β1-integrin protects keratinocyte stem cells (KSC) from cell-detachment apoptosis (`anoikis'). Here we show that caspase-8 active protein is detected in both young transit amplifying (TA) cells and TA cells, but not in KSC. On suspension, caspases are activated earlier in young TA than in KSC, whereas anti-β1-integrin neutralizing antibody accelerates caspase activation in both KSC and young TA. Caspases 8 and 10 are the first caspases to be activated whereas caspase-8 inhibitor zIETD-fmk delays the activation of Bid, caspase-9 and caspase-3. However, the caspase-9 inhibitor zLEDH-fmk does not block the activation of caspase-8, Bid, caspase-10 and caspase-3. Moreover, caspase-8, but not caspase-9 inhibitor partially prevents keratinocyte anoikis. As FLIP inhibits caspase-8 processing, we retrovirally infected HaCaT keratinocytes with c-FLIPL. Anti-β1-integrin fails to activate caspase-8, Bid, caspase-9 and to induce the release of cytochrome c in c-FLIPL overexpressing keratinocytes. Finally, overexpression of c-FLIPL partially prevents anoikis in both suspended and anti-β1 integrin-treated cells. Taken together, these results indicate that the extrinsic apoptotic pathway triggered by caspase-8 predominates in keratinocyte anoikis. However, the release of cytochrome c and the later activation of caspase-9 seem to suggest that the intrinsic mitochondrial pathway may intervene as a positive feedback loop of caspase activation.
Apoptosis that occurs following detachment of a cell from the extracellular matrix (ECM) has been termed anoikis (Frisch and Screaton, 2001). This mode of cell death has been observed in numerous adherent cells, such as endothelial cells (Meredith et al., 1993), fibroblasts (Tian et al., 2002), neurons (Banks and Noakes, 2002) and various epithelial cells, including intestinal and epidermal cells (Grossman et al., 2001; Tiberio et al., 2002), kidney cells (Frisch and Francis, 1994) and prostate cells (Bondar and McConkey, 2002). Disturbances of anoikis can be involved in the alteration of tissue physiology, such as in intestinal epithelial cell homeostasis (von Herbay and Rudi, 2000) and in several disease states (Grossman, 2002). The acquisition of anoikis resistance is regarded as a crucial step in the metastatic transformation of a tumor (Yawata et al., 1998; Windham et al., 2002). Cell attachment not only provides structural anchorage for a cell, but it can initiate a cascade of survival signals. The anchorage of cells to the ECM is mediated mainly by integrins, and integrin signaling controls cell death in many cell settings (Giancotti and Ruoslahti, 1999). Integrins mediate survival in nontransforming cells, ensuring tissue homeostasis (Howlett et al., 1995) and during neoplastic transformation (Manzotti et al., 2000). On integrin ligation, focal adhesion-kinase (FAK) is phosphorylated and initiates a complex cascade of signaling molecules (Cary et al., 1999). However, the molecular mechanisms involved in this process are still incompletely understood. Recent studies have indicated that anoikis is Bcl-2 dependent (Coll et al., 2002; Tiberio et al., 2002) and is controlled by FAK (Frisch et al., 1996), which in turn inactivates the pro-apoptotic protein Bad (Datta et al., 1997).
