The plasmacytoma growth inhibitor restrictin-P is an antagonist of interleukin 6 and interleukin 11. Identification as a stroma-derived activin A.

A stromal protein, designated restrictin-P, that specifically kills plasma-like cells was purified to homogeneity and shown to be identical with activin A. The specificity to plasma-like cells stemmed from the ability of restrictin-P/activin A to competitively antagonize the proliferation-inducing effects of interleukin (IL) 6 and IL-11. Restrictin-P further interfered with the IL-6-induced secretion of acute phase proteins by HepG2 human hepatoma cells and with the IL-6-mediated differentiation of M1 myeloblasts. A competition binding assay indicated that restrictin-P did not interfere with the binding of IL-6 to its receptor on plasma-like cells, suggesting that it may act by intervening in the signal transduction pathway of the growth factor. Indeed, concomitant addition of restrictin-P and IL-6 to cytokine-deprived B9 hybridoma cells was followed by sustained overexpression of junB gene until cell death occurred, while IL-6 alone caused a transient increase only. This altered response to IL-6 stimulation was accompanied by a moderate increase in STAT protein activation. Thus, in this study, we identified the plasmacytoma growth inhibitor, restrictin-P, as being activin A of stromal origin. It is shown that activin A is an antagonist of IL-6-induced functions and that it modifies the IL-6 signaling pattern.

Regulation of hemopoiesis is mediated by cytokines that act through distinct mechanisms. Some, like colony-stimulating factors (CSFs) 1 promote accumulation of hemopoietic cells by inducing proliferation coupled with differentiation (1). Others, like tumor necrosis factor, may cause cell cycle arrest and thus limit cell accumulation (2). The outcome of the interaction between the growth factor and the cell often depends on the nature of the target cell; as it is with transforming growth factor (TGF)-␤, the same cytokine may be stimulatory to one cell type and inhibitory to the other (reviewed in Ref. 3). Whereas some inhibitors operate by slowing down cell growth (4) or by induction of terminal differentiation (5), others cause cell death (6) by inducing apoptosis (7)(8)(9)(10). Restrictin-P has formerly been described as an inhibitor of plasmacytoma cell growth (11)(12). The biological activity of this factor was first noticed through the selective ability of primary stromal cells to slow down the proliferation of plasmacytoma cells (13,14). A similar function was exhibited by trypsin-released proteins obtained by mild treatment of a bone marrow-derived stromal cell line of mouse origin (MBA-2.1) (12). The released crude protein mixture inhibited the growth of a series of plasmacytomas and hybridomas but did not have significant effects on the growth of a variety of other leukemia cell lines of lymphoid, erythroid, and myeloid origin (12). Similarly, no effect was observed on normal cell populations such as bone marrow cells responding to colony-stimulating factors or spleen cells induced by mitogens (12). This unique specificity prompted us to isolate the active component. However, factor(s) mediating the growth inhibition were found to be produced by the stroma cell line in minuscule amounts, and it was necessary to establish conditions for large scale production of the factor. We found that the producer cell line MBA-2.1 could be propagated on a threedimensional carrier of nonwoven fabric of polyester loaded in a bioreactor system under complete protein-free conditions (15). The study of such bioreactors showed that the cells could be maintained under protein-free conditions for up to 10 months while producing restrictin-P activity along with TGF-␤, macrophage (M)-CSF, and IL-6. Restrictin-P obtained from the bioreactor system induced in its target cells early G 1 /G 0 arrest, morphological changes, and signs of cell damage characteristic of apoptosis (11,16) accompanied by intracellular ionic changes (17).
The present study was aimed at identifying restrictin-P by purifying it to homogeneity and at analyzing the mechanism by which restrictin-P exerts its specific inhibitory effect on plasma-like cells.
