The role of GM-CSF in B cell (patho)physiology is unclear. Although B cells can respond to GM-CSF, there is controversy concerning the extent to which various resting and activated B cell types can themselves produce this cytokine, and the possibility that it can function in an autocrine fashion has not previously been considered. The aim of the present study was to address these issues using hairy cells (HCs) and chronic lymphocytic leukemia cells, two intrinsically activated mature malignant B cell types (with activation being more uniform and more pronounced in HCs). Normal B cells were used for comparison. Using a number of techniques, we demonstrated the constitutive production of GM-CSF by all three cell types and showed that the cytokine was biologically active. GM-CSF mRNA and protein were increased after cell activation by PMA, and constitutive production of the cytokine was highest in HCs, suggesting that the level of GM-CSF production is influenced by cell activation. Because GM-CSF is known to be antiapoptotic for myeloid cells, we used blocking anti-GM-CSF Abs to examine the contribution of autocrinely produced cytokine to cell survival. The Abs produced marked reduction in the in vitro survival of HCs, chronic lymphocytic leukemia cells, and normal B cells by promoting apoptosis. Taken together, these findings suggest that, in combination with other known rescue factors, autocrinely produced GM-CSF may contribute to normal and malignant B cell survival in vivo.
The best-known functions of GM-CSF are its stimulatory effects on myelopoiesis and on mature myeloid cell function (1, 2). However, it has been known for some time that the effects of GM-CSF are not confined to myeloid cells and that, like most other cytokines, the agent has pleiotropic effects on multiple target cell types (1). In particular, it has become clear that GM-CSF can exert a number of different effects on both immature and mature B cells (3, 4, 5). For example, we have previously focused on the role of GM-CSF in mature B cell lymphoproliferative disorders, especially hairy-cell leukemia3 (HCL) and chronic lymphocytic leukemia (CLL), and have shown that exogenous GM-CSF can influence the adhesion and motility of hairy cells (HCs) (6).
Regarding the production of GM-CSF, T lymphocytes, fibroblasts, endothelial cells, and macrophages are considered the major cellular sources of the cytokine (7, 8). B cell production of GM-CSF has been less studied and remains a subject of some controversy (9, 10, 11, 12, 13, 14, 15, 16, 17, 18). For example, in one report, normal blood B cells were shown to produce GM-CSF constitutively (9), but in another no such production could be demonstrated (10). In CLL, one group has shown constitutive GM-CSF production in some cases but not in others; cell stimulation enhanced production (10). In direct contrast, in a recent report no GM-CSF message was detectable by RT-PCR in 10 of 10 cases of CLL (11). The ability of HCs to produce GM-CSF does not yet seem to have been specifically studied, but in a single report concerning T cells in HCL, RT-PCR of purified HCs failed to detect the cytokine in four cases (12).
Given these contradictory findings concerning B cell production of GM-CSF and the absence of specific studies of HCs, we examined the expression of the cytokine in HCs and CLL cells as compared with that in normal B cells. The results unequivocally demonstrated that all three cell types produce the cytokine constitutively and that the quantity produced and secreted is related to cell activation. These findings caused us to consider the function of this B cell-derived GM-CSF.
It has been suggested that B cell-derived GM-CSF may have a paracrine role in stimulating myelopoiesis and mature myeloid cell function (10) but, surprisingly, the possibility that GM-CSF produced by mature B cells might have autocrine effects does not seem to have been considered. During leukemic myelopoiesis, autocrine production of GM-CSF can contribute to malignant myeloid cell proliferation and survival (7, 19). Because HCs and CLL cells have a low proliferate capacity and because we have shown that exogenous GM-CSF has no effect on this (6), in this study we focus on the possible role of autocrine GM-CSF in enhancing the survival of these cells. We show that autocrinely produced GM-CSF is indeed a survival factor for HCL, CLL, and normal B cells.
Materials and Methods
Cell and tissue preparation
Mature malignant B cells.
