The Role of Soluble Growth Factors in Inducing Transient Growth

Home , Cell death

The role of soluble growth factors in inducing transient growth and clonal extinction of stroma cell dependent erythroblastic leukemia cells K Itoh1, J Friel1, C Laker1, W Zeller2, U Just1, S Bittner2, RJB Nibbs3, PR Harrison3, S-I Nishikawa4, KJ Mori5 and W Ostertag1

1Heinrich-Pette Institute for Experimental Virology and Immunology, Hamburg University; 2University Hospital Eppendorf, II Medical Clinics, Department of Oncology and Hematology, Hamburg, Germany; 3Cancer Research Campaign Beatson Laboratories, The Beatson Institute for Cancer Research, Glasgow, UK; 4Department of Clinical Molecular Biology, Faculty of Medicine, Kyoto University; and 5Department of Biology, Faculty of Science, Niigata University, Japan

A coculture system of a murine erythroblastic leukemia cell line mia which meets many of these requirements.10,11 The leu- (ELM-D) with its supportive stromal cell line (MS-5) was estab- kemic cells require contact with primary cultured stromal lay- lished. Long-term growth of ELM-D cells is strictly stroma cell dependent. Interaction between stem cell factor (SCF) and its ers for growth in vitro and differentiate in response to Epo. A receptor, c-kit, was demonstrated to be important for stroma stroma-dependent cell line, termed ELM-D, was isolated from 12 cell-dependent growth by anti c-kit neutralizing monoclonal this erythroblastic leukemia. Utilizing this novel in vitro cul- antibody (mAb) inhibition experiments. Significantly, soluble ture system, we report the effect of soluble growth factors or growth factors such as granulocyte–macrophage colony- stromal elements on growth, differentiation and the induction stimulating factor (GM-CSF), interleukin-3 (IL-3) or SCF of MS- of cell death of these erythroid precursor cells. 5 stromal cells (MS-5 CM) could replace the requirement of stroma cells for a considerable period. However, ELM-D cells maintained in these growth factors underwent clonal extinction after 3–6 weeks unless contact with stroma was re-established. Materials and methods Furthermore, IL-3 or GM-CSF acted in a dominant manner in inducing cell death in the presence of stroma cells. Cells show- Growth factors, CM ing clonal extinction undergo programmed cell death and do not differentiate. These altered growth properties of ELM-D cells exposed to soluble growth factors or to stroma cells rGM-CSF and rEpo were gifts from Boehringer Mannheim appear to be analogous to those described for T or B cells (Mannheim, Germany). Conditioned media of Wehi-3D and 13 primed by antigen presenting cells and then grown in growth RAT1 GM 4 cells were used as sources of IL-3 or GM-CSF factors. unless indicated. Routinely, GM-CSF (5 U/ml), rGM-CSF Keywords: apoptosis; clonal extinction; erythropoiesis; stem cell (10 U/ml) and IL-3 (5 U/ml) were used after titration on FDC- factor; stroma cells P1 cells and rEpo (5 U/ml) after titration by CFU-e formation. MS-5 CM, used as the source of SCF, was concentrated 10- fold by ultrafiltration (Amicon, Witten, Germany). The amount Introduction of MS-5 CM applied was five-fold the amount required for half maximal growth stimulation of MC-9 cells. Three-fold Differentiation and maintenance of many, if not all, hemopoi- lower and higher amounts were also tested to ensure optimal etic cell types in vivo requires direct cell contact with stroma growth conditions. cells.1,2 The nature of these stroma cell interactions is as yet neither functionally nor molecularly defined, despite the recent progress in understanding the role of adhesion mol- Cells and culture conditions ecules in such systems.3–5 Mice mutated at the white spotted (W) or the Steel (Sl) locus have been invaluable for the Cell lines: MS-5 stromal and ELM-D cell lines were main- advancement of our knowledge of cell-to-cell