F-actin as a functional target for retro-retinoids: A potential role in anhydroretinol-triggered cell death

Of the three classes of vitamin A (retinol) metabolites, i.e. retinaldehydes, retinoic acids and retro-retinoids, the mechanisms of action have been delineated for the first two. Retinaldehyde combines with opsin to form the universal visual pigment, rhodopsin (Wald, 1968), and retinoic acids carry signals from the extracellular environment to the transcriptional apparatus (Evans, 1988). Although all three retinoid classes are united under the rubric of signalling, to date it is unclear how retro-retinoids exert their regulation in the cell. The basic fact is that many cell types fail to grow in culture when deprived of vitamin A. The growth-promoting properties were traced to derivative 14-hydroxy-retro-retinol (14HRR); retinoic acid is incapable of supporting cell survival (Buck et al., 1991). The cells can under certain conditions also synthesize another retro-retinoid, anhydroretinol (AR). AR not only counteracts the effects of 14HRR but can bring about rapid apoptosis of the cells. Because of the structural similarities between 14HRR and AR, and the mutual reversibility of their effects, competition for a common receptor(s) has been suggested (O’Connell et al., 1996).

The AR-triggered cell death displays characteristics of a nonclassical form of apoptosis. Morphologically, the cells undergo cell shrinkage, surface blebbing and ballooning followed by irreversible volume increase. Nuclear damage is relatively mild, as indicated by absence of chromatin condensation and ladder pattern of DNA fragmentation, although DNA nicks are discernible by terminal deoxynucleotidyl transferase assay at late stage. The process is rapid (3-6 hours for T-cell lines) and occurs independently of the presence of transcription and translation blockers, suggesting that the primary target(s), and the mechanisms leading to cell death, are cytoplasmic(O’Connell et al., 1996). The most abundant cytosolic protein, actin, has been demonstrated to be damaged during apoptosis induced by different stimuli (Song et al., 1997; Guenal et al., 1997; Levee et al., 1996). Farnesol and geranylgeraniol, lipids comprising linked isoprene moieties that are structurally closely related to retinoids, also caused disorganization of actin filaments and apoptosis (Miquel et al., 1996). Recently, Fas-induced damage of actin filaments was connected to gelsolin cleavage by caspase-3. Expression of the gelsolin cleavage product induced nuclear fragmentation, suggesting that cleaved gelsolin may be one physiological effector of morphological change during apoptosis (Kothakota et al., 1997). However, it is still unclear whether actin cytoskeleton damage is a consequence of apoptosis induced by various stimuli or vice versa, i.e. whether cytoskeletal injury is a pre-apoptotic event (Levee et al., 1996; Guenal et al., 1997). We present data here indicating that F-actin is a functional target for retro-retinoids and show that disorganization of actin cytoskeleton is a part of the cell death program.

Retro-retinoids, 14-HRR and AR, were synthesized as described (Derguini et al., 1994a,b; Shantz et al., 1943), purified by reversed-phase HPLC and stored in methanol under argon at −70°C. All-trans retinol, antibodies and phalloidin-FITC were purchased from Sigma Chemical Co. (St Louis, MO). Electrophoresis reagents were from Bio-Rad Laboratories (Cambridge, MA). The NIH3T3 cell line was from ATCC (CRL 1658), EL-4 murine thymic lymphoma was obtained from the cell culture collection at Memorial Sloan-Kettering Cancer Center (New York, NY). FDC-P1 cells infected with a retroviral construct containing the human bcl-2 gene or a control retroviral construct were kindly provided by Dr D. L. Vaux (Hall Institute, Melbourne, Australia; Vaux et al., 1988). Cell culture reagents were from Gibco-BRL (Gaithersburg, MD). Antibodies and all other reagents were from Sigma Chemical Co.

