Induction of apoptosis in U937 human leukemia cells by ... - Nature

ã

Oncogene (1999) 18, 7016 ± 7025 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc

Induction of apoptosis in U937 human leukemia cells by suberoylanilide hydroxamic acid (SAHA) proceeds through pathways that are regulated by Bcl-2/Bcl-XL, c-Jun, and p21CIP1, but independent of p53 JA Vrana1, RH Decker3, CR Johnson1, Z Wang1, WD Jarvis1, VM Richon5, M Ehinger6, PB Fisher7,8 and S Grant*,1,2,3,4 1

Department of Medicine, Medical College of Virginia, Richmond, Virginia, USA; 2Department of Pharmacology, Medical College of Virginia, Richmond, Virginia, USA; 3Department of Biochemistry, Medical College of Virginia, Richmond, Virginia, USA; 4 Department of Microbiology, Medical College of Virginia, Richmond, Virginia, USA; 5Department of Developmental Cell Biology and Genetics, Memorial Sloan-Kettering Cancer Center, New York, NY, USA; 6Research Department-2, Lund University Hospital, Lund, Sweden; 7Department of Pathology, Columbia University College of Physicians and Surgeons, New York, NY, USA; 8 Department of Urology, Columbia University College of Physicians and Surgeons, New York, NY, USA

Determinants of dierentiation and apoptosis in myelomonocytic leukemia cells (U937) exposed to the novel hybrid polar compound SAHA (suberoylanilide hydroxamic acid) have been examined. In contrast to hexamethylenbisacetamide (HMBA), SAHA-related maturation was limited and accompanied by marked cytoxicity. SAHA-mediated apoptosis occurred within the G0G1 and S phase populations, and was associated with decreased mitochondrial membrane potential, caspase-3 activation, PARP degradation, hypophosphorylation/cleavage of pRB, and down-regulation of c-Myc, c-Myb, and B-Myb. Enforced expression of Bcl-2 or BclxL inhibited SAHA-induced apoptosis, but only modestly potentiated dierentiation. While SAHA induced the cyclin-dependent kinase inhibitor p21CIP1, antisense ablation of this CDKI increased, rather than decreased, SAHA-related lethality. In contrast, conditional expression of wild-type p53 failed to modify SAHA actions, but markedly potentiated HMBA-induced apoptosis. Finally, SAHA modestly increased expression/activation of the stress-activated protein kinase (SAPK/JNK); moreover, SAHA-related lethality was partially attenuated by a dominant-negative c-Jun mutant protein (TAM67). SAHA did not stimulate mitogen-activated protein kinase (MAPK), nor was lethality diminished by the speci®c MEK/MAPK inhibitor PD98059. These ®ndings indicate that SAHA potently induces apoptosis in human leukemia cells via a pathway that is p53-independent but at least partially regulated by Bcl-2/Bcl-xL, p21CIP1, and the c-Jun/AP-1 signaling cascade. Keywords: SAHA; HMBA; apoptosis; dierentiation; leukemia; p53

Introduction SAHA (suberoylanilide hydroxamic acid) is a second generation hybrid polar compound recently shown to be a potent inducer of dierentiation in murine-

*Correspondence: S Grant, Department of Medicine, Medical College of Virginia, MCV Station Box 230, Richmond, VA 23298, USA Received 6 May 1999; revised 19 August 1999; accepted 23 August 1999

erythroleukemia (MEL) cells (Richon et al., 1996). The prototype of this class of compounds, HMBA (hexamethylenebisacetamide), has undergone evaluation in humans and has exhibited clinical activity in patients with myelodysplastic syndromes and leukemia (Andreef et al., 1992). In MEL cells, SAHA is approximately 2000-fold more potent than HMBA in promoting dierentiation, but is less eective in downregulating the proto-oncogene c-Myb (Richon et al., 1996). More recently, treatment of MEL cells with SAHA (but not HMBA) has been found to inhibit histone deacetylases, enzymes implicated in the transcriptional regulation of genes involved in cellular maturation (Richon et al., 1998). Together, these ®ndings suggest that SAHA and HMBA engage separate programs of leukemic cell maturation. A complex relationship exists between dierentiation and apoptosis (Ra, 1992). HL-60 human promyelocytic leukemia cells induced to undergo maturation by retinoic acid ultimately die an apoptotic death, although as a relatively late event (Martin et al., 1990). However, dierentiation and apoptosis are mutually exclusive in other systems (Selvakumaran et al., 1994). Moreover, dysregulated maturation of U937 human leukemia cells (e.g., following enforced expression of protein kinase C-z) leads to apoptosis rather than dierentiation after treatment with phorbol myristate acetate (PMA) (de Vente et al., 1995). Collectively, these observations suggest that apoptosis and maturation represent alternative cellular fates, and that disruption of normal dierentiation predisposes cells to an apoptotic form of cell death. It is currently unknown whether, and to what extent, SAHA can elicit dierentiation in leukemia cells of human origin. The myelomonocytic cell line U937 (Sundstrom and Nilsson, 1976) matures along a monocyte/macrophage lineage in response to tumorpromoting phorboids such as PMA (Hass et al., 1991). In the present study, the capacities of SAHA and HMBA to induce dierentiation and apoptosis in U937 cells have been compared in order to identify determinants of these fundamentally dierent cellular fates. To this end, we have employed a series of U937 mutants exhibiting functional alterations in several genes implicated in cell cycle arrest, dierentiation, and cell death. Herein we report that when administered above threshold concentrations, SAHA potently

SAHA-mediated apoptosis in human leukemia cells JA Vrana et al

induces loss of U937 cell mitochondrial membrane potential and apoptosis rather than maturation, suggesting that the former processes may contribute to the limited capacity of this compound to induce dierentiation. Furthermore, SAHA-induced apoptosis proceeds via pathways independent of the cell cycle checkpoint gene p53, but that are at least partially dependent upon or in¯uenced by expression of Bcl-2/ Bcl-xL, p21CIP1, and an intact SAPK-JNK/AP1 cascade.

