Interleukin-1 β enhances and interferon-γ suppresses activin A actions by reciprocally regulating activin A and follistatin secretion from bone marrow stromal fibroblasts

Abstract

Activin A is a multi-functional cytokine with a potent stimulation on erythroid cell differentiation in the bone marrow. The actions of activin A are determined by a balance of the levels of activin A and its inhibitor, follistatin (FS). However, the regulation of its actions in the bone marrow has been unclear. Here we show that bone marrow-derived stromal fibroblasts are the major source of activin A and FS in the bone marrow, and that the production of activin A is enhanced by interleukin-1β (IL-1β) and lipopolysaccharide (LPS), whereas interferon-γ (IFN-γ) inhibits the secretion of activin A by stromal fibroblasts. Concomitantly, IL-1β as well as LPS inhibits and IFN-γ stimulates FS secretion from stromal fibroblasts. Thus, these cytokines potently regulate activin A actions by reciprocal modulation of activin A and FS secretion from stromal fibroblasts. Because activin A exhibits anti-inflammatory effects in various tissues, up-regulation of activin A actions by IL-1β and endotoxin in the bone marrow may play a protective role against inflammatory processes as well as anaemia. The present results also suggest that the inhibitory effect of IFN-γ on erythropoiesis is mediated at least in part by a suppression of activin A actions in bone marrow.

Introduction

Activin A was originally found as a cytokine that enhances the release of follicle-stimulating hormone from the pituitary [1,2]. Subsequent studies by Eto et al. have shown that activin A is identical to erythroid differentiation factor, that induces the differentiation of mouse Friend and human K562 erythroleukaemia cells into haemoglobin-containing cells [3,4]. Activin A is a covalently linked homodimer of the inhibin β A chain and belongs to the transforming growth factor β (TGF-β) superfamily [4]. Biological activities of activin A are neutralized by a surplus of an activin binding protein, follistatin (FS) [5,6]. FS is found in a variety of tissues [7,8] and acts as a local modulator of activin A action [9].

Bone marrow stromal cells exert significant effects on proliferation and differentiation of haematopoietic cells. Increasing attention has been focused on the role of bone marrow stromal cells in the production of growth factors for haematopoiesis. Bone marrow tissues express mRNA for both activin A and FS [7], and bone marrow stromal cells as well as monocytes have been shown to produce activin A among various bone marrow cells [10–14]. Bone marrow stromal cells were demonstrated further to produce a significant amount of activin A, while monocytes produced at minimal levels [12]. These findings suggest that stromal cells are a major source of activin A in bone marrow microenvironment and that activin A plays a physiological role in function of bone marrow stromal cells. Haematopoietic effects of activin A include the induction of erythroid progenitor cell differentiation [15,16], suppression of early myeloid progenitor cell growth [17], potentiation of megakaryocyte differentiation [18,19] and induction of apoptosis in some myeloma cell lines [20,21]. Therefore, activin A is regarded as an important regulator of haematopoiesis in the bone marrow microenvironment.

However, the regulation of activin A action in the haematopoietic microenvironment has been poorly understood. In order to clarify the regulation of activin A and FS production in the bone marrow, we examined the effect of cytokines, interleukin-1β (IL-1β) and interferon-γ (IFN-γ) as well as lipopolysaccharide (LPS) on the production of activin A and FS by bone marrow-derived stromal fibroblasts. The results demonstrate that IL-1β and IFN-γ reciprocally regulate activin A and FS production by stromal fibroblasts in opposite directions.

Materials and methods

Reagents

Recombinant human (rh) IL-1β, rhIL-2, rh interferon-α (IFN-α) and rhIFN-γ were provided by Otsuka Pharmaceutical Co. (Tokushima, Japan). Specific activity of rhIL-1β and rhIFN-γ were defined to be 2 × 10⁸ units (U)/mg by the cell proliferation assay for D 10 cells and 2 × 10⁷ units (U)/mg by the inhibition assay for the infection to WISH cells by vesicular stomatitis virus, respectively. rhIL-3, rh granulocyte-colony stimulating factor (G-CSF), rh stem cell factor (SCF) and neutralizing rabbit antihuman IL-3 were supplied by Kirin Brewery Co. (Tokyo, Japan). The following reagents were purchased from the indicated manufacturers: rhIL-4, rhIL-6, rh tumour necrosis factor-α (TNF-α), rhTGF-β, rh granulocyte-macrophage-colony stimulating factor (GM-CSF) and rh macrophage-colony stimulating factor (M-CSF) from Genzyme Techne (Cambridge, MA, USA); lipopolysaccharide (LPS) from Sigma (St Louis, MO, USA); mouse antihuman CD3, CD4, CD5, CD8, CD11b, CD14, CD19 and CD33 monoclonal antibodies (MoAb) from Nichirei (Tokyo, Japan). Rabbit antihuman activin A and FS polyclonal antibodies and mouse antihuman FS MoAb were raised in our laboratory as described previously [22,23].