Murine EL4 T lymphoma cells were grown in RPMI 1640 containing 7% fetal bovine serum. For retinoid starvation experiments EL4 cells were cultured in serum-free insulin, transferrin, linoleic acid and bovine albumin (ITLB) containing medium (Garbe et al., 1992). Promyelocytic FDC-P1 transfected cells were grown in RPMI/10% fetal bovine serum in the presence of interleukin 3 (IL-3). NIH3T3 cells were cultured in DMEM supplemented with 10% fetal bovine serum until they reached 40-60% confluency. They were washed extensively in serum-free medium prior to pharmacological treatment. Retinoids, anhydroretinol, 14-hydroxy-4,14-retro-retinol (14-HRR), and retinol (ROL), were added from stock solution in methanol. Cells were incubated in medium containing 0.05% BSA with retinoids at 37°C. At the indicated time points cell viability was determined by Trypan Blue exclusion. Alternatively, intact cells were probed with 1 μg/ml of propidium iodide (PI) with subsequent analysis of fluorescence intensity by flow cytometry.

Measurement of DNA nicking was performed according to Gorczyca et al. (1993), viz. incorporation of biotinylated-dUTP at free 3′ OH ends of nicked DNA by TdT followed by FITC-avidin staining. NIH 3T3 cells were collected using trypsin, fixed with 3% paraformaldehyde in PBS, and analyzed after staining by flow cytometry. The apoptotic index was expressed as per cent of TUNEL-positive cells.

The treated cells were fixed for 15 minutes with 3% paraformaldehyde in PBS. Cells were then washed with PBS and quenched in 0.1 mol/l glycine in PBS for 15 minutes. After an additional wash the cells were permeabilized with 0.2% Triton X-100 (w/v) in PBS containing 1% bovine serum albumin (BSA) for 10 minutes, and nonspecific binding sites were blocked in 2% BSA. The cells were then treated for 30 minutes with 0.5 μM FITC-conjugated phalloidin. All the above incubations were performed at room temperature. After washing suspended cells were analyzed by flow cytometry. For inspection by fluorescence microscopy, parallel cell samples were mounted on glass slides in Vectashield medium.

Analytic flow cytometric measurements were performed using a FACScan flow cytometer with argon laser excitation at 488 nm. Green and red fluorescence were detected through a 515-545 nm and 563-607 nm bandpass filters, respectively. Ten thousand cells in each sample were analyzed. The PI positive cells from ungated populations exhibited damaged membranes and were considered dead. For determination of the F-actin amount the gate was defined as corresponding to scatter parameters on a dot plot of control viable cells. Avidin-FITC and phalloidin-FITC fluorescence intensities from the gated population were presented as a histogram. The geometrical mean value was used as a measure of F-actin amount.

Cytosol, membrane and cytoskeletal fractions were separated from cell homogenate as described (Korichneva et al., 1995). The cells were washed twice with cold PBS and scraped from the flasks in the buffer containing 20 mM Hepes, 250 mM sucrose, 1 mM EDTA, pH 7.4. Cells were spun down and resuspended in glycerophosphate buffer to which 1 mM PMSF, 10 μg/ml aprotinin and 10 μg/ml leupeptin were added. After homogenization with a glass Dounce homogenizer, the homogenate was centrifuged for 30 minutes at 15,000 g. The supernatant was centrifuged for 90 minutes at 120,000 g to minimize contamination of the cytosolic fraction with light membranes. The pellet was resuspended in 1% Triton X-100-containing glycerophosphate buffer with the same protease inhibitors. After 30 minutes incubation on ice, the homogenate was spun for 30 minutes at 12,000 g. The supernatant was saved as membrane fraction and the pellet was further washed 3-4 times with the Triton X-100 containing buffer and saved as the detergent-insoluble cytoskeletal fraction. Nuclei were isolated from NIH 3T3 cells according to Allo et al. (1991). Briefly, after washing the cells with PBS they were rinsed once with the buffer containing 10 mM Tris, 10 mM NaCl, 2.5 mM MgCl₂, pH 7.9 (buffer I), scraped from the dishes in this buffer containing protease inhibitors and then 0.3 M sucrose and 0.3% Triton X-100 were added. The cells were homogenized, the homogenate laid on an equal volume of the buffer I containing 0.6 M sucrose and centrifugated for 10 minutes at 1500 g. The pellet was resuspended in the buffer I, washed and saved as nuclear fraction. Each fraction was mixed with 4× Laemmli buffer and boiled for 5 minutes prior to electrophoresis.