Heparinized peripheral blood (PB) samples were obtained from HCL and B-cell CLL patients after informed consent. All had typical disease as determined by clinical presentation, morphology, immunocytochemistry, and immunophenotyping (20, 21). Mononuclear cells were isolated by density centrifugation (Lymphoprep, Life Technologies, Paisley, U.K.) and depleted of T cells (CD3+ < 2%) and monocytes (CD14+ < 2%) by immunomagnetic bead separation (Minimax System, Miltenyi Biotech, Bergisch Gladbach, Germany) (CD19+ > 98%).
Normal B cells.
These (CD19+ > 95%) were purified from the PB mononuclear cells of healthy donors by positive selection (using immunomagnetic beads) of cells expressing CD19.
Paraffin-embedded or frozen, formaldehyde-fixed tissues were employed. The normal node and tonsil were reactive surgical specimens, whereas the “normal” splenic tissue was from patients with immune thrombocytopenic purpura.
Immunochemical detection of GM-CSF
Cytocentrifuged, air-dried, methanol-fixed (5 min) cell suspensions; rehydrated, xylene-cleared tissue sections; or frozen tissues were washed with PBS and incubated with freshly prepared 3% hydrogen peroxide/PBS for 60 min at room temperature to remove endogenous peroxidase activity. Material was incubated with first-layer Ab (diluted to an optimal concentration in 1% AB serum/PBS) for 10 min at room temperature.
After further washing with PBS, a 1:15 dilution of anti-mouse Ig-biotin conjugate was added, and slides were incubated for an additional 30 min at room temperature. An avidin-peroxidase conjugate third layer was then added, and enzyme activity was visualized with 3-amino-9-ethylcarbazol. The final substrate incubation was standardized at 8 min to permit signal intensity comparison. All samples were counterstained with hematoxylin and examined microscopically.
A triple-layer method was used to detect GM-CSF and its receptor. After staining with a given mAb, cells were washed and exposed to biotinylated horse anti-mouse second-layer Ab and then to streptavidin-PE as third layer. Class-specific controls were included in all instances. Fluorescence was then measured with a Becton Dickinson (Oxford, U.K.) flow cytometer.
For permeabilization, cells were first fixed in cold 2% paraformaldehyde containing 15 mmol glucose (60 min, 4°C) and then were exposed to 0.2% Tween 20/PBS (30 min, 37°C) and washed before staining with mAb.
Abs and other reagents
Several different anti-GM-CSF Abs were used. Two mAbs (clones 3092 and 1089 from Endogen, Boston, MA; both IgG1) were used extensively. Both of these Abs have been reported to block the binding of GM-CSF to its receptor (24). In addition, a blocking polyclonal sheep Ab (R&D Systems, Abingdon, U.K.) was used for the GM-CSF bioassay. The blocking activity of all three Abs was confirmed here in preliminary experiments in which each of the reagents blocked, in a concentration-dependent manner, the ability of exogenous GM-CSF to support the growth of TF-1 cells (data not shown). Equivalent concentrations of class-specific control Abs and nonimmune sheep Ig produced no such blocking. For each set of experiments in which these blocking Abs were used, optimal concentrations were first determined.
For the ELISA assays of GM-CSF, two rat polyclonal Abs (BVD2-23B6 and BVD2-21C11 from PharMingen, San Diego, CA) were used.
mAbs against the α and βc chains of the GM-CSF receptor were a kind gift from Professor A. Lopez (Institute of Medical and Veterinary Science, Adelaide, Australia).
The mAbs (CD3, CD14, and CD19) used for purification and characterization of the B cell preparations were obtained from Becton Dickinson.
GM-CSF and IL-3 were obtained from Schering-Plough (Kenilworth, NJ) and Sandoz (Leeds, U.K.), respectively.
Total RNA was isolated using the procedure of Chomczynski and Saachi (25). A total of 15 μg were treated with deionized glyoxal at 55°C for 60 min, electrophoresed, and blotted onto a nylon membrane. Membranes were hybridized to 32P-labeled probes using standard methodology (26).