Results SAHA induces mitochondrial permeability transition and triggers apoptosis in a dose- and schedule-dependent manner When cells were exposed to 1 mM SAHA for 24 h, the extent of apoptosis was modest (*10%), and increased only marginally thereafter (Figure 1a). However, at SAHA concentrations 52.5 mM, apoptosis increased markedly (e.g., to *40% at 24 h), and approached 100% after 72 h. In contrast, HMBA induced apoptosis in only *4% of cells over a similar time frame (not shown). Consistent with these results,

oligo-nucleosomal DNA fragments were readily apparent in cells exposed to 52.5 mM SAHA, but not in cells treated with HMBA (Figure 1b). Loss of mitochondrial membrane potential (DCm) was modest after a 12 h exposure to 2.5 mM SAHA, but increased to *80% after 24 h (Figure 1c). At all intervals, changes in DCm preceded the onset of apoptosis. Exposure of cells to 2.5 mM SAHA also resulted in time-dependent degradation of PARP to its 85-kDa fragment and cleavage/activation of full length 32-kDa procaspase-3 (Figure 1d), but did not appreciably alter levels of Bcl-2 or Bax (not shown). In separate studies, the dose-response pro®le of HL-60 promyelocytic leukemia cells to SAHA-mediated apoptosis was essentially identical to that of U937 cells (Figure 1c; % apoptosis at 2.5 mM shown). Bcl-2 and Bcl-xL oppose SAHA-induced apoptosis, but only marginally augment SAHA-mediated maturation Expression of the myelomonocytic maturation marker CD11b was observed in *20% of U937 cells exposed to 1 mM SAHA for 72 h (Figure 2a), vs 440% of cells treated with HMBA (not shown). However, when cells were exposed to lethal concentrations of SAHA (e.g.,

Figure 1 SAHA induces apoptosis in U937 cells. (a) Dose response and time course of U937 cell apoptosis following treatment with SAHA. Apoptotic cells were evaluated by morphological assessment as described in the text; values represent the means for triplicate experiments+s.d. & Control, ~ 0.1 mM, ! 0.25 mM, ^ 0.5 mM, * 1.0 mM, & 1.5 mM, ~ 2.5 mM, ! 5.0 mM, ^ 10.0 mM. (b) Agarose gel electrophoresis assay of internucleosomal DNA fragmentation. Cells were treated with the designated concentrations of SAHA or 3 mM HMBA and harvested at 24 h. Lanes (1) 100 bp DNA marker (GIBCO ± BRL), (2) Control, (3) 0.5 mM SAHA, (4) 1.0 mM SAHA, (5) 2.5 mM SAHA, (6) 5.0 mM SAHA, (7) 3.0 mM HMBA. (c) Temporal relationship between mitochondrial membrane potential and morphological assessment of apoptosis in cells exposed to SAHA. At the indicated intervals, cells treated with 2.5 mM SAHA, were harvested and either incubated with 40 nM DiOC6 and analysed by cyto¯uorometer or cytocentrifuged for morphological analysis. & % apoptosis, & (% apoptosis in HL-60 at 2.5 mM), * % shift in DiOC6; * 3 mM HMBA (% shift in DiOC6 at 24 h). (d) Western blots illustrating time dependent activation of caspase-3 and PARP cleavage after treatment with SAHA. Top panel: Time course of cleavage of the 115 kD native form of PARP to the 85 kD cleavage fragment. Middle panel: Activation of caspase-3 was assessed by the reduction in the full length 32 kD form

7017


7018

52.5 mM), CD11b expression was negligible. As observed in MEL cells (Richon et al., 1998), treatment of U937 cells with SAHA (but not HMBA) inhibited deacetylation of histones H3 and H4 (Figure 2a, inset). Enforced expression of either Bcl-2 or Bcl-xL aorded signi®cant protection against SAHA-induced apoptosis at 24 h (Figure 2b,d) as well as at 72 h (not shown). However, despite clear evidence of cytoprotection, these cells exhibited only a modest increase in sensitivity to SAHA-mediated dierentiation compared to their wild-type counterparts (2c,e). These data indicate that SAHA-mediated apoptosis proceeds along pathways that are at least partially regulated by Bcl-2 and Bcl-xL, and suggest that factors other than, or in addition to induction of cell death are responsible for the failure of SAHA to engage a normal dierentiation program.

SAHA induces apoptosis in the G0G1 and S phase cell populations A 24 h exposure to HMBA or 1 mM SAHA exerted minimal eects on U937 cell cycle traverse (Table 1). Treatment with 2.5 mM SAHA, sharply increased the subdiploid (apoptotic) population, and substantially decreased the G0G1 and S phase fractions; in contrast, the G2M population displayed a modest but signi®cant expansion. When apoptosis was inhibited (e.g., by enforced expression of Bcl-2 or Bcl-xL), SAHA-mediated reductions in the G0G1 and S phase populations were diminished; in fact, following a 72 h exposure to 2.5 mM SAHA, Bcl-2- and Bcl-xL-overexpressing cells experienced an increase in the G0G1 population. Finally, S phase synchronization of cells by aphidicolin did not significantly alter the apoptotic response to 2.5 mM SAHA (not

