Bmf: A Proapoptotic BH3-Only Protein Regulated by Interaction with the Myosin V Actin Motor Complex, Activated by Anoikis

See allHide authors and affiliations

Science  07 Sep 2001:
Vol. 293, Issue 5536, pp. 1829-1832
DOI: 10.1126/science.1062257

This article has a correction. Please see:


Bcl-2 family members bearing only the BH3 domain are essential inducers of apoptosis. We identified a BH3-only protein, Bmf, and show that its BH3 domain is required both for binding to prosurvival Bcl-2 proteins and for triggering apoptosis. In healthy cells, Bmf is sequestered to myosin V motors by association with dynein light chain 2. Certain damage signals, such as loss of cell attachment (anoikis), unleash Bmf, allowing it to translocate and bind prosurvival Bcl-2 proteins. Thus, at least two mammalian BH3-only proteins, Bmf and Bim, function to sense intracellular damage by their localization to distinct cytoskeletal structures.

The BH3-only proteins are proapoptotic members of the Bcl-2 family that share with their relatives only the short BH3 domain, which allows their interaction with prosurvival Bcl-2 family members to trigger apoptosis (1). Genetic experiments have demonstrated that they are essential for initiation of programmed cell death in species as distantly related as Caenorhabditis elegans and the mouse. EGL-1, the only C. elegans BH3-only protein identified thus far, is required for all developmentally programmed death of somatic cells in this organism (2). In contrast, multiple BH3-only proteins have been identified in mammals, including Blk, Bad, Bik, Hrk, Bid, Bim, Noxa, and Puma (1). Experiments with gene knockout mice have shown that pathways to apoptosis resulting from different damage signals require distinct BH3-only proteins for their initiation (1). For example, Bim (3) is essential for apoptosis of lymphocytes induced by cytokine withdrawal but dispensable for that induced by phorbol ester (4). In contrast, Bid is involved in Fas-induced killing of hepatocytes (5). Moreover, distinct BH3-only proteins appear to have tissue-specific roles in programmed cell death. Bim-deficient mice, for example, accumulate abnormal numbers of lymphoid and myeloid cells but have normal erythrocyte levels (4).

The proapoptotic activity of BH3-only proteins is controlled by several mechanisms. Expression of C. elegans EGL-1 is repressed by the transcription factor TRA-1A in neurons required for egg-laying (6). Certain mammalian BH3-only proteins are also subject to transcriptional regulation. For example,noxa and puma were discovered as p53-inducible genes and are therefore attractive candidates for mediating apoptosis in response to DNA damage (7–9). Several other mammalian BH3-only proteins can be regulated by different posttranslational mechanisms (1). Signaling pathways activated by certain growth factors induce phosphorylation of Bad, allowing 14-3-3 scaffold proteins to bind and sequester it from prosurvival Bcl-2 proteins (10). In contrast, Bim is normally sequestered to the microtubular dynein motor complex by binding to the 8-kD dynein light chain (DLC1, also called LC8 or PIN) (11). Certain apoptotic stimuli free Bim (together with DLC1), allowing it to translocate and bind to prosurvival Bcl-2 proteins, thereby activating the apoptotic program. This process occurs independently of the cell death executioner cysteine proteases (caspases) and therefore constitutes an upstream signaling event in apoptosis (11). These observations demonstrate that by their sequestration to specific sites within cells, different BH3-only proteins are uniquely placed to sense distinct forms of cellular stress.

