Survival Factor Withdrawal-induced Apoptosis of TF-1 Cells Involves a TRB2-Mcl-1 Axis-dependent Pathway*

Tribbles, an atypical protein kinase superfamily member, coordinates cell proliferation, migration, and morphogenesis during the development of Drosophila and Xenopus embryos. Although Tribbles are highly conserved throughout evolution, the physiological functions of mammalian Tribbles family remain largely unclear. Here we report that human TRB2 is a pro-apoptotic molecule that induces apoptosis of cells mainly of the hematopoietic origin. TRB2 mRNA is selectively induced by removal of granulocyte macrophage colony-stimulating factor (GM-CSF) or interleukin-2 from human erythroleukemia-derived TF-1 cell line or activated primary CD4+ T cells, respectively. It is, however, not induced by many other treatments that trigger apoptosis of these two cell types. Overexpression of TRB2 activates many apoptotic events observed in GM-CSF-deprived TF-1 cells, including loss of mitochondrial membrane potential, Mcl-1 cleavage/degradation, and activation of Bax and a number of caspases. Specific knockdown of TRB2 significantly suppresses GM-CSF deprivation-induced apoptosis and all apoptotic events mentioned above. Finally, we demonstrate that TRB2-induced cleavage and degradation of Mcl-1 are mediated via a caspase-dependent but proteasome-independent mechanism, and overexpression of Mcl-1 or its upstream activator Akt can markedly overcome the apoptogenic effect of TRB2. Altogether, these results suggest that the TRB2-Mcl-1 axis plays an important role in survival factor withdrawal-induced apoptosis of TF-1 cells.

Tribbles has been shown to coordinate cell proliferation, migration, and morphogenesis during the development of Drosophila and Xenopus embryos (1)(2)(3)(4)(5). The protein sequence of Tribbles suggests that it is a member of the protein kinase superfamily, but its sequence diverges from the conventional kinase consensus in subdomains I and II, which are essential for ATP binding (6). The mammalian orthologs of Tribbles, TRB1, TRB2, and TRB3, all appear to contain the consensus serine/ threonine kinase catalytic core, but lack a conserved ATP-binding pocket. Accordingly, no kinase activity has been demonstrated for these proteins by in vitro kinase assay (1,7). Recent studies on mammalian Tribbles proteins have revealed that they may play some important roles in metabolism and growth regulation. In 293 cells, TRB3 is induced by NF-B and functions to negatively regulate NF-B-dependent transcription (8). Under fasting conditions, TRB3 is induced through the PCG-1/peroxisome proliferator-activated receptor-␣ pathway and inhibits Akt/PKB 2 activation in liver (7,9). TRB1 and TRB3 were shown to either inhibit or activate the activities of mitogen-activated protein kinase kinase via protein-protein interaction (10). During endoplasmic reticulum stress-induced apoptosis, human TRB3 expression was up-regulated, and knockdown expression of TRB3 by RNA interference rescued cell viability (11). Conversely, whereas TRB3 mRNA was induced by nutrient starvation, overexpression of TRB3 prevented nutrient starvation-induced cell death (12). trb2 is recently reported to be an oncogene, which induces acute myelocytic leukemia in the chimeric mice bearing TRB2-transduced bone marrow, a result likely because of conversion of C/EBP␣ p42 to C/EBP␣ p30 in TRB2-transduced cells (13).
In the absence of a specific cytokine, cytokine-dependent hematopoietic cells actively undergo apoptosis, which can be blocked by treatment with actinomycin D or cycloheximide, implying a requirement for de novo RNA/protein synthesis in this apoptotic pathway (14,15). Several genes have been reported to be induced in mRNA levels following cytokine deprivation and may play some important roles in the cell-death pathway. For example, hrk, the BH3-only Bcl-2 family gene, is up-regulated during interleukin 3 (IL-3) deprivation of hematopoietic progenitors and is involved in the regulation of apoptosis (16). 24p3, which encodes a secreted lipocalin, is induced by IL-3 deprivation in several IL-3-dependent cell lines and primary bone marrow cells and is able to induce apoptosis of a variety of leukocytes through an autocrine mechanism (17). RC3, a calcium/calmodulin-binding protein, is induced by IL-2 deprivation in several IL-2-dependent cell lines and activated T cells, which is part of the apoptosis pathway (18). We also launched a gene expedition study by transcription profiling of a human granulocyte macrophage colony-stimulating factor (GM-CSF)-dependent cell line TF-1 cultured in the absence of GM-CSF. Here we report the identification of TRB2 as an important modulator of apoptosis in TF-1 cells.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture Conditions-TF-1 is a cytokine-dependent hematopoietic cell line whose normal growth in vitro requires culture medium supplemented with human GM-CSF or IL-3. TF1-bcl-2 is a TF-1 derivative ectopically overexpressing Bcl-2 (19). Ba/F3, 32D, and FDCP-1 are all murine IL3-dependent cell lines. Cytokine-independent suspension cell lines (Jurkat, WEHI-3, U937, and K562) were maintained in RPMI 1640 medium containing 10% fetal bovine serum and 55 M ␤-mercaptoethanol. All cytokine-dependent suspension cell lines were cultured in the same medium but supplemented with specific survival cytokines. Adherent cell lines (HeLa, 293, H1299, and HepG2) and MEL were all maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum.