The GM-CSF cDNA probe was obtained by RT-PCR cloning of a fragment corresponding to nucleotide positions 157–505 of the published sequence (accession no. E02975). To measure RNA loading, the filter was hybridized to L27, a probe for ribosomal RNA; this was a gift from Dr. O. Braissant (Institut de Biologie Animale, Lausanne, Switzerland). The insert corresponds to nucleotides 144–380 of the rat ribosomal RNA L27 subunit gene (accession no. X07427). The identities of both probes were confirmed by sequencing.
Total RNA was reverse transcribed with 25 U Moloney murine leukemia virus reverse transcriptase (Promega, Southampton, U.K.) at 37°C for 60 min in a 20-μl reaction mixture containing 5 μg total RNA, 4 μl 5× primary-strand buffer (0.25 M Tris-HCl (pH8), 0.75 M KCl, 50 mM MgCl2, 1 mM DTT, and 2.5 mM of each dNTP), 20 U RNasin (Promega), and 0.25 μg of an RT primer (5′-ACTCCCACCATGGCTGTGG) designed from the GM-CSF cDNA sequence. The reaction was terminated by incubation at 70°C for 10 min, and the volume was made up to 80 μl with double distilled H2O. Five-microliter aliquots of the resulting first-strand cDNA were subjected to PCR using PARR buffer (Cambio, Cambridge, U.K.), 2 mM of each dNTP, 0.25 μg of forward (5′CTGCTGAGATGAATGAAACAG) and reverse (5′TCCAAGATGACCATCCTGAG) primers, and 5 U Taq polymerase (Promega).
The mix was overlaid with mineral oil and amplified using a “touchdown” protocol (27) in which the annealing temperature was reduced from 65°C to 55°C in 1°C steps every second cycle; 15 cycles at 55°C were subsequently performed. Equal aliquots of the PCR product were electrophoresed, and Southern blotting (26) was performed using the GM-CSF cDNA probe described above. Molecular size was determined by coelectrophoresis of a mixture of λ and φX174 DNA restricted with HindIII and HaeIII, respectively.
The RT primer corresponded to nucleotides 566–584, the forward primer to nucleotides 181–201, and the reverse primer to nucleotides 530–549 of the published GM-CSF sequence (accession no. M11220). Because the forward primer flanked the exon l-exon 2 junction and because the reverse primer was from exon 4, amplification from potentially contaminating genomic DNA was prevented.
These were cultured in RPMI 1640 (Life Technologies) supplemented with 10% low endotoxin FCS (Globepharm, Esher, U.K.), l-glutamine, penicillin, and streptomycin (Life Technologies) and subcultured as appropriate. The TF-1 cell line was maintained with either GM-CSF (200 U/ml) or IL-3 (15 ng/ml).
Lymphoid cell culture.
To measure GM-CSF secretion, HCs, CLL cells, and normal B lymphocytes were cultured for 24 h in serum-free QBSF-51 (Sigma, Gillingham, U.K.) with or without PMA (1 μg/ml).
Detection of GM-CSF by ELISA
A sandwich technique using a capture/detection Ab pair (PharMingen) was used. Capture Ab (BVD2-23B6, 2 μg/ml in 0.1 M NaHCO3 (pH 8.2)) was coated onto a MicroElisa III 96-well plate (Becton Dickinson) by overnight incubation at 4°C. After three washes with PBS/Tween, the plate was blocked with 10% FCS in PBS for 2 h at RT. GM-CSF standards (0–500 pg/ml) or test samples were added after further washing, and the plates were incubated for 18 h at RT in a humidified atmosphere. GM-CSF was detected by addition of biotin-conjugated anti-GM-CSF Ab (BVD2-21C11) before avidin-peroxidase and 2,2′azino-bis(3- ethylbenzthiazoline-6-sulfonic acid) (Sigma) substrate development. Optical density was measured at 405 nm, and GM-CSF concentration was determined by extrapolation from a standard curve derived from known amounts of the cytokine.
Cell survival studies
A range of culture conditions were employed. Cells were cultured in either QBSF-51 or RPMI 1640 + 1 mg/ml BSA; in all experiments, serum was not used to avoid possible confounding effects of exogenous cytokines. Cells were cultured on plates that were either untreated or coated with polyHEMA (poly(2-hydroxyethyl methacrylate)) (Sigma), a nontoxic hydrophilic polymer that prevents cell adherence. PolyHEMA was employed to avoid losing from analysis any adherent cells and to eliminate possible confounding effects of adhesion on cell survival.