Figure 2 Relationship between dierentiation and apoptosis in SAHA treated U937, U937/Bcl-2, and U937/BclxL cells. (a) CD11b expression was evaluated in U937 cells after 72 h of treatment with the designated concentrations of SAHA. Cell suspensions were mixed with phycoerythrin-labeled antibody (CD11b or IgG2a) and analysed by ¯ow cytometry. Values represent the percentage of cells expressing CD11b positivity less the corresponding IgG control, and correspond to the means for nine experiments+s.e.m. (Inset): Western blots illustrating the presence of acetylated histone 3 and histone 4 in cell extracts from SAHA-(1.0 or 2.5 mM) but not in HMBA-(3 mM) treated cells. (b) and (d) U937 cells stably transfected with either Bcl-2 (b) or Bcl-xL (d) (corresponding controls, pCEP4 and pcDNA3.1) were exposed to the designated concentrations of SAHA (24 h) after which apoptosis was determined by morphological assessment. * U937/pCEP4, * U937/Bcl-2, & U937/pcDNA3.1, & U937/Bcl-xL. (c) and (e) U937 cells stably transfected with either Bcl-2 (c) or Bcl-xL (e) (corresponding controls, pCEP4 and pcDNA3.1) were exposed to the designated concentrations of SAHA (72 h) after which the percentage of cells expressing CD11b was assessed. Values represent the means for six separate experiments+s.e.m


shown), consistent with the notion that S phase cells are vulnerable to SAHA-mediated cell death. SAHA induces pRB dephosphorylation and cleavage, and extensive down-regulation of c-Myc, c-Myb, and B-Myb To determine whether SAHA-mediated apoptosis might be related to dysregulation of genes implicated in cellular maturation, the phosphorylation status of pRB and the expression of c-Myc, c-Myb, and B-Myb were monitored in U937 cells exposed to HMBA and either 1 or 2.5 mM SAHA. Western blot analysis employing an antibody speci®c for the hypophosphorylated form of pRb (Dunaef et al., 1994) revealed marked pRb dephosphorylation in response to HMBA, and, to a considerably lesser extent, to 1 mM SAHA (Figure 3a). Cells exposed to 2.5 mM SAHA displayed a prominent dephosphorylated pRB species migrating at *65 kDa, suggestive of a pRB cleavage product. A similar phenomenon was observed in blots probed with an antibody recognizing both phosphorylated and dephosphorylated pRB (not shown). HMBA induced c-Myc down-regulation as early as 12 h after treatment, which persisted throughout the 72 h exposure interval (Figure 3b). One mM SAHA elicited a similar response, although the extent of downregulation was slightly less than that seen with HMBA. However, 2.5 mM SAHA essentially abrogated c-Myc expression throughout the exposure period. Both HMBA and 1 mM SAHA induced modest down-regulation of c-Myb after 6 h exposure, followed by complete recovery by 48 h (Figure 3c). In contrast, 2.5 mM SAHA induced virtually complete c-Myb down-regulation by 6 h, and recovery did not occur over the ensuing 72 h. These ®ndings dier from earlier results obtained in MEL cells, in which HMBA, but not SAHA, down-regulated c-Myb after 48 h (Richon et al., 1996). Neither HMBA nor 1 mM SAHA down-regulated c-Myb, whereas 2.5 mM SAHA abrogated B-Myb expression after 24 h. The CDKI p21CIP1 is induced by SAHA but is not required for SAHA-induced apoptosis The CDKI p21CIP1, implicated in G1 arrest during exposure of leukemic cells to PMA (Jiang et al., 1994) was minimally expressed in cells treated with 1.0 mM SAHA, but potently induced in cells treated with HMBA (at 48 h) or 2.5 mM SAHA (at 24 h) (Figure 4a). To assess the functional role of p21CIP1 in SAHArelated toxicity, a U937 cell line stably transfected with a p21CIP1 antisense construct was employed (Wang et al., 1998). As anticipated, SAHA (2.5 mM) failed to induce p21CIP1 in antisense-expressing cells (U937/p21AS), but potently up-regulated p21CIP1 expression in controls (U937/pREP) (Figure 4b). However, p21CIP1-antisense cells were signi®cantly more sensitive to SAHA-mediated apoptosis throughout the concentration range tested (1 ± 5 mM). In contrast, dysregulation of p21CIP1 did not increase HMBA-mediated apoptosis (inset). These ®ndings indicate that while SAHA is a potent inducer of p21CIP1 expression in U937 cells, its ability to induce apoptosis does not require, but instead may be limited by expression of this CDKI.

7019

SAHA-mediated apoptosis proceeds through a p53independent pathway U937 cells are functionally p53 null as a consequence of a large deletion in the p53 gene (Sugimoto et al., 1992). Other studies assessed the functional role of p53 in SAHA- and HMBA-mediated apoptosis in U937 cell transfectants stably expressing temperature-sensitive p53 constructs (p53135val?ala). At the permissive temperature (328C), the wild-type p53 conformation is assumed, while the mutant conformation predominates at the non-permissive temperature (378C) (Michalovitz et al., 1990). When probed with an antibody recognizing the wild-type conformation, p53

Figure 3 Expression levels of pRB, c-Myc, c-Myb, and B-Myb following exposure to SAHA or HMBA. (a) Western blots illustrating the cleavage and phosphorylation status of the retinoblastoma protein (pRB) following exposure to either 3 mM HMBA, 1 or 2.5 mM SAHA for the indicated interval. For these studies, an antibody recognizing only the underphosphorylated form of the pRB protein was used as described in the text. Blots were reprobed with an anti ± actin antibody to insure equal loading and transfer. (b) Western blot illustrating down-regulation of c-Myc following exposure to either 3 mM HMBA or 1 mM and 2.5 mM SAHA for the indicated intervals. (c) Western blots illustrating c-Myb or B-Myb expression in cells exposed to HMBA (3 mM) or SAHA (1 or 2.5 mM) for the designated interval

Table 1 Treatment Control HMBA (3 mM) SAHA (1 mM) SAHA (2.5 mM)

U937 Cells ± 24 h cell cycle pro®le Subdiploid

G0/G1

S

G2/M

3+1 7+1 7+3 51+5

42+1 39+2 46+1 21+4

47+2 47+2 39+2 14+2

8+1 11+1 7+1 19+2

U937 cells were exposed to HMBA (3mM) or SAHA (1 or 2.5 mM) and cell cycle distribution analysed by ¯ow cytometry after 24 h. Consistent with previous reports, HMBA induced G0/G1 arrest (67% at 48 h and 78% at 72 h), whereas exposure to a sublethal concentration of SAHA (1 mM) induced only minimal G0/G1 arrest (50% at 48 h and 60% at 72 h). Values are expressed as the mean percentage (relative to the total cell population) for three independent experiments+s.d.