Cytokine Deprivation and Re-stimulation-For cytokine depletion experiments, factor-dependent cells were grown to a saturation density, washed three times in medium without cytokine, and seeded in cytokine-free medium with 0.5% fetal bovine serum for the indicated length of time. For cytokine re-stimulation experiments, GM-CSF was added to the starved TF-1 cells to a concentration of 20 ng/ml.
Isolation and Activation of Mouse Primary CD4 ϩ T Cells-Splenic CD4 ϩ T cells were isolated by incubating splenocytes with anti-CD4 (L3T4) microbeads followed by purification with an LS-positive selection column and a MACS separator (Miltenyi Biotec, Auburn, CA) according to the manufacturer's recommendations. Purified CD4 ϩ cells were activated in plates coated with anti-CD3 (2C11, 5 g/ml) and anti-CD28 antibody (37.51, 2.5 g/ml) for 3 days and then transferred to plates without antibody coating for further cultivation in the presence or absence of IL-2.
Expression Vectors-pTRB2-IRES-EGFP was generated by subcloning mouse trb2 cDNA into the BamHI site of the pIRES2-EGFP vector (Clontech). For construction of pEGFP-TRB2 (T2E), the coding region of human TRB2 cDNA without a stop codon was ligated into the BamHI sites of the pEGFP-N3 vector (Clontech). To construct expression plasmids encoding the TRB2-EGFP mutants, standard PCR-assisted mutagenesiscoupled cloning methods were carried out using primer 5Ј-GCAAGGTGTTTGATATCAGC-3Ј in combination with the following primers (underlined nucleotides differ from the wild-type sequence): TRB2-K177A, 5Ј-CTACTTTTTCCGCAG-CGCGAGGTCCCGCAGCA-3Ј, and TRB2-K177R, 5Ј-CTACT-TTTTCCGCAGCCTGAGGTCCCGCAGCA-3Ј. These PCR products and a third primer 5Ј-AGCTGGGTTCAATGT-CATGG-3Ј were used in a second PCR using pEGFP-TRB2 as template. The resultant PCR products were restricted with EcoRV and BstXI before they were used to replace the corresponding fragment of the pEGFP-TRB2. All mutated nucleotides were confirmed by sequencing. Expression plasmids encoding HA-tagged mouse mcl-1 (20) and HA-tagged myr-Akt were as described previously (21).
Gene Transfer-Gene transfers into hematopoietic cells were carried out by electroporation using a Bio-Rad Genepulser set at 180 V and 975 F (for TF-1 and TF-1-bcl-2 cells) or at 200 V and 975 F (for other hematopoietic cells) or as otherwise indicated. Transfections of adhesion cells were performed by using Lipofectamine (Invitrogen) according to the manufacturer's protocol.
Northern and Western Blot Analyses-For Northern blot analysis, total RNA was prepared using the TRIzol reagent kit (Invitrogen) and analyzed by the standard protocol using cDNA probes specific to TRB1, TRB2, TRB3, and G3PDH. For Western blot analysis, cells to be analyzed were lysed in a buffer containing 25 mM Tris (pH 7.6), 150 mM NaCl, 2.5 mM MgCl 2 , 0.5 mM EDTA, 0.5% Nonidet P-40, 5 mM ␤-glycerophosphate, 1 mM dithiothreitol, 5% glycerol, and protease inhibitors (7), and 100 g of cell lysates, unless otherwise indicated, were analyzed by Western blotting using antibodies as indicated in each figure.
Antibodies-Polyclonal antibody specifically recognizing TRB2 was generated by immunizing rabbits with bacterially produced recombinant TRB2 protein by a standard protocol, and was affinity-purified using specific antigen cross-linked to CNBr-activated Sepharose (GE Healthcare). The specificity of this antibody was reconfirmed by Western blot analysis with or without TRB2 antigen competition (data not shown). This antibody can recognize both human and murine TRB2. Other antibodies used in this study include those specific to Bcl-2, Bcl-xL, Bax, GFP, human Mcl-1, cytochrome c, p38 (all from Santa Cruz Biotechnology), activated Bax (Clontech), caspase-3, caspase-8, caspase-9, cleaved poly(ADP-ribose) polymerase (all from Cell Signaling Technologies, Boston), hemagglutinin (HA) tag (Roche Diagnostics), and actin (Sigma). Antibody specific to mouse mcl-1 (20) or mitochondrial import receptor Tom70 was as described previously (22). Antibody specific to p53 was a gift from Sheau-Yann Shieh, Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan.