We used mouse Mcl-1, an antiapoptotic Bcl-2 family member, as bait for a yeast two-hybrid screen of a mouse embryo cDNA library and isolated bmf (bcl-2–modifying factor; Fig. 1A) (12). Using mousebmf cDNA as the probe, we cloned its highly similar human counterpart (87% amino acid identity; Fig. 1A) from a human T lymphocyte cDNA library (12). Detailed sequence analysis (13) revealed that Bmf has a BH3 domain most similar to that found in Bim, Bik, and EGL-1 (Fig. 1, A and B). In the yeast two-hybrid system, Bmf interacted with Mcl-1 and other prosurvival Bcl-2 proteins (Bcl-2, Bcl-xL, and Bcl-w) but not with the proapoptotic family members tested (Bax, Bid, and Bad) (14). When transiently overexpressed in HEK293T cells, Bmf could be coimmunoprecipitated with prosurvival Bcl-2 family members Bcl-2 and Bcl-w (Fig. 1C), as well as with Bcl-xL and Mcl-1, but did not bind the proapoptotic protein Bax or the BH3-only protein Bim (14). The interaction of Bmf with Bcl-2 or Bcl-w was greatly diminished by mutating the invariant leucine within its BH3 domain to alanine (L138A; Fig. 1C) (15). Mutations of conserved residues within the BH1 (G145E) or BH2 (W188A) domain of Bcl-2, which abolish its binding to Bim (3) or Bax (16), also disrupted binding to Bmf (14). Endogenous Bmf could be coimmunoprecipitated with endogenous Bcl-2 from detergent-lysed MCF-7 human breast carcinoma cells (Fig. 1D), excluding the possibility that these proteins associate only when overexpressed.

Figure 1

Bmf, a novel mammalian BH3-only protein. (A) Predicted amino acid sequence of mouse and human Bmf. The conserved dynein light chain-binding motif is indicated by a box marked with a single asterisk (*). The short BH3 region, identified by hidden Markov modeling (13), is indicated by a box marked with two asterisks (**). (B) Alignment of the BH3 region of Bmf with that of some other proapoptotic Bcl-2 family members. Black boxes indicate identical amino acids; gray boxes indicate similar residues. (C) Wild-type Bmf, but not a BH3 mutant, binds to Bcl-2 and Bcl-w. HEK293T cells were transiently cotransfected with expression constructs for FLAG-tagged Bcl-2 (or Bcl-w) and EE (Glu-Glu)-tagged Bmf or L138A mutant Bmf. Twenty-four hours after transfection, the cells were metabolically labeled overnight with [35S]methionine. Volumes of cell lysates with equivalent trichloroacetic acid (TCA)–precipitable 35S counts were used for immunoprecipitations with mAbs to the FLAG or EE epitope tags. (D) Interaction of endogenous Bmf with Bcl-2 in MCF-7 cells. Lysates from 107 MCF-7 cells were immunoprecipitated with either Bcl-2-100 mAb (antibody against human Bcl-2) or an isotype-matched control mAb coupled to Sepharose. Bound proteins were eluted, size fractionated on SDS–polyacrylamide gel, subjected to electrophoresis (SDS-PAGE), and transferred onto nitrocellulose filters. Protein immunoblotting was performed with human Bcl-2 mAb or a rat Bmf mAb (9G10) (18). The asterisk (*) indicates the light chain of the mAb used for immunoprecipitation. (E) Wild-type Bmf, but not a BH3 mutant, kills L929 fibroblasts. L929 fibroblasts were transfected with empty vector, expression constructs for hygromycin resistance alone, or with wild-type Bmf, a BH3 mutant (L138A) of Bmf, or Bmf lacking its BH3 domain. Transfected cells were plated in medium containing hygromycin, and the resulting drug-resistant colonies were counted after 10 to 14 days. Values are means (±SD) of three independent experiments.