7020

Figure 4 The apoptotic response elicited by SAHA is not dependent on p21CIP1 expression. (a) Western blot illustrating p21CIP1 expression following treatment with either 3 mM HMBA or 2.5 mM SAHA for the indicated intervals. (b) Western blot illustrating p21CIP1 expression levels following treatment of U937 cells stably transfected with an p21CIP1 antisense construct (p21AS) or its corresponding control, pREP4, with 2.5 mM SAHA (24 h). (c) Induction of apoptosis in p21AS cells (*) and corresponding controls (pREP4) (&) following exposure to the designated concentrations of SAHA for 24 h. The extent of apoptosis in cells exposed to 3 mM HMBA (24 h) is shown in the inset. Values for apoptosis are expressed as the means+s.e.m. for triplicate experiments; **P50.001 for p21AS versus vector (pREP4) control

was expressed to varying degrees at 328C in three separate clones (B3, A2, and A5), but not at 378C or in empty-vector controls (M2) (not shown). Notably, HMBA-mediated apoptosis was signi®cantly enhanced in each subline expressing wild-type p53 (Figure 5). In contrast, wild-type p53 expression failed to potentiate SAHA-mediated apoptosis in any of the clones evaluated, suggesting that the responses to HMBA and SAHA respectively proceed along p53-dependent and -independent pathways. At 378C, only SAHA (2.5 mM) induced p21CIP1 after 24 h (Figure 5, lower panel), consistent with results shown in Figure 4. However, at 328C, constitutive expression of p21CIP1 was observed in the three temperature-sensitive mutants, but not in vector controls. A similar phenomenon has been reported in Saos-2 osteosarcoma cells (Tang et al., 1998). Nevertheless, increases in expression of p21CIP1 over basal levels were most prominent in HMBA-treated cells, which also displayed evidence of a cross-reactive *17-kDa species, possibly representing a p21CIP1 cleavage product (Gervais et al., 1998). SAHA-mediated apoptosis is partially dependent upon an intact SAPK-JNK and c-Jun/AP1 pathway, but is independent of MEK/MAPK Leukemic cell maturation has been linked to activation of c-Jun and the AP1 transcription factor (Franklin and Kraft., 1995) as well as mitogen-activated protein kinase (MAPK) (Kharbanda et al., 1994). To assess the eects of SAHA on these pathways, expression of the phosphorylated forms of stress-activated protein kinase (SAPK) and MAPK was monitored. Following

exposure of cells to HMBA or SAHA (2.5 mM), activated phospho-JNK was modestly increased over basal levels, as were c-Jun (Figure 6a) and phospho-cJun (not shown). However, the extent of these responses was clearly less than that observed in cells exposed to 10 nM PMA (Figure 6b). Neither HMBA nor SAHA increased expression of phospho-ERK over basal levels, in contrast to PMA, which induced a robust increase (not shown). To assess the functional contribution of c-Jun/AP1 to SAHA-mediated apoptosis, a c-Jun transactivation de®cient mutant (TAM67) was employed (Brown et al., 1994, Grant et al., 1996a). Two separate TAM67-expressing clones exhibited modest, but statistically signi®cant, reductions in SAHA-induced apoptosis compared to vector controls (Figure 6c). This ®nding indicates that SAHAmediated lethality toward U937 cells depends, at least in part, upon an intact c-Jun/AP1 axis. Consistent with the failure of SAHA to stimulate ERK activity, SAHA-related apoptosis was not attenuated by the MEK1 inhibitor PD98059 nor by the p38 reactivating kinase (RK) inhibitor SB 203580 (Figure 6d). Discussion The ®ndings described herein indicate that SAHA, a second generation hybrid bipolar compound that stimulates dierentiation in MEL cells (Richon et al., 1996), is a highly potent inducer of apoptosis in human leukemia cells of myeloid origin (U937 and HL-60). SAHA, like the four-carbon short-chain fatty acid butyrate (Archer et al., 1998), functions as an inhibitor of histone deacetylase (HDI) (Richon et al., 1998), an enzyme implicated in regulation of histone acetylation and genes implicated in cellular maturation (Ogryzko et al., 1996). Given the close relationship that exists between dierentiation and apoptosis (Martin et al., 1990), it is plausible that SAHA also induces genes involved in the cell death program. An alternative possibility is that SAHA triggers an aberrant differentiation program, a phenomenon known to lead to apoptosis (de Vente et al., 1995). Lastly, SAHA may act through an unrelated mechanism to lower the cell death threshold. In this regard, sodium butyrate has been reported to promote mitochondrial damage and release of cytochrome c into the cytosol (Medina et al., 1997); moreover, butyrate and another HDI, trichostatin, have been shown to induce apoptosis in malignant cells of epithelial and neuronal origin (Dangond and Gullans, 1998; Salminen et al., 1998; McBain et al., 1997). The ability of SAHA to initiate an early loss in mitochondrial membrane potential (DCm) (Figure 1) supports the premise that HDIs exert their lethal eects, at least in part, by triggering loss of mitochondrial membrane potential. It is noteworthy that butyrate-induced dierentiation and apoptosis in human colon tumor cells have recently been associated with down-regulation of c-Myb as well as Bcl-2 (Thompson et al., 1998). However, while SAHA potently down-regulated c-Myb (as well as B-Myb) expression in U937 cells, it did not reduce Bcl-2 levels, rendering this mechanism of apoptosis induction unlikely. Results obtained with Bcl-2- and Bcl-xL-overexpressors argue against the possibility that induction of