interaction. tained in ␣-MEM medium supplemented with 20% horse These mice are defective for either the c-kit receptor6 or its serum of selected batches. ELM-D cells were maintained on ligand, SCF,7 and are severely anemic. This indicates that SCF irradiated MS-5 cell layers (16000 rad) by passaging every 2– acts selectively on primitive and erythroid progenitors. SCF 3 weeks and then cloned by serial transfers. One clone, ELM- induces in synergy with erythropoietin (Epo) extensive pro- D No. 6, was routinely used in further experiments. liferation of erythroid progenitor cells.8 Extensive studies have been carried out in order to under- stand the mechanisms of growth and differentiation of Evaluation of cloning efficiency: Cells were plated into erythroid cells utilizing erythroleukemia cell lines established 96-well plates by limiting dilution and cloning efficiencies with the Friend virus.9 One major problem is that these cells were evaluated after 2–3 weeks according to the Poisson have lost their normal requirements for soluble growth factors distribution. or stromal interactions and their ability to differentiate in response to physiological growth factors such as Epo in vivo. Thus, an in vitro culture system retaining these features of nor- Determination of the frequency of stroma cell independent mal erythropoiesis has been keenly sought for a long time. growth: Cells were cloned in semi-solid medium (103–105 Recently, we reported a novel murine erythroblastic leuke- cells/ml) or seeded into 96-well plates (0.5–104 cells/well) without stroma. After 3–4 weeks of culture, the frequency of independent growth was evaluated. Most of the colonies, Correspondence: W Ostertag, Heinrich-Pette Institute for Experi- found after 2–3 weeks of culture, degenerated. mental Virology and Immunology, Hamburg University, Martinistrasse 52, 20251 Hamburg, Germany First and second authors contributed equally Received 18 October 1996; accepted 18 May 1997 Semi-solid culture of two layers: The two layer method of Ligand-induced dominant extinction of myeloid cells K Itoh et al 1754 agar (0.3%) was performed as described by Sibley and evaluated by limiting dilution. In contrast to the high cloning Tomkins.14 Cloning efficiencies were evaluated after 2 weeks efficiency of ELM-D cells in coculture with irradiated stroma by counting colonies. (30–50%), a significantly reduced cloning efficiency (2–3%) was observed when ELM-D cells were cultured in growth media without stroma or growth factors (Table 1). Cloning Assay for long-term growth: Cells were cloned by limiting efficiencies in the presence of MS-5 CM or growth factors such dilution and cloned cells were transferred at a 1/10 dilution as GM-CSF or IL-3 were comparable to those obtained with into new wells every week. The percentage of positive wells stroma (Table 1). with growing cells was corrected by initial cloning efficiencies To analyze the growth dependency in more detail, the and presented as ‘% clones surviving’. To analyze the long- mutation frequency of ELM-D cells to stroma-independent term growth in more detail, changes of the cloning efficiencies growth was determined. In both semi-solid cloning and limit- were also examined. After cloning by limiting dilution, cloned ing dilution experiments in liquid culture, the survival rate of cells were recloned every 2–3 weeks and changes in the ELM-D cells was extremely low after 3–4 weeks of culture, cloning efficiencies were recorded. estimated to be Ͻ3 × 10−5 in liquid and Ͻ3.5 × 10−6 in semi- solid culture.

Growth inhibition assay using ACK-2 mAb: Cells were plated into Terasaki plates at a concentration of 50 cells/well with 1:3 serial dilution of ACK-2 mAb15 (maximum concen- Conditioned media (CM) cannot support the long-term tration: 30 ␮g/ml). The number of cells/well were recorded growth of ELM-D cells after 5 days of culture.