Cell preparations and cultures

Peripheral blood mononuclear cells (PBMC) and bone marrow mononuclear cells (BMMC) were isolated by Ficoll-Hypaque density gradient centrifugation (Pharmasia LKB Biotechnology, Uppsala, Sweden) from heparinized blood drawn from healthy volunteers after informed consent had been received and used immediately. All procedures involving human specimens were performed according to the protocol approved by the institutional review board for human protection. In order to purify BMMC, monocytes and myeloid cells were depleted by incubation with anti-CD11b, CD14 and CD33 MoAb and subsequent addition of Dynabeads M-450 goat antimouse IgG (Dynal, Great Neck, NY, USA) according to the manufacturer's instructions. BMMC thus obtained were resuspended in 75-cm² tissue flasks in Iscove's modified Dulbecco's medium (IMDM) supplemented with 12·5% fetal calf serum (FCS: Whittaker Bioproducts, Inc, Walkersville, MA, USA), 12·5% horse serum (Whittaker), 50 U/ml penicillin and 50 μg/ml streptomycin (Gibco BRL, Rockville, MD, USA). At weekly intervals, cultures were fed by replacing culture medium. Adherent cells were serially passaged at confluence, using 0·05% trypsin/0·53 mm EDTA (Gibco BRL) to achieve a homogeneous population of spindle-shaped cells. These cells were defined to be fibroblasts by their expression of the parenchymal cell antigens including vimentin, collagen type I and III without expression of endothelial cell antigens including collagen type IV and factor VIII antigen or the haematopoietic cell antigens of various cell lineages. These cells were further subcultured at 2 × 10⁵ cells/ml onto 24-well culture plates. After being expanded, the culture medium was changed to a serum-free medium containing a 1 : 1 mixture of Ham's F-12 medium and Dulbecco's modified Eagle's medium (GIT medium, Wako Pure Chemicals, Osaka, Japan). Monocytes were obtained from PBMC or BMMC by removal of non-adherent cells and subsequent depletion of T and B cells using Dynabeads M-450 CD4, CD8 and CD19 (Dynal). Adherent cells were prepared according to the adherence technique as described previously [24].

Stromal fibroblasts were cultured in the serum-free medium with addition of various cytokines and LPS. Culture supernatants were harvested after 2 days. IL-1-β, INF-γ and LPS showed no effect on the proliferation of cells at day 2, confirmed by a colourimetric assay for cell growth using a Cell Counting Kit (Dojindo, Kumamoto, Japan).

Activin A and FS assays

Immunoreactive activin A protein in culture supernatants was measured with RIA as described previously [22]. Briefly, supernatants or standards were incubated in tubes with rabbit antihuman activin A polyclonal antibody, before adding ¹²⁵I-labelled activin A. After overnight incubation of the mixtures, goat antirabbit IgG was added. The tubes were centrifuged at 2000 g for 30 min at 4°C, followed by decanting supernatants. Radioactivity was counted in a well-type gamma counter. Immunoreactive FS in supernatants was measured using an enzyme-linked immunosorbent assay (ELISA) with some modifications of our previously described method [23]. In brief, mouse antihuman FS MoAb was used to coat round-bottomed 96-well ELISA plates before loading samples or standards. Biotin-labelled goat antirabbit IgG and an avidin–biotin peroxidase system (Vectastain ABC kit, Vector, Burlingame, CA, USA) were used to detect the second-stage rabbit antihuman FS polyclonal antibody. The detection limit of this assay was 0·1 ng/ml. Intra- and interassay coefficient of variations were 3·4 and 8·7%, respectively.

