Caspase Remodeling of the Spectrin Membrane Skeleton during Lens Development and Aging*

Terminal differentiation of lens fiber cells resembles the apoptotic process in that organelles are lost, DNA is fragmented, and changes in membrane morphology occur. However, unlike classically apoptotic cells, which are disintegrated by membrane blebbing and vesiculation, aging lens fiber cells are compressed into the center of the lens, where they undergo cell-cell fusion and the formation of specialized membrane interdigitations. In classically apoptotic cells, caspase cleavage of the cytoskeletal protein α-spectrin to ∼150-kDa fragments is believed to be important for membrane blebbing. We report that caspase(s) cleave α-spectrin to ∼150-kDa fragments and β-spectrin to ∼120- and ∼80-kDa fragments during late embryonic chick lens development. These fragments continue to accumulate with age so that in the oldest fiber cells of the adult lens, most, if not all, of the spectrin is cleaved to discrete fragments. Thus, unlike classical apoptosis, where caspase-cleaved spectrin is short lived, lens fiber cells contain spectrin fragments that appear to be stable for the lifetime of the organism. Moreover, fragmentation of spectrin results in reduced membrane association and thus may lead to permanent remodeling of the membrane skeleton. Partial and specific proteolysis of membrane skeleton components by caspases may be important for age-related membrane changes in the lens.

The spectrin-actin membrane skeleton underlies the plasma membranes of all cells and is important for cellular shape, membrane stability and deformability, as well as the formation of membrane subdomains (1). The major component of the membrane skeleton, spectrin, is composed of an ␣/␤ heterodimer that self-associates head-to-head to form a 200-nm extended tetramer filament. Spectrin cross-links actin filaments into an isotropic meshwork. This spectrin-actin meshwork is attached to the membrane by direct interactions of ␤-spectrin with membrane proteins and indirect interactions of ␤-spectrin with membrane attachment proteins such as ankyrin (2).
Proteolysis of ␣-spectrin (␣II-spectrin, non-erythroid spectrin, or fodrin) to discrete fragments is implicated in changes in cell shape and membrane morphology which occur in many cell types. During platelet activation, which includes a cell shape transformation from discs into irregular spheres, spectrin is cleaved to ϳ150-kDa fragments by the calcium-dependent protease, calpain (3). ␣-Spectrin cleavage by calpain has also been implicated in cellular hypoxia (4), neuronal injury and degeneration (5), and neuronal growth cone formation (6). However, in apoptotic cells, ␣-spectrin proteolysis to ϳ150-kDa fragments is mediated by caspases; in these cells, spectrin proteolysis is thought to be important for the disintegration of the plasma membranes via formation of vesicular "apoptotic bodies" (7)(8)(9)(10)(11)(12). Although calpain cleavage of spectrin is known to affect its ability to bind membranes or actin filaments (13,14), the detailed consequences of caspase cleavage of spectrin have not been studied.
The terminal differentiation and aging of lens fiber cells are marked by dramatic membrane morphological changes. As new cells arise on the outside of the lens, older cells are pushed inward, where organelles are lost, and cells fuse to form a syncytium (15,16). Because these cells are never lost from the lens, the most central cells are as old as the organism. In addition, as lens fiber cells mature, specialized membrane interdigitations develop, which are distributed regularly on the lateral membranes (17). These age-related changes in membrane morphology have been compared with the formation of apoptotic bodies (18). In addition, they are believed to be important for lens transparency by reducing light scattering at cell boundaries and by allowing for protein turnover and ion homeostasis (16).
We have shown previously that spectrin and other components of the membrane skeleton are associated with the plasma membranes of young and old lens fiber cells (19). Here, we report that ␣-spectrin is cleaved to ϳ150-kDa fragments during terminal differentiation and aging of lens fiber cells. In addition, ␤-spectrin, which dimerizes with ␣-spectrin, is also proteolyzed to ϳ120and ϳ80-kDa fragments. Fragmentation of spectrin progresses with lens and fiber cell age so that in the oldest fiber cells of the adult lens, most, if not all, of the spectrin is fragmented. The spectrin-binding and membrane-binding protein, ankyrin, is also partially proteolyzed with lens fiber cell age. N-terminal amino acid sequencing of the spectrin fragments reveals that caspase cleavage is responsible for lens spectrin proteolysis. Moreover, subcellular fractionation of lens fiber cells indicates that caspase-cleaved spectrin fragments display reduced association with lens membranes. The specific proteolysis of membrane skeleton components by caspases may be important for age-related membrane changes in the lens.

