Filamin (280-kDa Actin-binding Protein) Is a Caspase Substrate and Is Also Cleaved Directly by the Cytotoxic T Lymphocyte Protease Granzyme B during Apoptosis*

We used yeast two-hybrid screening to identify the cytoskeletal protein filamin as a ligand for the proapoptotic protease granzyme B, produced by cytotoxic T lymphocytes. Filamin was directly cleaved by granzyme B when target cells were exposed to granzyme B and the lytic protein perforin, but it was also cleaved in a caspase-dependent manner following the ligation of Fas receptors. A similar pattern of filamin cleavage to polypeptides of ∼110 and 95 kDa was observed in Jurkat cells killed by either mechanism. However, filamin cleavage in response to granzyme B was not inhibited by the caspase inhibitor z-Val-Ala-Asp-fluoromethylketone at concentrations that abolished DNA fragmentation. Filamin staining was redistributed from the cell membrane into the cytoplasm of Jurkat cells exposed to granzyme B and perforin and following ligation of Fas receptors, coincident with the morphological changes of apoptosis. Filamin-deficient human melanoma cells were significantly (although not completely) protected from granzyme B-mediated death compared with isogenic filamin-expressing cells, both in clonogenic survival and51Cr release assays, whereas death from multiple other stimuli was not affected by filamin deficiency. Thus, filamin is a functionally important substrate for granzyme B, as its cleavage may account at least partly for caspase-independent cell death mediated by the granzyme.

Cytotoxic T lymphocytes (CTL) 1 and natural killer (NK) cells induce target cell apoptosis through either of two contact-dependent mechanisms. The first requires a receptor-ligand interaction between a molecule of the TNF family on the surface of the killer cell and its death receptor on the target cell (1). The second mechanism operates through contact-dependent release of the granule-bound apoptotic mediators perforin and granzyme B (grB) and is chiefly responsible for eliminating virusinfected cells (2).
Fas ligation leads to the recruitment of a death-signaling complex at the inner leaflet of the cell membrane (3). This results in activation of an evolutionarily conserved, caspasedependent cell death pathway present in virtually every cell. The subsequent cleavage of vital proteins and DNA leads to cell death. The precise mechanism of collaboration between perforin and grB is unknown; it is generally accepted that perforin, a pore-forming protein, enables proapoptotic serine proteases (granzymes) to access their substrates in the target cell, cleavage of which leads to apoptosis (4 -6). Thus, perforin deficiency inactivates the killing mechanisms that operate through granzymes (7,8). Although several different granzymes are capable of inducing cell death in a perforin-dependent manner (9), the most potent and best characterized is grB, with an ability to cleave proteins after specific Asp residues that enables it to directly activate procaspases, commencing with the cleavage of procaspase-3 (10). The congenital absence of grB from the CTL of mice delays target cell apoptosis and is manifested in their reduced ability to provoke nuclear damage including DNA fragmentation (11).
Although caspase activation is clearly an important function of grB, we (12) and others (13) have shown that grB-mediated apoptosis can occur despite the absence of active caspases. However, the molecular targets responsible for caspase-independent death stimulated by grB have so far proven elusive. The cumulative data from several laboratories have indicated strongly that additional pathways to target cell death arose to overcome the effects of viral anti-apoptotic molecules that delay the suicide of infected cells by blocking caspases (14). These antiviral pathways probably represent the cumulative response to viral strategies that delay apoptosis, which have coevolved with the immune system (14). Among the many viral apoptosis antagonists identified, prototypes include the cowpox protein CrmA (15) and baculovirus p35 (16), both direct caspase inhibitors.
