-crystallin protein as the reference. NA, not applicable.

␣-Crystallin Membrane Association
conjugated ␣-crystallin proteins (compare lane 4 with 5 and 6 with 7), although a slight blurring effect of the protein band is seen with the Alexa350 ® conjugates. General Membrane Binding Characterization-To understand fully the mechanism by which ␣-crystallin binds to the fiber cell plasma membranes, it is important to elucidate the solution conditions that affect binding. To this end, characterization of the membrane binding event under a variety of conditions was carried out using ␣Aand ␣B-crystallin homopolymers as well as a reconstituted 3:1 heteromeric complex.
To examine whether membrane association is a saturable event, binding assays were performed where the amount of membrane was held constant at 500 g, and the amount of ␣-crystallin complex was varied. In all cases, saturable binding was observed, which then allowed calculation of the binding capacity for these membranes (Fig. 2A). For ␣A-crystallin, a capacity of 6.59 Ϯ 0.13 ng of ␣A/g of membrane was observed, whereas ␣B-crystallin had a lower capacity at 4.23 Ϯ 0.17 ng of ␣B/g of membrane. Interestingly, the binding capacity of the 3:1 heteromeric complex (3.45 Ϯ 0.11 ng of ␣/g of membrane) was lower than either of the homopolymeric complexes. In addition, control binding assays were performed in the presence of either BSA or human ␥D-crystallin and showed no change in the binding capacities of either ␣Aor ␣B-crystallin (Fig. 2B). These results agree with estimates made by quantitative gel analysis, indicating that the quantum yield of the fluorescent probe was not affected significantly by the local environment.
Next, the time dependence of association was examined by incubating 8 g of Alexa350 ® -conjugated ␣-crystallin protein with a fixed amount of plasma membrane at 37°C for up to 16 h. All three protein complexes bind with similar rates, with the calculated t1 ⁄2 values (time to reach 50% completion) of 50 Ϯ 5 min for ␣A-crystallin, 34 Ϯ 5 min for ␣B-crystallin, and 31 Ϯ 4 min for the 3:1 heteromeric complex (Fig. 3A). Although ␣A-crystallin binding does appear slightly slower than the other two, it is considered not significantly different based on 95% confidence analysis of each curve fit.
Dependence upon ionic strength was then determined using a similar series of binding assays for each of the three protein complexes as described above, using NaCl concentrations ranging from 0 to 1 M (Fig. 3B). Our data, normalized with no salt as the reference, show essentially no effect of salt on the membrane association of all three ␣-crystallin complexes, even at 1 M NaCl.
The pH dependence was also analyzed for each protein complex with a series of assays containing 8 g of ␣-crystallin complex and a fixed amount of plasma membranes in the Tri-buffered system at various pH values, with neutral pH serving as the reference. We show that binding is increased markedly at low pH values (Fig. 3C). For the ␣A-crystallin homopolymer, the binding was increased nearly 2-fold, whereas the ␣B-crystallin homopolymer was increased nearly 3-fold. Interestingly, the 3:1 heteromeric complex was affected by low pH to a greater extent than either homopolymeric complex.
Finally, membrane binding was assayed for each complex at a variety of temperatures between 4 and 45°C (Fig. 3D). Binding by both ␣Aand ␣B-crystallin homopolymers shows a high sensitivity to temperature, such that a change from room temperature to 37°C caused a 4-fold increase in binding. The change in binding increases another 50% by raising the temperature to 45°C for the two homopolymers. The 3:1 heteromeric complex, however, shows a much larger increase between 37 and 45°C, similar to the additive effect seen with the pH sensitivity.
