β₂-Microglobulin Identified as an Apoptosis-Inducing Factor and Its Characterization

The human cell lines used in this study were nonlymphocytic cell lines (K562 [chronic myelocytic], HL-60 and NB-4 [acute promyelocytic], THP-1 [acute monocytic], U937 [myelomonocytic], and KG-1 [acute erythrocytic]), B-lymphocytic cell lines (SKW, BALL, and Daudi), and T-lymphocytic cell lines (Jurkat, TALL, and CCRF-CEM). All of these cell lines were cultured in GIT media (Wako Co, Tokyo, Japan), which does not contain proteins such as β₂m or cytokines, as demonstrated in our previous studies.40,41 Human β₂m was purified from human urine to homogeneity (its sodium dodecyl sulfate-polyacrylamide gel electrophoresis [SDS-PAGE] analysis showed a single band) and was used in the early experiments (data not shown). Endotoxin was not detectable by Limulus assay (data not shown). Recombinant human β₂m consisting of 99 amino acids (Oriental Yeast Co, LTD, Shiga, Japan) was used for the same purposes as urine-derived β₂m. Reducing SDS-PAGE analysis of the recombinant product showed homogeneity with a single band. The concentrations of β₂m in supernatants of PDBu (Sigma, St Louis, MO) -treated HL-60 cells and in several maintained cell lines were measured by enzyme-linked immunosorbent assay (ELISA). In addition, anti-Fas MoAb (IgM, clone CH-11; MBL, Nagoya, Japan), recombinant human TNF-α (Calbiochem, La Jolla, CA), and endothelial interleukin-8 (IL-8; R&D System, Minneapolis, MN) were used as apoptosis-inducing factors in control studies.

The HL60 cells were resuspended at 2 × 10⁵ cells/mL in HamF12/Dulbecco’s modified Eagle’s medium (DMEM; GIBCO, Rockville, MD), which contains 10% GIT medium and 50 nmol/L PDBu and were incubated at 37°C for 3 days as in our previous studies.40,41 After 3 days, the supernatant was collected and precipitated in ammonium sulfate. The preparation was then purified further in a process using cation exchange fast protein liquid chromatography (FPLC) and gel filtration high performance liquid chromatography (HPLC). Additionally, the pool of active fraction eluted from the gel filtration HPLC column was added to reverse-phase HPLC. When the concentration of acetonitrile reached approximately 40%, the highest peak of AIF activity was eluted and added to the reverse-phase HPLC. The N-terminal sequence of the protein was determined by Edman degradation on an automated Applied Biosystems 473A sequencer (Applied Biosystems, Foster City, CA), as in our previous study.40,41 

Each of the human cell lines was seeded at 1 × 10⁵cells/mL in HamF12/DMEM with 10% GIT medium, and β₂m was added to the media at 10 μg/mL. After 2 days, the capacity to reduce 3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide (MTT) was determined. After adding 10 μL of MTT solution (5 mg/mL MTT in phosphate-buffered saline [PBS]), the preparation was incubated at 37°C for 3 hours. Cells with MTT formazan were dissolved in a solution of 2-propanol and 0.04 mol/L of HCl, and color absorbance was measured at 595 nm by a microplate reader (Bio-Rad, Hercules, CA). Additionally, residual cells were incubated with a digoxigenin-UTP terminal deoxynucleotide transferase mixture. Subsequently, they were stained with a peroxidase-conjugated antibody for digoxigenin by using an in situ Apoptosis Detection Kit (Apop Tag Plus; Oncor, Gaithersburg, MD); they were then counter-stained with 1% methyl green in sodium acetate (pH 4.0) and finally mounted. Specimens were examined and photographed with a microscope as in our previous study.40,41 Statistical analysis was performed using a Student’s t-test.