Annexin V and TUNEL Staining and Caspase Activity Assays-To analyze cells that have undergone apoptosis, annexin V staining and terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) assays were carried out using the annexin V-cy3 apoptosis detection kit (BioVision, CA) and the in situ cell death detection kit (TMR red, Roche Diagnostics), respectively. In some cases, cells were stained with 4Ј,6diamidino-2-phenylindole to visualize nuclei and examined by confocal microscopy. The activated caspase-3, caspase-8, and caspase-9 were evaluated with the CaspGLOW TM Red Active caspase-3, caspase-8, and caspase-9 staining kits (BioVision, CA) by flow cytometry. Flow cytometric analysis was performed with a BD FACSCalibur system.
Cytochrome c Release and Detection of Oligomerized Bax-Staining of cytochrome c was carried out essentially as described by Willis et al. (23). In some cases, cells to be analyzed were fractionated into cytosolic and mitochondrial fractions (24), and cytochrome c partitioned in these two fractions was analyzed by Western blotting using anti-cytochrome c antibody. To detect oligomerized Bax, cells to be analyzed were cross-linked with disuccinimidyl suberate essentially as described by Sundararajan and White (25). Following crosslinking, cells were lysed and cell lysates analyzed by Western blotting using Bax-specific antibody.

TRB2 Is an Immediate Early Gene Induced by GM-CSF
Deprivation-In the TF-1 cell line, deprivation of human GM-CSF induces profound apoptosis within 24 h. This apoptotic process can be suppressed significantly by the protein synthesis inhibitor cycloheximide, suggesting that de novo gene expression is required to mediate this death program. To identify genes whose expression is up-regulated after cytokine deprivation, we studied the expression profiles of 8000 genes using a cDNA microarray (Taiwan Genome Sciences, Inc., Taipei) with mRNA isolated from TF-1 cells cultured in medium with or without GM-CSF. Among several candidates, we confirmed by Northern blot analysis that the TRB2 gene was highly inducible upon GM-CSF deprivation. After removal of GM-CSF, the mRNA levels of TRB2 were up-regulated by 3 h and reached a peak at 24 h (Fig. 1A, lanes 2-5). However, when the cytokine-deprived cells were re-stimulated with GM-CSF, the TRB2 mRNA returned to a base-line level at 3 h (Fig. 1B, lanes 3-5). Furthermore, the induction of TRB2 mRNA by GM-CSF deprivation did not require new protein synthesis, as cycloheximide could not suppress this induction (Fig. 1C, compare lanes 6 and 9). In fact, cycloheximide alone slightly induced TRB2 expression (Fig. 1C, lane 3). This result suggests that TRB2 is an immediate early gene activated by GM-CSF deprivation. Consistent with the increased expression of TRB2 mRNA, the increased TRB2 protein level following GM-CSF deprivation of TF-1 was also observed (Fig. 1D, lanes 2 and  3). Induction of TRB2 expression was also observed in TF-1 cells that had been previously maintained in another survival cytokine IL-3 or in TF-1 derivatives stably overexpressing IL5R ␣ chain (19) maintained in IL-5 (data not shown). We next examined whether TRB2 could also be induced in other cytokine-dependent cells upon removal of their dependent cytokines. To our surprise, most cell lines examined, including murine GM-CSF-dependent C2GM cells, IL-2-dependent HT-2 cell line, and IL-3-dependent Ba/F3, 32D, and FDCP-1 cells do not manifest this property (data not shown). However, the trb2 mRNA and the TRB2 protein were clearly induced in the activated primary murine (Fig. 1, E and F, respectively) or human (data not shown) CD4 ϩ T cells upon IL-2 deprivation.
TRB2 Induction Is Highly Selective-To explore whether TRB2 is a general apoptosis-responsive gene, TF-1 cells were treated with various apoptotic stimuli, including actinomycin D, camptothecin, cycloheximide, dexamethasone, etoposide, staurosporine, and UV irradiation ( Fig. 2A). These apoptotic agents, except cycloheximide and dexamethasone, could induce a substantial degree of apoptosis, which was comparable with that induced by GM-CSF deprivation (ϪGM, Fig. 2A). Intriguingly, the results of Northern blot analysis demonstrated that TRB2 expression was induced profoundly by GM-CSF starvation, and induced slightly by cycloheximide and dexamethasone, but not by other apoptotic treatments ( Fig. 2A, lanes [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]. Because another human tribbles ortholog, TRB3 (also named NIPK or SKIP3), could be induced in PC6-3 cells by nerve growth factor deprivation (26), we investigated whether other members of tribbles are inducible in our experiments. Same as in A except that protein lysates were analyzed by Western blot using antibodies specific to TRB2 or ␤-actin. E, induction of murine trb2 mRNA; F, protein by removal of IL-2 from activated mouse primary CD4 ϩ T cells. Activated splenic CD4 ϩ T cells (see "Experimental Procedures") were deprived of IL-2 for the indicated times before total RNA or protein lysates were analyzed by Northern (E) or Western blotting (F), respectively. Molecular weight markers and ribosomal RNAs are as indicated.