Membrane-free control experiments confirmed that the Al-exa350 ® conjugates remained soluble at all times in all experiments described above (data not shown).
pH Dependence of Ionic Strength Sensitivity-To investigate further the increased binding seen at low pH, assays were performed at pH 5.0 using the ␣B-crystallin homopolymeric complex with varied NaCl (0 -1 M) and compared with similar assays performed at pH 7.3 (Fig. 4). At pH values Ͼ6.5, binding of ␣B-crystallin is insensitive to NaCl. However, at pH 5.0, binding is reduced nearly 2-fold in the presence of 1.0 M NaCl.
Reversibility of Membrane Association-To assess the reversibility of membrane binding, we set up a series of identical binding assays using the standard protocol, then challenged the bound Alexa350 ® -conjugated protein using a range of concentrations of nonconjugated ␣-crystallin with and without control proteins in a second incubation. As shown in the inset of Fig. 5, virtually no ␣-crystallin is released into the soluble phase when the membrane-␣-crystallin complex (the pellet) is resuspended in binding buffer alone (the y axis), but addition of nonconjugated ␣-crystallin to the incubation effectively re-

FIG. 2. Binding capacity for ␣-crystallin membrane binding.
Panel A, binding capacity curves. Assays were performed using 500 g of membranes and a varied amount of ␣-crystallin. The amount bound was calculated using the fluorescence of the membrane pellet and the specific activity for each ␣-crystallin complex. The binding capacity (B) was calculated using the asymptote of each curve and was 6.59 Ϯ 0.13 ng/g of membrane for ␣A-crystallin (q), 4.23 Ϯ 0.17 ng/g of membrane for ␣B-crystallin (E), and 3.45 Ϯ 0.11 ng/g of membrane for the 3:1 heteropolymer (ϫ ). (n ϭ 3) Panel B, effects of BSA and ␥D-crystallin. Control assays were performed with 500 g of plasma membranes and 25 g of conjugated ␣-crystallin with and without 25 g of control protein (either BSA or ␥D-crystallin). Samples A-C show the effect of the control proteins on the binding of ␣A-crystallin; samples D-F show their effects on ␣B-crystallin binding. Data were normalized to the "no addition" data. (n ϭ 3).

␣-Crystallin Membrane Association
moved a portion of the membrane-associated ␣-crystallin. However, even at Ͼ50 fold molar excess of nonconjugated ␣-crystallin, only about 55% of the bound Alexa350 ® -conjugated ␣Acrystallin could be removed by competition with nonconjugated ␣A-crystallin, whereas 75% of the bound ␣B-crystallin could be removed with nonconjugated ␣B-crystallin. Addition of control proteins (BSA or ␥D-crystallin) in the second incubation had no effect on release of bound ␣-crystallin (Fig. 5).
␣-Crystallin Membrane Binding Affinity-To calculate the binding constants (K) for each ␣-crystallin complex, we defined the plasma membrane to be the ligand. This facilitated the mathematical treatment of membrane binding as a function of membrane concentration. Therefore, binding assays were performed using a range of membrane amounts (ϳ50 g to greater than 8,000 g) at 37°C. Fig. 6 shows the data obtained in these assays transformed into an affinity plot to illustrate best the binding constant for each complex. The plot for the ␣A-crystallin homopolymer data is horizontal and linear, indicative of noncooperative binding with a K value of 5.9 Ϯ 0.3 ϫ 10 Ϫ4 g Ϫ1 of membrane. The 3:1 heterocomplex data also indicate noncooperative binding with a binding affinity of 4.6 Ϯ 0.5 ϫ 10 Ϫ4 g Ϫ1 of membrane, which is essentially the same as obtained for the ␣A-crystallin homopolymer. However, with the ␣B-crystallin homopolymer the affinity plot is characterized by a sigmoidal curve with positive slope, which indicates apparent positive cooperativity in the binding to plasma membranes. As membrane concentration increases, the affinity, or binding constant, increases approximately 3-fold. Based on the two horizontal asymptotes of this affinity curve, a binding constant, K low , of 5.6 Ϯ 0.5 ϫ 10 Ϫ4 g Ϫ1 of membrane is seen at low membrane concentration, but at high membrane concentrations this changes to 15 Ϯ 3 ϫ 10 Ϫ4 g Ϫ1 of membranes (K high ).