Northern blot analysis

Cultured stromal fibroblasts and monocytes were lysed with guanidinium–isothiocyanate lysis buffer. Total cytoplasmic RNA was prepared using the single step method of guanidinium/phenol–chloroform extraction as described previously [25]. Fifteen μg RNA was subjected to electrophoresis on a 1% agarose–formaldehyde gel and transferred onto a nylon membrane (Hybond-N, Amersham Buchler). cDNA probe specific for the human βA chain was amplified by reverse transcription-polymerase chain reaction amplification using specific primers for the βA chain, which resulted in a 222-bp DNA segment containing exons 1 and 2. The primer sequences were ATGGAAT TCTCGGGGAGAACGGGTATGTG and ACGAAGCTTGTCCT GGTCCTGTTGGCCTT as 5′ sense and 3′ antisense, respectively. The cDNA probe specific for human FS was provided by Dr H. Sugino (Center for Enzymology, University of Tokushima, Tokushima, Japan). Blots were hybridized to α³²P cDNA probes using a random primer DNA labelling kit (Boehringer Mannheim, Germany). After hybridization, membranes were washed under conditions of high stringency and exposed to Cronex-4 autoradiography films (DuPont, Bad Homburg, Germany) at − 70°C. RNA loading was assessed using ethidium bromide stain and UV light.

Erythroid cell differentiation assay

Activin A bioactivity in the same culture supernatants used for both the activin A and FS protein assays was determined by differentiation of mouse Friend cells into haemoglobin-containing cells as reported previously [3]. Test samples were added to Friend cells cultured in 96-well plates. After 5 days of incubation, the cells containing haemoglobin were detected using dianisidine staining. The percentage of haemoglobin-containing cells was compared with that obtained after incubation with a series of known concentrations of rh activin A.

Results

Activin A and FS production by bone marrow stromal fibroblasts

Bone marrow stromal fibroblasts concomitantly secreted activin A and FS. In contrast, most of haematopoietic cells secreted negligible levels of activin A and follistatin, and only monocytes secreted activin A, but not follistatin, with much lower concentration compared to stromal cells (0·61 ± 0·08 ng/10⁵ cells). Therefore, stromal cells appear to be a major source of activin A in the bone marrow. Interestingly, IL-1β augmented activin A secretion and suppressed FS secretion by stromal fibroblasts, while IFN-γ suppressed activin A and augmented FS secretion (Fig. 1). No significant effect was observed, however, by other cytokines including IL-3, IL-4, IL-6, IFN-α, TNF-α, TGF-β, G-CSF, GM-CSF, M-CSF and SCF. The effects of IL-1β and IFN-γ were dose-dependent, and LPS showed an effect similar to IL-1β on activin A and FS production (Fig. 2).

Fig. 1.

Effect of cytokines on activin A and FS production from bone marrow stromal fibroblasts. The cells were cultured in triplicate for 2 days in serum-free medium with or without addition of cytokines including 100 U/ml IL-1β, 10 ng/ml IL-3, 100 ng/ml IL-4, 10 ng/ml IL-6, 1000 U/ml IFN-α, 1000 U/ml IFN-γ, 10 ng/ml G-CSF, 100 U/ml GM-CSF, 10 ng/ml M-CSF, 100 ng/ml SCF, 10 ng/ml TGF-β and 10 ng/ml TNF-α. Activin A and FS levels in CSNs were quantified. The data are representative of three independent experiments. Results are expressed as ng per 10⁵ cells (mean ± s.d., n = 4). *Indicates significant difference from the values in non-stimulated cultures according to the Mann–Whitney U-test (P < 0·05). □, Activin A; , FS.

Fig. 2.

Dose–response of activin A and FS production from human stromal fibroblasts. The cells were cultured in quadruplicate for 2 days in serum-free medium containing increasing concentrations of IL-1β (U/ml) (a), IFN-γ (U/ml) (b) and LPS (μg/ml) (c). The data are representative of three independent experiments. Results are expressed as ng per 10⁵ cells (mean ± s.d., n = 4). *Indicates significant difference from the values in non-stimulated cultures according to the Mann–Whitney U-test (P < 0·05).

mRNA expression of the βA chain and FS in stromal fibroblasts and monocytes

Northern blot analyses using the βA chain and FS oligonucleotide probes revealed constitutive expression of the βA chain and FS mRNA in stromal fibroblasts (Fig. 3). The expression of the βA chain and FS mRNA in stromal fibroblasts was modulated by IL-1β and IFN-γ in the same manner as the protein secretion. Although the βA chain is a subunit of inhibin (αβA) and activin AB (βAβB), neither the α chain nor the βB chain mRNA were expressed in bone marrow tissues [7,26]. Therefore, activin A and FS synthesis in stromal fibroblasts was controlled by these cytokines at a mRNA level. Monocytes constitutively expressed the βA chain mRNA, but not FS mRNA.