EXPERIMENTAL PROCEDURES
Antibodies-Affinity-purified rabbit polyclonal antibodies to bovine brain ␣-spectrin (fodrin) were prepared as described (R6017) (20). Rabbit polyclonal antibodies (pAb10D) were raised against a recombinant peptide representing residues 1676 -2204 of human ␤II-spectrin, span-* This work was supported in part by National Institutes of Health Grants EY10814 (to V. M. F.) and NS32578 and DK43812 (to J. S. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  ning from repeat unit 13 to the COOH terminus (7,21). Monoclonal antibodies to actin (C4) were a generous gift from J. Lessard (Children's Hospital Research Foundation, Cincinnati, OH). Rabbit polyclonal antibodies to repeats 13-24 in the N-terminal ankyrin repeats domain of human red blood cell ankyrin were a generous gift from P. S. Low (Purdue University, West Lafayette, IN).
Isolation and Subcellular Fractionation of Lens Cells-Whole lenses were harvested from chickens or rats, and contaminating ciliary epithelium was removed by careful dissection. For Fig. 3, cortical fiber cells were isolated from nuclear fiber cells, which were compacted more tightly, by dissection of adult chicken lenses (6 -8 weeks old). Using fine forceps on the nuclear fiber cells, the outer nuclear fiber cells were peeled away in layers from the inner nuclear fiber cells. Whole lenses or isolated fiber cells were rinsed in phosphate-buffered saline with 10 mM EGTA and homogenized in lens buffer (100 mM NaCl, 25 mM Hepes, pH 7.4, 4 mM MgCl 2 , 10 mM EGTA, 1 mM dithiothreitol) at 30 mg/ml, using a Dounce homogenizer (8 -10 strokes with the tight pestle). Subcellular fractionation was performed as described in Ref. 19. Briefly, lens fiber cell homogenates were centrifuged at 30,000 ϫ g for 20 min at 4°C to separate the cytosol supernatant from the membrane pellet. The pellet was washed by two more rounds of resuspension and centrifugation to prepare washed membranes, which were either resuspended in lens buffer or extracted in lens buffer with 1% Triton for 1 h on ice. Triton extracts were subsequently centrifuged at 30,000 ϫ g for 20 min at 4°C (supernatant 2 and pellet 2) (see Fig. 4).
Electrophoresis and Western Blotting-SDS-polyacrylamide gel electrophoresis was performed on large pore 10% polyacrylamide gels according to Dreyfuss et al. (22), and gels were transferred to nitrocellulose in Tris-glycine transfer buffer (23) with the addition of 0.01% SDS and the omission of methanol (this was optimal for efficient transfer of spectrin). Broad range molecular mass standards were purchased from Bio-Rad. The relative mobility of each fragment was estimated based on standard R F analysis. Unless otherwise indicated, Western blotting was performed as described (24) except that antibodies were detected with protein A-horseradish peroxidase (Sigma) followed by standard chemiluminescence detection methods. To reprobe blots using a different antibody, blots were first stripped in 62.5 mM Tris, pH 6.5, 2% SDS, and 100 mM 2-mercaptoethanol at 65°C for 30 min. Stripped blots were washed extensively with phosphate-buffered saline with 1% Triton and then blocked and probed as usual. For the quantitation of the ratio of full-length ␣-spectrin to fragments, autoradiographic films were scanned into NIH Image. The total number of pixels in each band was quantified in arbitrary units. To correct for nonspecific binding, pixels from an unlabeled part of the film were subtracted. Similar results were obtained from direct labeling of the blots with 125 I-protein A followed by ␥-counting (data not shown).
Purification of Fragments and N-terminal Sequencing-Adult chicken lenses (200) were homogenized using a Dounce homogenizer at 500 mg/ml in 10 mM NaHPO 4 , pH 7.4, 100 mM KCl, 5 mM EDTA, 5 mM EGTA, 0.5 mM dithiothreitol (lens buffer). Homogenates were centrifuged at 30,000 ϫ g to obtain a membrane pellet (water-insoluble fraction), which was washed twice in lens buffer and subsequently extracted in 10 mM NaHPO 4 , pH 7.4, 1.5 M KCl, 5 mM EDTA, 5 mM EGTA, 0.5% Triton, 0.5 mM dithiothreitol. Extracts were then centrifuged at 30,000 ϫ g, and the supernatant, containing fragments and full-length spectrin, was retained (high salt extract). The high salt extract was dialyzed into 10 mM Tris, pH 8.0, 20 mM NaCl, 5 mM EDTA, 5 mM EGTA, 1 mM dithiothreitol and then loaded onto a 5-ml Resource Q anion exchange column (Amersham Pharmacia Biotech). Bound proteins were eluted using a 94-ml linear gradient from 20 to 500 mM NaCl; the elution position of fragments was determined by SDS-polyacrylamide gel electrophoresis and Coomassie staining. Fractions enriched in ␣-spectrin fragments were selected, and the proteins were trichloroacetic acid precipitated, and run on two-dimensional gels (25), followed by transfer to polyvinylidene difluoride membranes in 10 mM CAPS, 1 10% methanol, 0.01% SDS. Anion exchange fractions containing the ␤-spectrin 120-kDa fragment were pooled separately, trichloroacetic acid precipitated, solubilized in 0.1% SDS (with 20 mM Tris, pH 8.0, 2.5 mM EDTA, 2.5 mM EGTA), and loaded onto a Sepharose CL-4B gel filtration column. Bound proteins were loaded and eluted in the presence of 0.1% SDS as described in Ref. 26. Fractions enriched in the ␤-spectrin 120-kDa fragment were loaded on a one-dimensional SDSpolyacrylamide gel and transferred to polyvinylidene difluoride membranes. Proteins were eluted from polyvinylidene difluoride and subjected to standard N-terminal sequencing by J. Leszyk (University of Massachusetts Medical School, Shrewsbury, MA). Molecular masses of fragments were estimated based on amino acid composition using the ProtParam tool of the ExPASY proteomics server (59).