The calculated binding constants for all three proteins are summarized in Table II. This table illustrates that the binding affinity is essentially the same at low concentrations of lens plasma membrane for both homopolymers as well as the heteromeric complex. However, a considerable difference is seen in the binding affinity of ␣B-crystallin at high membrane concentrations.
To determine if the high affinity observed for ␣B-crystallin (K high ) at high membrane concentration correlated with a change in the observed binding capacity (B), assays were performed as before with constant membrane concentration and varied ␣B-crystallin concentration. However, in these experiments, 5,000 g of membrane was used in each sample rather than 500 g (curve not shown). The measured capacity at high membrane concentration (3.1 Ϯ 0.09 ng/g of membrane) was reduced marginally compared with low membrane concentration, 4.2 Ϯ 0.2 ng/g of membrane (Table II).
Control experiments were performed to assess the effect of membranes on the fluorescence measurements to ensure accuracy. By measuring the specific activity of Alexa350 ® conju- gates in the presence of a wide range of membrane concentrations added just prior to each fluorescence reading, a standard curve was created and curve fitted (data not shown). These experiments verified that at membrane concentrations less than 1,000 g (a majority of the reported measurements fall in this category), the fluorescence was not affected significantly by either light scattering or inner filter effects (Ͻ1% error).
However, above 1,000 g of membranes, the measured fluorescence was increased by approximately 2-10%, depending on the amount of membranes (data not shown). To correct for this, each fluorescence measurement containing more than 1,000 g of membranes was adjusted according to the standard curve equation where F is the measured fluorescence, M is the amount of membranes (g), and F c is the corrected fluorescence.

Effects of Membrane Digestion by Trypsin on Membrane
Binding-The role of membrane proteins in mediating ␣-crystallin association with the plasma membrane is uncertain. To evaluate the effects of membrane-associated proteins on ␣-crystallin binding, fractionated plasma membranes were digested with trypsin. Once the digestion was complete, the binding affinity (K t ) and binding capacity (B t ) of both homopolymers and the 3:1 complex for trypsinized membranes were measured.
Saturation curves obtained for trypsinized and nontrypsinized membranes using ␣A, ␣B, and the 3:1 heterocomplex, respectively, are shown in Fig. 7. In the case of ␣Acrystallin (Fig. 7A), little or no significant change in capacity is observed when using the trypsinized membranes (p ϭ 0.813). However, for the ␣B-crystallin homopolymer (p ϭ 0.015) (Fig.  7B) and especially the heteromeric complex (p ϭ 0.016) (Fig.  7C), the binding capacity is changed significantly (summarized in Table II). In fact, the 3:1 complex binding capacity increases by 2-fold when using trypsinized membranes.
The binding affinities (K t ) for trypsinized membranes were determined by holding the amount of ␣-crystallin constant at 8 g and varying the amount of membranes added. In all cases, the binding constants were reduced when using trypsinized membranes ( Fig. 8 and summarized in Table II). For the ␣A homopolymer complex and the 3:1 heteromeric complex, the affinity is reduced by about 50%, from 5.9 Ϯ 0.03 ϫ 10 Ϫ4 g Ϫ1 to 2.9 Ϯ 0.7 ϫ 10 Ϫ4 g Ϫ1 and 4.6 Ϯ 0.5 ϫ 10 Ϫ4 g Ϫ1 to 1.8 Ϯ 0.3 ϫ 10 Ϫ4 g Ϫ1 membrane, respectively (Fig. 8, A and C). However, the most striking result is seen with the ␣B-crystallin homopolymer (Fig. 8B). Without trypsin treatment, the binding affinity increases from 5.6 Ϯ 0.5 ϫ 10 Ϫ4 to 15 Ϯ 3 ϫ 10 Ϫ4 g Ϫ1 of membrane with increasing membrane concentration. However, the apparent positive cooperativity observed with native membrane preparations was lost when the trypsinized membranes were used, giving one binding constant of 2.7 Ϯ 0.2 ϫ 10 Ϫ4 g Ϫ1 of membrane. DISCUSSION ␣-Crystallin is known to associate with lenticular plasma membranes in vivo as well as in vitro, and the amount bound increases with both age and the onset of cataract (29,30). ␣-Crystallin also shows an increasing tendency to become insoluble during aging and/or at the onset of cataract (24 -28). Given this correlation between cataract formation and increased binding of ␣-crystallin to the plasma membrane complex, it is important to understand how this process is regulated in lens cells. In the present report, we have examined some fundamental binding properties of lens ␣Aand ␣B-crystallins.