Cell death in the absence of active caspases implies that other proteases can cleave key proteins in the target cell. We have previously reported that grB is rapidly targeted to the nucleus of target cells (17)(18)(19)(20), suggesting that grB might induce nuclear collapse when caspases are inactive (21,22). GrB can directly cleave some nuclear proteins such as nuclear matrix antigen and poly(ADP-ribose) polymerase (22), and granzyme A can activate DNA fragmentation in a caspase-independent manner in cell extracts (23). We also recently demonstrated a marked reduction of nuclear apoptotic changes in response to granzyme B when caspases were inactivated; however, the cells still died through a caspase-independent pathway manifested by cytoplasmic and cell membrane damage (12). These non-nuclear events were a reliable predictor of cell death, as they correlated closely with colony numbers in long-term clonogenic assays (12). In another study in which multiple parameters of CTL-mediated apoptosis were examined, non-nuclear manifestations such as phosphoserine exteriorization, cell membrane blebbing, and changes in mitochondrial membrane potential did not require active caspases, whereas nuclear changes such as chromatin condensation and DNA fragmentation did (24). It was recently reported that grB can induce activation of caspase-activated DNase (CAD) by cleaving its inhibitor (ICAD), thus leading to DNA fragmentation (25). Some direct activation of CAD-induced DNA damage persisted despite caspase inhibition; however, the concentrations of grB used was 40 -80 times higher than needed to induce grB-mediated apoptosis, so the physiological significance of this pathway remains in question. Moreover, the nucleus is dispensable for CTL-mediated cell death (26), indicating that damage to other parts of the cell are more crucial for determining apoptotic outcome.
These results strongly suggested that direct cleavage of nonnuclear substrates by granzymes can lead to caspase-independent cell death. However, no direct granzyme substrates (other than the caspases and ICAD) have been described outside the nucleus. To shed light on unidentified grB substrates, we used yeast two-hybrid screening to identify the cytoskeletal protein filamin (280-kDa actin-binding protein, ABP-280) as a ligand for grB. Herein we demonstrate that filamin can be directly cleaved by grB independently of caspase activation and can also be cleaved following Fas ligation in a caspase-dependent manner. Cells deficient in wild-type filamin are less sensitive to grB-mediated cell death, indicating that filamin is a candidate for the mediation of a caspase-independent non-nuclear pathway to apoptosis in response to grB.

MATERIALS AND METHODS
Cell Culture-The human leukemia cell line Jurkat was maintained in RPMI medium supplemented with 10% fetal calf serum at 37°C in air containing 5% CO 2 . The human melanoma cell line M2, which lacks expression of filamin mRNA, and the isogenic cell line A7, in which expression was re-established by gene transfection, were maintained as above. Both cell lines were a kind gift of Dr. John Hartwig (Harvard Medical School).
Chemicals and Reagents-Human perforin, purified as described, was kindly supplied by Dr. Chris Froelich, Evanston University, Evanston, IL (27). Pneumococcal pneumolysin (PLO) (a gift from Dr. James Paton, University of Adelaide, Australia) was activated in phosphatebuffered saline containing 2-mercapto-ethanol for 30 min. at room temperature. A sublytic dose of the membranolytic agents was defined as producing Ͻ10% specific release of 51 Cr in a 4-h assay at 37°C. Neither membranolytic agent was inhibitory for the Asp-ase activity of grB, nor was membrane perforation negatively affected by the caspase inhibitors described below (data not shown). Immunoaffinity purification of human grB from nuclear lysates of YT (human NK leukemia) cells was performed as described (28). The grB was free of grA and Met-ase activities, as well as perforin, as demonstrated by Western blotting and functional assays (28). The oligopeptide caspase inhibitors z-Val-Ala-Asp-fluoromethylketone (z-VAD-fmk) and Phe-Ala-fluoromethylketone (z-FA-fmk) were purchased from Enzyme Systems Products (San Diego, CA), dissolved in Me 2 SO, and stored in aliquots at Ϫ20°C. Final concentrations of Me 2 SO did not exceed 0.5% in any of the assays. Polyclonal antisera detecting human ABP-280 were raised in rabbits immunized with a fusion protein consisting of glutathione S-transferase and the C-terminal 476 amino acids of filamin, which was purified by chromatography on glutathione-agarose.