RESULTS
Spectrin Is Cleaved to Discrete Fragments during Lens Development and Aging-We have shown previously that ␣-spectrin is partially proteolyzed to three ϳ150 -160-kDa fragments in the nuclear (oldest) fiber cells of the adult chicken lens (19). These results suggested either that spectrin fragmentation occurs only during terminal differentiation (i.e. organelle loss) of fiber cells in the embryonic lens or that spectrin fragmentation progresses with aging of fiber cells in post-hatched lenses.
To determine when ␣-spectrin proteolysis occurs, Western blotting was performed on proteins from chicken lenses at different stages of embryonic and post-hatched development (Fig. 1A). Of particular interest were the day 6 embryo, when the lens first appears transparent (27), the day 8 -10 embryo, when organelle breakdown begins (28), and the day 12 embryo, when an organelle-free zone is first observed. When non-crystallin proteins were loaded equivalently (Fig. 1A, actin), ␣-spectrin fragments were barely detected at day 17 of embryonic lens development; moreover, no fragments smaller than ϳ150 kDa were detected. Strikingly, scanning quantitation of blots indicated that the ratio of ␣-spectrin fragments to full-length ␣-spectrin increased dramatically with age after hatching (Fig.  1B). In the representative experiment shown, fragments were first detected at day 17 of embryonic development (Fig. 1A). However, when proteins from twice as much tissue (wet weight) were loaded from day 12 embryonic lens proteins, ␣-spectrin fragments were barely detected (data not shown). Nonetheless, fragments of ␣-spectrin were not detected from samples of day 6, 9, or 10 embryonic lens proteins, even when 3 ϫ wet weight of tissue was loaded (data not shown). These results suggest that ␣-spectrin fragmentation may be initiated simultaneously with the formation of an organelle-free zone. However, the extent of fragmentation of ␣-spectrin continues to increase in post-hatched lenses relative to late embryonic lenses, suggesting that proteolytic processing of ␣-spectrin progresses with lens age.
To determine whether ␤-spectrin, which forms a heterodimer with ␣-spectrin in most mammalian cells, is also proteolyzed, blots were stripped and reprobed with a ␤-spectrin antibody (7, 21) (Fig. 1A). Two fragments were detected with the ␤-spectrin antibody, which was raised against a peptide from the C-terminal half of the protein (see "Experimental Procedures"). A ϳ120-kDa fragment first appeared in day 17 embryonic lenses, when ␣-spectrin fragments were first detected. In contrast, a smaller ϳ80-kDa fragment was not detectable in day 17 embryonic lenses and was first detected in older lenses. These results suggest either that the ϳ80-kDa fragment arises from cleavage at a site that is digested less readily than the site for the ϳ120-kDa fragment or that the ϳ80-kDa fragment arises from cleavage of the ϳ120-kDa fragment. An additional band at ϳ70 kDa was also barely detectable in the 18-month lens. We also observed that, similar to ␣-spectrin, more ␤-spectrin fragments relative to full-length were observed with increasing age of the lens.
Fragments of spectrin were also found to be present in the rat lens (Fig. 2). As for the chicken lens, the amount of fragmentation increased progressively with the postnatal age of the rat lens. The ratio of intensities of ␣-spectrin fragments to full-length ␣-spectrin at postnatal day 3 in the rat lens was similar to that of the day 17 embryonic chicken lens (data not shown; see Fig. 1A); this is consistent with the later development of an organelle-free zone in rodent lenses as compared with chicken lenses (28,29). However, in addition to 150 -160-kDa ␣-spectrin fragments, an additional ␣-spectrin fragment at ϳ110 kDa was found in the adult rat lens. Moreover, in addi-tion to ϳ120and ϳ80-kDa ␤-spectrin fragments, another fragment at ϳ50-kDa was also detected. Similar-sized ␣-spectrin fragments were also observed in the newborn bovine lens (data not shown). No ␣-spectrin fragments smaller than ϳ110 kDa or ␤-spectrin fragments smaller than ϳ50 kDa were detected in bovine or rat lenses (data not shown). These results indicate that both ␣and ␤-spectrin are cleaved to discrete fragments during development and aging of the chicken, rat, and bovine lens.