Cell survival was measured by FACS using staining either with propidium iodide (PI; 5 μg/ml) or with 40 nM 3,3′-dihexolyloxacarbocyanine iodide (DiOC6; Sigma). Dead cells become permeable to PI, and therefore they fluoresce bright red. DiOC6 is a cell-permeable green fluorochrome that is taken up by charged but not depolarised mitochondria and which therefore stains live but not dead cells (28). Apoptosis was specifically detected by double staining cells with annexin V-FITC and PI (28). Briefly, the cells were washed in PBS and incubated for 15 min in 50 μl of a 1:20 dilution of annexin V-FITC (PharMingen), added to 350 μl PI (10 μg/ml), and analyzed by flow cytometry. Annexin V specifically binds to phosphatidylserine, a phospholipid expressed on the surface of apoptotic but not live cells.
Detection of GM-CSF protein
Immunocytochemistry of isolated cells.
All hairy cells (n = 12) displayed moderate or strong GM-CSF positivity (Fig. 1). CLL lymphocytes (n = 20) were also positive, although expression was variable and generally weaker than in HCL. Normal PB B cells (n = 6) displayed weak positivity (Fig. 1).
To quantitate GM-CSF expression and to determine its cellular location, triple-layer FACS was used to analyze highly pure permeabilized (detects surface and cytoplasmic cytokine) and nonpermeabilized (detects surface cytokine only) B cell populations (Fig. 2). HCs and CLL cells possessed comparable amounts of GM-CSF at the cell surface. HCs also expressed large amounts of the cytokine in their cytoplasm, whereas in CLL expression was largely confined to the cell surface. Normal B cells had only small amounts of GM-CSF.
Immunohistochemistry of lymphoreticular tissue.
The majority of HCs in bone marrow (n = 6) and spleen (n = 3) contained moderate or strong reactivity. Most CLL cells in node (n = 2) and spleen (n = 1) displayed moderate cytoplasmic positivity (data not shown). Therefore, these studies are in accord with our findings for PB cells.
In normal node (n = 1), tonsil (n = 4), and spleen (n = 4), most lymphocytes were weakly stained except in follicle centers and in the area of tonsil immediately adjacent to the reticulated epithelium of the crypts where the expression was stronger. Similar results were obtained for both frozen and paraffin-embedded sections (data not shown).
Bioassay of lymphoid GM-CSF.
As a measure of biological activity, we examined the ability of cell lysates to support survival/proliferation of a GM-CSF-dependent cell line (TF-1 cells). As shown in Fig. 3, HCL and CLL cell lysates supported the proliferation of TF-1 cells. This proliferation was consistently reduced in the presence of optimal concentrations of the three different blocking anti-GM-CSF Abs. Blocking by the specific Abs was incomplete, indicating that factors in the lysates other than GM-CSF were also able to support the TF-1 cells.
Detection of GM-CSF mRNA
In all instances, a single band of the correct molecular size (368 bp) was demonstrated by RT-PCR (Fig. 4). Hybridization to a GM-CSF cDNA probe confirmed that the RT-PCR product was indeed derived from GM-CSF mRNA (Fig. 4).
The expression of GM-CSF protein and message after cell stimulation
Having demonstrated by immunocytochemistry and FACS that HCs contain more GM-CSF protein than do CLL cells, which in turn contain more than normal B cells, we postulated that the amount of GM-CSF present might be related to intrinsic cell activation. Thus, it is well established that HCs are a form of highly activated B cell (20), whereas in CLL activation is less prominent (29).
To test this postulate, CLL and normal B cells were stimulated with PMA, and their GM-CSF content was measured by FACS (Fig. 5, A and B). In addition, normal B cell lysates were also analyzed by ELISA (Fig. 5 C). By both methodologies, cell stimulation led to an increase in GM-CSF protein production. Brefeldin A, which blocks the translocation and release of secretory vesicles (30), also increased the level of GM-CSF, as measured by permeabilized FACS, and further enhanced the increase in GM-CSF detected after PMA stimulation. Because we show later that unstimulated CLL and normal B cells secrete GM-CSF, blocking of this secretion by brefeldin would be expected to increase intracellular GM-CSF levels.