The biological activity of Bmf was investigated in transient transfectants of human Jurkat T lymphoma cells and stable transfectants of L929 mouse fibroblasts (Fig. 1E) or interleukin 3 (IL-3)–dependent FDC-P1 mouse promyelocytic cells (Fig. 2B). Expression of Bmf triggered apoptosis in ∼80% of Jurkat cells within 24 hours (14) and reduced formation of L929 fibroblast colonies by about 65% (Fig. 1E). Bmf-induced apoptosis of Jurkat cells could be blocked by the caspase inhibitor, baculovirus p35, or by coexpression of Bcl-2 or its homologs (Bcl-xL, Bcl-w, Mcl-1) but not by BH1 (G145E) or BH2 (W188A) mutants of Bcl-2 (14). Consistent with its proapoptotic activity, large amounts of Bmf could be expressed stably in FDC-P1 cells only when Bcl-2 (or one of its functional homologs) was also expressed. Such FDC-P1 cells coexpressing Bmf and Bcl-2 died more rapidly than did Bcl-2–expressing cells in response to cytokine withdrawal (Fig. 2B), γ-irradiation, or treatment with etoposide. In all the cell death assays performed, Bmf mutants that lacked the BH3 domain or had the L138A mutation were inert (Figs. 1E and 2B). These results establish that Bmf is a BH3-only protein that binds prosurvival Bcl-2 family members to initiate apoptosis.

Figure 2

Regulation of Bmf by interaction with DLC2. (A) Binding of Bmf to DLC2 through its dynein light chain–binding region. Coimmunoprecipitation experiments were done as described (Fig. 1C) with lysates of HEK293T cells transiently expressing FLAG-tagged DLC2 and EE-tagged wild-type (wt) Bmf, a BH3 mutant (L138A) of Bmf, dynein light chain–binding region mutants of Bmf (A69P or D67A,K68A), Bid, or Bax. The asterisk (*) indicates the light chain of the mAb used for immunoprecipitation. (B) Interaction with DLC2 regulates the proapoptotic potency of Bmf. FDC-P1 cells stably expressing Bcl-2 plus EE-tagged wt Bmf, a BH3 mutant (L138A) of Bmf or dynein light chain–binding region mutants of Bmf (A69P or D67A,K68A) were deprived of IL-3 for 1 to 6 days. Cell viability was assessed by propidium iodide staining and flow cytometric analysis. Values are means (±SD) of three independent experiments done with four independent clones of each genotype.

The expression pattern of Bmf was investigated by Northern blotting, reverse transcriptase–polymerase chain reaction (RT-PCR) analysis, and by protein immunoblotting. bmf mRNA was found in many cell lines of B and T lymphoid, myeloid, or fibroblastoid origin and in mouse embryos at all developmental stages from embryonic day 9 (E9) to birth [Web fig. 1A (17)]. Protein immunoblotting of cell lysates with specific antibodies (18) detected a single band corresponding to Bmf in many organs, with abundant proteins in pancreas, liver, kidney, and hematopoietic tissues [Web fig. 1B (17)]. To assess whether bmf expression was induced by apoptotic stimuli, RT-PCR analyses were performed of mRNA from thymocytes exposed to various forms of stress, including cytokine deprivation, γ-irradiation, or treatment with dexamethasone or ionomycin. None of these stimuli had any impact on bmfexpression [Web fig. 1C (17)].

The lack of evidence for transcriptional control prompted us to investigate whether Bmf is regulated posttranslationally, perhaps by interacting with other proteins. A yeast two-hybrid screen of a mouse embryo cDNA library with Bmf as bait isolated 14 independent clones of Mcl-1 and more than 40 clones encoding dynein light chain (DLC clones) (12). In a previous screen, Bim had isolated exclusively DLC1 (11). In contrast, most dynein light chain clones interacting with Bmf encoded the closely related protein DLC2 (19). Coimmunoprecipitation experiments in transiently transfected HEK293T cells confirmed the interaction of Bmf with DLC2 (Fig. 2A). Sequence comparison (Fig. 1A) revealed a motif in Bmf (amino acids D67KATQTLSP) that closely resembles the region in Bim (amino acids D51KSTQTPSP) that mediates its binding to DLC1 (11, 20). This appears to be the DLC-binding motif of Bmf, because mutations within it (A69P or D67A,K68A) abrogated the interaction of Bmf with DLC2 in yeast and in mammalian cells (Fig. 2A). Upon IL-3 deprivation or γ-irradiation, FDC-P1 cells coexpressing Bcl-2 and mutants of Bmf that do not bind DCL-2 died much more rapidly than those expressing both Bcl-2 and wild-type Bmf (Fig. 2B). These Bmf mutants also suppressed the formation of L929 fibroblast colonies more potently than did wild-type Bmf (14). Hence, interaction with DLC2 negatively regulates the proapoptotic activity of Bmf.