The mouse anti-β₂m MoAbs (IgG, clone NK1, NK2, and NK3) were derived from hybridoma cells and purified in a PBS together with ammonium sulfate. K562 cells and CCRF-CEM cells, at a concentration of 1 × 10⁵ cells/mL were exposed in 1 mL of HamF12/DMEM to a 10% GIT medium for 48 hours at 37°C after preincubation of 100 μg of anti-β₂m MoAb with 10 μg of β₂m for 15 minutes at room temperature (RT). Inhibition of the β₂m-induced apoptosis in these cells were detected using the MTT assays and TUNEL method as described above. The anti-Fas MoAb (IgG2b, clone SM1/23; Chemicon Int Inc, Temecula, CA) and anti-TNFR(p55) MoAb (IgG; AUSTRAL Biologicals, San Ramon, CA) at a concentration of 1 μg/mL were added to show whether β₂m-induced apoptosis was inhibited. In addition, the anti–IL-8 MoAb (IgG1, clone #6217.11; R&D Systems), anti–IL-8 receptor A (CDW128), MoAb (IgG2b, clone 5A12; Pharmingen, San Diego, CA), and anti–IL-8 receptor B (CXCR2) MoAb (IgG1, clone 6C6; Pharmingen) were added at a concentration of 5μg/mL to show whether β₂m-induced apoptosis was inhibited. Ten micrograms per milliliter each of TU149 MoAb (IgG2a, anti-HLA-B, C, and some A; Caltag, Burlingame, CA), YTH862 MoAb (IgG2b, anti-HLA-A, B, and C; Serotec, Oxford, UK), and W6/32 MoAb (IgG2a, anti-HLA-A, B, and C; Serotec) were added at 10 μg/mL to study the correlation with apoptosis induced by HLA-class I. The mouse anti-RecA MoAb (IgG2b, clone ARM414; MBL) was used as a negative control. These anti-HLA class I MoAbs and anti-RecA MoAb were used after dialysis against PBS.

The IL-1β–converting enzyme (ICE)-like protease inhibitors Ac-Tyr-Val-Ala-Asp-aldehyde (Ac-YVAD-CHO) and Z-Val-Ala-Asp(OMe)-fluoromethylketone (Z-VAD-FMK) as well as the CPP32/Apopain-like protease inhibitor Ac-Asp-Glu-Val-Asp-aldehyde (Ac-DEVD-CHO) were purchased from Calbiochem-Novabiochem Co (San Diego, CA). These peptides were dissolved separately in dimethylsulfoxide (DMSO; Wako Co, Tokyo, Japan) and 10 or 100 μmol/L of each peptide was added to the cell suspensions with final concentrations of 0.1% or 0.3% DMSO as performed in our previous study.42 

Fluorescein isothiocyanate (FITC)-conjugated mouse IgG1 (DAKO, Glostrup, Denmark) was used as a negative control. FITC-conjugated anti-HLA class I MoAb (clone W6/32; Serotec) was added to the solution to detect the presence of HLA class I on the cell surface membrane. β₂m was biotinilated with an ECL protein biotinylation module (Amersham, Little Chalfont, Buckinghamshire, UK). Avidin R-PE (Caltag) was added. Cells from various cell lines were centrifuged at 2,000 rpm for 1 minute and washed once in PBS. The cells were exposed to the biotinilated-β₂m (2μg) for 30 minutes, put on ice, washed twice in PBS, and exposed to avidin R-PE- and FITC-conjugated mouse anti-HLA class I MoAb simultaneously for 30 minutes on ice. The cells were analyzed using the FACScan (Nippon Becton Dickinson Co, Tokyo, Japan) as performed in our previous study.43 Each data point was derived from an analysis of 1 × 10⁴ cells.

The results of our binding study are shown briefly. Biotinilated β₂m was incubated with 5 × 10⁶ K562 cells or CCRF-CEM cells for 30 minutes at 4°C. After resuspension in 450 μL PBS, cells were washed twice. After removal of PBS, 1 μL of 120 μg/mL biotinilated β₂m was added. As a cold competitor, 100× β₂m was added. Ten microliters of 1 mmol/L d-biotin (Sigma) was added and incubated at 4°C for 1 hour with occasional shaking. As for the manufacturer’s protocol, 1 × 10⁶ cells of K562 cells were suspended in a total volume of 100 μL PBS and solubilized with 2.0% (vol/vol) Nonidet P 40.44,45 After removing insoluble material by centrifugation, 0.5 mL of the detergent extract was mixed with 5 μL disuccinimidyl suberate (DSS; Pierce, Rockford, IL) solution (20 mmol/L in DMSO; final concentration, 1 mmol/L) and allowed to react for 30 minutes at 14°C. The reaction was stopped by adding 2 μL of 1 mol/L Tris-HCl, pH 7.5.