Yeast Two-hybrid Screening-The cDNA encoding human grB was the template in spliced overlap extension polymerase chain reactions to generate a clone lacking the leader sequence and in which the active site Ser 183 was mutated to Ala. This fragment was cloned in frame into the EcoRI site of the plasmid vector pAS2-1 (CLONTECH), downstream of sequences encoding the GAL4 DNA binding domain. This "bait" construct was introduced into the yeast strain PJ69-4A and transformants selected by growth on minimal medium deficient in Trp. Yeast containing the bait construct were then cotransfected with cDNA from human peripheral blood lymphocytes stimulated with phytohemagglutinin fused to the transcription activation domain of GAL4 ("prey" construct library) in the plasmid vector pGAD10 (CLONTECH). Clones with putative interactions between bait and prey were selected by growth on medium lacking Trp/Leu/Arg (TLA Ϫ ) and Trp/Leu/His (TLH Ϫ ). DNA from clones of interest was used to transform electrocompetent Escherichia coli TOP10F cells, and cDNA inserts of interest were sequenced.
Assays of Apoptosis-51 Cr release (a measure of non-nuclear cell damage) and TUNEL assays were performed as described (12). Fasmediated apoptosis of Jurkat cells was accomplished by incubating the cells in the presence of CH11 anti-Fas antibody (100 ng/ml) for 4 h at 37°C. Cells undergoing apoptosis in response to anti-Fas antibody or perforin and grB were analyzed by confocal laser scanning microscopy, as described previously in detail (17)(18)(19). M2 and A7 cells were induced to undergo apoptosis by exposure to medium containing staurosporine for 24 h. Loss of viability was assessed by trypan blue exclusion.
Assays of Clonogenic Survival-A7 and M2 cells in logarithmic growth phase were incubated for 2 h with a sublytic quantity of porin alone and/or with grB (30 -120 nM). Cells (5 l from a stock of 4 ϫ 10 4 cells/ml) treated with the various apoptotic stimuli were plated out in triplicate in individual wells of a 48-well plate and then incubated at 37°C for 4 -5 days when discrete colonies were counted under an inverted microscope.

The C Terminus of Filamin Contains a grB Binding Site-To
identify potential new ligands for grB, we expressed mature human grB as a chimeric protein with the GAL4 DNA binding domain for yeast two-hybrid studies. The same approach was successful in identifying substrates of caspases (apoptotic cysteine proteases) (29). As in the previous study, we engineered an amino acid substitution at the active site residue (Ser to Ala) to favor more stable protease-substrate interactions. Inactive variants of grA have also been used to purify its putative ligands (30). The mutated grB was coexpressed in yeast cells with a library of proteins expressed as fusions with the GAL4 transcription activation domain. Serial rounds of screening identified clones in which growth on selective media suggested the presence of bona fide interactors for grB. Sequencing revealed that one clone encoded in frame the 476 amino acids (residues 2172-2647) at the C terminus of the cytoskeletal protein, filamin (ABP-280). The interaction with grB was specific, as the filamin domain did not bind to control proteins p53, CD46, or IFI 16 (data not shown). Filamin is a ubiquitous phosphoprotein located in the peripheral cytoplasm, where its N terminus binds to and promotes orthogonal branching of actin filaments (31), contributing to membrane stability and the ability to withstand mechanical stress (32) and enabling cell migration (33).