Proteins were separated by 8.5% SDS-PAGE and electrophoretically transferred to PVDF membranes (70 V, 1.5 hours). Non-specific binding sites of PVDF sheets were then blocked for 1 hour at 37°C with 5% nonfat milk in PBS, pH 7.4, 0.1% Tween 20. After washing with PBS, 0.1% Tween 20, the sheets were incubated for 1 hour at room temperature with the primary antibody (e.g. mouse anti-actin, 1/500). Then, the membranes were washed and probed with the secondary antibody coupled to peroxidase (anti-mouse IgG, 1/4000) for 1 hour at room temperature. The sheets were washed and the proteins were revealed using the ECL detection system (New England Biolabs) according to the manufacturers’ instructions. The ECL blots were exposed on Kodak XAR film for 10 seconds or up to several minutes.

Results are expressed as mean ± s.d. and evaluated by Student’s t test.

Upon addition of AR, lymphocytes and fibroblasts undergo morphological alterations typified by a change from their characteristic irregular shape to a rounded shape with widespread surface blebbing. As these changes suggested damage to the cytoskeleton, the relative amount of filamentous actin (F-actin) was monitored using fluorescence of FITC-conjugated phalloidin, phallotoxin from Amanita phalloides, a selective label of F-actin. The population of EL4 lymphocytes was homogeneous in their polymerization state of actin, as deduced from the relatively narrow symmetrical peak revealed by FACS analysis (Fig. 1A). After 1 hour of AR treatment, however, the distribution broadened and the mean fluorescence intensity decreased, indicating an increase in the cells with low amount of F-actin (Fig. 1B). The effective doses of AR ranged from 10⁻⁵ to 10⁻⁷ M (Fig. 2A). Accordingly, micromolar doses were used in further experiments because they caused a maximal effect yet without having non-specific toxicity, as determined by reversibility with 14HRR (discussed later). The kinetics of F-actin loss were compared to the kinetics of cell death. The result of a representative experiment is shown in Fig. 2B. A significant loss of F-actin developed in the majority of lymphocytes within 1 hour of AR treatment, when no dead cells had yet appeared. The mean fluorescence intensity, indicating the quantity of F-actin, decreased by up to 39% within 1 hour and 71% in 3 hours. It should be noted that gating on a small population of live cells according to scatter parameters (Darzynkiewicz et al., 1997), rather than the entire population, gave identical results. The comparison of the time course of decrease in mean fluorescence intensity with that of Trypan Blue uptake showed that F-actin loss preceded cell death (Fig. 2B).

Since it is difficult to distinguish morphological details of actin structures in EL4 lymphocytes due to their scant cytoplasm, we used NIH3T3 fibroblasts for microscopic inspection. Importantly, the onset of cell death is significantly delayed in these cells compared to lymphoma or promyelocytic cells. However, the criteria applied to judge the extent of apoptosis in AR-treated NIH3T3 cells are similar to those in lymphocytes (O’Connell et al., 1996) and include DNA damage assessed by TUNEL technique. Fig. 3 shows that DNA nicks develop in fibroblasts treated with increasing doses of AR for 48 hours. More that 40% of the cells become TUNEL-positive at 10 μM AR by this time.