TRB2 Modulates Apoptosis in TF-1 Cells
Northern blot analysis showed that in the absence of GM-CSF TRB1 mRNA was only slightly induced ( Fig 14). On the other hand, TRB3 was clearly induced by dexamethasone ( Fig. 2A, 3rd panel, lanes 11 and 12) but not by other treatments. In contrast, camptothecin and staurosporine could reduce the expression of TRB3, like the effect of GM-CSF deprivation ( Fig. 2A, lanes 3, 4, 7, 8, 15, and 16).
To extend our observation in TF-1 cells, activated mouse primary CD4 ϩ T cells were subjected to the same treatments and analyzed. As shown in Fig. 2B, murine trb2 was induced strongly by IL-2 starvation (lanes 3 and 4), induced slightly by cycloheximide and etoposide, but not induced by other treatments (lanes 5-16), suggesting that a similar induction mechanism may apply to both human and mouse TRB2 genes. We further tested whether TRB2 can be induced by stimuli that activate the extrinsic death pathway. To address this issue, TF-1 cells were treated with cycloheximide plus tumor necrosis factor-␣ (TNF-␣, Fig. 2C) to activate a TNF receptor 1-dependent death signal, or the activated murine CD4 ϩ T cells were treated with IL-2 and CD3 antibody ( Fig. 2D) to elicit a CD95-dependent activation-induced cell death. As shown in Fig. 2, C and D, in neither case did we observe any significant induction of TRB2 (lanes 3, 4, 7, and 8 of Fig. 2C and lanes 4 and 6 -8 of Fig. 2D). Altogether, these results suggest that the induction of mammalian tribbles orthologs is highly selective.
Ectopic Expression of TRB2 Induces Apoptosis of Many Cell Lines of Hematopoietic Origin-The correlation between induction of TRB2 and apoptosis after cytokine deprivation in TF-1 and activated primary CD4 ϩ T cells prompted us to investigate whether ectopic expression of TRB2 is sufficient to induce apoptosis. To address this issue, a bicistronic expression plasmid pTRB2-IRES-EGFP containing an internal ribosomeentering site (IRES) sequence, which can simultaneously produce the authentic TRB2 and the EGFP (Clontech) (Fig. 3A,  panel a), was constructed and introduced into rapidly growing Ba/F3 cells. As shown in Fig. 3B, the expression of TRB2 was readily detectable with anti-TRB2 antibody in Western blot analysis (lane 2). In IL-3-free culture medium overexpression of EGFP alone (Vec) induced apoptosis of 31.6% (Ϯ3.4) (n ϭ 5) of transfected cells (i.e. GFP(ϩ) population), whereas overexpression of TRB2 (T2) increased apoptosis to 65.6% (Ϯ4.9) (n ϭ 5).
In these experiments, we observed consistently that TRB2 expression significantly enhanced IL-3 withdrawal-induced

TRB2 Modulates Apoptosis in TF-1 Cells
apoptosis of Ba/F3 cells. However, no significant apoptosis-inducing effect was observed for TRB2 when cells were cultured in medium containing IL-3 (data not shown).
We next examined whether a similar apoptosis-enhancing effect could be observed for the TRB2-EGFP fusion protein (T2E, Fig. 3D, lane 2), which is encoded by the plasmid pTRB2-EGFP (Fig. 3A, panel b). In the IL-3-containing medium, transfection of Ba/F3 cells with 30 g of pEGFP-N3 vector usually resulted in 20 -30% GFP(ϩ) cells (12 or 20 h post-transfection), whereas under the same conditions, transfection with the pTRB2-EGFP plasmid resulted in only 5-15% GFP(ϩ) cells (see one example shown in supplemental Fig. S1). Of note, although transfection of cells with lower amounts of EGFP or T2E expression vectors (5-15 g) resulted in lower percentages of GFP(ϩ) cells, the same trend was always observed, i.e. significantly less GFP(ϩ) cells were observed in cells transfected with T2E than with EGFP expression vector (supplemental Fig. S1). Furthermore, we noticed that for cells transfected with the control vector, a majority of GFP(ϩ) cells were TUNEL(Ϫ) under confocal microscopy. However, the TUNEL(ϩ)/ GFP(ϩ) population in pTRB2-EGFP-transfected cells was significantly increased, even when cells were cultured in the presence of IL-3 (Fig. 3E). Flow cytometric analysis revealed that in one representative experiment, expression of T2E in Ba/F3 cells resulted in 61% of GFP(ϩ) cells to be TUNEL-positive, whereas overexpression of EGFP proteins resulted in only 3% cell death (Fig. 3F, top panels, T2E versus E). Similar percentages of cell death were observed when the apoptotic cells were quantified by the annexin V staining assay (Fig. 3F, bottom panels). Furthermore, T2Einduced cell death was also consistently observed in experiments performed with TF-1 cells (Fig. 3G).