Filamin Is Cleaved during Apoptosis Induced through Fas Ligation and by grB-The C-terminal 476 amino acids of human filamin contains 16 Asp residues, but none conformed closely to a tetrapeptide consensus sequence for preferred cleavage by grB or caspases (34,35). To establish whether filamin is cleaved during apoptosis induced by grB, or when cells are killed through Fas ligation (the two mechanisms used by cytotoxic lymphocytes), lysates from dying Jurkat cells were probed in Western blots with an antiserum raised against the filamin C terminus (Fig. 1). Cells killed by Fas ligation with anti-Fas IgM demonstrated cleavage of filamin into fragments of ϳ110 and 95 kDa, consistent with cleavage at a minimum of two sites. A larger protein possibly corresponding to a partially cleaved species of ϳ205 kDa was also seen. As the antiserum reacted poorly with the N terminus of filamin, it is likely that at least one other cleavage product was not identified by Western blotting. Pre-incubation of the Jurkat cells in 20 M z-VADfmk (but not the control inhibitor z-FA-fmk) prior to anti-Fas treatment completely abolished filamin cleavage, indicating that activated caspases are responsible, either directly or indirectly, for filamin proteolysis. When Jurkat cells were exposed to grB with sublytic quantities of perforin, a similar pattern of filamin cleavage was seen; however, the cleavage was not blocked by z-VAD-fmk (Fig. 1). In contrast, this concentration of z-VAD-fmk completely abolished DNA fragmentation as measured in the TUNEL assay and cleavage of poly(ADPribose) polymerase (see Refs. 4 and 12, and data not shown). Indeed, grB-mediated cleavage of filamin was not inhibited by pretreatment of the cells with concentrations of z-VAD-fmk up to 50 M, and filamin cleavage also persisted in MCF-7 human breast cancer cells, which lack caspase-3 (data not shown). Proteolytically inactive (in) grB (with its the active site Ser residue mutated to Ala) expressed in baculovirus-infected insect Sf21 cells was unable to induce filamin cleavage (Fig. 1) or cell death (data not shown), indicating the need for Asp-ase activity for both phenomena. When immunoprecipitated filamin was incubated with purified grB, filamin was cleaved into a number of peptides (Fig. 2), confirming that grB can directly cleave filamin in the absence of other proteases. Several polypeptides (205, 200, and 110 kDa) were of similar size to those observed after intact cells were incubated with perforin and grB. However, the 95-kDa polypeptide seen when intact cells were exposed to perforin/grB was not observed. Collectively, our results suggested that (i) certain sites in filamin are cleaved far more efficiently by grB than are others, and (ii) proteases other than grB may also contribute to filamin processing in the absence of caspases. Filamin is therefore the target of several classes of protease during apoptotic death.
Intracellular Redistribution of Filamin Is Coincident with Cell Death Induced by Perforin and grB-We next used confocal microscopy to examine the subcellular distribution of filamin in Jurkat cells undergoing apoptosis in response to grB. Filamin is normally located subjacent to the plasma membrane, where it binds transmembrane proteins including the ␤ 2 -integrin CD18 (36) or the platelet glycoprotein Gp1b/9 (37), linking the plasma membrane and the actin cytoskeleton. Cells were exposed either to anti-Fas antibody for 4 h or to perforin (Pfp) or the bacterial pore-forming toxin, PLO, alone or in combination with grB for 2 h (Fig. 3). Like perforin, PLO can also deliver grB to induce apoptosis in target cells (4). In cells that were mock-treated or exposed to Pfp or PLO alone (Fig. 3, left) or to grB alone (data not shown), filamin staining was seen only immediately adjacent to the cell membrane or within pseudopodal membrane projections. In each of these controls, the Jurkat cells remained free from apoptotic changes. However, the filamin staining pattern was markedly altered in cells exposed to both the Pfp/PLO and active grB or to anti-Fas antibody (Fig. 3, right). Filamin became dissociated from the membrane, and was diffusely distributed through the cytoplasm of the dying cells. Filamin continued to be excluded from the nucleus, causing the outlines of apoptotic nuclei to become easily seen. Many of the dying cells also demonstrated the cytoplasmic shrinkage characteristic of apoptotic death.