Having demonstrated these marked increases in GM-CSF protein after cell stimulation, we then examined by Northern analysis whether such stimulation led to a comparable increase in message production. PMA caused GM-CSF mRNA to be readily detectable in HCs, CLL cells, and normal B cells although in all three cell types it was undetectable by Northern blotting before stimulation (data not shown).
Secretion of GM-CSF
Culture supernatants from HCL, CLL, and normal B cells (±PMA stimulation) were analyzed by ELISA for the presence of GM-CSF (Fig. 6).
GM-CSF was detected in the supernatants of all three cell types in the absence of stimulation (Fig. 6). Levels of secreted cytokine were highest in the CLL cell supernatants. In the case of HCs, the low levels of GM-CSF detected in the supernatant were not the result of cytokine binding to its surface receptor because a blocking anti-receptor Ab did not increase the levels of GM-CSF detected (n = 4; data not shown).
For all three cell types, PMA produced an increase in GM-CSF secretion (Fig. 6).
FACS analysis of GM-CSF receptor expression by B cells
Before considering the functional effect of GM-CSF on B cells, it seemed important to examine receptor expression. This was done by a triple-layer FACS method (Fig. 7) identical with that employed for detection of cell-surface GM-CSF. Both α and βc chains of the GM-CSF receptor were readily demonstrable on HCs. CLL and normal B cells also expressed both receptor chains but at lower levels.
Given our demonstration above (Fig. 6) that unstimulated CLL and normal B cells secrete more GM-CSF than do HCs, it was necessary to consider the possibility that secreted ligand was interfering with Ab detection of the receptor. Therefore, cells were submitted to acid washing to remove any bound GM-CSF (31) and were retested for α-chain expression; little or no increase in staining was observed (data not shown).
We next considered the functional implications of our demonstration that B cells express both GM-CSF and its receptor.
Effect of GM-CSF on lymphoid cell survival
It is well-known that GM-CSF enhances the survival of myeloid cells (1). Therefore, we postulated that the cytokine might also have an antiapoptotic effect on lymphoid cells.
We first tested exogenously added cytokine (10, 100, or 1000 ng/ml) and demonstrated that it did not enhance the survival of HCL, CLL, or normal B cells (data not shown).
We next examined the possibility that endogenously produced GM-CSF might be having an antiapoptotic effect. To do this we employed the two anti-GM-CSF mAbs (clones 3092 and 1089) used earlier in this study; both block binding of the cytokine to its receptor (Ref. 24 and Materials and Methods).
Both Abs but not the isotypic control produced a marked reduction in the survival of HCL, CLL, and normal B cells (Fig. 8).
To provide further evidence that GM-CSF can act as an autocrine rescue factor for mature B cells, the effect of anti-GM-CSF blocking mAb was studied using different culture conditions and a range of cell densities and employing a different method (mitochondrial depolarization) of measuring cell death. It seemed important to employ a range of cell densities because we have shown that, on a nonadherent surface, CLL cell survival is density dependent. Blocking autocrine GM-CSF induced lymphoid cell killing over a wide range of cell densities (Fig. 9).
Finally, to determine the mode of cell death, we examined the effect of blocking endogenous GM-CSF on cell survival as measured by double staining of cells with annexin V-FITC/PI. As shown in Fig. 10, cell death clearly occurred by apoptosis as indicated by a marked increase in the proportion of annexin-bright/PI-dim cells.
Although there have been a number of reports describing effects of GM-CSF on B cells of different maturities (3, 4, 5, 6), the true role of this cytokine in B cell physiology and pathology remains unclear. In particular, there is still controversy concerning the ability of different activated and nonactivated B cell types to produce this growth factor. Furthermore, the possibility that GM-CSF might function as an autocrine B cell cytokine has not previously been addressed.