Next, we examined whether the apoptotic effect of T2E was mediated through caspase-dependent pathways. As shown in Fig. 3G, T2Einduced apoptosis of either Ba/F3 or TF-1 cells cultured in cytokine-containing medium could be markedly suppressed when cells were treated with the pan-caspase inhibitor Z-VAD-fmk (Fig. 3G, white versus gray columns). These data suggest that overexpression of the TRB2-EGFP fusion protein could cause cell death with typical apoptotic characteristics.
Because TRB2 is a kinase-like protein, we wondered whether any potential kinase activity of TRB2 would be required for its pro-apoptotic function. To address this issue, the lysine residue at the putative catalytic center of subdomain VIB of TRB2, which has the motif RDLKL, was mutated into either arginine (K177R) or alanine (K177A) to create a kinase-dead mutant, and the apoptogenic activity of these two kinase-dead mutants was then compared with that of the wild-type protein. As shown in Fig. 3, H and I, under the same experimental conditions when wild-type and both mutants (all expressed as EGFP

TRB2 Modulates Apoptosis in TF-1 Cells
fusion proteins, i.e. T2E, K177A-E and K177R-E) were transiently expressed at a similar level in Ba/F cells (Fig. 3H), they all resulted in a similar degree of apoptosis (Fig. 3I), suggesting that the pro-apoptotic function of TRB2 does not require its putative kinase activity.
The ability of the TRB2 fusion proteins (T2E) to induce cell death was further investigated in a few other murine and human cell lines. As shown in Fig. 4, T2E induced apoptosis in all cytokine-dependent hematopoietic cell lines tested, including 32D, FDCP-1, and HT-2 (lanes 1-3), although these cell lines did not express TRB2 upon cytokine deprivation (data not shown). Furthermore, the primary murine CD4 ϩ T cells were very sensitive to T2E (Fig. 4, lane 4). On the other hand, some cytokine-independent leukemic cell lines, e.g. WEHI3 and Jurkat (Fig. 4, lanes 5 and 6), were very sensitive to T2E-induced apoptosis, whereas other cell lines, including K562 and U937 (lanes 8 and 9), were highly resistant. Interestingly, most nonhematopoietic cell lines tested, including HeLa, 293, H1299, and HepG2, were highly resistant to T2E-induced apoptosis (Fig. 4, lanes 10 -13). These data suggest that the apoptotic activity of T2E is largely restricted to hematopoietic cells.
TRB2-EGFP Triggers Apoptosis via Mitochondria Dysfunction-A variety of external and internal signals converge on mitochondria to trigger or inhibit apoptosis (27). To understand the mechanism of the apoptotic effect of TRB2, we investigated the ability of T2E to damage mitochondria. The mitochondrial dye MitoTracker Red was used to stain the viable cells, because apoptotic cells lost membrane potential and lost the staining of MitoTracker Red. As shown in one representative result (Fig. 5A), expression of T2E in Ba/F3 cells increased dramatically the proportion of cells that lost the membrane potential, from 10 to 78% (see ⌬ m L). Loss of mitochondrial membrane potential (MMP, ⌬ m ) correlated strongly with activation of caspases, including caspase-3, caspase-8, and caspase-9 (Fig. 5A), and the percentage of cells that have undergone apoptosis as revealed by the annexin V (An-V) staining method. T2E-induced activation of caspase-3, caspase-8, and caspase-9, as evident from the generation of specific cleavage products, was also demonstrated in TF-1 cells by Western blotting with antibody specifically recognizing each caspase (Fig. 5B). Of note, under our experimental conditions, although the cleaved form of caspase-3 was only slightly detectable, the production of cleaved poly(ADP-ribose) polymerase, a typical caspase 3 substrate, was quite prominent (Fig. 5B, 2nd to the bottom panel), suggesting that caspase-3 was also highly activated by T2E in TF-1 cells.
We next examined whether T2E could induce activation of Bax, a proapoptotic molecule that, in response to apoptotic stimuli, will change conformation and oligomerize, which leads to mitochondria dysfunction (28,29). To address this issue, Ba/F3 cells were transfected with vectors expressing control (EGFP) or T2E, and transfected cells (GFP ϩ ) were sorted out and treated with DSS before cell lysates were prepared and analyzed by Western blotting using Bax-specific antibody. As shown in Fig. 5C, dimerized Bax was identified only in the presence of DSS in T2E-expressing cells (Fig. 5C,  lane 4), but not in cells expressing the control protein EGFP (lane 2). Alternatively, T2E-induced activation of Bax was demonstrated in TF-1 cells using antibody specifically recognizing the active conformation of human Bax (a-Bax). As shown in Fig.  5D, in one representative experiment expression of T2E markedly increased the percentage of cells stained by such an antibody (66% of T2E versus 16% of EGFP-expressing cells). T2Einduced activation of Bax and loss of MMP (⌬ m L) could not be blocked by the pan-caspase inhibitor Z-VAD-fmk (Fig. 5E), albeit such a peptide inhibitor inhibited T2E-induced apoptosis quite well (Fig. 5E, compare bars labeled with An-V). These data indicate that Bax activation and mitochondria dysfunction induced by T2E overexpression do not require prior activation of caspases.