Spectrin Is Completely Cleaved to Discrete Fragments in the Oldest Fiber Cells of the Adult Lens-The progressive increase in spectrin fragmentation in older chicken and rat lenses suggested that spectrin might be fragmented progressively with the age of the lens fiber cell. To investigate this possibility, we compared the amount of spectrin fragmentation in fiber cells of different ages in the 6 -8-week-old adult chicken lens (Fig. 3). The youngest (ϳ0 -4 weeks old), newly differentiated cortical fiber cells constitute the outermost shell of the lens fiber cell mass, and are easily peeled away from the remaining, older (nuclear) fiber cells. The outer nuclear fiber cells (ϳ4 -6 weeks old) were then separated from the inner nuclear fiber cells, the oldest fiber cells of the lens. Thus, for this experiment, the inner nuclear cells were ϳ6 -8 weeks old. We estimate that our inner nuclear preparation (ϳ 1-mm diameter core) contained ϳ8,000 cells out of 1.5 million total lens cells (30) based on the average cross-sectional area of the human embryonic nuclear fiber cell (31).
As expected, in newly differentiated fiber cells, spectrin fragments were barely detectable relative to full-length by Western blotting (Fig. 3). In older fiber cells (outer nuclear) the ratio of fragments to full-length spectrin was ϳ1:1. In contrast, fulllength ␣or ␤-spectrin was barely detectable in the oldest, inner nuclear fiber cells (Fig. 3, left and middle panels). Strikingly, ϳ150-kDa ␣-spectrin and ϳ120-kDa ␤-spectrin fragments were easily detectable in the inner nuclear cells, suggesting that spectrin fragments were not further degraded and lost from the lens. In some inner nuclear preparations, which presumably contained a smaller core of central fiber cells, only spectrin fragments and no full-length spectrin at all could be detected by Western blotting or silver staining (data not FIG. 1. ␣and ␤-spectrin are cleaved to discrete fragments during development and aging of the chicken lens. A, Western blotting of lens proteins harvested from chickens at different days of embryonic development (day 6, 9, 10, 12, 15, or 17) or ages after hatching (1 day, 6 weeks, 5.5 months, 18 months). ␣-Spectrin is partially proteolyzed to ϳ150 -160-kDa fragments in the older embryonic and posthatched chicken lens. Simultaneously, ␤-spectrin fragments at ϳ120 and ϳ80 kDa are also detected during late embryonic development. Western blotting for actin indicates that lanes were loaded relatively evenly for non-crystallin proteins (the concentration of non-crystallin proteins decreases with lens age (56)). Wet weight of tissue loaded is as follows: 0.7 mg (6E, 9E, 10E, 12E), 1.2 mg (15E, 17E, 1D, 6 weeks, 5.5 months), and 0.92 mg (18 months). B, NIH Image quantitation of the ratio of intensities of the full-length ␣-spectrin band relative to the intensities of the ␣-spectrin fragments. The average and standard error of two experiments are shown. Comparison of Coomassie staining (data not shown) and Western blotting indicates that our transfer and Western blotting procedures slightly overestimate the amount of fragment to full-length by 1.1-fold.
FIG. 2. Spectrin fragmentation occurs during rat lens development and aging. Western blotting of lens proteins from postnatal day 3, day 10, and adult (6 weeks) rats using antibodies raised against ␣and ␤-spectrin is shown. Note the presence of an ␣-spectrin fragment at 110 kDa. ␣-Spectrin is partially proteolyzed to one ϳ120-kDa and three ϳ150 -160-kDa fragments. Simultaneously, ␤-spectrin fragments at ϳ120, ϳ80, and ϳ50 kDa are also detected during late embryonic development.
shown). In contrast, in cortical fiber cells, Western blotting revealed that the majority of the ␣and ␤-spectrin was fulllength ( Fig. 3; also see Ref. 19). Although it is technically difficult to isolate the undifferentiated epithelial cell layer from adult chicken lenses, in the undifferentiated epithelial cells of day 15 embryonic lenses, most of the ␣-spectrin was full-length (data not shown).
To investigate whether the spectrin-binding and membranebinding protein, ankyrin, is also proteolyzed during lens fiber cell aging, Western blotting was performed on cortical and nuclear fiber cells of the adult chicken lens (Fig. 3, right panel). In both cortical and nuclear fiber cells, several ankyrin antibodies recognized bands corresponding to the full-length ankyrin (220 kDa). However, in nuclear fiber cells, an additional doublet at ϳ190 kDa was recognized by antibodies raised against brain ankyrin, human erythrocyte ankyrin (data not shown), or an N-terminal peptide of human erythrocyte ankyrin (Fig. 3, right panel). These results suggest that ankyrin is cleaved during lens fiber cell aging. Furthermore, because the antibodies raised against an N-terminal peptide of ankyrin recognize the 190-kDa cleavage product, the cleavage site is likely to be located near the C terminus.