Human ␣Aand ␣B-crystallin as homopolymers and in a reconstituted 3:1 heteromeric complex show saturable binding to lenticular fiber cell plasma membranes (Fig. 2). These observations are in contrast to previous reports that suggested that ␣A-crystallin but not ␣B-crystallin was the only subunit involved with membrane association (20,23). Our results also show that this interaction is sensitive to time and temperature, FIG. 5. Reversibility of ␣-crystallin membrane binding. Binding assays were performed using a constant membrane concentration and 8 g of Alexa350 ® -conjugated ␣-crystallin protein under standard conditions. Once formed, the ␣-crystallin-membrane complexes were resuspended in binding buffer. For competition experiments, membranebound ␣-crystallin was incubated for Ͼ24 h at 37°C with 100 g of nonconjugated ␣-crystallin and 100 g of control protein (BSA or ␥Dcrystallin) or buffer alone (n ϭ 3). After centrifugation, the ␣-crystallinmembrane complex (pellet) and supernatant were analyzed for fluorescence and compared with the amount bound during the first incubation. There was no observable effect of either BSA or ␥D-crystallin on ␣A-(samples A-C) and ␣B-(samples D-F) crystallin reversibility. In titration experiments (inset), a maximum of 44 Ϯ 2% and 25 Ϯ 1% of bound ␣A-(q) and ␣B-(E) crystallins, respectively, could not be removed after treatment of membranes with up to a 50-fold molar excess of the corresponding nonconjugated protein subunit (n ϭ 2).
FIG. 6. Affinity plots of ␣-crystallin membrane binding. Binding assays were performed using 8 g of conjugated ␣-crystallin with a varied amount of membranes under standard conditions. The apparent binding constant K at each point was calculated as described under "Experimental Procedures." For ␣A-crystallin (q), a linear horizontal plot was observed with a binding constant of 5.89 Ϯ 0.27 ϫ 10 Ϫ4 g Ϫ1 of membrane (n ϭ 5). For ␣B-crystallin (E), a sigmoidal curve was found, with two binding constants, K high ϭ 14.9 Ϯ 2.5 ϫ 10 Ϫ4 g Ϫ1 of membrane and K low ϭ 5.60 Ϯ 0.50 ϫ 10 Ϫ4 g Ϫ1 of membrane (n ϭ 5). The 3:1 heteropolymer (ϫ) has a linear horizontal plot and a binding constant of 4.57 Ϯ 0.50 ϫ 10 Ϫ4 g Ϫ1 of membrane (n ϭ 3). collectively suggesting that binding is specific in nature (Fig. 3,  A and D). In fact, we observed a significant enhancement of binding at temperatures above 37°C. It is unclear whether this increase is due to ␣-crystallin's strong affinity for heat-stressed proteins, presumably including intrinsic membrane proteins, or if it results from a phase transition of lens lipids at high temperatures.