The fluorescent images of NIH3T3 cells stained with phalloidin-FITC (Fig. 4A-I) displayed substantial reorganization and disassembly of actin filaments. In control conditions (serum-free medium) most of the cells had abundant and well-organized stress fibers at 3, 6 and12 hour time points (Fig. 4A,C,E). After 24 hours in serum-free medium stress fibers were still prevalent, although a mild disorganization of actin filaments was noticeable (Fig. 4G). This damage could be ascribed to the retinoid starvation (see Fig. 5). At 3 hours of AR treatment, however, phalloidin-FITC decorated the cell periphery with images of membrane ruffles, and the decrease in organized arrays of stress fibers became pronounced. Some cells displayed membrane ruffles and stress fibers simultaneously. Later, at 6 hours of treatment, fluorescent images displayed short, thick, stick-like and punctate fluorescent areas at the expense of organized stress fibers, notably in the nuclear area (Fig. 4D). These images are reminiscent of the morphological changes documented in cells induced into apoptosis by serum deprivation (Brancolini et al., 1997; Kulkarni and McCulloch, 1994). Further along, the organized actin structures disappeared completely and were replaced by an increasingly diffuse staining pattern. Fig. 4I shows the cell at a late stage of AR response (26 hours). While F-actin was completely depolymerized the nucleus remained intact. The condensed nuclei appeared at the time of plasma membrane dissolution (not shown). Therefore, the changes in actin structures preceded nuclear alterations, thus supporting the notion that AR-triggered apoptosis is independent of nuclear processes. The fluorescence experiments were corroborated by western blotting experiments with an anti-actin antibody. They revealed no reduction in the intensity of the 42 kDa actin band in the membrane fraction, and a slight increase in the cytosolic fraction, and no change in the nuclear fraction at 6 hours after AR administration, whereas the relative proportion of actin diminished substantially in the detergent-insoluble cytoskeletal fraction (Fig. 4H).

Since 14HRR and AR represent an agonist/antagonist pair of regulators it was of interest to test whether 14HRR could reverse or prevent the action of AR. This was the case; mixing AR with an equimolar concentration of 14HRR prevented F-actin destruction. Fig. 1C,D shows the FITC-phalloidin fluorescence distribution of F actin in EL4 cells treated with AR alone or in combination with retinoids. Incubation of the cells with 14HRR or retinol alone gave results similar to the control (not shown). Increasing the AR concentration over that of 14HRR re-instituted apoptotic conditions.

AR-induced cell death has striking similarities to cell cycle-independent cell death consequent to retinol deficiency (Buck et al., 1991). In either case, the cell morphology displayed vacuolization and disintegration of the cytoplasm while nuclei remained intact. It is likely that in both cases the critical issue is availability of 14HRR (or retinol) to a putative receptor. We measured the F-actin content in serum-starved lymphocytes cultured in the presence or in the absence of exogenous retinol. Fig. 5 shows that the EL4 cells partitioned into two populations according to their F-actin content. Following serum starvation the cells with low levels of F-actin were the majority. Retinol (line b) significantly preserved the F-actin content in serum-starved cells and enriched the cell population with high F-actin. Together these findings demonstrate that F-actin is a common target for regulation by retinoids.

AR-triggered cell death was characterized as a non-classical form of apoptosis (O’Connell et al., 1996). Since most forms of apoptosis are counteracted by bcl-2 it was desirable to check whether AR-induced cell death was also controlled by the bcl-2 gene and whether actin reorganization in turn depended on bcl-2. We used isogenic FDC-P1 myelocytic cell lines expressing bcl-2-containing or empty retroviral constructs as described previously (Vaux et al., 1988). Control transfected cells elicited the same effect of AR on actin content as EL4 lymphocytes, whereas F-actin in bcl-2-overexpressing cells was resistant to the damaging influence of AR (Fig. 6). Concomitantly, overexpression of bcl-2 protected myelocytic cells from apoptosis as assessed by Trypan Blue exclusion (data not shown).