Since the initial cloning efficiency in the presence of MS-5 Morphological examination and detection of differentiated CM was comparable to that on stroma (Table 1), we examined cells: Cells were concentrated on slides and stained by whether the coculture with stroma is required for the long- May–Gru¨ nwald Giemsa and benzidine. Differentiated cells term growth of ELM-D or, alternatively, if CM is sufficient. All were determined by morphological examination. of 115 clones transferred in the presence of CM lost their self- renewal capacity and showed clonal extinction within 6 weeks, whereas all the clones transferred on irradiated MS-5 Reverse transcriptase (RT)-PCR analysis survived during the same time period (Figure 1). This demonstrates that although ELM-D cells can clone initially at high Detection of mRNA for SCF: Expression of mRNA for SCF efficiency in CM, CM is not sufficient to support long-term was examined by RT-PCR as described previously.16 Specific growth of ELM-D cells, suggesting the importance of direct DNA fragments were amplified by PCR for 25 cycles, separ- contact with stroma. ated, transferred, hybridized and exposed to X-ray films. Primer sets used in this study were SCF1: position 126–145 and 1041–1060 and SCF 2: position 845–863 and 1041– 17 1060. Table 1 ELM-D cells can be cloned in the presence of MS-5 CM, GM-CSF or IL-3 or MS-5 CM for short time only ␣ Semi-quantified assay of -globin mRNA: Semi-quantitat- Condition Cloning efficiency (%) ive RT-PCR for ␣-globin was additionally performed. Various Mean (No. of experiments) cycle numbers had been previously tested to ensure a linear range of amplification and cDNAs were amplified for 15 ELM-Da ELM-D No. 6b cycles. Oligos used for ␣-globin expression were position 460–480 and position 930–950. After densitometric scanning On MS-5 layer of the autoradiographs, ratios of ␣-globin to ␤-actin18 were unirradiated 47 ± 20c (4) 54 ± 21 (2) calculated. irradiatedd 32 ± 12 (4) 36 ± 7 (20) MS-5 CMe 50 (1) 44 ± 11 (8) GM-CSF CMf NT 28 ± 12 (2) Tdt fluorescence assay: Tdt fluorescence assay was perfor- hg± 19 recombinant NT 34 9 (7) med according to the method described by Gorczyca et al. IL-3 Cells were fixed in 1% paraformaldehyde, dehydrated, rehy- CMi NT 26 ± 7 (4) drated and reacted with 1 nM biotin-16-dUTP in the presence Medium only 0.0032 ± 0.002 (6) 2.6 ± 1 (22) of 0.1 U/ml Tdt at 37°C for 30 min, followed by avidin-conju- gated phycoerythrin. Washed cells were analyzed on a FAC- aCloning efficiency after 2 weeks (only clones with viable cells Scan cytometer (Becton Dickinson, Heidelberg, Germany) are recorded). after gating out contaminated MS-5. bCloning efficiency after 3 weeks. cStandard deviation. d16 000 rad. e Results MS-5 CM was concentrated 10-fold and added at a concentration of 10%. fRAT1 GM 4 cell supernatant used at a final concentration of ELM-D cells are stroma dependent 10 U/ml. gNot tested. To define the growth dependency of the ELM-D cells11,12 on hAt a final concentration of 10 U/ml. MS5 stroma20 or in growth factors, cloning efficiencies were iWehi 3D cell CM used at a final concentration of 10 U/ml. Ligand-induced dominant extinction of myeloid cells K Itoh et al 1755

Figure 1 Loss of self-renewal capacity of ELM-D cells cultured with CMs. After removal of stroma, ELM-D cells were cloned in CM and clones were transferred in the presence of CM every week. The percentage of positive wells was corrected by initial cloning efﬁciencies and presented as ‘% clones surviving’. On irradiated MS- 5(̅), 50 clones; in growth medium without stroma (ᮀ), 55 clones; in 50% MS-5 CM (᭺), 41 clones; in 100% MS-5 CM (᭿), 38 clones; in 10% MS-5+ELM-D CM (᭡), 36 clones.