As in most apoptotic pathways, anoikis is mediated by the activation of caspases. From a functional perspective, caspases are viewed as either upstream (initiator) caspases or downstream (executioner) caspases. Two pathways for activating caspases have been elucidated (Reed, 2000). The extrinsic pathway, which is induced by `death receptors', centers on FLICE/caspase-8 (Juo et al., 1998). The stimulation of death receptors such as Fas and TRAIL-R results in clustering of the receptor, which in turn leads to the recruitment of the Fas-associated death domain (FADD) and causes caspase-8 activation (Ashkenazi and Dixit, 1998). The extrinsic apoptotic pathway is inhibited, among others, by FLICE/caspase-8 inhibitory protein (FLIP), which prevents caspase-8 processing (Irmler et al., 1997). The intrinsic pathway involves mitochondria with the release of cytochrome c, which complex apoptosis protease activating factor (Apaf-1) resulting in the activation of caspase-9 (Martinou and Green, 2001). In certain conditions, crosstalk between the two apoptosis networks exists (Fisher et al., 2003): Bid, a pro-apoptotic Bcl-2 family member, is cleaved by active caspase-8 following death receptor stimulation and translocates to the mitochondria, where it initiates the mitochondrial apoptotic pathway (Yin, 2000). Initiator caspases stimulate the intracellular activation of executioner caspases 3, 6 and 7, which in turn cleave distinct substrates, eventually leading to the activation of endonucleases that complete cellular suicide by DNA fragmentation (Enari et al., 1998).
A population of keratinocyte stem cells (KSC) governs the renewal of mammalian epidermis (Watt, 2001). KSC generate transient amplifying (TA) cells that terminally differentiate after a discrete number of cell divisions (Fuchs and Segre, 2000). KSC express highest levels of β1-integrin which strikingly decrease as keratinocytes leave the KSC compartment (Jones et al., 1995), implying that this receptor is required for maintenance of KSC and for holding them in the right place. Loss or alteration of integrin expression ensures departure from stem cell niche through differentiation or anoikis (Watt and Hogan, 2000). Loss of adherence of KSC leading to anoikis seems to be a valuable and physiological approach to study this type of cell death and to dissect its underlying mechanisms. Using this experimental design, we have shown recently that KSC, but not TA cells, are protected from anoikis via an integrin signaling pathway, in a Bcl-2-dependent manner (Tiberio et al., 2002).
In the present study, we extend the analysis of the mechanisms involved in KSC anoikis: in particular, we evaluate the sequential activation of caspases to understand whether the extrinsic or the intrinsic apoptotic pathway prevails in this type of cell death. We report that, on blockade of β1 integrin, caspase activation occurs in TA cells much more rapidly than in KSC. Furthermore, the caspase-8 pathway seems to predominate in KSC anoikis.
Materials and Methods
Cell culture and reagents
Normal human keratinocytes were obtained from foreskin and cultured as described previously (Pincelli et al., 1997). Keratinocytes were divided into three populations and plated onto plastic dishes, coated for 2 hours at 37°C with type IV collagen 100 μg/ml (Sigma, St Louis, MO). They were first allowed to adhere to type IV collagen for 5 minutes, and the nonadherent cells were then transferred to fresh collagen-coated dishes and allowed to attach overnight. Finally, keratinocytes not yet attached after one night were plated onto type IV collagen to obtain a third population. The three keratinocyte populations were characterized as previously described (Tiberio et al., 2002) in order to obtain a population enriched KSC, a population consisting of young TA cells and a third population of TA cells. Keratinocytes were then cultured in serum-free medium (KGM, Clonetics, San Diego, CA) and used for further experiments. The spontaneously transformed keratinocyte line HaCaT was kindly provided by N. Fusenig (DKFZ Heidelberg, Heidelberg, Germany) and cultured as described previously (Boukamp et al., 1988).
For blocking adhesion experiments, normal keratinocytes and HaCaT cells were trypsinized and suspended at high density in polypropilene at 37°C with or without the addition of anti-β1-integrin neutralizing antibody (1:250) (Immunotech) or medium alone for different time points. For caspase inhibition experiments, adhering young TA keratinocytes were pre-incubated at 37°C for 1 hour with caspase-9 inhibitor zLEDH-fmk or caspase-8 inhibitor zIETD-fmk 100 μM.