These results indicate that spectrin and ankyrin are progressively cleaved during fiber cell aging. Moreover, the oldest fiber cells of the adult lens are likely to contain only spectrin fragments and no full-length spectrin.
A Caspase Is Responsible for Lens Spectrin Cleavage-Both calpains and caspases have been shown to cleave ␣-spectrin to ϳ150-kDa fragments in vitro (7,32). However, the lens ␤-spectrin fragment sizes are similar to that reported for in vitro caspase cleavage of bovine brain ␤-spectrin (ϳ110 and ϳ85 kDa) (7). In contrast, calpain cleavage of ␤-spectrin results in 165-, 125-, and 120-kDa fragments (32). The sequences of caspase-3 cleavage sites of ␣IIand ␤II-spectrin have been identified (7) and are distinct from the sites of calpain cleavage of ␣IIand ␤II-spectrin (32).
To determine directly whether a caspase is responsible for lens spectrin proteolysis, spectrin fragments were purified from adult chicken lens membranes (see "Experimental Proce-dures") and subjected to N-terminal sequencing (Fig. 4B). No sequence could be obtained from the ␣-spectrin 160-kDa fragment, suggesting that this fragment contained the blocked N terminus (7,32). The N-terminal sequence of both of the smaller ␣-spectrin fragments (␣155 and ␣150) mapped to the sequence DETD 1185 *SKTASP (with * representing the site of cleavage and the beginning of the amino acid sequence obtained). This sequence is the previously reported site of cleavage of bovine brain ␣II-spectrin by caspase-3 in vitro (7). However, this amino acid sequence differs slightly from the GenBank sequence for chicken ␣-spectrin, which reports the sequence S 1186 KTVSP (33). Reverse transcription-polymerase chain reaction amplification of chicken stomach and lens spectrin cDNA followed by DNA sequencing confirmed that the correct DNA and amino acid sequences, respectively, for that region are TCTAAGACAGCCSTCTCCT 3489 and KTASP 1190 . Interestingly, the caspase cleavage site in ␣-spectrin is located in repeat 11, just nine residues upstream of the calmodulin binding site (32) (Fig. 4B).
The N-terminal sequence of the ␤-spectrin 120-kDa fragment was XKRLTVEKKFLE (Fig. 4A). Although the DNA or amino acid sequence of chicken ␤II-spectrin is not known, this Nterminal sequence of the 120-kDa fragment is homologous to the human ␤II-spectrin sequence DEVD 1457 *SKRLTVQT-KFME in repeat 11 (34). Importantly, this sequence is also the same site as that reported for caspase-3 cleavage of bovine brain ␤II-spectrin in vitro (7). Interestingly, this site was not located near regions important for ankyrin, ankyrin-independent membrane (MAD), or actin filament binding (1) (Fig. 4B). Unfortunately, we were unable to obtain sufficient amounts of the ␤-spectrin 80-kDa fragment for N-terminal sequencing.
In conclusion, the cleavage sites of lens spectrin, as well as the sizes of the ␣and ␤-spectrin fragments, indicate that a caspase is responsible for spectrin cleavage in the lens. In particular, the sequences of the cleavage sites strongly implicate caspase-3.