In addition, we show that binding is not affected by high ionic strength at physiological pH. This suggests that binding is mediated mostly by hydrophobic interactions presumably involving either the fatty acid core of the membrane bilayer or the buried hydrophobic regions of the intrinsic membrane proteins still present in the membrane preparation (Fig. 3B). We also demonstrate that binding increases at pH values below pH 7.0, which is in contrast to a previous study in which native ␣-crystallin was reported to have optimum binding at pH 7.5 ( Fig. 3C) (22). We observed a significant enhancement of binding, especially with the reconstituted 3:1 heteromeric complex, at pH 5.0, and this increased binding can be eliminated by increasing the salt concentration. This suggests that the salt sensitivity at low pH is due to either the introduction of electrostatic interactions or the elimination of electrostatic repulsion with some part of the membrane. The degree of enhancement is different for the three ␣-crystallin complexes, which further implies that the ionizable group is within the protein complex itself rather than on the membrane. In addition, the point at which pH sensitivity is observed for ␣Aand ␣Bcrystallin can be predicted by their theoretical isoelectric pH values of 5.77 and 6.76, respectively. Despite the sensitivity to ionic strength at low pH, we believe that the interaction between ␣-crystallin and the plasma membrane is mediated mostly through hydrophobic interactions rather than electrostatic interactions in vivo.
Our data also suggest that intact ␣-crystallin complexes are not in equilibrium between the membrane-bound and soluble phases. If this were the case, Ͼ99% removal of the bound fluorescent ␣-crystallin should be possible when competed with a large excess of nonconjugated ␣-crystallin. Rather, it appears that once on the membrane, some fraction of the subunits is not exchangeable with subunits comprising soluble ␣-crystallin complexes (Fig. 5). We propose that the conjugated ␣-crystallin removed by competition with unlabeled ␣-crystallin reflects exchange of subunits between Alexa350 ® -conjugated complexes associated with the membrane and nonconjugated complexes in solution. According to this model, the nonexchangeable subunits become buried in the membrane bilayer such that they are irreversibly associated. Another interesting observation is that different percentages of ␣Aand ␣B-crystallin are removable. If subunit exchange is the mechanism by which fluorescent ␣-crystallin is being released from the membrane, then these results suggest that bound ␣Aand ␣B-crystallin have a different number of subunits in direct contact with the membrane. This could occur if either the quaternary structures of the complexes are significantly different while bound or if ␣A-crystallin simply embeds itself deeper into the membrane than ␣B-crystallin, resulting in fewer exchangeable subunits.
It is difficult to compare the calculated binding capacities of each homopolymer and the reconstituted 3:1 heteromeric com-  7. Effect of membrane trypsinization on ␣-crystallin binding capacity. Plasma membranes were treated with trypsin for 3 h at 37°C. These membranes were then used in binding assays in which the amount of membrane was held constant at 500 g, and the ␣-crystallin concentration was varied from 0 to 64 g. For all three graphs, n ϭ 3; E, trypsinized membranes; and q, nontreated membranes. Panel A, ␣A-crystallin; panel B, ␣B-crystallin; panel C, 3:1 heteropolymer.
plex for lens plasma membrane in the present study with those reported previously. In protein-free synthetic vesicles containing 50 mol % cholesterol, the binding capacity was approximately 40 ng of native ␣-crystallin/g of lipid (33). In addition, Ifeanyi and Takemoto (20) measured a binding capacity of 44 ng/g of membrane protein using purified lens ␣A-crystallin and lens membranes prepared in a manner similar to that used in the present study but quantified by the Bradford dye binding assay rather than total dry weight. Using the Bradford assay on the membranes used in the present study, we calculate a binding capacity of 87 ng of ␣A-crystallin/g of membrane protein (data not shown). Because previous studies utilized ␣-crystallins purified from adult cow lenses, differences in binding capacity could reflect post-translational changes known to accumulate in lens proteins which would not be represented in the recombinant human proteins used in the present study (43)(44)(45).