Cells from KSC and young TA were fixed in situ in paraformaldehyde (4% in PBS) and cytospun onto slides precoated with 0.01% poly-L-lysine and air dried at 48 hours. Slides were washed in PBS and put into 0.3% H2O2 for 15 minutes to remove endogenous peroxidase activity. Slides were incubated with anti-cytokeratin 10 (1:25) (Dako, Glostrup, Denmark) for 1 hour. They were then incubated with secondary rabbit biotin-conjugated anti-mouse antibody (1:100) (Dako) for 1 hour followed by incubation with ABC complex (1:100) (Biospa Division, Milano, Italy) for 1 hour. Finally, 3-amino-9-ethylcarbazole (Sigma) was used for visualization. Slides were fixed in acetone and incubated with anti-keratinocyte transglutaminase (type I) at a dilution of 1:20 (Biomedical Technologies, Stoughton, MA). Slides were then incubated with a secondary anti-mouse antibody (1:20) (Dako). They were finally incubated with APAAP complex (1:20) (Dako) for 45 minutes. Nephthol-AS-B1 phosphate in new fuchsin stain (Sigma) was then added for 30 minutes for color development.
Retroviral infection and generation of stable cFLIPL-overexpressing HaCaT
For overexpression of cFLIPL the retroviral vector PINCO containing cFLIPL cDNA (kindly provided by Francesco Grignani and Giorgio Stassi) was used (Grignani et al., 1998; Stassi et al., 2000). Retroviral infection of HaCaT cells was performed as described previously (Leverkus et al., 2003b; Leverkus et al., 2003a). Briefly, the transient transfection of PINCO into an amphotropic (Phoenix) packaging cell line was done using 10 μg of DNA by calcium-phosphate precipitation. Transfected Phoenix were selected using 2.5 μg/ml puromycin (Sigma) for 7-14 days. Viral particles containing cell culture supernatants were generated by cultivating transfected Phoenix cells with DMEM containing 10% FCS for 24 hours, filtered (0.45 μm), and added to HaCaT cultures in the presence of polybrene (1 μg/ml). HaCaT were centrifuged for 3 hours at 21°C and viral particle containing supernatant was subsequently replaced by fresh medium. After a culture period of 10-21 days bulk-infected cells were subcloned to yield monoclonal cell lines overexpressing cFLIPL or vector alone (Wachter et al., 2004). FACS analysis confirmed >95% GFP positivity (data not shown). Western blotting confirmed overexpression of cFLIPL and was performed as reported (Leverkus et al., 2003b) using antibodies to cFLIP (NF-6) or caspase-8 (C15). These antibodies are available from Alexis (Gruenberg, Germany) and were kindly provided by P. H. Krammer. Horseradish peroxidase (HRP)-tagged goat antimouse IgG1 and IgG2b were obtained from Southern Biotechnology (Birmingham, AL). Cytotoxicity analysis showed relative resistance of the cFLIPL-overexpressing line to death ligands like TRAIL when compared with vector-infected control lines (Wachter et al., 2004).
Cell death analysis
Young TA keratinocytes were pre-incubated at 37°C for 1 hour with caspase-9 inhibitor zLEDH-fmk or caspase-8 inhibitor zIETD-fmk, 100 μM, suspended in polypropylene at 37°C with or without anti-β1-integrin antibody for 12 hours. Cells were collected, fixed in paraformaldehyde (4% in PBS), cytospun onto slides precoated with 0.01% poly-L-lysine and air dried. Adhering HacaT cells were fixed in situ in paraformaldehyde, as before, whereas suspended cells were treated with anti-β1-integrin antibody or medium alone for 16 hours. All cells were then stained with TUNEL method, as described previously (Pincelli et al., 1997). Fluorescent specimens were analyzed by Confocal Scanning Laser Microscopy (Leica TCS4D) in conjunction with a conventional optical microscope (Leica DM IRBE). For flow cytometric analysis (subG1 peak), cells were trypsinized after 48 hours and resuspended (1×106 cells) in 400 μl hypotonic fluorochrome solution: propidium iodide (PI) 50 mg/ml in 0.1% sodium citrate containing 0.1% Triton X-100, Sigma. After a 60 minute incubation in this solution, cells were analyzed using a FACScan.