Spectrin Fragments Are Partially Dissociated from Lens Plasma Membranes-To investigate the biochemical consequences of caspase fragmentation of lens spectrin, subcellular FIG. 3. Spectrin is cleaved to discrete fragments during lens fiber cell aging. Left and center panels, Western blotting for spectrin of lens proteins from cortical (C), outer nuclear (oN), and inner nuclear (iN) fiber cells of the adult chicken lens (6 -8 weeks). We were unable to use actin as a loading control because the high amounts of ␦-crystallin in inner nuclear fiber cells interfered with the detection of actin. However, proteins corresponding to 0.6 mg of cortex (1ϫ), 1.2 or 1.7 mg of outer nucleus (2ϫ, 3ϫ), and 1.2 mg or 4.7 mg of inner nucleus (2ϫ, 7ϫ) were loaded. These are roughly equivalent estimates of non-crystallin protein because ␦-crystallin protein concentration increases (and thus, non-crystallin concentration decreases) by ϳ3-fold from cortex to outer nucleus and an additional ϳ1.5-fold from outer nucleus to inner nucleus (57). Moreover, the ratio of intensities of the full-length to fragment bands for each sample did not change when different amounts of sample were loaded. This experiment was repeated three times with similar results; a representative experiment is shown. Right panel, Western blotting for ankyrin of lens proteins from cortical and nuclear fiber cells of the adult chicken lens (6 -8 weeks). fractionation of adult lens fiber cells was performed (Fig. 5). Immunoblotting of cytosol and membrane fractions indicated that ␣-spectrin fragments were considerably more abundant in the 30,000 ϫ g supernatant fractions (66.3 Ϯ 5.97% in S1) than in the membrane pellet. In contrast, full-length ␣-spectrin was more abundant in the membrane pellet (only 15.6 Ϯ 10.0% in S1), as expected (19,35). Similarly, the majority of ␤-spectrin fragments remained in the 30,000 ϫ g supernatant (58.1 Ϯ 3.82% in S1), whereas full-length ␤-spectrin was more abundant in the membrane pellet (only 16.8 Ϯ 2.56% in S1), as shown previously (19,35). However, although all of the ϳ80-kDa ␤-spectrin fragment remained in the supernatant (99.5 Ϯ 1.53% in S1), a significant proportion (ϳ40%) of the ϳ120-kDa ␤-spectrin fragment was pelleted (58.1 Ϯ 3.82% in S1). This difference in membrane association of the two fragments may be a result of the 80-kDa fragment missing the C-terminal membrane association domain (MAD2), as suggested in Fig. 4B.
However, the small proportion of ␣and ␤-spectrin fragments that did pellet with the membranes appeared to be tightly associated with the membranes. These fragments were not extracted in 1% Triton (S2) and were only partially extracted by 1 M NaCl or by 1% Triton with 1 M NaCl (data not shown). Thus, spectrin fragments that are membrane-associated might constitute a different population from the cytosolic fragments.
Interestingly, all three of the ␣-spectrin fragments in the supernatant appeared to fractionate together on gel filtration (data not shown), suggesting that they were tightly associated in a complex. On the other hand, the ␤-spectrin fragments did not fractionate together, suggesting that they had dissociated from one another and from the remaining full-length ␣-spectrin. Thus, our results collectively indicate that caspase fragmentation of spectrin leads to partial dissociation of both spectrin subunits from lens membranes. DISCUSSION This is the first report of caspase cleavage of membrane skeleton proteins to discrete and stable fragments during cellular maturation and aging. We have shown that ␣-spectrin is cleaved to ϳ150-kDa fragments and ␤-spectrin to ϳ120and ϳ80-kDa fragments during lens fiber cell maturation and aging. The spectrin-binding protein, ankyrin, is also cleaved to ϳ190-kDa fragments. These fragments appear to be extremely FIG. 5. Spectrin fragmentation results in partial dissociation from lens membranes. A, adult chicken lenses (6 -8 weeks) were homogenized in a physiological buffer followed by centrifugation at 30,000 ϫ g to obtain cytosol-enriched supernatants (S1) and membrane pellets (P1). Membranes were then extracted to obtain 1% Tritonsoluble (S2) extracts and Triton-insoluble pellets (P2). Proteins were subjected to Western blotting with anti-␣and anti-␤-spectrin antibodies. B, histogram indicates fraction of total full-length spectrin or spectrin fragments in S1. The percentage of soluble spectrin or spectrin fragments was calculated by dividing the amounts in S1 by the sum of S1 and P1. For ␣-spectrin and ␤-spectrin, respectively, the average percentage and standard deviation of six or three experiments is shown .   FIG. 4. N-terminal sequencing indicates that a caspase is responsible for lens spectrin cleavage. A, ␣and ␤-spectrin fragments were purified from adult chicken lens membranes and subjected to N-terminal sequencing. For ␣-spectrin, the sequence across the predicted cleavage site was determined from reverse transcription-polymerase chain reaction followed by DNA sequencing of chicken lens and stomach ␣-spectrin RNA. For ␤-spectrin, the human (34) and bovine (7) sequences were used. B, caspase cleavage sites for ␣and ␤-spectrin are indicated. Depicted is the domain structure of spectrin, with boxes identifying each ϳ106-residue repeat. The placement of unknown cleavage sites (?) was estimated based on the fragment size and known sites for caspase-3 digestion of spectrin in vitro (7). The location of binding sites for calmodulin (CaM), ankyrin, actin, and ankyrin-independent membrane binding sites 1-3 (MAD1, 2, 3) are also shown (1). The ␤-spectrin antibody used (pAb10D) was raised against a peptide spanning repeat 13 to the COOH terminus (7). stable and indeed, accumulate with age. In contrast, in other cell types, caspase cleavage of spectrin precedes cell death, thus ensuring a short half-life for the spectrin fragments (7)(8)(9)(10)(11)(12). Moreover, when cleavage of spectrin does not lead to cell death, (i.e. neuronal remodeling), a brief period of accumulation of fragments is followed by a decrease in the proportion of fragments to full-length spectrin polypeptides, suggesting replacement of spectrin fragments with newly synthesized full-length spectrin (36). Our results indicate that in lens fiber cells, which do not undergo cell death, the accumulation of caspase cleavage products of spectrin and ankyrin may lead to a permanent remodeling of the membrane skeleton.