Through association with DLC1 or DLC2, respectively, Bim and Bmf may be sequestered to different sites within the cell to sense distinct stress signals. Because DLC forms part of the actin-based myosin V motor, as well as the microtubular dynein motor complex (21), we wondered whether Bmf and Bim might reside in separate cytoskeletal compartments. Indeed, when we separated cellular proteins into the filamentous actin and the paclitaxel (taxol)-polymerizable microtubular fractions (22), Bim and dynein intermediate chain (IC74) largely comigrated with microtubular components (P2), whereas Bmf and myosin V were confined to the filamentous actin–containing P1 fraction (Fig. 3A). Furthermore, treating cells with actin-depolymerizing agents, such as cytochalasin D orClostridium difficile toxin B, released Bmf from the filamentous actin–containing P1 fraction whereas the fractionation of Bim was unaffected (Fig. 3B).

Figure 3

Association of Bmf with the actin-based myosin V motor complex via DLC2. (A) Lysates from 107 MCF-7 cells were separated into F-actin–enriched P1, dynein-enriched P2, and supernatant fractions (22). Proteins from each fraction were size-fractionated by SDS-PAGE, transferred onto nitrocellulose, and probed with mAbs specific to Bmf (18), BimL (27), myosin V (29), or dynein intermediate chain IC74 (Sigma). (B) MCF-7 cells were treated for 3 hours with cytochalasin D (10 μM), or toxin B (10 ng/ml), then fractionated and processed as described in (A). Presence of IC74 and BimL in the F-actin–enriched P1 fraction represents contamination with microtubular components as evidenced by the fact that some BimL can still be found in the P1 fraction even after treatment with cytochalasin D or toxin B. (C) Myosin V is associated mostly with DLC2. Cytoplasmic dynein was enriched from MCF-7 cells (28) and myosin V from mouse spleen (m) or chicken brain (c) (30). These fractions were analyzed by protein immunoblotting using rat mAb 11F7 or 10D6, which can be used to distinguish between DLC1 and DLC2 [Web fig. 2 (17)]. Nitrocellulose membranes were probed with antibodies to myosin V or IC74 (Sigma) to demonstrate purity of the myosin and dynein motor fractions. (D) Extracts from mouse spleen cells (200 μg of protein) were incubated for 3 hours at 4°C with recombinant glutathione S-transferase (GST) or GST-tagged FADD, Bmf, or BimL proteins. Bound proteins were recovered on glutathione beads, eluted, size-fractionated by SDS-PAGE, and electroblotted onto nitrocellulose membranes, which were probed with an antibody to myosin V (29). The nitrocellulose membrane was stained with amido black (bottom panel) to document that comparable amounts of proteins were used in the pull-down experiments. (E) Lysates from 107 MCF-7 cells were fractionated through a 5 to 20% sucrose gradient (11). The pellet and soluble fractions were analyzed by protein immunoblotting for the presence of Bmf, Bim, DLC1, or DLC2.

The distinct localization of Bmf and Bim may be determined largely by their preferred dynein light chain partners. Contrary to a previous report (21), we found that highly enriched myosin V motor complexes (22) contained DLC2 but not DLC1 (Fig. 3C). Thus Bmf, by being preferentially bound to DLC2, might be complexed with myosin V on filamentous actin rather than forming part of the dynein motor complex. Consistent with this notion, incubation of extracts from mouse spleen cells with recombinant Bmf and Bim confirmed that only Bmf associated with myosin V (Fig. 3D). Furthermore, Bmf and Bim showed distinct migration patterns after subcellular fractionation (22) of lysates from MCF-7 cells on sucrose gradients (Fig. 3E). Because DLC1 forms homodimers avidly and because it binds Bim and IC74 through the same region (20), one partner of a DLC1 homodimer probably interacts with IC74, whereas the other binds Bim, thereby sequestering Bim to the microtubular dynein motor complex. It is likely that DLC2 homodimers similarly sequester Bmf to filamentous actin by binding with one arm to Bmf and with the other to myosin V.