After further centrifugation to remove precipitates from this sample, the supernatant was loaded onto an SDS-PAGE and immunoblotted. Protein samples were submitted to SDS-PAGE analysis on a 4% to 20% gradient gel under standard conditions using a mini-Protean II system (Bio-Rad), followed by silver staining. For immunoblotting, the proteins were transferred to polyvinylidene difluoride (Bio-Rad) membrane filters and were blocked in 2% (wt/vol) bovine serum albumin in washing buffer Tween 20 in PBS.

HL-60 cells at a concentration of 2 × 10⁵ cells/mL were cultured with 50 nmol/L of PDBu at 37°C for 3 days and differentiated into the monocyte/macrophage lineage. We investigated whether apoptosis was induced in K562 cells when the supernatant from PDBu-treated HL-60 cells was added to a culture media of K562 cells. PDBu itself did not induce apoptosis in K562 cells (data not shown), and we did not detect any AIF activity in conditioned media from unstimulated HL-60 cells. To detect whether proteins were the AIFs, the protein in the supernatant was precipitated from a solution of 95% ammonium sulfate. The preparation was then purified further in a process using cation exchange (FPLC; Fig1A) and gel filtration (HPLC; Fig 1B). Additionally, the pool of active fractions eluted from the gel filtration column was applied to reverse-phase HPLC. The purified AIF activity was eluted when the concentration of acetonitrile reached approximately 40% in reverse-phase HPLC (Fig 1C). We obtained specific activity of 1 × 10⁵ U/mg of protein (1 U AIF inhibited 50% of maximum reduction, as indicated by MTT reducing activity) and with a final purification of 270-fold, calculated from the first supernatant concentration.40 The overall yield was 0.05%.

We sequenced the protein that showed a single peak with AIF activity. From an analysis of the N-terminal sequence of the AIF, we found that it was strikingly (97%) homologous to the N-terminal sequence of human β₂m (Fig 1D). In addition, using ELISA, the concentrations of β₂m in the supernatants from the PDBu-treated HL-60 were measured and found to increase in relation to time (Fig 1E). We did not detect β₂m in unstimulated HL-60 cell-conditioned media (data not shown). The expression of β₂m mRNA in PDBu-treated HL-60 cells correlated with the concentration of β₂m in the supernatant (data not shown). Therefore, β₂m was recognized as a candidate AIF.

We initially examined whether β₂m would induce apoptosis in K562 cells. The proliferation of K562 cells was suppressed in a dose-dependent manner in the presence of β₂m at concentrations of more than 10 μg/mL, as detected by the MTT assay (Fig 2A). Simultaneously, we detected the induction of apoptosis in K562 cells with the TUNEL assay (Fig 2C and D), and we found that β₂m induced apoptosis in these cells in a dose-dependent manner (Fig 2B). In the presence of β₂m for 48 hours, the fraction of apoptotic K562 cells increased from 21.0% ± 2.2% (at 10 μg/mL) to 42.7% ± 4.5% (at 20 μg/mL) and to 43.3% ± 1.7% (at 100 μg/mL). Time course studies showed both inhibition of proliferation and induction of apoptosis most markedly between 48 and 96 hours (Fig 2E and F).

Based on the detection of β₂m-induced apoptosis in K562 cells, various human leukemic and lymphoma cell lines were cultured at 1 × 10⁵/mL in the presence of 10 μg/mL of β₂m for 48 hours at 37°C. Proliferation and percentage of apoptotic cells were studied. We found that β₂m induced apoptosis in more than 10% of U937, BALL, and CCRF-CEM cells (Fig 3A and B) as well as K562 cells (Fig 3C). Moreover, apoptosis both in K562 cells and in CCRF-CEM cells was induced by recombinant human β₂m (Fig3D and E). FasL/Fas interaction is known to strongly and rapidly initiate apoptosis in human leukemic cell lines such as Jurkat cells. β₂m does so more slowly, eg, over a few days instead of a few hours. K562 cells and CCRF-CEM cells were especially useful in this study of the mechanism of β₂m-induced apoptosis, because apoptosis is not regularly induced in their natural in vitro growth patterns.