The cleavage of membrane skeleton components in the lens is specific. In contrast to spectrin and ankyrin, other components of the membrane skeleton, tropomodulin, tropomyosin, and actin, do not appear to be proteolyzed during lens fiber cell aging (19). However, there are reports of other membraneassociated lens proteins being proteolyzed to discrete and stable fragments. Partial proteolysis of the major lens membrane protein, MP-26 from a ϳ26-kDa to a ϳ20-kDa protein, has been reported to occur during fiber cell maturation (37). The gap junction protein connexin 50 (␣8) has been shown to be cleaved by calpain, leading to removal of the C-terminal tail of the protein from the plasma membranes (38). Connexin 46 (␣3) may also be cleaved in mature fiber cells (39). The lens-specific intermediate filament protein filensin is also proteolyzed to a discrete ϳ53-kDa band in maturing fiber cells in the bovine lens (40) and multiple bands in the chicken lens (19). Finally, ϳ150-kDa ␣-spectrin fragments have been observed in the rabbit and guinea pig lens (41) in addition to our observations in the chicken, rat, and cow lens. Partial proteolysis of particular membrane-associated proteins (and not others) to discrete and stable fragments may be important for membrane remodeling during lens fiber cell aging.
The timing of spectrin fragmentation also implies roles in mediating membrane morphological changes during lens fiber cell aging. Spectrin fragments are first detected late during embryonic development, coincident with formation of the organelle-free zone (28), but continue to accumulate during development and after hatching. In the adult lens, spectrin fragmentation appears to be restricted to older fiber cells, and the amount of spectrin fragmentation increases with the age of the lens fiber cell. Unfortunately, it is difficult to compare the timing of spectrin fragmentation with that of other membraneassociated proteins because a similar developmental analysis has not been performed with other proteins. However, a functional syncytium of cells within the organelle-free zone is first detected at day 12 of embryonic development and expands with age (16). In addition, membrane protrusions are not present in the day 7 embryonic lens but are first observed later in the day 10 embryonic lens; these membrane protrusions become more elaborate with age (42). Moreover, in the adult chicken lens, newly differentiated cortical fiber cells exhibit smooth profiles, whereas older fiber cells display numerous membrane protrusions (43). In support of a role for spectrin fragmentation in the development of membrane protrusions, antibodies, which recognize both full-length and fragments of ␣-spectrin stain protrusions of nuclear fiber cells (19) as well as blebs decorating differentiated lens cells in culture. 2 This is the first report of caspase-mediated cleavage of membrane-associated components in the lens. Our data suggests that caspase-3 may be involved in proteolysis of spectrin during lens fiber cell maturation and aging. The sequence of the cleavage site for chicken lens ␣-spectrin is identical to that obtained by in vitro cleavage of bovine brain ␣II-spectrin by caspase-3. In addition, the sequences of cleavage for both chicken lens ␣and ␤-spectrins (DETD*S, DEVD*S) match well with the consensus DXXD*S cleavage site identified for other caspase-3 substrates (44). In rat and cow lenses, the presence of the ϳ110 -120-kDa ␣-spectrin fragment generated by caspase-3 cleavage in vitro also suggests cleavage of endogenous lens ␣-spectrin by caspase-3 (7,45). The lack of a ϳ120-kDa fragment of ␣-spectrin in the chicken lens may be due to conformational differences between chicken and mammalian spectrins.
Caspase activity has previously been shown to be necessary for loss of nuclei during lens fiber cell differentiation (46 -48). In addition, members of the caspase family (1)(2)(3)(4)6) and the caspase substrates, DNA fragmentation factor and poly(ADPribose) polymerase, have been identified in the lens (46). Like spectrin, poly(ADP-ribose polymerase has also been reported to be cleaved late in lens development, after organelle loss (46).
Although our data suggest that caspase cleavage of the membrane skeleton occurs after organelle loss, others have reported that caspase activation is required for organelle loss. This apparent discrepancy may be explained by the idea that different caspases may be active at different times during lens fiber cell maturation. According to Wride et al. (46), particular caspases (i.e. caspase-3) are present early during lens fiber cell differentiation, whereas others (i.e. caspase-1) are predominant in older lens fiber cells. Moreover, although inhibitors of caspases-1, -2, -6, and -9 appeared to inhibit nuclear loss in lens cell cultures, inhibitors of caspase-3 and -8 were ineffective. Both of these observations suggest that although the caspase pathway has not been delineated precisely for lens cells, particular caspases are important at different times during lens fiber cell differentiation. Thus, it is possible that specific caspases (particularly caspase-3) are important for selective proteolysis of the membrane skeleton after organelle loss, whereas other caspases are important for initiating organelle loss.