We investigated whether Bmf and Bim are activated by distinct apoptotic stimuli in cells that express both proteins endogenously. Ultraviolet (UV) irradiation of MCF-7 cells released Bim from association with the dynein motor complex. In response to UV irradiation, Bmf translocated from denser to lighter fractions (22) in sucrose-gradient centrifugation (Fig. 4A). Treatment of cells with paclitaxel, which polymerizes microtubules, released Bim but not Bmf (Fig. 4A). Consistent with a critical role for Bim in this pathway to apoptosis, Bim-deficient thymocytes are abnormally resistant to the cytotoxic effects of paclitaxel (4). On the other hand, anoikis (absence of cell attachment and integrin signaling), an apoptotic stimulus that affects the actin cytoskeleton (23), resulted in the release of Bmf but not Bim (Fig. 4A). Because these experiments were conducted in the presence of the broad-spectrum caspase inhibitor zVAD-fmk at a concentration (50 μM) sufficient to block caspase activation, the release of Bmf or Bim, or both, is likely to represent an initiating event in apoptosis signaling rather than a consequence of apoptotic changes. Most important, endogenous Bmf (together with DLC2) released during anoikis could be coimmunoprecipitated with endogenous Bcl-2 isolated from mitochondria (Fig. 4B). In contrast, little if any Bmf was found complexed with Bcl-2 isolated from mitochondria of healthy cells.

Figure 4

Release of Bmf and Bim from their sequestration sites in response to distinct apoptotic stimuli. (A) MCF-7 cells were cultured in the presence of the broad-spectrum caspase inhibitor zVAD-fmk (50 μM). Lysates from control (untreated) cells were compared with those from cells subjected to various apoptotic stimuli, including anoikis (culturing cells for 24 hours in suspension on poly-hema–coated bacteriological Petri dishes), UV irradiation (100 J/m2), 1 μM paclitaxel. Lysates of 107 cells were fractionated through sucrose gradients (22). The sedimented (pellet) and soluble fractions were collected and analyzed by protein immunoblotting for Bmf and Bim. (B) During anoikis, Bmf translocates to mitochondria and binds to Bcl-2. Mitochondria were purified as described (11) from 2 × 108 healthy MCF-7 cells or cells subjected to anoikis. Mitochondrial proteins were extracted in lysis buffer containing 1% Triton X-100 (11). Immunoprecipitations were performed with mAb against human Bcl-2 bound to Sepharose beads. Bound proteins were eluted, size-fractionated by SDS-PAGE, electroblotted onto nitrocellulose membranes, and probed with mAbs to Bcl-2, Bmf, DLC1, or DLC1 plus DLC2.

Our data demonstrate that Bmf and Bim transduce distinct death signals caused by different forms of cell stress. They seem to represent sentinels mounted on the main cytoskeletal structures to monitor the well-being of the cell. Deregulated expression of Bcl-2 can promote tumorigenesis (24), thus it is possible that abnormalities in proapoptotic BH3-only proteins might also cause cancer. It is intriguing that the gene for human bmf is located on chromosome 15q14, the site of a candidate tumor suppressor gene lost in many metastatic, but not primary, carcinomas (25). Anoikis has been implicated as a barrier against metastatic tumor growth (26), raising the possibility that metastatic tumors harboring 15q14 mutations might have abnormalities in the expression or function of Bmf.

  • * To whom correspondence should be addressed. E-mail: strasser{at}


View Abstract

Navigate This Article