Trypsinization of the membranes was performed to determine the relative contribution of exposed protein domains on the binding of ␣-crystallin to the membrane. Previous studies have shown that brief treatment of lens membrane preparations with trypsin results in cleavage of the presumed cytoplasmic loops of MP26, the major intrinsic membrane protein (46). With both ␣B-crystallin and especially with the 3:1 heteromeric complex, the binding capacity to trypsinized membranes was increased markedly (Fig. 7, B and C), suggesting that exposed intrinsic membrane protein domains actually decrease the amount of ␣-crystallin that can bind to the membrane. As membrane proteins in lens fiber cells are thought to contribute a large percentage of the lens membrane mass, we consider it likely that the apparent increase in binding capacity of trypsinized compared with untreated membranes reflects enhanced access of ␣-crystallin complexes to the lipid bilayer, although we cannot rule out the possibility that trypsinization creates damaged proteins to which ␣-crystallin could potentially bind. Given the insensitivity of ␣-crystallin binding to increased ionic strength at physiologic pH, the apparent enhancement of binding capacity following trypsinization of membranes, and the published reports on the saturable association of ␣-crystallin with protein-free synthetic phospholipid vesicles, we postulate that ␣-crystallin binding to the membrane occurs predominantly through hydrophobic interactions rather than through protein-protein interactions as proposed previously (22,(31)(32)(33)47).
For each homopolymer and the 3:1 heteropolymer, binding constants were determined to measure the affinity of the interaction. All three complexes show similar affinity at low membrane concentration; however, the ␣B homopolymer shows a marked increase in apparent affinity at high membrane concentrations ( Fig. 6 and Table II). It is possible that this apparent positive cooperativity results from polysteric linkage, whereby the binding of a ligand influences the aggregation state of the macromolecule (48). In the case of ␣B-crystallin, it could be that binding to the membrane alters its quaternary structure such that additional ␣B subunits can insert into the existing bound complexes. Another possible explanation for the positive cooperativity behavior is that at high membrane concentrations, interactions between ␣B-crystallin and some other intrinsic membrane protein begin to add to the interactions between ␣B and the lipid bilayer. Consistent with this idea, we observed that the affinity plot of ␣B-crystallin measured with trypsinized membranes did not show the same positive cooperativity as seen with the nontreated membranes (Fig. 8B). Therefore we believe that the apparent increase in binding affinity seen with ␣B-crystallin at high membrane concentrations likely reflects the additive effects of protein-lipid and protein-protein interactions with MP26 or other membrane proteins.
One unexpected trend in our data was the properties of the reconstituted 3:1 heteromeric complex. In the cases of the pH and temperature dependence (Fig. 3, C and D), this heteropolymer did not behave as an average of the component subunits, as would be expected if they were functionally and structurally equivalent. Indeed, the heteromeric complex had the lowest binding capacity measured with the fractionated lens membranes, and this capacity was the most dramatically affected by trypsinization of the membrane (2-fold increase) (Fig. 7C). These results support the hypothesis that the mixed aggregate may hold particular functional significance in the lens. Based on the observation that enhanced membrane association is correlated with aging and the onset of cataract, it is reasonable to suggest that irreversible membrane binding is an event that negatively affects the transparency of the lens. By virtue of its relative resistance to membrane binding, the heteromeric complex may represent a more favorable state than homomeric complexes derived from either ␣Aand ␣B-crystallin subunits. The formation of cataract in the ␣A-crystallin knockout mouse lens adds support to this hypothesis (8).
Our studies suggest that ␣-crystallin is able to bind membranes through hydrophobic interactions in such a way as to embed a portion of the complex into the hydrophobic fatty acid core of the bilayer. This interaction appears to be irreversible for the subunits in direct contact with the membrane, but it does not appear to prevent subunit exchange between the exposed subunits with soluble ␣-crystallin aggregates. The present data suggest particular biological significance for the heteromeric complex comprised of both ␣Aand ␣B-crystallin, as found in the lens.