The mouse antihuman β₂m MoAbs (clone NK1, NK2, and NK3) were derived from hybridoma cells and purified in PBS together with ammonium sulfate (Kubota et al, manuscript in preparation). K562 cells were grown in HamF12/DMEM to a 10% GIT medium for 48 hours at 37°C after preincubation of 100 μg of anti-β₂m MoAb with 10 μg of β₂m for 15 minutes at RT. The anti-β₂m MoAb (clone NK2) completely blocked both the suppression of cell proliferation (data not shown) and the induction of apoptosis in K562 cells (Fig 4A) and coordinately blocked both the suppression of cell proliferation (data not shown) and the induction of apoptosis in CCRF-CEM cells (Fig 4B). The anti-β₂m MoAb (NK2) inhibited any interaction with either biotinilated β₂m or the cell surface of CCRF-CEM cells (Fig 4C). This result suggests that β₂m induces apoptosis through an interaction with a receptor on the cell’s surface.

To investigate whether β₂m-induced apoptosis is mediated by the FasL/Fas or TNF-α/TNFR systems, we examined the effect of an anti-Fas neutralizing MoAb and an anti-TNFR neutralizing MoAb each at a concentration of 1 μg/mL. Blocking Fas/FasL interaction by an antagonistic anti-Fas MoAb did not interfere with β₂m-induced apoptosis in either K562 or CCRF-CEM cells (Fig 5A and B). Similarly, blocking TNF-α/TNFR interaction also did not interfere in either cell line (Fig 5A and B). We propose that β₂m-induced apoptosis is dependent on neither the FasL/Fas system nor the TNF-α/TNFR system.

We previously reported that endothelial IL-8 is purified as an AIF from PDBu-conditioned media of HL-60 cells and induces apoptosis in human leukemic cell lines.40,41 To confirm that β₂m induces apoptosis in human leukemic cell lines distinct from the pathway of IL-8–induced apoptosis, we examined the effect of an anti–IL-8 neutralizing MoAb and 2 anti–IL-8 receptors neutralizing MoAbs, each at a concentration of 5 μg/mL. These 3 antibodies could not inhibit β₂m-induced apoptosis in either K562 (Fig 5C) or CCRF-CEM cells (Fig 5D). In addition, IL-8 was not detected in the purified human urine-derived β₂m by ELISA (data not shown).

Recent studies have proposed that various types of apoptosis, such as Fas-mediated apoptosis, depend on the activation of caspases. To investigate whether a caspase cascade participates in β₂m-induced apoptosis, we studied the inhibition of apoptosis through pretreatment with caspase inhibitors. Induction of apoptosis in K562 cells was blocked by Ac-DEVD-CHO, Ac-YVAD-CHO, and z-VAD-fmk 3 hours before adding the β₂m (Fig 6A). The reduction of β₂m-induced apoptosis in K562 cells was efficient in the presence of each of the 3 caspase inhibitors at concentrations of 10 μmol/L (data not shown). Ac-YVAD-CHO significantly inhibited apoptosis at 100 μmol/L, achieving more than 90% inhibition. Simultaneously, preincubation of CCRF-CEM cells with each of the 3 kinds of caspase inhibitors inhibited β₂m-induced apoptosis (Fig 6B). The reduction of apoptosis in CCRF-CEM cells from our control was insufficient with concentrations of 10 μmol/L of each inhibitor (data not shown), but effective with 100 μmol/L of each. Z-VAD-fmk almost completely inhibited apoptosis at 100 μmol/L. We concluded that β₂m-induced apoptosis depends on the activation of both ICE-like protease and CPP32-like protease.

Addition of the amine-reactive, cross-linker, DSS, to Triton-X extracts of K562 cell membranes resulted in the formation of high molecular broad bands (its average complex showed a molecular weight [MW] of 150 kD) that were specifically inhibited by the addition of 100× unlabeled human purified β₂m (Fig 7). Using anti-β₂m MoAb, Western blot showed an approximately 150-kD β₂m-binding protein complex (data not shown).