7021

Figure 5 Ectopic p53 expression sensitizes U937 cells to HMBA- but not SAHA-induced apoptosis. (a) (bar graphs): U937 cells expressing the murine p53 temperature sensitive Val1354 Ala mutation (ptsp53)(Clones B3, A2, A5) and the corresponding empty vector control (Clone M2) were exposed to HMBA (3 mM) or SAHA (1 or 2.5 mM) for 24 h at either 37 or 328C. Apoptosis was measured by morphological assessment; values are expressed as the means+s.e.m. for a representative experiment. Two additional experiments yielded equivalent results. (b) Cell extracts from these experiments were monitored for p21CIP1 expression by Western blotting. Blots reprobed with an anti-actin antibody con®rmed equal loading and transfer

apoptosis is solely responsible for the relatively limited capacity of SAHA to induce U937 cell maturation. In wild-type U937 cells, administration of sub-toxic concentrations of SAHA (e.g., 1 mM) induced expression of the myelomonocytic dierentiation antigen CD11b in *20% of cells, a value signi®cantly less than that observed in cells exposed to HMBA. At higher SAHA concentrations, however, maturation declined dramatically, accompanied by a marked increase in apoptosis. It is worth noting that SAHA exhibited a similar biphasic dose-response curve with respect to MEL-cell maturation (Richon et al., 1996), despite triggering dierentiation in the large majority of these cells at lower concentrations. These ®ndings raise the possibility that the lethal eects of high concentrations of SAHA interfere with normal cellular maturation.

However, when U937 cells overexpressing Bcl-2 or BclxL were exposed to SAHA, dierentiation was only modestly greater than that observed in empty-vector controls, despite protection from apoptosis and potentiation of G0G1 arrest (Table 2). In this context, it has previously been shown that enforced expression of Bcl-2 in HL-60 cells does not modify the maturation response to retinoic acid, but instead delays the onset of dierentiation-related apoptosis (Naumovski and Cleary, 1994). Taken together, these ®ndings suggest that aberrant dierentiation triggered by SAHA contributes to lethality, and suggest the notion that apoptosis represents an alternative fate for cells unable to engage a normal dierentiation program (de Vente et al., 1995; Wang et al., 1998). Another possible interpretation is that as yet unde®ned upstream events


7022

Figure 6 Role of the Stress-Activated Kinase cascade in SAHA-induced apoptosis. (a) Western blots illustrating the induction of cJun expression in U937 cells treated with HMBA or SAHA for the indicated intervals. (b) Western blots illustrating low levels of JNK activation in U937 cells treated with HMBA or SAHA for the indicated intervals. Extract from cells treated with 50 nM PMA for 15 min. serve as a control. To con®rm equal loading, blots were probed with an antibody that recognizes both non and phospho forms of JNK and an anti-actin antibody. (c) Expression of the c-Jun dominant negative mutant (TAM67) in U937 cells partially protects against SAHA-induced apoptosis. Two clonal U937 cell lines stably transfected with the c-Jun dominant-negative mutant (TAM67/2-1 and 2-2) and the corresponding control line (Control) were exposed to 2.5 mM or 5 mM SAHA for 24 h. The cells were scored for induction of apoptosis as above; values represent the means for triplicate experiments+s.d. Treatment with 3 mM HMBA induced 56% apoptosis in all cell lines (not shown). SAHA treated cells versus control; *P50.01 and **P50.001. Both TAM67 cell lines used in these studies displayed 470% reduction in apoptosis following exposure to sphingomyelinase (100 mU/ml) compared to controls (not shown). (d) The extent of apoptosis was determined in U937 cells exposed to 2.5 mM SAHA for 24 h +pretreatment with 20 mM SB203580 or 10 mM PD98059. Values represent the means+s.d. for a representative experiment; two additional studies yielded equivalent results

induced by SAHA are not inhibitable by Bcl-2/Bcl-xL, and remain capable of interfering with cellular maturation. The cell cycle checkpoint genes p53 and p21CIP1, both of which are involved in cell cycle arrest following DNA damage (Bunz et al., 1998), have also been shown to modulate cellular dierentiation and apoptotic responses in a reciprocal manner. For example, enforced expression of p53 in human leukemia cells increases their susceptibility to vitamin D3-mediated maturation (Ehinger et al., 1996). Similarly, expression of p21CIP1 has been linked to dierentiation and inhibition of apoptosis (Wang and Walsh, 1996). In human leukemia cells such as HL-60 and U937 which are p53-null, dierentiation induction (e.g., by PMA) leads to induction of p21CIP1 (Jiang et al., 1994; Steinman et al., 1994; Vrana et al., 1998); conversely, dysregulation of p21CIP1 opposes PMAmediated maturation, and reciprocally promotes apoptosis (Wang et al; 1998). The present results

indicate that functional alterations in p53 and p21CIP1 status in U937 cells exert diametrically opposed eects on dierentiation and apoptotic responses to SAHA and HMBA. Thus, restoration of functional p53 in U937 cells increased HMBA- but not SAHA-mediated apoptosis, whereas dysregulation of p21CIP1 increased SAHA- but not HMBA-induced apoptosis. These ®ndings provide functional evidence that the hybrid polar compounds SAHA and HMBA trigger leukemic cell death through fundamentally dierent cell cycle checkpoint-related mechanisms. In this regard, the histone deacetylase inhibitor sodium butyrate has recently been shown to induce G1 arrest and pRb dephosphorylation in 3T3 cells lacking p21CIP1 (Vaziri et al., 1998). Whether histone deacetylase inhibitors speci®cally exert their antiproliferative eects through a p21CIP1-independent mechanism remains an unresolved issue. Induction of leukemic cell dierentiation is associated with activation of both SAPK-associated