We next examined whether release of cytochrome c from mitochondria into the cytosolic compartment, a hallmark of mitochondria dysfunction, occurred in T2E-expressing cells. We first addressed this issue by comparing the cytochrome c levels in mitochondria and cytosolic fractions prepared from Ba/F3 cells expressing T2E or control protein EGFP. As shown in Fig. 5F, in control cells cytochrome c was only detected in the mitochondrial (pellet) fraction (compare lanes 1 and 3). How-  2 and 4). In another assay carried out in TF-1 cells, we observed that expression of T2E, but not the control protein EGFP, markedly reduced the staining of transfected cells with anti-cytochrome c antibody (Fig. 5G, C.c in TF-1). Together, these two approaches demonstrated that T2E expression caused cytochrome c release from mitochondria to cytosol.
Involvement of TRB2 in GM-CSF Deprivation-induced Apoptosis of Human TF-1 Cells-Next, we wanted to examine whether TRB2 is involved in GM-CSF deprivation-induced apoptosis of TF-1 cells. To address this issue, we carried out experiments using siRNA that could specifically knock down TRB2 in TF-1 cells (Fig. 6A). As shown in Fig. 6, B-D, in the absence of GM-CSF, knockdown of TRB2 significantly suppressed GM-CSF deprivation-induced Bax activation (Fig. 6B,  31.0% (Ϯ2.7) (n ϭ 3) in control versus 18.9% (Ϯ0.7) (n ϭ 3) in TRB2 siRNA), loss of MMP (Fig. 6C, ⌬ m L), activation of caspase-3, caspase-8, and caspase-9 (Fig. 6C), and apoptosis (ϳ20% reduction at 24 h and ϳ25% reduction at 36 h, Fig. 6D, p ϭ 0.0005). Altogether, these data suggest that TRB2 is involved in GM-CSF withdrawal-induced apoptosis of TF-1 Cell lysates from each fraction were then analyzed by Western blotting using antibodies specific to each protein as indicated. G, reduction of T2E-induced apoptosis and related events by Bcl-2. Both TF-1-and Bcl-2-overexpressing TF1-bcl2 cells were transfected and analyzed essentially as described in A. The data shown here are the averages (ϮS.D.) of three independent experiments. C.c, cytochrome c release; determined by reduction of staining with anti-cytochrome c antibody. The transfection efficiency for TF-1 and its derivatives (judged from GFP ϩ cell numbers) was 1.5-4% (for T2E) and 10 -12% (for E). cells and that the apoptotic signaling pathway triggered by the T2E fusion protein to some extent, if not all, reflects that induced by the TRB2 protein itself.
TRB2 Expression Triggers Mcl-1 Degradation-Bcl-2 family members are known to be the major players in MMP homeostasis and can intricately regulate mitochondria-dependent apoptosis. GM-CSF-deprivation of TF-1 cells was also shown to down-regulate the expression of pro-survival Mcl-1 protein (30). Therefore, we examined whether TRB2 can modulate the protein levels of any Bcl-2 family proteins. To address this issue, GFP(ϩ)-Ba/F3 cells expressing EGFP or T2E were sorted out by flow cytometry, and their cell lysates were analyzed by Western blot analysis using antibodies specifically recognizing Bcl-2, Bcl-xL, Mcl-1, or Bax. As shown in Fig. 7A, overexpression of T2E reduced the amount of the full-length Mcl-1 protein while generating a truncated product of the Mcl-1 molecule, which was implicated to have a pro-apoptotic activity in previous studies (31,32). However, under the same conditions, the protein levels of Bcl-2, Bcl-xL, and Bax (Fig. 7A) were not altered. On the other hand, reduction of full-length Mcl-1 and a concomitant increase of its cleavage product were observed in TF-1 cells deprived of GM-CSF (Fig. 7B, compare lanes 1 and 2). However, such a phenomenon was significantly reversed in cells whose TRB2 were specifically knock downed (Fig. 7B,  compare lanes 2 and 3).