Cytokine Biological Assays-Restrictin-P was monitored using either MPC-11 plasmacytoma or B9 hybridoma cells. MPC-11 cells were seeded at 8 ϫ 10 3 cells/ml in 96-well microtiter plates (100 l/well) (Costar, Cambridge, MA) in RPMI supplemented with 7.5% FCS in the presence of serial dilutions of the restrictin-P-containing samples or buffer (20 mM Tris-HCl, pH 7.8, or 20 mM Hepes, pH 7.8). Cell viability was determined following 4 days of incubation using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay, which measures cell viability via mitochondrial activity (30). One unit of activity was designated as the amount of protein which, under the above conditions, caused 50% growth inhibition relative to the control. The assay was essentially the same using B9 cells except that the culture conditions were as indicated above for the B9 hybridoma. The latter was also used to titrate IL-6 levels. Briefly, B9 cells (5 ϫ 10 3 cells) were cultured in 96-well plates (200 l/well) in the presence of test samples and recombinant human IL-6 controls. At 64 h of incubation, the cells were pulsed with [ 3 H]thymidine (Rotem Industries, Israel, 1 Ci/well) for 16 h. Unit IL-6 was determined by relating to IL-6 1st international standard (code 89/548). The biological effect of IL-6 was monitored in two additional assays. HepG2 cells were seeded at 10 6 /ml in 24-well plates (Costar, Cambridge, MA), washed twice in modified Eagle's medium, and incubated with IL-6, at the concentrations indicated. The acute phase proteins ␣-acid glycoprotein and haptoglobin were monitored in the conditioned medium by Western blotting using corresponding polyclonal antibodies (Sigma, Israel). The ECL Western blotting kit (Amersham International plc, United Kingdom) was used according to the recommended instructions. M1 myeloid leukemia cells were seeded at 2 ϫ 10 4 /ml in microtiter plates (100 l/well) and were induced to differentiate with the indicated concentration of IL-6. Mitochondrial activity was assayed by MTT as above.
Cytokines and Corresponding Neutralizing Antibodies-The antibodies to TGF-␤ that were used were rabbit anti-native porcine platelet TGF-␤1, neutralizing for both TGF-␤1 and TGF-␤2. These antibodies were purchased from British Biotechnology Ltd. (Abingdon, Oxon, UK). Human TGF-␤1 was obtained from the same source. Hamster antimurine interferon ␥ antibodies were obtained from Genzyme Corp. IL-3 was purchased from Peprotec (Rocky Hill, NJ). Recombinant human and monoclonal rat anti-mouse IL-6 neutralizing antibodies were purchased from Genzyme Corp. Recombinant, N-terminally truncated, human IL-6 (mutein) (31) and basic fibroblast growth factor were kindly provided by Pharmacia Biocenter (Nerviano, Italy). Crude concentrated murine IL-6 was kindly provided by Dr. J. Lotem, Weizmann Institute, and murine IL-6 was obtained from Dr. J. Van Snick, Ludwig Institute for Cancer Research, Brussels. Goat anti-M-CSF neutralizing antiserum was kindly provided by Dr. R. E. Stanley from the Albert Einstein University. Recombinant mouse IL-1␣, IL-2, IL-4, and IL-7 and recombinant human IL-10, IL-11, and platelet-derived growth factor were purchased from Genzyme Corp. Human G-CSF was kindly provided by Dr. S. Gillis of Immunex Corp., Seattle. Recombinant bovine activin A was purchased from Innogenetics (Belgium).
IL-6 Competition Binding on B9 Cells-IL-6 labeling and competition binding were performed as described previously (32). The amount of r-murine (Mu)-125 I-IL-6 or r-human (Hu)-125 I-IL-6 used was 30 -40% of the amount that gave saturation binding. Under these conditions, there were 2,070 Ϯ 530 high affinity binding sites and the Mu-IL-6 and Hu-IL-6 bound with apparent K d values of 5.5 ϫ 10 Ϫ10 and 1.1 ϫ 10 Ϫ10 M, respectively, as determined by the LIGAND program (33).