Direct stroma-cell contact is important for the growth of ELM-D

Semi-solid culture of ELM-D cells using two layers of agar was performed in order to determine whether direct cell-to-cell contact is necessary for its proliferation or not. As shown in Figure 2, a cloning efﬁciency of 28% was obtained when ELM-D cells were plated directly on MS-5 cell layers to permit direct contact with stroma. In contrast, their cloning efﬁciency was drastically decreased to about 3% when ELM-D cells were separated from stroma by semi-solid agar, demonstrating the importance of direct contact with stroma for the growth of ELM-D cells, even for short-term growth.

Growth of ELM-D cells can be inhibited by blocking the c-kit receptor

Since SCF has been reported to be one of the crucial factors in cell-to-cell interaction,21,22 a series of experiments were performed to determine if the SCF and c-kit interaction plays some role in coculture of ELM-D with MS-5 cells. The first experiment was set up to determine whether the Figure 2 Growth of ELM-D cells is dependent on direct cell-to- proliferative response of ELM-D cells to MS-5 CM observed cell contact with stroma. ELM-D cells were cloned in soft agar separ- in the cloning efficiency experiment (Table 1) could be ated from stroma in various distances as indicated. Cloning blocked by the addition of an anti-c-kit neutralizing mono- efficiencies were evaluated after 2 weeks of culture. clonal antibody (mAb), ACK-2.15 Data presented in Figure 3b confirm (1) that ELM-D cells proliferate initially in response to the soluble form of SCF derived from MS-5 and also show (2) (Figure 3), we checked the expression of SCF. As expected, mRNA for m1 type of SCF, that expresses both the membrane- that this proliferative response was completely inhibited by 23 the addition of ACK-2 mAb. In a second experiment, the effect bound and soluble form of SCF, was detected in MS-5 cells of ACK-2mAb was tested on the growth of ELM-D cells cocul- (Figure 4). Together with the anti-kit antibody experiments tured with MS-5. Addition of ACK-2, as expected, drastically described above (Figure 3), these data suggest that growth inhibits the growth of ELM-D cocultured with MS-5 in a con- stimulated by MS-5 CM corresponds to that of soluble SCF centration-dependent manner. These results demonstrate that and of direct contact with MS-5 cells to membrane-bound SCF and its receptor c-kit are involved in the growth of ELM- SCF. This implies that membrane-bound SCF is important for D cells cocultured with MS-5. the long-term growth of ELM-D.

MS-5 cells express the m1 type mRNA encoding both GM-CSF or IL-3 induce only transient growth of the membrane-bound and soluble form of SCF ELM-D

Since SCF and c-kit interaction was shown to be crucial in A series of experiments was performed to determine if the the interaction of ELM-D cells with MS-5 stromal cells long-term supporting activity provided by MS-5 coculture can Ligand-induced dominant extinction of myeloid cells K Itoh et al 1756