Western blotting analysis
Cells were washed with PBS and lysed on ice in RIPA buffer, pH 7.5. Total protein (25 μg) were analyzed on 10% polyacrylamide gels and blotted onto nitrocellulose membranes, whereas for PARP and Bid detection, 40 μg of total protein were analyzed under reducing conditions on 7% and 15% polyacrylamide gels, respectively. For cytosolic and mitochondrial fractions, 60 μg of protein were analyzed on 18% polyacrylamide gels under reducing conditions. The blots were blocked for 2 hours in blocking buffer (PBS buffer, pH 7.4 with 0.2% Tween 20 and 5% non-fat milk) and incubated with anti-human PARP monoclonal antibody (1:5000) (Enzyme Systems Production, Livermore, CA), anti-human procaspase 8 monoclonal antibody (1:1000) (BD Biosciences Pharmingen, San Diego, CA), anti-human caspase 8 (1:1000), anti-human caspase 10 polyclonal antibody (1:1000), anti-human caspase 9 polyclonal antibody (1:3000), anti-human caspase 6 polyclonal antibody (1:3000) and anti-human caspase 3 polyclonal antibody (1:3000). The above mentioned antibodies were kindly provided by John Reed (The Burnham Institute, La Jolla, CA), and recognize both the zymogen and the active form. Blots were also incubated with anti-human Bid polyclonal antibody (1:1000) (BD Biosciences Pharmingen), anti-human cytochrome c monoclonal antibody (1:1000) (BD Biosciences Pharmingen) and anti-human β-actin monoclonal antibody (1:1000) (Sigma), overnight at 4°C. Then membranes were washed in PBS/Tween 20, incubated with peroxidase-conjugated goat anti-mouse antibodies (1:800) (Bio-Rad, Hercules, CA) for PARP, procaspase 8, cytochrome c and actin for 45 minutes at room temperature, and with peroxidase-conjugated goat anti-rabbit (1:3000) (Bio-Rad) for Bid, procaspase-10, procaspase-9, procaspase-6 and procaspase-3. Finally, membranes were washed and developed using the ECL chemiluminescent detection system (Amersham Biosciences Ltd, Little Chalfont, UK).
Preparation of cytosolic and mitochondrial extraction for cytochrome c
Cells were washed with PBS and lysed on ice in 100 μl of lysis buffer (mannitol 210 mM, sucrose 70 mM, Hepes 5 mM, EGTA 1 mM, BSA 0.05%, DTT 1 mM) containing protease inhibitors for 10 minutes. Cells were homogenized by passing the suspension through a 25-gauge needle and sonicated. The homogenates were centrifuged at 1000 g for 5 minutes, than the supernatants were centrifuged again at 10000 g for 20 minutes. The supernatants were collected and stored as cytosolic fractions, whereas the pellets were resuspended in lysis buffer and stored as mitochondrial fractions.
KSC, but not TA cells are protected from apoptosis
We have shown previously, by means of TUNEL and propidium iodide staining, that young TA and TA cells, but not KSC, undergo anoikis (Tiberio et al., 2002). Whereas both techniques indicate that TA cells present with DNA fragmentation, we now confirm this finding by poly(ADP-ribose)polymerase (PARP) cleavage. PARP is a specific substrate for caspases and it is considered a hallmark of the execution stage of apoptosis (Pacini et al., 1999). We present evidence that PARP is cleaved in TA and, to a lesser extent, in young TA cells, but not in KSC, as shown by the decrease of the full length 116 kDa protein. Consistently, we show the appearance of the cleaved fragment of 85 kDa in young TA and, to a higher degree, in TA cells (Fig. 1).