Protein Purification-Crude restrictin-P, 3.2 g of protein (2.5 ϫ 10 6 units in 1.4 liters) was prepared as previously reported (15). Aliquots (200 mg of protein) were further purified by ion exchange chromatography on Q-Sepharose, using a fast protein liquid chromotography (FPLC) system. After loading the sample and appropriate washing in buffer A (20 mM Tris-Cl, pH 7.8), restrictin-P was eluted with 0.05 M NaCl in buffer A. Elution of proteins was followed at 280 nm. The salt-eluted material was desalted and concentrated using reversed phase high performance liquid chromatography (RP-HPLC) on an Aquapore RP-300 column. Proteins were eluted with a nonlinear gradient of aqueous acetonitrile, 5-80%, in 0.1% trifluoroacetic acid. Elution of proteins was followed at 214 nm, protein content was determined according to Bradford (Bio-Rad, Munich), and restrictin-P content was assayed as described above. Fractions with restrictin-P activity (kept at Ϫ20°C) were further purified on Superdex 75 (in batches of about 10 mg of protein each) in 2 ϫ phosphate-buffered saline. Elution pattern was followed at 280 nm. Final purification of restrictin-P was achieved by RP-HPLC on RP-300 using a multistep linear gradient of aqueous acetonitrile in 0.1% trifluoroacetic acid. Elution of proteins was followed at 214 nm. Biologically active fractions were pooled and rechromatographed (RP-HPLC) under essentially identical conditions. Fractions containing restrictin-P (eluted at 38% acetonitrile) were vacuum-dried and kept frozen at Ϫ20°C. Polyacrylamide gel electrophoresis (PAGE) analyses of different protein fractions were performed using a Mini-Protean II gel apparatus (Bio-Rad Laboratories). Silver staining was performed with a Quick-Silver kit (Amersham) or a Silver Stain Plus Kit (Bio-Rad Laboratories). Coomassie Brilliant Blue staining was performed using the Serva blue G stain. Sequencing of the purified protein was performed at Perkin-Elmer.
Northern Analysis-The junB probe used, RSVjunB, was a 1.2-kilobase XhoI-SmaI fragment cloned into the pUC-18 vector (34). DNA probes labeled with a Random Primed DNA Labeling Kit (Boehringer Mannheim) were passed over a Sephadex G-50 minispin column. A minimum of 10 6 cpm/ml was used for hybridization. Signal intensities were measured by a 300A computing densitometer (Molecular Dynamics, Tampa, FL).
Gel Retardation Assay-HepG2 cells were treated with IL-6 and/or restrictin-P, and nuclei were isolated and extracted. Gel retardation analysis was carried out as described previously (35) except that 50 mM Tris, 41.5 mM boric acid, 0.5 mM EDTA, pH 8.3 was used as electrophoresis buffer. A DNA oligonucleotide containing mutant 67 of c-fos promoter sis-induced element (36) was labeled with 32 P by filling in with Klenow and used as a probe.

RESULTS
The inhibitory activity of restrictin-P, as detected in conditioned media form MBA-2.1 cells, was specific to plasma-like tumor cell lines (Table I). A variety of other cell lines representing different hemopoietic lineages and stages of maturation were only slightly inhibited or were totally unaffected by this factor. To rule out the possibility that restrictin-P activity could be ascribed to one of the cytokines known to affect hemopoietic cells, we searched for factors that might have restrictin-P-like activity. IL-1, IL-2, IL-3, IL-4, IL-6, IL-7, IL-10, IL-11, M-CSF, G-CSF, TGF-␤,platelet-derived growth factor, bovine fibroblast growth factor, interferon ␥, and leukemia inhibitory factor were tested over a range of concentrations. These cytokines were found to be devoid of the ability to inhibit the growth of the MPC-11 plasmacytoma which is highly sensitive to restrictin-P (12). In addition, neutralizing antibodies to TGF-␤1 and -2, tumor necrosis factor, IL-6, interferon ␥, and M-CSF did not reduce restrictin-P-like activity in media conditioned by MBA-2.1 cells (results not shown).
We used the above-conditioned media in an attempt to obtain some clue as to the mechanism by which the inhibition of plasma-like cells is mediated. The MBA-2.1 cell-conditioned medium inhibited the growth of the MPC-11 plasmacytoma and, as detailed below, it also interfered with the growthpromoting effect of IL-6 on B9 cells. To determine whether these two functions were mediated by the same molecule, it was necessary to purify restrictin-P to homogeneity. We therefore constructed a bioreactor production system wherein restrictin-P activity could be observed in media conditioned by the cells in absolute protein-free conditions (15). A batch of 600 liters of conditioned medium was concentrated by diafiltration and subjected to a further step of concentration by Amicon ultrafiltration followed by fractionation using an automated FPLC column. The fractionation included anion exchange chro-matography, gel filtration, and, finally, two steps of purification to homogeneity by reverse phase HPLC. Fig. 1 summarizes the above purification steps. As can be seen, the last stage yielded a single peak of protein that coincided with the biological activity of restrictin-P, i.e. the capacity to inhibit the growth of the plasmacytoma cell line, MPC-11. PAGE analysis of this product, under reducing conditions, revealed a single protein band at 15 kDa ( Fig. 2A). This purified peptide was N-terminally sequenced, and the first 36 residues were indistinguishable from those of the precursor of the ␤A subunit of inhibin which gives rise to a 15-kDa polypeptide (37-39) (Fig.  2B). The dimer of this subunit is known as activin A. Fig. 3A shows a comparison between restrictin-P and recombinant activin A of bovine origin. It is clear that these factors have an equal ability to suppress the growth of MPC-11 cells. Considered together, these results suggest that restrictin-P is identical with activin A.