Although filamentous actin was reported to depolymerize during apoptosis (Endersen et al., 1995; Levee et al., 1996; Van de Water et al., 1996), no causal link to cell death was established. Pharmacological tools that affect specifically F-actin were applied to clarify this question. Cytochalasin B has been long known to disturb actin filament structures (MacLean-Fletcher and Pollard, 1980), and it is widely used to study the role of cytoskeletal reorganizations in different cell functions (Manie et al., 1997; Zhang et al., 1997). On the other hand, phallotoxins, which bind to specific sites on F-actin, promote actin polymerization and stabilize microfilaments against the action of depolymerizing factors (Miki et al., 1987). Recently a macrocyclic peptide, jasplakinolide, isolated from the marine sponge Jaspis johnstoni, became available as a potent inducer of actin polymerization (Bubb et al., 1994) and its application has furthered our understanding of the involvement of the cytoskeleton in different physiological functions, e.g. adhesion (Stewart et al., 1998), endocytosis (Shurety et al., 1998) and ion transport (Matthews et al., 1997). The availability of specific drugs with opposite effects on the cytoskeleton offered the opportunity to question whether alternately weakening, or reinforcing, actin filaments would predispose, or armor, cells against apoptosis. This was clearly the case. We used cytochalasin B to disrupt filamentous actin, and observed a heightened sensitivity to apoptosis induction by AR (Fig. 7). Disorganization of the actin cytoskeleton by preincubation with cytochalasin B changed the cell morphology from angular to rounded cells, and it accelerated the rate of cell death triggered by AR. Conversely, when jasplakinolide was applied to stabilize actin filaments the onset of apoptosis was significantly delayed.

The cytoskeleton not only provides cells with mechanical stability but also serves as a major scaffold to organize parts of the signalling machinery that monitors and regulates adhesion and intercellular communications. The assembly of a complex array of signalling molecules at focal adhesion points is well documented (Clark and Brugge, 1995). If one accepts the central organizational role of the cytoskeleton in signal transduction it becomes clear that disruption of the cytoskeleton would dilute survival signals and thus contribute to apoptosis. The literature contains a few examples where both positive and negative correlations between the state of actin cytoskeleton and survival/death response have been observed. For example, IL-4 has been found to activate cytoskeletal arrangement and at the same time delayed neutrophil apoptosis (Girard et al., 1997). Conversely, a human melanoma cell line unable to express the actin-binding protein 280 was found to be defective in the proper assembly of actin in response to force application and underwent apoptosis (Glogauer et al., 1998). We report here that F-actin reorganization emerged as an early event in non-classical apoptosis triggered by AR, and that stabilized actin filaments promote cell resistance to apoptosis. AR belongs chemically to the class of retinoids, vitamin A and its metabolites, known to play a role in many fundamental physiological processes, e.g. embryogenesis, cell growth and differentiation, reproduction, vision etc. A dichotomy exists in the function of retinoid metabolites in that the retinoic acids regulate gene expression and differentiation (Szondy et al., 1998; Nagy et al., 1998), whereas the class of hydroxylated retinoids, comprising 14HRR and 13,14-dihydroxy-retinol perform a regulatory function in cell survival/cell death decisions (Buck et al., 1993; Derguini et al., 1995; O’Connell et al., 1996; Chen et al., 1997).

The view that retinoids follow a strict dichotomy is not unopposed since RA purportedly either induces, supports or inhibits apoptosis in various systems, although the available evidence sheds little information on how this regulation occurs. Moreover, the biological relevance is unclear since the doses at which retinoic acid elicited cell death were not physiological, and the effects themselves, at best, were modest (Fesus et al., 1995; Iwata et al., 1992). In our hands micromolar doses of RA triggered apoptosis but not actin depolymerization (our unpublished results). Recently, induction of apoptosis by retinoids through a mechanism not involving nuclear receptors has been documented with the synthetic compound 6-{3-(1-adamantyl)-4-hydroxyphenyl}-2-naphthalene carboxylic acid (Hsu et al., 1997). This is reminiscent of the apoptotic cell death induced by AR that we had previously recognized as a fast, non-nuclear form of apoptosis (O’Connell et al., 1996). The molecular mechanisms of the apoptotic process induced by AR have not been established. DNA nicks were detected in both AR-treated lymphocytes and fibroblasts. However, in earlier studies retro-retinoids failed to bind to RAR and RXR nuclear receptors (Mangelsdorf et al., 1994). More recently our unpublished studies suggest that these retinoids bind with high affinity to serine/threonine kinases (Hoyos, B. and Hämmerling, U., unpublished), lending credence to the idea that the target of interaction, and the primary site of apoptotic injury, resides in the cytoplasm.