Caspase-8 is activated in young TA and TA cells, but not in KSC
Caspases are present as inactive zymogens (pro-caspases) and, on apoptotic stimuli, are cleaved to active states (Boatright and Salvesen, 2003). We used antibodies against caspases to evaluate both the zymogen and the active form in KSC, young TA and TA cells. Caspase-8 active protein is detected in both young TA and TA cells, but not in KSC. However, whereas procaspases 9, 6 and 3 are slightly decreased in young TA as compared with KSC, the active fragment of these caspases is clearly visible only in TA cells (Fig. 2).
Anti-β1-integrin induces anoikis and not differentiation in human keratinocytes
We have shown previously that anti-β1-integrin induces anoikis in human keratinocytes (Tiberio et al., 2002). To rule out the possibility that blocking β1-integrin signal induces differentiation (Levy et al., 2000) in our system, we treated a pool of KSC and young TA cells with anti-β1-integrin antibody and evaluated apoptosis by FACS analysis. Anti-β1-integrin induced a significant increase in the percentage of cells in SubG1 peak at 18 hours (Fig. 3A). At the same time point, anti-β1-integrin failed to induce the expression of cytokeratin 10 and keratinocyte transglutaminase I in keratinocytes (Fig. 3B). These results indicate that blocking β1-integrin signal induces anoikis and not differentiation in human keratinocytes.
Caspases are activated earlier in young TA than in KSC
As TA cells consist of a post-mitotic population of keratinocytes undergoing terminal differentiation and/or apoptosis, we focused on the difference between KSC and young TA cells endowed with a greater proliferative capacity than TA cells and more similar to KSC. To evaluate cell behavior in regard to apoptosis, we suspended cells in polypropylene and analyzed caspase activation at different time points. Both initiator and effector caspases, except for caspase-9, are activated earlier in young TA cells than in KSC. In particular, 3 hours after suspension, caspase-8 activation is less pronounced in KSC than in young TA, whereas at 6 hours it is completely cleaved in young TA, but not in KSC. Whereas caspase-6 is activated at 6 hours in young TA, it becomes activated at 9 hours in KSC. Caspase-3 starts to be activated at 9 hours in young TA cells and at 12 hours in KSC. Finally, caspase-9 is activated at 9 hours both in young TA and in KSC (Fig. 4).
Anti-β1-integrin antibody accelerates caspase activation
As inhibition of β1-integrin by antibodies to this molecule induces anoikis in keratinocytes (Tiberio et al., 2002), we wanted to evaluate the sequence of caspase activation in human keratinocytes after blocking β1 integrin, in relation to mere cell suspension. To this purpose, we used young TA because of their higher sensitivity to anoikis. Caspase-8, caspase-10, caspase-6, caspase-3 and caspase-9 are activated earlier in cells treated with anti-β1-integrin than in suspended cells. In particular, on anti-β1-integrin treatment, caspase-8 and caspase-10 are activated at 3 hours, whereas Bid and caspase-9 are activated at 6 hours. Caspase-6 and caspase-3 are activated at 3 and 9 hours, respectively (Fig. 5A). Cytochrome c is released from mitochondria and detected in the cytoplasm at 6 hours. Cytochrome c release is higher after anti-β1-integrin treatment than in merely suspended cells (Fig. 5B).