To study the mechanism by which restrictin-P/stromal activin A inhibits plasma-like cell growth, we examined whether its effect was mediated by reversible cytostasis or through a mechanism that involves cell destruction. Restrictin-P in its unpurified form appeared to induce ionic changes that are associated with apoptosis (16). This mode of cell death is known to occur in hemopoietic cells deprived of their specific growth factor. The B9 hybridoma is dependent for growth on IL-6 and was therefore used to examine the possibility that restrictin-P causes growth cessation by interfering with the action of the growth factor. Extensive proliferation of B9 cells was induced by 0.1 IU/well of IL-6, and only moderate growth stimulation is observed upon addition of increasing concentrations of the purified cytokine. Addition of restrictin-P to B9 cell cultures stimulated by 0.01-0.1 IU/well of IL-6 caused almost complete growth inhibition. This inhibitory effect was gradually reduced with increasing the concentration of IL-6 and was almost abolished at 200 IU/ml of IL-6 ( Fig. 4A). In contrast to the ability of restrictin-P to antagonize the growth-stimulating effect of IL-6 it had no effect on the growth of 14M1.1, NFS, and MC/9 cells which are dependent for growth on M-CSF, GM-CSF, and IL-3, respectively (Fig. 3B).
IL-11 is an additional stimulator of plasma-like cells (40). As shown in Fig. 4B, restrictin-P inhibited the growth of IL-11stimulated B9 cells, and this inhibition was competed out by increasing the titer of IL-11. Thus, restrictin-P counteracted the growth-stimulating effect of both IL-6 and IL-11, and these cytokines at high titers overcame the effect of restrictin-P.
Plasmacytomas are but one target cell type that responds to IL-6 signaling. The HepG2 hepatoma release acute phase proteins under the influence of IL-6 (41). As shown in Fig. 5, the secretion of both ␣-acid glycoprotein and haptoglobin induced in HepG2 cells by IL-6 was markedly reduced by addition of restrictin-P (Fig. 5). M1 (clone 11) myeloblastic cells differentiate into adherent monocytes under the influence of IL-6 (28,29). Following IL-6 induction, M1 cells exhibit high mitochondrial activity and growth inhibition. As shown in Table II, restrictin-P abolished the IL-6-induced effect.
It was concluded, therefore, that IL-6 and restrictin-P are  competing on some target machinery used to generate a signaling pathway in at least 3 completely different target cell types. A candidate target molecule for restrictin-P action was the IL-6 receptor complex. We studied the possibility that restrictin-P is a receptor antagonist by testing its ability to compete with radiolabeled IL-6 for binding to its receptor on the surface of B9 cells. Fig. 6 shows that "cold" IL-6 competed out the binding of radiolabeled IL-6 to its receptor as expected. On the other hand, restrictin-P, at a concentration that would completely abolish the growth-stimulating effect of IL-6, failed to reduce the binding of radiolabeled IL-6 to its receptor. Thus, restrictin-P does not seem to interfere with ligand binding and may therefore interfere with postreceptor event(s) within the IL-6 signaling pathway. B9 cells stimulated by IL-6 following a period of cytokine deprivation showed a transient increase in expression of the early response gene junB (Fig. 7). Restrictin-P added to such cells caused increased and sustained expression of the junB gene until 24 h post-treatment at the time cell death already occurs. A similar augmented expression of junB mRNA was observed in the MPC-11 cell line. The expression of another early response gene TIS11 was increased in B9 cells incubated with restrictin-P but was unaffected in MPC-11 cells.