Figure 4 Only the m1 type of SCF is detected in MS-5 stroma cells. (a) Schematic representation of SCF cDNAs and the positions recog- nized by primers. The boxed, hatched and solid bars represent the Figure 3 Growth of ELM-D cells can be inhibited by anti-c-kit neu- coding region, signal sequence and the transmembrane region. Trans- tralizing mAb. Cells were cultured in the presence of serial dilution membrane coding region of SCF is deleted in the Sld mutant (– – –).23 of ACK-2 mAb (᭺,᭹) (30 ␮g/ml) or without (᭺,᭝). Data are mean (b) Analysis of SCF cDNAs by RT-PCR. SCF1 primers amplify the pro- values of three to five independent experiments. (a) Antibody controls: ducts of 925 bp from m1, 841 bp from m2 and 750 bp from Sld type SCF unresponsive FDC-P1 cells cocultured with unirradiated MS-5 in of SCF. SCF2 primers amplify the product of 627 bp from both the the presence of ACK-2 c-kit antibody (᭺). ELM-D cells cultured in m1 and m2 type SCF, but not from Sld. MS5 CM in the presence of an isotype-matched irrelevant antibody (YTH) (̅) (30 ␮g/ml). (b) ELM-D cells cultured with MS-5 CM without ᭺ ᭹ stroma ( ), or cultured on unirradiated MS-5 ( ). Taken together, when GM-CSF or IL-3 are added to ELM- D cell cultures transient growth of the cells is stimulated for be replaced by soluble growth factors such as recombinant several weeks but cell death is induced by growth factors in GM-CSF (rGM-CSF) or IL-3, since cloning efficiencies in the a dominant manner in the presence of stromal cells. presence of these factors without stroma were very high (Table 1). ELM-D cells were cloned in the presence of these factors without stroma, and 116 clones (rGM-CSF: 53 clones; GM-CSF induces the programming of ‘commitment’ to IL-3: 63 clones) were transferred weekly in the presence of cell death these factors. As is shown in Figure 5a, rGM-CSF or IL-3 was able to delay the onset of clonal extinction up to 4 weeks, but To determine when commitment to cell death occurs, a more could not replace stromal contact for long-term self-renewal detailed analysis of the pattern of clonal extinction was perfor- of the clones. Cell numbers were also measured at each trans- med using rGM-CSF. Initially, three subclones were isolated fer and the average increase in cell numbers for six clones is from ELM-D cells in the presence of rGM-CSF, ie subclone shown in Figure 5b. These data suggest that in the presence Nos 3, 2 and 6 (Figure 7a). These subclones were serially of GM-CSF or IL-3 ELM-D cells can grow logarithmically for recloned every 2–3 weeks until cell death occurred in the up to 2–3 weeks, even without stroma, but stopped proliferat- presence of rGM-CSF. Figure 7a shows (1) that all the clones ing after 3–4 weeks, at which time they undergo a severe derived from the same initial subclone underwent cell death at decline in cell numbers. In contrast, cells passaged on MS-5 the same stage, ie all six clones derived from subclone No. 3 stroma continued to grow logarithmically for more than 8 underwent cell death during the fifth subcloning, two clones weeks. Thus, GM-CSF or IL-3 induce the growth of ELM-D from subclone No. 2 during the second subcloning, and all cells only transiently and cannot support permanent growth. seven clones from subclone No. 6 during the fourth subcloning. The data also demonstrate (2) that clonal extinction occurred at different times depending on the origin of the GM-CSF or IL-3 act in a dominant manner to induce cell death clone, ie progeny of subclone No. 3 survived until the fifth of ELM-D cells To examine whether clonal extinction of subcloning and that of subclone No. 6 until the fourth sub- ELM-D cells also occurred in coculture when growth factors cloning, while the progeny of subclone No. 2 survived only are present, long-term growth of ELM-D cells was examined until the third subcloning. These data imply (1) that clonal in the presence of both MS-5 stroma and rGM-CSF or IL-3. extinction is not simply due to a stochastic loss of growth Interestingly, all 47 clones with GM-CSF and 19 clones with capacity in culture but due to an endogenous programming IL-3 underwent cell death after 3–6 weeks even when trans- of single cells and (2) that ‘commitment’ to cell death is pro- ferred on MS-5 stroma (Figure 6). grammed at the single cell level for each progeny of these Ligand-induced dominant extinction of myeloid cells K Itoh et al 1757

Figure 6 Loss of self-renewal capacity of ELM-D cells cocultured with irradiated MS-5 and soluble growth factors. In the presence of stroma, ELM-D cells were cloned in GM-CSF. Clones were transferred in the presence of both stroma and rGM-CSF every week. The percentage of wells containing viable cells was calculated as in Figure 5. On irradiated MS-5 without growth factors (̅), 50 clones; in rGM-CSF with irradiated MS-5 (ᮀ), 47 clones; in IL-3 with irradiated MS-5 (᭡), 19 clones.

ferred back on to MS-5 in the second subcloning procedure, contact with stroma can prevent the clonal extinction of these clones, whereas the clones transferred back on to MS-5 during the third or fourth subcloning procedure either undergo cell death or survive after a severe reduction of cloning efﬁciency.