Mcl-1 degradation has been reported to be mediated by the proteasome-dependent pathway (20,33). Furthermore, caspase 3 was shown to cleave Mcl-1 at Asp-127 and Asp-157, and the resultant products containing only the C-terminal domain of

TRB2 Modulates Apoptosis in TF-1 Cells
Mcl-1 turned out to be proapoptotic (32). Therefore, we next examined whether proteasome or caspases might be involved in TRB2-induced Mcl-1 degradation and/or cleavage. To address this issue, Ba/F3 cells transiently transfected with EGFP or T2E-expressing vector were cultured in medium containing MG132 (proteasome inhibitor) or Z-VAD-fmk (a pan-caspase inhibitor). As shown in Fig. 7C, T2E-induced degradation of full-length Mcl-1 and generation of the ϳ22-kDa cleavage product were markedly prevented by treatment of cells with Z-VAD-fmk (compare lanes 1 and 3). In contrast, in the presence of MG132, T2E still efficiently induced the degradation of MG132-stabilized full-length Mcl-1 and enhanced Mcl-1 cleavage (compare lanes 2 and 5 of Fig. 7C). Of note, MG132 alone under our experimental conditions (treatment for 16 h) slightly activated caspase-3, which might lead to the generation of the background level of the ϳ22-kDa cleavage product of Mcl-1. The T2E-induced degradation/cleavage of Mcl-1 in the presence of MG132 was specific, as such an effect was not observed for the tumor suppressor protein p53, whose stability was also markedly enhanced by MG132 (Fig. 7C, compare lanes 2 and 5 of the p53 panel). Taken together, these results suggest that TRB2 expression leads to cleavage and degradation of Mcl-1 via a caspase-dependent but proteasome-independent mechanism. Finally, we examined whether overexpression of Mcl-1 or a constitutively active form of Akt (myrAkt), which could upregulate the expression of Mcl-1 (21), would protect cells from TRB2-induced cell death. As shown in Fig. 7D, transfection of pTRB2-EGFP alone into Ba/F3 cells caused apoptosis in 77.6% (Ϯ6.7) (n ϭ 6) of GFP(ϩ) cells in IL-3-containing medium. However, under the same conditions co-expression of myrAkt or mMcl-1 reduced the extent of apoptosis to 31.2% (Ϯ2.3) (n ϭ 6) (Fig. 7D, lane 3)

DISCUSSION
In this study, we demonstrate that the TRB2 gene is induced in TF-1 cells in the absence of their survival cytokines. However, removal of cytokines from most other cytokine-dependent cell lines tested in this study does not induce TRB2, although under such conditions they all undergo apoptosis. This result suggests that the cellular context plays an important role in determining TRB2 induction. It further suggests that other factors functionally similar to TRB2 may be involved in the factor deprivation-induced apoptotic program in other factor-dependent cells. In fact, earlier studies on two factor deprivation-induced pro-apoptotic molecules, 24p3 and RC3 (17,18), support this notion. Although 24p3 is induced by cytokine deprivation of some IL-3-dependent cells, it is not induced by withdrawal of survival factor from IL-2 dependent HT-2 or IL-7-dependent D1-F4 cells (17). On the other hand, although rc3 is induced preferentially by IL-2 deprivation of IL-2-dependent primary human T lymphocytes and HT-2 and CTLL-2 cell lines, it is not induced upon removal of IL-3 from FL5.12 and 32D or IL-7 from D1-F4 cells (18). These results suggest that expression of TRB2, 24p3, or RC3 is likely to be regulated in a cytokine-and cell context-specific manner.
The induction of TRB2 upon IL-2 withdrawal from the activated primary CD4 ϩ T lymphocytes is intriguing, as it may suggest that TRB2 is involved in the apoptosis regulation of activated T cells and may thus play some roles in homeostasis regulation of some T lymphoid progenitor cells. Along this line of reasoning, the observed induction of TRB2 in TF-1 cells may simply be due to the possibility that TF-1 cells still retain some properties of T lymphoid progenitors, because it was originally isolated from the bone marrow of patients with erythroleukemia (34).
To further address the role of TRB2 in the T cell system, we have generated the TRB2-deficient mice using the conventional gene targeting approach (which will be reported elsewhere). Our initial characterization of these mutant mice revealed that lack of TRB2 expression does not significantly affect the survival of activated T cells following IL-2 deprivation. 3 Several possibilities may account for this negative result. For example, TRB2 does not play a role in T cell apoptosis in vivo or the existence of a redundant apoptosis pathway after IL-2 deprivation of the activated T cells, e.g. the RC3-dependent pathway (17,18). Alternatively, a compensatory increase of a TRB2-related protein, e.g. TRB3 in the TRB2 knock-out mice, could have accounted for the observed negative result. More experiments are required to distinguish these possibilities.
Of note, the TRB2-EGFP fusion protein (T2E) appeared to have a better apoptotic activity than the wild-type TRB2 molecule, as the fusion protein can overcome the survival activity of IL-3 and trigger apoptosis of Ba/F3 cells cultured in IL-3-containing medium, whereas the wild-type TRB2 molecule can only enhance IL-3 withdrawal-induced apoptosis of the same cells. Comparison of the expression levels of these two proteins revealed that, under the same transfection conditions, the wildtype TRB2 protein was expressed at a comparable (without IL-3) or slightly higher level (with IL-3) than the EGFP fusion protein (data not shown). Although these results suggest that the T2E fusion protein may have gained some extra functions than the wild-type protein, our TRB2 siRNA knockdown experiments strongly suggest that the majority, if not all, of the observed apoptotic effects of the T2E fusion protein come qualitatively from the TRB2 molecule itself. Based on the fact that wt-TRB2 manifests no activity or marginal apoptotic activity in the presence of survival cytokine, but can enhance apoptosis under conditions when cell death is initiated by removal of these survival cytokines, we propose the following "latencyand-activation" model. In this model, we propose that TRB2 may exist as a latent apoptotic molecule when survival cytokines are present in the growth environment. Upon deprivation of these survival cytokines, the latent molecule then becomes activated via a yet-to-be identified mechanism. Given that the T2E fusion protein can trigger apoptosis in the presence of survival cytokines, we propose that a conformational alteration of TRB2 has been generated as a result of protein fusion, which prematurely converts the latent state of TRB2 into an active conformation. One possible active conformation of TRB2 is to become a dimer or oligomer, as EGFP has been reported to have a dimerization ability (35). More experiments are required to test this model.