The effect of restrictin-P on junB expression was also observed in HepG2 hepatoma cells (not shown). In these cells, restrictin-P did not interfere with the JAK/STAT pathway (42) (Fig. 8). The data further suggest that restrictin-P moderately increased STAT activation (Fig. 8). This is surprising in view of the fact FIG. 2. A, SDS-PAGE analysis of highly purified restrictin-P. An aliquot from the active fraction (RP-HPLC-II, Fig. 1D) was loaded on 15% SDS-polyacrylamide gel under reducing conditions (lane 1). Protein bands were viewed by silver staining. Molecular mass markers, shown on the right-hand side, are: carbonic anhydrase, 31 kDa; soybean trypsin inhibitor, 21.5 kDa; lysozyme, 14.4 kDa; and aprotinin, 6.5 kDa. B, N terminus amino acid sequence of purified restrictin-P compared to the known sequence of activin A monomer. FIG. 3. A, inhibition of MPC-11 plasmacytoma cell growth by purified restrictin-P and by recombinant activin A. Cells (4 ϫ 10 2 /well) were seeded in 96-well Falcon microtiter plates, in 100 l of RPMI containing 10% FCS with serial dilutions of restrictin-P (E), activin A (f), or control growth medium (q). Cells were incubated for 4 days, and their viability was estimated by the MTT assay. The bar lines represent the mean of duplicate determinations Ϯ S.D. B, lack of effect of restrictin-P on MC/9 ( ), 14M1.1 (Ⅺ), and NFS ( ) cell lines dependent for growth on IL-3, M-CSF, and GM-CSF, respectively. Cells were seeded at 5 ϫ 10 3 , 10 3 , and 2.5 ϫ 10 3 /well, respectively, with the indicated amount of restrictin-P and specific growth factor, and their viability was monitored as in A.

FIG. 4. Restrictin-P antagonizes IL-6-and IL-11-induced proliferation of B9 cells.
Growth factor-deprived B9 cells were washed (3 times) with growth medium (RPMI ϩ 5% FCS) and were seeded in 96-microtiter well plates (5 ϫ HepG2 hepatoma cells were grown to confluence and stimulated with 100 units/ml IL-6 or an equal amount of IL-6 with 312 units/ml restrictin-P (RP). Conditioned media were collected at 24 h, subjected to PAGE (25 l/well, 10% and 7% gels for AGP and HP, respectively) and were tested by Western blotting for the acute phase proteins using the corresponding antibodies.  that restrictin-P abrogated the IL-6-induced secretion of acute phase proteins by HepG2 hepatoma (Fig. 5). It is implied therefore that a separate, possibly unknown pathway exists which allows cross-talk between the restrictin-P and IL-6 signaling cascades. DISCUSSION We investigated the nature of the activity, designated as restrictin-P (11), found in media conditioned by stromal cells (12), which causes growth arrest and subsequent cell death of mouse plasmacytomas and hybridomas. This activity was mediated by a protein that was purified to homogeneity from medium conditioned by the stromal cell line MBA-2.1 and was found to have an N-terminal amino acid sequence indistinguishable from that of activin A (37)(38)(39) which is also known as follicle-stimulating hormone releasing protein or erythroid differentiation factor (39,(43)(44)(45)(46). Activin A was found to be expressed by stromal cells (47). The molecular mass of monomeric restrictin-P, as deduced from PAGE, was 15 kDa, a size similar to that of monomeric activin A (␤A-inhibin). Like activin A, restrictin-P is a dimer of 25 kDa under nonreducing conditions and loses its biological activity upon reduction (not shown). Furthermore, recombinant activin A was inhibitory to the MPC-11 plasmacytoma to the same extent as was restrictin-P. It is therefore concluded that these two molecules are identical.
Restrictin-P in its purified form killed the factor-dependent hybridoma cell line B9 by competing with externally added IL-6 or IL-11. On the basis of the inability of a 270-fold excess of partially purified restrictin-P to compete with r-Mu-125 I-IL-6 binding (Fig. 6A) or a 340-fold excess of highly purified restrictin-P to compete with r-Hu-125 I-IL-6 (mutein) (Fig. 6B) binding, it is concluded that restrictin-P does not exert its effect by competing with IL-6 for high affinity IL-6 ligand binding sites.   8. Effect of restrictin-P on STAT activation. HepG2 cells were incubated with human recombinant IL-6 (10 units/ml) and restrictin-P as indicated. Either IL-6 and restrictin-P were added simultaneously or restrictin-P was added 30 min or 16 h prior to IL-6. 15 min after addition of IL-6 to the medium, the cells were harvested and nuclear extracts were prepared. 10 g of protein was then analyzed in a gel retardation assay using a 32 P-labeled oligonucleotide probe that contained a high affinity mutant of the c-fos promoter sis-induced element (SIE). The positions of DNA-protein complexes containing either Stat3 and Stat1␣ homodimers or Stat3/Stat1␣ heterodimers are indicated.