MHC class I antigens, or HLA antigens, play an important role in tumor immunity. Loss of HLA molecules in B-cell lymphomas is associated with an aggressive clinical effect, such as lung cancer.16,17,46In melanoma, whose tumor cells often lack HLA class I antigen expression, distinct mutations in the β₂m gene have been identified.47 Antibodies against MHC class I antigen complexes induced apoptosis in activated T cells and T-cell leukemia cell lines. However, it is still not clear which epitope is responsible for apoptosis and which ligand binds to the responsible molecule on HLA class I antigens. We observed that human β₂m itself induced apoptosis in various kinds of leukemic cells and β₂m-enhanced antibody induced MHC complex-apoptosis in the T-cell leukemia cell line CCRF-CEM. We used K562 cells as target cells because they showed low background in MTT and TUNEL assay. Although K562 cells have no MHC class I antigen complex (data not shown),48,49 β₂m induced apoptosis even in these cells. We strongly suggest that human β₂m uses an apoptosis pathway in both T cells and myeloid leukemia cells distinct from the MHC class I antigen complex-induced apoptosis and Fas- or TNFR-mediated apoptosis pathways. Our report is the first evidence demonstrating the existence of human β₂m-induced apoptosis. Neutralizing antibodies against β₂m resulted in β₂m-induced apoptosis being completely blocked. So far, we have not determined the epitopes that might be involved in cell death signaling. Our cross-linking study suggested the presence of a MW 150-kD complex including presumably a binding protein and β₂m. Altered expressions of those genes that control apoptosis may lead to autoimmunity, malignant cell growth, neurodegenerative diseases, prolonged survival of cells during latent viral infections, and enhanced cell death during acquired immunodeficiency syndrome (AIDS). For example, Fas-ligand and Fas antigen, which control apoptosis and may lead to autoimmunity, are known to be mutations of lpr and gld, respectively, in mice.21,50-52 Christianson et al53reported that the phenotype in β₂m null mice was predisposed to develop chronic lupus erythematosus and to have defective antibody responses. Thus, in addition to the role oflpr and gld in autoimmunity, gene knock-out technology showed the importance of the β₂m gene in this disease. Obviously, a defective mechanism in apoptosis leads to autoimmunity, and β₂m may be involved. As for the relationship between FasL and Fas antigen, the binding protein for β₂m will be important. β₂m-deficient and TAP-1–deficient mice show self-tolerance of NK cells due to low expression levels of MHC class I protein.54 

TCR antigen-induced cell death occurs from the late G₁phase of the cell cycle studies55 and is dependent on pRb (retinoblastoma).55 In the future, we will investigate whether pRb or cdk-2 (cyclin-dependent kinase) activation is also involved in our system.

Compared with the action of the Fas- or TNFR-dependent pathway, the dose and time frame required for β₂m-induced apoptosis is large. β₂m actions require a specific physiological concentration measured in micrograms (which corresponds to serum levels) to induce apoptosis.

In our system, β₂m’s binding protein in the apoptosis pathway or its signaling pathway remains to be clarified. Ceramide, which is known as a second messenger, is produced by caspase-dependent apoptosis after activation by Fas or HLA class I antigen.56In Fas- and HLA class I-mediated peripheral T-cell apoptosis, caspase-dependent ceramide production, which acts downstream of the mitochondria, is important.32 Like the HLA class I-mediated pathway, our system also involves at least 2 different caspases. Table 1 shows the characterization of β₂m-induced apoptosis and various types of HLA class I-induced apoptosis. The presence of antibody for HLA class I antigen induced more apoptosis in CCRF-CEM cells by β₂m shown in TUNEL assay (data not shown). A recent report suggests that Fas-dependent apoptosis is dependent on ceramide.57 

Using specific MoAbs against HLA class I, α2 domain, which binds to the TCR, the results of several studies suggest that β₂m induces Fas-independent cell death.58 This pathway does not involve ICE-like proteases or CPP32-like proteases.

The cell surface regulator Toso has been discovered to be a regulator of Fas-induced apoptosis in T cells and blocks the Fas pathway.59 In β₂m-induced apoptosis, the molecule that blocks the β₂m pathway has not yet been identified. A recent study using primary embryonal fibroblasts from transgenic mice expressing murine H-2 and a swine class I transgene, transformed with the highly oncogenic Ad12, showed abnormal turnover of β₂m and suggests the existence of a novel β₂m-binding molecule that sequesters β₂m.60 One or more specific regulators or inhibitors for β₂m-induced apoptosis remain to be discovered. However, the signaling pathway is addressed in our system to the extent that β₂m-induced apoptosis is a distinct pathway from the Fas-dependent or MHC class I-dependent and the TNFR-dependent pathways of apoptosis.

The authors thank S. Kurokawa, K. Sato, and H. Ishikawa for technical assistance and we appreciate Dr Y. Miura.