Table 2 Treatment pCEP Control SAHA Bcl2 Control SAHA

U937-Bcl2/BclXLcells ± cell cycle pro®le Subdiploid

G0/G1

S

G2/M

-24 -72 -24 -72

h h h h

3+1 2+1 48+5 84+7

40+2 45+2 21+1 10+2

44+1 44+2 15+4 1+1

12+1 8+1 15+1 5+2

-24 -72 -24 -72

h h h h

3+1 1+2 17+1 8+2

45+5 54+3 32+3 62+3

41+3 37+1 24+4 10+1

13+3 8+2 27+2 20+2

h h h h

3+2 1+1 46+6 88+8

41+3 48+3 27+6 6+1

44+3 43+3 16+3 1+1

11+2 8+1 13+2 5+1

h h h h

6+2 3+2 15+3 26+4

41+2 50+2 43+2 50+2

42+2 40+2 28+4 13+1

10+1 8+2 15+2 11+2

pcDNA3.1 Control -24 -72 SAHA -24 -72 Bcl-xL Control -24 -72 SAHA -24 -72

U937 cells stably transfected with either Bcl-2 or Bcl-xL and their corresponding controls, pCEP4 and pcDNA3.1, were exposed to 2.5 mM SAHA and cell cycle distribution analyses after 24 and 72 h. Values are expressed as the mean percentage relative to the total cell population for three experiments+s.d.

(Franklin and Kraft, 1995) as well as MAPK-related (Kharbanda et al., 1994) signaling pathways. Furthermore, pharmacologic blockade of the MEK/ERK module (by PD98059) or interruption of the SAPKJNK/AP-1 pathway (by TAM67) interferes with retinoic acid- and PMA-related actions respectively (Yen et al., 1998; Dong et al., 1997). On the other hand, JNK and ERK exert opposing eects on apoptosis, at least in response to certain stimuli (e.g., growth factor deprivation) (Xia et al., 1995). In this regard, SAHA failed to activate ERK, and its capacity to induce apoptosis was not inhibited by the selective MEK1 inhibitor PD98059. These ®ndings indicate that SAHA-mediated toxicity proceeds independently of the cytoprotective MAPK-ERK pathway. In contrast, SAHA treatment activated JNK, albeit modestly, and associated lethal eects were partially attenuated by interference with c-Jun transactivation of AP-1. In view of evidence that (1) an intact SAPK-JNK/AP-1 axis is required for the lethal actions of ceramide (Verheij et al., 1996), and (2) perturbations in ceramide and sphingosine metabolism accompany leukemic cell maturation and apoptosis (Ohta et al., 1995), such ®ndings raise the possibility that these lipid messengers may be involved in SAHA-induced apoptosis. Studies designed to test this hypothesis are underway. In summary, the present ®ndings demonstrate that SAHA, a hybrid polar compound and histone deacetylase inhibitor *2000-fold more active than HMBA as an inducer of MEL cell maturation, potently triggers apoptosis in human myeloid leukemia cells. Furthermore, SAHA-mediated cell death proceeds via a pathway that is p53-independent, but regulated, at least in part, by p21CIP1, Bcl-2/Bcl-xL, and c-Jun/AP-1. Finally, inhibition of SAHA-induced apoptosis (e.g., by Bcl-2 or Bcl-xL overexpression) only modestly enhances cellular dierentiation, suggesting that induction of an aberrant maturation program accounts for at least some of the lethal actions of this agent. Issues remaining to be resolved

include clari®cation of the mechanism responsible for SAHA-mediated mitochondrial damage, de®nition of the relationship between SAHA-related lethality and inhibition of histone deacetylase activity, and identification of the speci®c factor(s) that induce cell death rather than maturation in cells exposed to SAHA. In this regard, studies examining the ability of other histone deacetylase inhibitors such as butyrate and trichostatin to induce apoptosis (Dangond and Gullans, 1998; Salminen et al., 1998; McBain et al., 1997) may provide useful insights. Given evidence of activity of polar-planar compounds such as HMBA in some hematologic malignancies (Andreef et al., 1992), as well as the considerably greater potency of SAHA as an inducer of dierentiation (in murine leukemia cells) (Richon et al., 1996) and apoptosis (in human leukemia cells), further eorts to elucidate the molecular determinants of action of SAHA and related agents could have therapeutic implications.

Materials and methods Cell lines The U937 (Sundstrom and Nilsson, 1976) and HL-60 (Collins et al., 1977) cell lines were obtained from ATCC and maintained as described (Vrana et al., 1998). Transfectant U937 cells stably overexpressing Bcl-2 or Bcl-xL were generated by electroporation of the parental line with a plasmid (pCEP4 or pcDNA3.1; Invitrogen) containing Bcl-2 or Bcl-xL cDNA (Wang et al., 1997). These cells, designated either U937/Bcl-2 or U937/Bcl-xL, express approximately a seven- or tenfold increase in Bcl-2 or Bcl-xL protein respectively, compared to their empty-vector controls. The U937/p21AS cell line that exhibits dysregulation of the cyclin-dependent kinase inhibitor (CDKI) p21WAF1/CIP1 through expression of p21WAF1/CIP1 in the antisense con®guration were used as described previously (Wang et al., 1998). p53 temperature-sensitive cells, designated ptsp53, were generated by electroporation of U937 cells with an expression vector containing the cDNA for mutant murine p53 as described (Ehinger et al., 1996). We have previously shown that at 328C, the exogenously expressed p53 protein assumes the wild-type conformation and the cells exhibit increased maturation, cell cycle arrest, and apoptosis compared to cells incubated at 378C (Ehinger et al., 1996). Stable U937 transfectants expressing TAM67 (a c-Jun transactivation domain-de®cient mutant that retains normal DNA binding and dimerization functions) along with their empty-vector counterparts, were used as described (Grant et al., 1996a). All transfectants were maintained under appropriate selection pressure (200 mg/ml hygromycin or 400 mg/ml G418). Drugs and reagents The preparation and characterization of SAHA has previously been described (Richon et al., 1996). HMBA (Sigma) was freshly prepared before use by dilution in sterile medium and in all experiments was used at a concentration of 3 mM. PMA (Sigma), PD98959 and SB203580 (Calbiochem) were dissolved in DMSO. For all experiments, cells in log-phase growth were suspended at 1 ± 26105 cells/ml and exposed to test agents for the indicated intervals. Assessment of apoptosis Apoptotic morphology was assessed in cytocentrifuge preparations stained with the Di-Quik stain set (Dade Diagnostics) and viewed by light microscopy as previously