Caspase-8 inhibitor delays caspase activation and prevents keratinocyte anoikis
To evaluate the apoptotic pathway in anoikis, we pretreated keratinocytes with caspase-9 inhibitor zLEDH-fmk before adding anti-β1-integrin antibody. First, caspase-8 activation was not affected by zLEDH-fmk. Caspase-10, an initiator caspase of the death receptor apoptotic signaling, is an ortholog of caspase-8, but it can function independently from it (Wang et al., 2001). Here we show that caspase-9 inhibitor fails to block caspase-10 activation. In particular, caspases 8 and 10 were activated at 3 hours, whereas Bid started to be activated at 6 hours and caspase-3 at 9 hours. These results seem to indicate that caspase-9 is not crucial to the induction of anoikis in keratinocytes (Fig. 6A). If caspase-9 doesn't function as initiator caspase in keratinocyte anoikis, we reasoned that caspase-8 pathway should predominate. To test this assumption, young TA keratinocytes were pretreated with caspase-8 inhibitor zIETD-fmk and ant-β1-integrin antibody was added. Indeed, caspase-8 inhibitor caused a delay in the activation of caspase-9 and blocked the activation of Bid at least up to 6 hours, whereas caspase-3 was not yet active at 12 hours (Fig. 6B). These data seem to indicate that caspase-8 acts as the initiator caspase in keratinocyte anoikis.
To confirm that caspase-8 pathway prevails in the initiation of anoikis in keratinocytes, cells were pretreated with zIETD-fmk and zLEDH-fmk, and apoptosis evaluated by TUNEL. First, anti-β1-integrin antibody at 12 hours induces a significantly higher percentage of apoptosis, as compared with suspended cells, in agreement with the earlier activation of caspases during β1-integrin blockade. Caspase-8 inhibitor prevented anoikis in both suspended and anti-β1-treated cells, as compared with controls. β1-integrin blockade partially prevents caspase-8 inhibitor effect, confirming that β1-integrin is crucial for keratinocyte survival. By contrast, caspase-9 inhibitor failed to protect keratinocytes from anoikis (Fig. 6C).
Anti-β1-integrin fails to cleave Bid, to release cytochrome c and to activate caspase-9 in c-FLIPL overexpressing keratinocytes
cFLIP is a potent inhibitor of death receptor-mediated apoptosis and is able to interfere with caspase-8 activation at the death-inducing signaling complex (DISC) (Krueger et al., 2001). To further clarify the role of the extrinsic and intrinsic pathway of apoptosis during anoikis, we asked if inhibition of the extrinsic pathway of apoptosis may interfere with anoikis induction in keratinocytes. To this end we ectopically expressed cFLIPL in human keratinocytes. In line with our previous results, cFLIPL-overexpressing keratinocytes were highly resistant to the death ligand TRAIL when compared with control cells (Fig. 7A,B). These data show that cFLIPL is able to functionally block death receptor signaling in human keratinocytes. It has been shown that the extrinsic, `death receptor-mediated' apoptotic pathway is controlled by a balance between caspase-8 and c-FLIP expression levels. Two different isoforms of c-FLIP, namely FLIP long (FLIPL) and Flip short (FLIPS), are known to interfere with caspase-8 activation at the DISC (Scaffidi et al., 1999). Whereas primary keratinocytes express high levels of c-FLIP, the transformed keratinocyte line HaCaT has undetectable levels of c-FLIP, and we have shown previously that levels of c-FLIP correlate with sensitivity to TRAIL-induced apoptosis (Leverkus et al., 2000). To explore the response of c-FLIP overexpressing HaCaT, we have treated HaCaT cells overexpressing c-FLIPL with anti-β1-integrin and measured the activation of caspase-8, Bid and caspase-9. Caspase-8 is cleaved at 3 hours, at the time when t-Bid fragments start to be visible, in agreement with Bid cleavage by caspase-8 (Li et al., 1998). Bid was not activated in c-FLIPL overexpressing keratinocytes, whereas caspase-9 was activated at 6 hours in mock but not in c-FLIPL overexpressing keratinocytes (Fig. 8A). Moreover, cytochrome c was released from mitochondria and detected in the cytosol 3 hours after anti-β1-integrin treatment in mock but not in c-FLIPL overexpressing cells (Fig. 8B), suggesting that c-FLIPL interferes with anoikis induction at the level of caspase-8 activation.
c-FLIPL overexpression prevents anoikis in HaCaT keratinocytes
If caspase-8 is the initiator caspase in anoikis, we argued that inhibition by c-FLIP should block the anti-β1-integrin apoptotic effect in keratinocytes. Indeed, overexpression of c-FLIPL partially prevented anoikis in both suspended and anti-β1 integrin-treated cells (Fig. 9). Taken together, these results indicate that caspase-8 is a key caspase triggering anoikis in keratinocytes.