Given this background, the aims of the present study were to clarify the question of whether or not B cells express GM-CSF and to elucidate the functional significance of any such GM-CSF expression. Because we have been interested in chronic lymphoproliferative disorders for a number of years (32, 33, 34, 35, 36) and because we had already demonstrated some functional effects of GM-CSF on HCs (6), in the present study we focused on the malignant B lymphocytes of HCL and CLL and compared them with normal B cells.
Using a range of techniques, we showed that both malignant and normal mature B lymphocytes unequivocally produce GM-CSF. Thus, both cell-associated and secreted GM-CSF protein were demonstrated by a number of methods (immunocytochemistry and FACS of isolated cells, immunohistochemistry of lymphoid tissues, and bioassay and ELISA of cell lysates and culture supernatants) and was increased after cell stimulation. Furthermore, the HCL, CLL, and normal B cells all expressed GM-CSF mRNA, which was increased after stimulation. HCs and CLL cells constitutively expressed more GM-CSF than did normal B cells. Because HCs, and to a lesser and different extent CLL cells, are known to be constitutively activated (20, 29) (presumably by intrinsic oncogenic events), it seems likely that their baseline GM-CSF production is influenced by this constitutive activation.
As already reported by us and others, we confirmed in this study that HCs and CLL cells possess both chains of the GM-CSF receptor (6, 37). Receptor expression, although low on both cell types, was higher on HCs than on CLL and normal B cells. GM-CSF receptor is known to internalize after ligand binding (38). Because we show in this study that CLL cells constitutively secrete more cytokine than HCs do, the lower levels of receptor observed on CLL cells may be a consequence of this higher rate of GM-CSF secretion.
Given our demonstration that mature B cells express both GM-CSF and its receptor, we next considered the possibility that the cytokine might serve an autocrine function. From the reported paracrine effects of the cytokine, the principal candidate functions included proliferation, differentiation, and cell survival (1). Because neither HCs nor CLL cells spontaneously proliferate or differentiate in vitro, we chose to concentrate on cell survival. This seemed further justified by our previous demonstration that exogenous GM-CSF does not induce the proliferation or differentiation of HCs in vitro (6).
To establish the effects of GM-CSF on spontaneous apoptosis, we first tested the effect of exogenously added cytokine and showed that it did not influence cell survival. We next employed three different blocking anti-GM-CSF Abs to neutralize endogenously produced GM-CSF; each Ab reduced the viability of HCL, CLL, and normal B cells by promoting apoptosis. Furthermore, the effect was observed using three different methods of measuring cell survival/apoptosis and employing cells cultured under different conditions over a range of cell densities.
In conclusion, the work presented here fully supports those previous studies demonstrating that both normal and malignant mature B cells produce GM-CSF, particularly after activation. In addition, our studies of HCL and CLL cells suggest that not only external stimulation but also disease-related intrinsic activation may provide the stimulus for cytokine production in these cells. Autocrine GM-CSF then clearly has the potential to contribute to the malignant behavior of the cells. Our functional studies suggest that the most likely contribution of the cytokine is to provide protection from apoptosis. Although a range of cytokines are known to enhance CLL cell survival (39, 40, 41), only IL-8 and basic fibroblast growth factor (bFGF) have been shown to be consistently produced by these cells in the absence of stimulation (11, 42). Furthermore, only in the case of IL-8 has it been demonstrated that blocking of endogenous cytokine results in apoptosis (41). In HCL, only TNF has been implicated as an autocrine survival factor (43). Therefore, the present study adds GM-CSF to IL-8 and TNF as cytokines capable of regulating the survival of mature malignant B cells in an autocrine fashion.
We thank the Cancer Tissue Bank Research Centre, University of Liverpool, for providing many of the tissue sections used in this study.
This work was supported by the Leukaemia Research Fund, U.K., and the North West Cancer Research Fund, U.K.
Abbreviations used in this paper: HCL, hairy-cell leukemia; CLL, chronic lymphocytic leukemia; HC, hairy cell; PB, peripheral blood; polyHEMA, poly(2-hydroxyethyl methacrylate); PI, propidium iodide; DiOC6, 3,3′-dihexolyloxacarbocyanine iodide.