Deficiency of Filamin Is Associated with Resistance to grBmediated Apoptosis-In previous work, we postulated that caspase-independent pathways to apoptosis are likely to be localized in the cytoplasm (12,14). Our demonstration that filamin cleavage is not dependent on caspases raised the possibility that filamin is part of a non-nuclear pathway to cell death mediated by grB, prompting us to evaluate the effect of filamin cleavage on cell survival. The only cells suitable (and available) for such a comparison are a pair of nonclonal isogenic human melanoma cell lines, A7 and M2, that have long served as the best available model of functional filamin deficiency (Refs. 38 and 39). 2 M2 cells fail to synthesize filamin and have no filamin mRNA; however, expression has been re-established in A7 cells by gene-mediated transfection, resulting in appro-2 T. Stossel (Harvard University), personal communication. Western blot (using rabbit anti-filamin antiserum) following digestion at 37°C with grB (300 nM) for 120 min of filamin immunoprecipitated from a lysate of healthy Jurkat cells and whole Jurkat cell lysates following exposure to sublytic Pfp or in combination with active grB (120 nm) for 120 min. The filled arrowheads indicate filamin fragments generated by direct cleavage of immunoprecipitated filamin and during grB-mediated apoptosis of whole Jurkat cells migrating at approximately the same molecular weight. The open arrowheads indicate filamin fragments generated that are unique to either direct grB cleavage or grB mediated apoptosis of whole cells. The migration of Ig heavy chain (ϳ50 kDa, detected by the anti-rabbit Ig reagent) is also shown. Pre-immune rabbit serum used in the same experiment immunoprecipitated no detectable filamin (data not shown).

FIG. 3. Redistribution of filamin staining in Jurkat cells exposed to anti-Fas antibody or sublytic Pfp or PLO and grB in combination.
Jurkat cells were exposed to sublytic quantities of Pfp or PLO Ϯ grB (120 nM) for 120 min at 37°C or to anti-Fas antibody (100 ng/ml) for 240 min at 37°C, fixed, and stained with rabbit anti-filamin antiserum.
priate targeting of filamin subjacent to the cell membrane and correction of defects in cell motility and fragility (39).
Significantly greater cell death was observed in the A7 cells compared with the mutant cell line when the cells were exposed to grB, as judged by specific release of 51 Cr, the standard assay for CTL-mediated cell death (Fig. 4A). Experiments 1 and 2 were deliberately carried out with different amounts of PLO. In experiment 1, 51 Cr release increased ϳ3-fold for M2 cells but 11-fold for A7 cells when the cells were additionally exposed to grB. In experiment 2, where the 51 Cr release with lysin alone was intentionally set higher, the increase in response to grB addition was 1.5-fold (30 -45%) for M2 and Ͼ3-fold (22-70%) for reconstituted A7 cells. To determine whether these differences in Cr release were indicative of long-term protection of M2 cells, both cell lines were treated with grB and lysin for 60 min and then washed and plated out to evaluate their clonogenic potential (Fig. 4B). A7 filamin-reconstituted cells were significantly and reproducibly more sensitive than filamindeficient M2 cells to grB-mediated cell death, as indicated by a reduction in colony numbers by ϳ40% compared with M2 cells.
Our results collectively indicated that filamin is an apoptotic substrate of grB, in which cleavage contributes significantly to the demise of grB-treated cells. Importantly, M2 cells lacking filamin were not resistant to other forms of apoptosis, as indicated by their equivalent sensitivity to staurosporine (Fig. 4C) when compared with A7 cells. M2 cells were also not protected from ␥-irradiation-induced death (data not shown); nor could their resistance to grB be accounted for by reduced susceptibility to perforin/PLO lysis or to intrinsic differences in clonogenic capacity (data not shown). The only death stimulus we have found to which M2 cells were resistant was grB, in the context of lysin-mediated delivery. Unfortunately, A7 and M2 cells are resistant to Fas-and TNF-mediated apoptosis due to poor receptor expression, so it was not possible to evaluate the effects of filamin deficiency on receptor-mediated cell death. We also found no evidence of defective grB trafficking in filamin-deficient cells; the kinetics of uptake of fluoresceinated grB into M2 and A7 cells was indistinguishable, as measured by confocal microscopy and image analysis (see Ref. 20; data not shown). We thus concluded that filamin deficiency can specifically confer partial protection from grB-mediated apoptotic death.