Other than the latency-and-activation property as described above, we also noticed that manipulation of the wild-type TRB2 levels does not cause a dramatic effect on cellular apoptosis, i.e. enforced expression of TRB2 in Ba/F3 cells increases cell death about 2-fold, whereas siRNA knockdown of TRB2 in TF-1 cells decreases apoptosis by ϳ50%. These data suggest that TRB2 is likely to be a modulator of apoptosis, rather than a key component in the induction of the apoptosis cascade.
Bcl-2 (36 -38) and Bcl-xL (39) were reported to be cleaved during cell death. However, TRB2 selectively induces degradation/cleavage of Mcl-1 in our system. How TRB2 triggers such a response is not clear. Given that Drosophila Tribbles target String/CDC25 and Slbo for ubiquitin-dependent proteasome degradation (3,4), it is possible that human TRB2 induces Mcl-1 degradation via a similar pathway. However, in this study we found that TRB-2-induced degradation of full-length Mcl-1 and generation of an ϳ22-kDa cleavage product is mediated via a caspase-dependent but proteasome-independent pathway. More experiments are needed to delineate the molecular mechanism by which TRB2 expression leads to caspase activation and Mcl-1 cleavage/degradation. On the other hand, one common feature of Tribbles proteins is that many of them have been shown to bind to and inhibit specific kinases, including Akt/ PKB (7), MEK-1, MKK4, and MKK7 (10). Akt/PKB and MEK-1 have been shown to regulate Mcl-1 protein levels and play an important role in mediating survival signals of many cytokines in hematopoietic cells (30,40). Moreover, Du et al. (7) reported that TRB2 can suppress the Akt kinase activity. It is therefore possible that TRB2 may modulate cellular levels of Mcl-1 indirectly via inhibiting a kinase like Akt. More experiments are required to test this possibility.
Intriguingly, although TRB2 has a very potent cytotoxic effect in most factor-dependent cells, it causes no harm to most adherent cells or some leukemic cell lines tested in this study. This cell type-restricted cytotoxicity of TRB2 has a strong resemblance to that of a secreted lipocalin, 24p3 (17), which promotes apoptosis in many leukocytic cell lines, primary lymphocytes, and neutrophils, but not in nonhematopoietic cells or macrophages (17). Given that the oncogenic proteins, including Bcl-2 and the constitutively active form of Akt, are able to alleviate the strong apoptotic signals of TRB2, it is possible that in many cancer cells Bcl-2 and Akt oncoproteins are often overexpressed, which are able to confer a strong anti-apoptotic effect to reverse the adverse effect of TRB2. Likewise, focal adhesion through the ␤1 integrin provides strong survival signals in fibroblasts via activation of the integrin-linked kinase and Akt/PKB pathway (41) or protects keratinocytic epithelial cells from apoptosis by activation of the focal adhesion kinase and ERK (extracellular signal-regulated kinase) pathway (42). Therefore, in this study, fibroblastic and epithelial cell lines are likely protected from the intrinsic cell death signal of TRB2 by cell adhesion to the extracellular matrix. These data suggest a recessive nature of the apoptosis signal induced by TRB2, as it appears to be suppressed by dominant survival activities generated by intracellular oncogenic proteins or by cell adhesion to various extracellular matrixes.
In contrast to the pro-apoptotic property reported in this study, Keeshan et al. (13) demonstrated that trb2 could function as an oncogene that induces acute myelogenous leukemia in hematopoietic progenitors. It is not clear what has caused these two seemingly opposite functions of TRB2, but below is one likely interpretation. In the report by Keeshan et al. (13), wild-type murine TRB2 was expressed from an IRES-EGFPcontaining MigR1 vector. In this study, we demonstrate that wt-TRB2 enhances apoptosis only in the absence of survival cytokine as described in our latency and activation model. Bone marrow (BM) is in an environment that is rich in many survival cytokines. BM progenitors may thus be protected from TRB2induced apoptosis. Under such conditions, other functions of TRB2 may then be revealed. Furthermore, TRB2 may manifest distinct activities in a cellular context-specific manner. In our experimental system, TRB2 down-regulates Mcl-1 levels and induces apoptosis. However, in BM progenitors as used in the study by Keeshan et al. (13), TRB2 may predominantly affect the processing of C/EBP␣ from the p42 to the p30 form, which may confer proliferation advantage on the TRB2-transduced BM progenitor cells. More experiments are certainly required to test this possibility.