In keratinocytes, the β1 integrins mediate adhesion to the ECM and also regulate the initiation of terminal differentiation (Watt, 2002). Although the correlation between terminal differentiation and apoptosis in normal epidermis is still a matter of debate (Gandarillas, 2000), here we show that lack of adhesion by inhibition of β1-integrin induces keratinocyte anoikis through a distinct molecular mechanism which is initiated by the activation of caspase-8. It is known that integrins are crucial mediators of cell survival in many cell systems (Illario et al., 2003), including keratinocytes (Tiberio et al., 2002). It appears that when KSC are deprived of integrin signaling, they loose adhesion, leave the niche and undergo anoikis. This indicates that blocking β1-integrin triggers a peculiar pathway which ultimately leads to cell death. Whereas, under certain circumstances, caspase-8 induces only a partial form of apoptosis (Kuwana et al., 1998), the activation of caspase-3 and mostly the cleavage of its substrate PARP, as shown in this paper, clearly show that, on β1-integrin blockade, keratinocytes undergo a complete apoptotic program. Moreover, caspases, the key regulators of the apoptotic machinery, are activated earlier in TA than in KSC, probably by virtue of the different expression levels of β1-integrin (Jones and Watt, 1993; Tiberio et al., 2002).
Molecular events in anoikis are rather complex and far from being fully understood. However, as in most apoptotic pathways, also anoikis is mediated by the activation of caspases (Grossman et al., 2002). On the basis of which caspase signal prevails, the extrinsic or the intrinsic apoptotic pathway will characterize anoikis in each cell type. The temporal and hierarchical order of the activation of caspase cascade during anoikis has been a matter of intensive debate. FAK protects epithelial cells from anoikis by inactivating caspase-9 (Cardone, 1998), whereas caspase-8 is cleaved downstream from cytochrome c release as well as from caspase-9 and -2 activation during anoikis in intestinal epithelial cells (Grossman et al., 2001). Furthermore, it was recently reported that early events in mammalian epithelial cells are independent from caspase-8 activation (Wang et al., 2003). Conversely, it has been shown recently that loss of anchorage in endothelial cells leads to an increase in Fas and Fas ligand expression, whereas FLICE-inhibitory protein is downregulated (Aoudjit and Vuori, 2001). Moreover, expression of dominant-negative Fas-associated death domain protein (FADD-dn), blocks Fas-associated caspase-8-induced anoikis (Frisch, 1999; Rytomaa et al., 1999). The present work reports for the first time the complete caspase cascade during anoikis in human keratinocytes. In particular, the paper shows that caspase-8 triggers the apoptotic machinery, once β1-integrin signal is inactivated. This result is supported by the following data: caspase-8 activates earlier than caspase-9; caspase-8, but not caspase-9 inhibitor blocks anti β1 integrin-induced anoikis. Finally, anti-β1-integrin fails to induce anoikis in keratinocytes overexpressing c-FLIPL. The latter is in line with anoikis observed in endothelial cells where c-FLIP levels are highest in adherent and lowest in detached cells (Aoudjit and Vuori, 2001). Although it is clear that c-FLIP acts at the level of the DISC as a central regulator of death-receptor-mediated apoptosis (Thome and Tschopp, 2001), c-FLIP may have different cellular targets outside of the DISC. Ectopic overexpression may lead to the activation of other signaling pathways as reported (Kataoka et al., 2000), thereby inhibiting anoikis by autocrine release of factor(s) leading to resistance to anoikis. Further studies are required to delineate this point.