7023


7024

described (Grant et al., 1996b). Triplicate experiments were performed in which 15 randomly selected ®elds were evaluated for each condition, encompassing a total of 51500 cells. In some cases the extent of apoptosis, manifested by the percentage of propidium iodide-treated cells containing sub-diploid quantities of DNA, was con®rmed by ¯ow cytometry (see below). Dierentiation and cell cycle studies Induction of cellular maturation and analysis of cell cycle pro®les were measured as described (Vrana et al., 1998). Brie¯y, dierentiation was monitored by assessing CD11b positivity in cells following 72 h exposure to test agents using phycoerythrin-labeled antibody (CD11b or IgG2a, BectonDickinson). Cells were analysed on a Becton-Dickinson FACScan ¯ow cytometer and CyCLOPS 2000 (v/4.0) software. For cell cycle analysis, cells were treated for either 24 or 72 h as indicated, and analysed as described (Vrana et al., 1998). Cells were ®xed in ethanol, pelleted, resuspended in buer (3.8 mM sodium citrate, 0.5 mg/ml RNAse A, and 0.01 mg/ml propidium iodide), incubated on ice for 4 h, pelleted, and resuspended in PBS before analysis. Cell cycle distribution was determined by cyto¯uorometry using the ModFit LT program (v/2.0; Verity Software). For S phase synchronization, cells were exposed to 0.1 mg/ml aphidicolin (Sigma) for 24 h, washed 63, and resuspended in fresh medium before further studies. This treatment arrests 85 ± 90% of cells in S phase without inducing toxicity. Western analysis Treated cells were washed once in cold PBS and either lysed, separated by 10% polyacrylamide gel electrophoresis, and immunoblotted as previously described in the case of p21CIP1 (Transduction Laboratories; 1 : 500), pRB (Pharmingen; 1 : 1000), c-Myc (provided by Dr JP Cleveland, St Jude Research Hospital, Memphis, TN, USA; 1 : 5), c-Myb (provided by Dr JL Sleeman, CRC Institute, Cambridge, UK; 1 : 200), B-Myb (Santa Cruz; 1 : 500), Bcl-2 (Dako; 1 : 2000), Bax (Pharmingen; 1 : 1000) caspase-3 (Transduction Laboratories; 1 : 2000), PARP (Oncogene Research Products; 1 : 1000) and actin (Sigma; 1 : 1000) (Vrana et al., 1998; Wang et al., 1998) or as per the manufacturer's instructions in the

case of phospho-JNK, JNK, c-Jun, phospho-ERK, and ERK (all 1 : 2000; New England Biolabs). For the former, 25 mg of cell extracts were loaded in each lane, and 56105 cell equivalents for the latter studies. Assessment of mitochondrial function/integrity At the indicated intervals, cells were harvested and incubated with 40 nM 3,3-dihexyloxacarbocyanine (DiOC6; Molecular Probes Inc) for 15 min at RT as previously described in detail (Wang et al., 1998). Cells (26105 cells/treatment) were analysed by ¯uorocytometry and the percentage of cells exhibiting low levels of DiOC6, re¯ecting loss of mitochondrial membrane potential (DCm), was determined using CyCLOPS 2000 software. Histone deacetylase (HDAC) assay Acid extraction of proteins from treated cells and detection of acetylated Histone H3 and H4 by Western blot analysis was performed per the manufacturer's instructions (Upstate Biotechnology). 10 mg of extract was loaded per lane and equal loading was con®rmed by Coomassie blue staining of a duplicate gel. Histones were separated on precast 4 ± 20% Biorad Gradient Gels and immunoblotted on 0.2 mm Optitran nitrocellulose.

Abbreviations ERK, extracellular receptor kinase; HMBA, hexamethylenebisacetamide; JNK, Jun N-terminal kinase; MAPK, mitogen-activated protein kinase; PMA, phorbol myristate acetate; SAHA, suberoylanilide hydroxamic acid. Acknowledgments We would like to thank E Freeman, A Nambiar, G Schaefer and J-H Chen for their technical assistance. This work was supported by awards CA63753, CA72955, CA77141, CA35675 from the NIH, award 6405-97 from the Leukemia Society of America, and the Chernow Endowment Trust. PB Fisher is the Michael and Stella Chernow Urological Research Scientist.