Lymphocytes and fibroblasts damaged by AR displayed rapid and widespread blebbing of the surface membrane followed by changes in cell shape. Volume measurements that we performed show cell shrinking within the first hour of treatment followed by dramatic swelling (our unpublished results), thus indicating again active processes occurring in the cytosol after AR administration. Irreversible, uncontrolled swelling would lead to membrane rupture without disturbing the nucleus, a likely scenario in AR-triggered apoptosis. Cell shape and volume relate to cytoskeletal rearrangements. Thus the observed changes in filamentous actin and F-actin depolymerization are consistent with apoptosis initiated primarily in the cytoplasm. It is noteworthy that a delayed and muted AR effect on F-actin was observed in cells pretreated with herbimycin A (our unpublished results) suggesting that regulation and possible interference with kinase function rather than direct interaction of retinoids with actin, were responsible. Actin cytoskeleton reorganization, concomitant with nuclear fragmentation, was recently described in NIH3T3 cells after serum deprivation (Brancolini et al., 1997). In our experiments, AR treatment of NIH3T3 cells lead similarly to accumulation of actin in the nuclear area but also to membrane ruffle formation, and to depolymerization at a later stage of apoptosis. In its time course the reorganization of the cytoskeleton precedes nuclear damage. First, phalloidin binding was significantly reduced much earlier than the development of membrane permeability and, second, gating on a small population of live cells according to scatter parameters (Darzynkiewicz et al., 1997) also documented the loss of F-actin.

As mentioned earlier an antagonistic relationship between AR and 14HRR was noted in pharmacological studies of cell growth (Buck et al., 1993). The mutual reversibility of effects of these retro-retinoids is reflected once again in their action on F-actin. Morphological changes, depolymerization of filamentous actin and cell death were all reversed when 14HRR was in excess of AR. The best explanation for reversible inhibition is competition for the same binding site(s). Receptor blockade by AR may equal the physiological situation of vitamin A deprivation and the consequent decline in the endogeneous 14HRR. Indeed, vitamin A withdrawal from T and B cells causes apoptosis in a manner indistinguishable from that induced by AR, except for a more gradual onset that is probably related to the consumption over time of stored retinyl esters (Eppinger et al., 1993).

Bcl-2 plays a central role in guarding the gateway to the common core execution pathway. Thus it was unsurprising that overexpression of bcl-2 protected the cells from apoptosis induced by AR. However, it was unexpected that actin depolymerisation was also inhibited in the cells overexpressing bcl-2, since it implied that actin reorganization is operationally downstream of bcl-2-related processes. This assumption is only valid if one subscribes to a strictly linear relationship. It might be more appropriate to consider a complex communication network, including feedback loops.

The role of cytoskeleton reorganization in programmed cell death has not yet been deciphered. Our experimental data suggest that actin stability is a positive factor in the vitality of the cells. Enforced stabilization of actin filaments by jasplakinolide renders cells resistant to apoptotic signals, while weakening the cytoskeleton with cytochalasin B promotes apoptosis. The consequences of AR treatment or retinol deprivation are pleiotropic and affect, besides the cytoskeleton, the mitochondria (Korichneva, I. and Hämmerling, U., unpublished results). Other mechanisms may also contribute to retinoid cytotoxicity. Thus, F-actin destruction, in a non-classical form of apoptosis induced by AR, appears to be an important part in a plurality of events triggering programmed cell death. Dissecting the pathways by which retinoids control cell survival will provide valuable insights on how decisions are made by cells whether to live or die.

We thank Dr Fadila Derguini for providing anhydroretinol and for helpful discussion. This work has been supported by grants from the National Institute of Health (GM 45799) and the Johnson & Johnson Focused Giving Program.