What are the molecular consequences of caspase cleavage of spectrin and ankyrin? We have shown that spectrin fragments are partially dissociated from lens membranes. This may be due to cleavage of ankyrin, which binds to both ␤-spectrin and integral membrane proteins. Calpain-cleaved ankyrin, which, like lens ankyrin, is also a ϳ190-kDa fragment derived from cleavage near the C terminus, exhibits an 8-fold weaker affinity for erythrocyte membranes (49). Alternatively, cleavage of ␤-spectrin could lead to reduced membrane association. However, the initial site of cleavage near the middle of the protein is not within its ankyrin binding site, actin binding site, or within any of the ankyrin-independent MADs (1). Thus, ␤-spectrin cleavage is unlikely to interfere with membrane association by direct interference with its membrane binding sites, but perhaps indirectly by affecting the conformation of these sites.
Cleavage of lens spectrin by caspase may lead to lowered membrane affinity as well as calmodulin down-regulation of spectrin activities. The caspase cleavage site on ␣-spectrin is within 9 amino acids of the calpain cleavage site and is proximal to the calmodulin binding site (Fig. 4B). When calmodulin is bound to ␣-spectrin during the action of calpain, both ␣and ␤-spectrin subunits are cleaved, and the heterotetramer dissociates irreversibly into its component fragments (14). This results in complete loss of tetramer formation, F-actin binding, or membrane binding. Such a dissociation of spectrin fragments is reminiscent of the dissociation reported here which results from caspase-mediated cleavage of both subunits in the intact heterodimer.
Interestingly, ␣and ␤-spectrin are both cleaved at repeat 11, at sites that interact in the ␣-␤-spectrin heterodimer (50), suggesting that spectrin may be fully assembled before cleav-age. In contrast, in other types of cells (1), calpain may either cleave the ␣or ␤-spectrin subunits separately or together in the intact heterotetramer, thus leading to targeted loss of tetramer assembly, F-actin binding, or membrane binding (13,14). Thus, although other cell types may use a multistep process involving calpain and calmodulin to accomplish an incrementally regulated disassembly of the spectrin cortical skeleton, lens cells may use predominately caspase and calmodulin to disassemble their membrane skeleton in a single step process.
We have developed a speculative model (Fig. 6) to describe how specific and partial proteolysis of membrane skeleton proteins could lead to membrane morphological changes in the lens. In young (cortical) fiber cells (Fig. 6A), as in other nonerythroid cells, ␣-␤-spectrin tetramers are likely cross-linked to short actin filaments. This two-dimensional meshwork is then anchored to the membranes via spectrin interactions with ankyrin and other membrane proteins (1,51). In older fiber cells (Fig. 6B), localized caspase activity could result in proteolytic cleavage of specific regions of the membrane skeleton, loosening constraints on the membranes and thus allowing for membrane blebbing in specific sites. The extent and placement of membrane blebbing could be regulated by whether ankyrin, ␣-spectrin, and/or ␤-spectrin are cleaved, or by calmodulin binding. With age, more caspase activity, and thus, more proteolytic processing, would occur. This could result in the increased density of membrane protrusions with fiber cell age (17). Membrane blebbing could lead to repositioning of integral membrane proteins, such as ion channels, which might be necessary for age-related changes in ion and water flow (52). Moreover, membrane blebbing might lead to cell-cell fusion events that have been observed in maturing lenses (16,17).
Interestingly, cleavage of ␣-spectrin by calpain, not by caspase, has been associated with a number of cataract models (41,53). Inhibition of calpain inhibits cataract formation and spectrin cleavage (54), and human lenses with age-related nu-clear cataracts display a higher density of finger-like membrane projections than transparent lenses of the same age (55). Thus, it is possible that aberrant cleavage of spectrin by calpain may lead to uncontrolled or incomplete membrane furrowing and opacification of the lens, whereas specific cleavage of spectrin by caspase may be important for normal physiological functioning of the lens.
In conclusion, partial and specific proteolysis of spectrin and ankyrin by caspases appear to effect an apoptosis-like program of membrane changes during lens aging. Further characterization of lens membrane morphology and membrane skeleton proteolysis in caspase knockouts, as well as additional biochemical analysis of the lens membrane skeleton, will lead to greater insight in the importance of caspase cleavage of membrane skeleton components during lens development and function. It is likely that the transparency of the aging lens nucleus may depend not only on organelle loss but also on membrane skeleton remodeling. We anticipate that these changes in the membrane skeleton, induced by caspase-mediated but limited proteolysis, will modulate ion homeostasis, the positioning of ion channels, the frequency of intercellular fusion events, and lens deformability.