The antagonistic effect of restrictin-P is specific to IL-6 and IL-11 since restrictin-P did not affect the growth of other cytokine-dependent cell lines such as 14M1.4 macrophages that depend on M-CSF for growth, MC/9 mastocytoma which are IL-3-dependent or NFS-60, GM-CSF-dependent cells (Fig. 3B).
The strict specificity of killing by restrictin-P of plasmacytomas and hybridomas suggested that the factor detects some molecular machinery characteristic to this cell type. The growth dependence on IL-6 is common to many plasmacytomas and hybridomas. We show here that restrictin-P inhibits the growth of B9 cells by competing with the growth factors obligatory for the survival of the hybridoma. However, some cells, like the MPC-11 clone, are cytokine-independent, but are nonetheless growth-inhibited by restrictin-P. The question raised is whether the mechanism of action of restrictin-P in the case of IL-6-dependent B9 cells is different from that in MPC-11 cells. An alternative possibility is that the restrictin-P/activin A receptor (48) transduces a signal that interferes with IL-6 signaling downstream in the pathway. It has been shown that human myelomas possess intracellular IL-6 and IL-6 receptor, and their growth is triggered by an internal autocrine loop (49). Restrictin-P may interfere with such a hypothetical internal loop in the MPC-11 cell. One piece of evidence seems to support this notion, i.e. the increased junB expression induced in both B9 and in MPC-11 cells by restrictin-P. We further utilized the HepG2 cells in an attempt to identify a possible interference of restrictin-P in the JAK/STAT pathways involved in IL-6 signaling (42). Restrictin-P did not have such an effect, and we further noted an increase in STAT activation.
IL-6 is a pleiotropic cytokine that affects cells in different tissues and organs (50 -54). It is therefore expected that the activity of IL-6 would be tightly regulated. This may occur on the level of expression of the IL-6 gene as a result of activity of other cytokines or a variety of mediators (55)(56)(57). A more refined control involves interference with the biological activity of IL-6. This may occur due to inactivation of the protein. In U937 cells, membranal peptidyltransferases inactivated the IL-6 fragments by dimerizing them into 16-kDa complexes (58). A different regulatory mechanism involves a variety of agents that diminish the ability of cells to respond to IL-6 (59 -63). Natural receptor antagonists to IL-6 have not been isolated, to the best of our knowledge. However, it has been reported that oncostatin M, at high doses, is an IL-6 antagonist in a hepatoma cell model (64). These cells do not harbor oncostatin M receptors and the factor binds through the gp130 signal transduction transmembrane receptor which is part of the IL-6 receptor complex. A recent study of alveolar macrophages from smokers showed that these cells release upon lipopolysaccharide stimulation a mediator that antagonized the ability of IL-6 to support the growth of B9 cells (65). This activity was not biochemically isolated. Our results indicate that the pleiotropic cytokine activin A is identical with restrictin-P and that this molecule is an antagonist of IL-6 and IL-11. Since in our experiments restrictin-P acted in subnanogram amounts, it is likely that the function we describe has physiological significance.
Unrelated cell systems were studied, i.e. proliferating hybridomas, differentiating myeloblasts, and activated hepatoma cells. In these different cell systems, the inducer molecule IL-6 caused different biological outcomes, but, nonetheless, restrictin-P similarly antagonized the IL-6 functions implying that restrictin-P may be a universal antagonist of IL-6.
Cytokines that affect hemopoiesis are pleiotropic both from the point of view of target cells and in the biological consequence of their actions. Lineage-specific inducers are rare, and the explanation for the formation of tissue sites in which only one cell type accumulates requires an alternative explanation. It is conceivable that the interplay between inducer molecules and corresponding inhibitors would determine which cell type predominates in a particular tissue site. The theory of restrictins maintains that the growth of each hemopoietic cell type is negatively regulated by a lineage-specific inhibitor (66 -69). This may be a mechanism to prevent specifically the survival of cells in nondesirable sites. The identification of restrictin-P as a stromal activin A and the demonstration that it is an antagonist of IL-6 and IL-11, which are growth factors for plasmalike cells, provides support to this notion. The antagonist is specific for the growth factor and through this property it would not be restrictive to other cells that depend on alternative growth factors.