References Andreef M, Stone R, Michaeli J, Young CW, Tong W, Sogolo H, Ervin T, Kufe D, Rifkind RA and Marks PA. (1992). Blood, 80, 2604 ± 2609. Archer SY, Meng S, Shei A and Hodin RA. (1998). Proc. Natl. Acad. Sci. USA, 95, 6791 ± 6796. Brown PH, Chen TK and Birrer MJ. (1994). Oncogene, 9, 791 ± 799. Bunz F, Dutriaux A, Lengauer C, Waldman T, Zhou S, Brown JP, Sedivy JM, Kinzler KW and Vogelstein B. (1998). Science, 282, 1497 ± 1501. Collins SJ, Gallo RC and Gallegher RE. (1977). Nature, 270, 347 ± 349. Dangond F and Gullans SR. (1998). Biochem. Biophys. Res. Commun., 247, 833 ± 837. de Vente J, Kiley S, Garris T, Bryant W, Hooker J, Posekany K, Parker P, Cook P, Fletcher D and Ways DK. (1995). Cell Growth Di., 6, 371 ± 382. Dong Z, Crawford HC, Lavrovsky V, Taub D, Watts R, Matrisian LM and Colburn NH. (1997). Mol. Carcinogenesis, 19, 204 ± 212. Dunaef JL, Stober BE, Guha S, Khavari PA, Alin K, Luban J, Begemann M, Brabtree GR and Go SP. (1994). Cell, 79, 119 ± 130.

Ehinger M, Bergh G, Olofsson T, Baldetorp B, Olsson I and Gullberg U. (1996). Blood, 87, 1064 ± 1074. Franklin CC and Kraft AS. (1995). Oncogene, 11, 2365 ± 2374. Gervais JL, Seth P and Zhang H. (1998). J. Biol. Chem., 273, 19207 ± 19212. Grant S, Freemerman AJ, Martin H, Turner AJ, Birrer MJ and Jarvis WD. (1996a). Cell Growth Di., 7, 603 ± 613. Grant S, Turner AJ, Freemerman AJ, Wang Z, Kramer L and Jarvis WD. (1996b). Exp. Cell Res., 228, 65 ± 75. Hass R, Pfannkuche H-J, Kharbanda S, Gunji H, Meyer G, Harmann A, Hidaka H, Resch K, Kufe D and GoppeltStrube M. (1991). Cell Growth Di., 2, 541 ± 548. Jiang H, Lin J, Su Z, Collart FR, Huberman E and Fisher PB. (1994). Oncogene, 9, 3397 ± 3406. Kharbanda S, Saleem A, Emoto Y, Stone R, Rapp U and Kufe D. (1994). J. Biol. Chem., 269, 872 ± 878. Martin SJ, Bradley JG and Cotter TG. (1990). Clin. Exp. Immunol., 79, 448 ± 453. McBain JA, Eastman A, Nobel CS and Mueller GC. (1997). Biochem. Pharmacol., 53, 1357 ± 1368. Medina V, Edmonds B, Young GP, James R, Appleton S and Zalewski PD. (1997). Cancer Res., 57, 3697 ± 3707.


Michalovitz D, Halevy O and Oren M. (1990). Cell, 62, 671 ± 680. Naumovski L and Cleary ML. (1994). Blood, 83, 2261 ± 2267. Ogryzko VV, Hirai TH, Russanova VR, Barbie DA and Howard BH. (1996). Mol. Cell Biol., 16, 5210 ± 5218. Ohta H, Sweeney EA, Masamune A, Yatomi Y, Hakamori S and Igarashi Y. (1995). Cancer Res., 55, 691 ± 697. Ra MC. (1992). Nature, 356, 397 ± 400. Richon VM, Webb Y, Merger R, Sheppard T, Jursic B, Ngo L, Civoli F, Breslow R, Rifkind RA and Marks PA. (1996). Proc. Natl. Acad. Sci. USA, 93, 5705 ± 5708. Richon VM, Emiliani S, Verdin E, Webb Y, Breslow R, Rifkind RA and Marks PA. (1998). Proc. Natl. Acad. Sci. USA, 95, 3003 ± 3007. Salminen A, Tapiola T, Korhonen P and Suuronen T. (1998). Mol. Brain Res., 61, 203 ± 206. Selvakumaran M, Reed JC, Liebermann D and Homan B. (1994). Blood, 84, 1036 ± 1042. Steinman RA, Homan B, Iro A, Guillouf C, Liebermann DA, El-Houseini ME. (1994). Oncogene, 9, 3389 ± 3396. Sugimoto K, Toyoshima H, Sakai R, Miyagawa K, Hagiwara K, Takaku F, Yazaki Y and Hirai H. (1992). Blood, 79, 2378 ± 2383. Sundstrom C and Nilsson K. (1976). Int. J. Cancer, 17, 565 ± 577.

Tang HY, Zhao K, Pizzolato JF, Fonarev M, Langer JC and Manfredi JJ. (1998). J. Biol. Chem., 273, 29156 ± 29163. Thompson MA, Rosenthal MA, Ellis SL, Friend AJ, Zorbas MI, Whitehead RH and Ramsay RG. (1998). Cancer Res., 58, 5168 ± 5175. Vaziri C, Stice L, Faller DV. (1998). Cell Growth Di., 9, 465 ± 474. Verheij M, Bose R, Lin XH, Yao B, Jarvis WD, Grant S, Birrer MJ, Szabo E, Zon LI. Kyriakis JM, HaimovitzFriedman A, Fuks Z and Kolesnick RN. (1996). Nature, 380, 75 ± 79. Vrana JA, Saunders AM, Chellapan SP and Grant S. (1998). Dierentiation, 63, 33 ± 42. Wang J and Walsh K. (1996). Science, 273, 359 ± 361. Wang S, Vrana JA, Bartimole TM, Freemerman AJ, Jarvis WD, Kramer LB, Krystal G, Dent P and Grant S. (1997). Mol. Pharm., 52, 1000 ± 1009. Wang Z, Su Z-Z, Fisher PB, Van Tuyle G, Wang S and Grant S. (1998). Exp. Cell Res., 244, 105 ± 116. Xia Z, Dickens M, Raingeaud J, Davis RJ and Greenberg ME. (1995). Science, 270, 1326 ± 1331. Yen A, Roberson MS, Varvayanis S and Lee AT. (1998). Cancer Res., 58, 3163 ± 3172.

7025