DISCUSSION
Filamin Is a Caspase-independent, Non-nuclear Cell Death Substrate of grB-Filamin is the first non-nuclear substrate of grB identified whose direct cleavage influences cell survival. Although filamin is a substrate for caspases, grB can also cleave filamin directly when caspases are inactivated. Under most circumstances, grB is a powerful apoptotic mediator as it can efficiently activate the caspase cascade, commencing with the cleavage of procaspase-3 (10). However, cell death through grB still occurs when caspases are inactivated by caspase inhibitors such as p35 or by z-VAD-fmk. Interestingly, commitment to death under these circumstances is accompanied by minimal initial nuclear damage, consistent with the fact that nuclear events are dispensable for CTL-mediated cell death (26). In contrast, non-nuclear manifestations of cell death persist despite caspase inactivation; and despite their normal nuclear morphology early in apoptosis, the cells do not survive in clonogenic assays, indicating they are already destined to die (12). We have therefore favored the likelihood that cytoplasmic, not nuclear, substrates are crucial in delivering the lethal hit from grB. Our present findings that cleavage of a cytoplasmic cytoskeletal element, filamin, can influence cell survival in response to grB is consistent with this hypothesis.
Cytoskeletal Elements Can Be Critical Substrates for Regulating Cell Death-Although the mechanism by which filamin deficiency protects cells against grB is not yet understood at the molecular level, there are several other instances of the cytoskeleton controlling apoptotic outcome. For example, the light chain of the dynein motor complex sequesters the proapoptotic BH3-only Bcl-2 family member, Bim, until an apoptotic stimulus triggers Bim release, thus permitting it to access and disrupt the integrity of the mitochondrial membrane (40). By analogy, filamin might sequester a key grB substrate or even grB itself by direct binding; however, the prediction of this model is that the levels of a free grB should increase in M2 cells, thereby increasing their susceptibility to apoptosis. This prediction is clearly at odds with our findings, which, if anything, predict the opposite scenario, that binding of grB to filamin might potentiate its ability to bind and cleave other proapoptotic substrates. An alternative possibility is that toxic cleavage products may be generated from filamin by grB and possibly the caspases. This mechanism would be reminiscent of the cleavage of another actin-binding protein, gelsolin, by caspases (41). Upon cleavage by caspase-3, it has been proposed that a 41-kDa N-terminal fragment of gelsolin induces cleavage of F-actin filaments, leading to cytoskeletal collapse (42). However, this mechanism remains controversial, as gelsolin overexpression did not result in increased sensitivity of lymphocytes to apoptotic stimuli (43). To date, there is no evidence that filamin cleavage products are toxic, but if this were the case one would expect M2 cells to also be resistant to receptormediated killing, as Fas ligation also resulted in filamin cleavage in Jurkat cells (Fig. 1). Unfortunately, M2 and A7 cells are resistant to Fas-and TNF-mediated cell death, making it impossible to test this prediction. We are currently mapping filamin cleavage sites and testing whether the cleavage products are toxic.
The findings presented in this paper represent a consider- FIG. 4. M2 filamin-deficient human melanoma cells are protected from grB-mediated apoptosis. A, release of 51 Cr from M2 and A7 cells exposed to PLO Ϯ grB (120 nM). The results of two independent experiments (of six performed) are shown in which equivalent PLOmediated 51 Cr release in M2 and A7 cells was induced using 3 units/ml and 6 units/ml (top) or 24 units/ml and 48 units/ml PLO (bottom), respectively, either alone or in combination with grB. Data points are for the mean of triplicate assays Ϯ standard errors. B, clonogenic survival of M2 and A7 cells 7 days after exposure to either PLO alone (3 or 12 units/ml) or PLO in combination with grB (120 nM). The data points represent the percentage of clones observed relative to treatment with PLO alone and are shown as the mean of triplicate assays Ϯ standard error. The experiment is representative of six performed. C, viability of M2 and A7 cells following incubation in medium containing staurosporine (0.06 -0.5 M) for 24 h, evaluated by trypan blue exclusion. able step forward in our understanding of the impressive redundancy in proapoptotic mechanisms within cytolytic granules. We have defined a novel pathway used by grB to overcome potential inhibition of caspase-mediated cell death pathways, involving the proteolysis of a cytoskeletal protein, filamin, the cleavage of which influences cell survival. The ability of grB to directly induce cell death in the cytoplasm without invoking caspase-mediated nuclear damage represents an evolutionary adaptation of the immune system to counter the ability of viruses to block the intrinsic cell suicide pathway.