We have also shown that caspase-8 cleaves Bid, which is known to stimulate the release of cytochrome c. In addition, cytochrome c release is observed at 3 hours, and caspase-9 is also activated in our system, though at later time points. This seems to indicate that the intrinsic apoptotic pathway could participate in keratinocyte anoikis as a secondary positive feedback mechanism. This is in good agreement with previous work that shows an amplification of caspase-8-induced apoptosis through the mitochondrial release of cytochrome c (Kuwana et al., 1998). The authors also show that caspase-6 synergizes with caspase-8 in order to produce full nuclear apoptosis. In other words, in the presence of low concentrations of caspase-8, the mitochondrial pathway would take over, but, when caspase-6 is activated, it would cooperate with caspase-8 to induce a complete apoptosis. Similarly, our study shows that caspase-6 is activated at 6 hours, possibly directly activating caspase-3 (Allsop et al., 2000), thus increasing the apoptotic response before the mitochondrial pathway intervenes. However, according to the temporal cascade of events in our study, cytochrome c release appears to be directly stimulated by caspase-8, rather than induced by caspase-6, as proposed in other cell systems (Cowling and Downward, 2002). Why cytochrome c release occurs early and caspase-9 pathway is delayed remains to be determined.
Whereas in the case of executioner caspases (3, 6, 7) the activating event is proteolysis by an initiator caspase (8, 9, 10), there are no upstream proteases that activate the apical caspases. It has been hypothesized in an induced proximity model that caspases possess intrinsic enzymatic activity (Muzio et al., 1998) and are activated by dimerization of monomeric zymogens (Boatright et al., 2003). Yet, it remains to be determined how the inhibition of integrin signal can activate caspase-8. Whereas caspase-8 is normally activated by binding of a death receptor to its ligand, a direct link between integrins and caspase-8 in the induction of apoptosis has been proposed. Indeed, cells that are exposed to integrin antagonists may undergo a particular type of cell death, the so-called `integrin-mediated death'. This type of apoptosis appears to be dependent on the activation of caspase-8 in a manner that is independent of death receptors (Stupack et al., 2001). This form of cell death is different from anoikis which is caused by the loss of adhesion per se. We could therefore speculate that in our system, keratinocytes in suspension undergo anoikis through the extrinsic apoptotic pathway. As the addition of anti-β1-integrin increases the rate of apoptosis, antagonized integrins recruit additional caspase-8 and amplify cell death through the integrin-mediated cell death.
Psoriasis not only is characterized by an alteration in keratinocyte differentiation and proliferation, but also the apoptotic machinery is defective during the development of a psoriatic lesion (Qin et al., 2002; Takahashi et al., 2002). It has been proposed that in psoriasis, the hyperproliferative defect is localized in stem cells expressing β1 integrin. In addition, T-lymphocytes promote the proliferation of stem cells expressing high levels of β1-integrin from psoriatic keratinocytes but not from normal cells (Bata-Csorgo et al., 1995). Whereas β1-integrin expression is confined to the basal layer in normal epidermis, it is also expressed in suprabasal layers in psoriasis (Hertle et al., 1992; Pellegrini et al., 1992). Moreover, transgenic mice in which overexpression of integrin in suprabasal layers is induced have a psoriatic phenotype (Carroll et al., 1995; Haase et al., 2001). One could speculate that in psoriasis, the high levels of β1-integrin would keep stem cells in their niche and protect them from anoikis, thus favoring the development of psoriatic lesions. This makes pharmacological targeting of integrins a promising idea for the treatment of psoriasis.
The authors wish to thank John Reed, P. H. Krammer and H. Walczak for providing part of the reagents. The project was partially supported by a grant from the Italian MIUR (2003061751_004). Martin Leverkus was supported by grants from the Deutsche Krebshilfe (project 10-1951) and Wilhelm-Sander-Stiftung (project 2000.092.2). We also thank the Associazione Angela Serra for funding part of the project.