Cell death induced by GM-CSF is correlated with double-stranded DNA breaks

Figure 5 ELM-D cells, in absence of stroma, proliferate extensively To delineate the type of cell death induced by GM-CSF, DNA in GM-CSF or IL-3 but undergo clonal extinction as a late event. (a) Loss of self-renewal capacity of ELM-D cells cultured in soluble breaks were detected by the terminal transferase (Tdt) fluor- 19 growth factors. After removal of stroma, ELM-D cells were cloned in escence assay at different time-points in different conditions. rGM-CSF or IL-3. Clones were transferred every week either in the Figure 8c shows that ELM-D cells cultured for 18 days without presence of rGM-CSF or IL-3. The percentage of wells containing stroma in the presence of rGM-CSF were highly Tdt fluor- growing cells was corrected for the initial cloning efficiencies and ̅ escence positive. In contrast, ELM-D cells cocultured on presented as ‘% clones surviving’. On irradiated MS-5 ( ), 50 clones; irradiated MS-5 were negative for Tdt fluorescence even after in growth medium without stroma and without growth factors (ᮀ), 55 clones; in rGM-CSF (᭿), 53 clones; in IL-3 (̆), 63 clones. (b) The 21 days of culture (Figure 8a). ELM-D cocultured on stroma in absolute increase of cell numbers was calculated by counting at the the presence of rGM-CSF also showed highly Tdt fluorescence indicated time-points. These numbers were related to the cell numbers positive at day 18 (Figure 8d). These data clearly show that seeded on day 0. The cell numbers of six clones of each condition cell death induced by GM-CSF is correlated with DNA breaks, described in (a) were plotted. ie apoptosis. Table 2 shows detailed data from the Tdt fluorescence assay performed at different time-points using different clones iso- subclones just after the initial subcloning. Such ‘committed’ lated on irradiated stroma. This Table demonstrates that an progeny of the subclones can survive up to 10 weeks before increase in the percentage of Tdt fluorescence-positive cells clonal extinction. This is consistent with the data shown in correlated with clonal extinction of subclones. Figure 5 which demonstrated that all of 116 subclones transferred in growth factors without stroma underwent cell death after about 8 weeks. Clonal extinction of ELM-D cells is not correlated with differentiation

The programming of ‘commitment’ to cell death can To analyze if induction of apoptosis was coupled to differen- only be partly prevented by contact with stroma tiation, we determined the morphology of the cells and the amount of ␣-globin transcripts. ELM-D cells grown in GM- We then examined whether the programming of ‘commit- CSF either with or without stroma cells showed an immature ment’ to cell death induced by rGM-CSF could be prevented erythroblast morphology similar to ELM-D control clones by renewed contact with stroma. In these studies, clones cocultured on stroma cells (Figure 9a, c and 9d). In contrast, derived from the ﬁrst three subclones were transferred back erythropoietin (Epo) induced erythroid differentiation of ELM- on to MS-5 cells without rGM-CSF and successive recloning D cells, as shown by the appearance of benzidine-positive was done with stroma in the absence of rGM-CSF (dashed cells (Figure 9b). lines, Figure 7b). The data indicate that if clones are trans- To conﬁrm that apoptosis induced by GM-CSF is not Ligand-induced dominant extinction of myeloid cells K Itoh et al 1758

Figure 8 Clonal extinction of ELM-D cells as shown by Tdt analysis. Clonal extinction occurs either for ELM-D cells cultured without stroma or with GM-CSF alone, or in the presence of stroma and GM- CSF. (a) ELM-D cells cocultured on irradiated MS-5 for 21 days. (b) ELM-D cells without stroma or GM-CSF for 18 days. (c) ELM-D cells in GM-CSF without stroma for 18 days. (d) ELM-D cells in GM-CSF on irradiated MS-5 for 18 days.