Fig. 3.

Expression of the βA chain and FS mRNA in stromal fibroblasts and monocytes. Stromal fibroblasts were exposed to either 100 U/ml IL-1β or 1000 U/ml IFN-γ for 3, 6 or 12 h (lanes 1, 4, 6 and 2, 5, 7, respectively) or neither IL-1β nor IFN-γ for 6 h (lane 3) before RNA extraction. RNA extracted from non-stimulated monocytes was applied at lane 8. The lowest panel shows ethidium bromide staining of 28S and 18S ribosomal RNAs.

Erythroid cell differentiation assay

Bioactive activin A levels were measured as erythroid cell differentiation activity in the same culture supernatants used for the protein assays. Bioactive activin A was found to be 2·9 ng per 10⁵ cells in non-stimulated supernatants. Addition of IL-1β at 1 and 10 U/ml induced bioactive activin A levels to 8·5 and 9·5 ng per 10⁵ cells after 2 days, respectively, while IFN-γ at 10 U/ml suppressed to 0·8 ng per 10⁵ cells after 2 days of culture. No bioactivity was detected in the culture with IFN-γ at 100 U/ml. A surplus of FS (100 ng/ml) completely abrogated the erythroid cell differentiation. Thus, bioactive activin A levels were lower than the protein levels, which suggested that concomitantly secreted FS binds to activin A to neutralize its bioactivity. IL-1β at 100 U/ml or IFN-γ at 1000 U/ml alone showed no effect on the activin A bioassay.

Discussion

The growth and differentiation of haematopoietic cells require exposure to growth factors/cytokines released in bone marrow microenvironment. Activin A has been demonstrated to act directly on erythroid progenitors to promote differentiation into haemoglobin-containing erythroid cells. Thus, activin A is regarded as one of the most important cytokines for erythroid differentiation. In the present study we explored the mechanism of regulation of activin A bioactivity by proinflammatory cytokines using cultured bone marrow stromal fibroblasts. Immunoreactive activin A and FS are constitutively synthesized and secreted by stromal fibroblasts, and stromal fibroblasts appear to be predominant cells to produce activin A and FS in the bone marrow. IL-1β and IFN-γ are found to be positive and negative regulators for activin A production by stromal fibroblasts, respectively. Interestingly, these cytokines inversely modulate FS production and secretion by stromal fibroblasts. Because FS binds to activin A to block an interaction with its cognate receptor, the unique regulation of activin A and FS production from stromal fibroblasts in opposite directions accentuates the modulation of activin A bioactivity and can dynamically regulate activin A actions in the bone marrow microenvironment.

LPS and IL-1β inhibit the synthesis of erythropoietin by renal cells [27], which is thought to play a major role in the development of anaemia in chronic disorder. The enhancement of activin A bioactivity in the bone marrow by LPS and IL-1β may serve as a compensatory mechanism by stimulating erythroid precursors to help the recovery of erythropoiesis after inflammation. In addition, activin A expression is also shown to be strongly induced in the skin after cutaneous injury [28], in the gut of patients with inflammatory bowel diseases [29], and in the synovial tissue of patients with inflammatory arthropathies [30]. Furthermore, activin A suppresses major IL-6-induced inflammatory processes, including B cell proliferation, phagocytic activities of monocytes and fibrinogen production in HepG2 cells [30]. Activin A also inhibits the production of IL-1β and enhances that of IL-1 receptor antagonist by monocytes [31]. Thus, there is a possibility that activin A provides anti-inflammatory effects in various tissues. The potent up-regulation of activin A bioactivity in the bone marrow by endotoxin and IL-1β may play a protective role against inflammatory processes as well as anaemia. IFN-γ is shown to inhibit haematopoiesis by its direct effect on haematopoietic progenitor cells [32]. In the light of the present results, the inhibitory effect of IFN-γ on erythropoiesis may be mediated at least in part by a suppression of activin A bioactivity provided by stromal fibroblasts.

Thus, the present studies provide evidence for the regulation of activin A bioactivity by proinflammatory cytokines in the bone marrow, and suggest a protective role of activin A against inflammatory processes as well as anaemia.