Discussion

Using the coculture system of ELM-D and MS-5 cells, we observed (1) that direct contact with stroma is essential for the long-term growth of ELM-D cells, (2) that soluble growth factors such as GM-CSF or IL-3 can induce only transient cell growth of ELM-D and (3) that these soluble factors dominantly induce apoptosis, even in the presence of stroma cells. Various lines of evidence suggest that direct cell contact is obligatory for the long-term growth of ELM-D by stroma cells. Although MS-5 CM can support short-term proliferation of ELM-D cells, it is not sufficient to prevent clonal extinction of these cells (Figure 1). Furthermore, permanent growth is Figure 7 Clonal extinction and recovery of ELM-D cells main- severely reduced if ELM-D cells are separated from stroma by tained in GM-CSF by repeated clonal transfer of clones. (a) Clonal an agar layer (Figure 2). These results demonstrate that cell extinction of ELM-D cells maintained in GM-CSF during repeated sub- contact is necessary for long-term proliferation. Even if our cloning. The Figure shows the methods used for repeated cloning in data have shown that soluble SCF is responsible for short-term GM-CSF without stroma, and indicates when cells undergo clonal growth of ELM-D in MS-5 CM, we demonstrate that the trans- extinction (+). (b) Clonal extinction of ELM-D cells, maintained in GM- CSF, and their recovery after transfer back to stroma. The numbers in membrane form of SCF is necessary for long-term growth on brackets are the cloning efficiencies of individual subclones at various MS-5 stroma and, in contrast to soluble SCF, can block clonal cloning steps, either when cloned in GM-CSF without stroma (—), extinction. Thus, cell–cell contact is necessary for interaction or after retransferred onto MS-5 without GM-CSF (– – –). (i) indicates between membrane-bound SCF expressed by MS-5 cells and surviving clones shown to be stroma independent by further clonal the c-kit receptor on ELM-D cells. The importance of trans- + analysis. ( ), clonal extinction. Data displayed in a and b represent membrane SCF in supporting erythropoiesis is also shown by the same set of data. Sld/Sld mice that produce only the soluble form of SCF and are severely anemic, similarly as Sl deficient mice.7 Not unex- pectedly, stroma isolated from these mice are incapable of coupled with differentiation, semi-quantitative RT-PCR was supporting long-term growth of ELM-D cells.24 However, we performed at different time-points in different conditions cannot currently exclude that the other cell–cell interactions together with the detection of Tdt fluorescence assay. In con- may also be involved in preventing clonal extinction. trast to the increase of Tdt fluorescence-positive population, Other growth factors such as GM-CSF or IL-3, are also inef- no significant change in ␣-globin synthesis was seen in ELM- ficient in replacing the stroma interaction requirement for D clones cultured in the presence of GM-CSF as compared to long-term growth of ELM-D cells. Proliferation induced by ELM-D grown on stroma cells (Table 3). Therefore, it is clear these factors was supported for a considerable, but only finite by morphological study and semi-quantitative assay of ␣-glo- period. This was confirmed by long-term clonal transfer bin that differentiation is not induced by GM-CSF in ELM-D experiments (Figure 5) and by serial passage of single ELM-D cells under conditions that lead to apoptosis. clones in the presence of GM-CSF that led to clonal extinction Ligand-induced dominant extinction of myeloid cells K Itoh et al 1759 Table 2 Induction of cell death is correlated with DNA breaks

Condition ELM-D clonesa % Tdt ﬂuorescence-positive cells (days of culture) (clone No.) 14 18 21 25 28 32 36

On irradiated MS-5 3 3 5 2 4 3 4 6246342 Medium 2 60 70 +b 64071+ rGM-CSF 4 12 20 17 14 4 2 53960+ 9163853+ 15 22 53 + On irradiated MS-5 + rGM-CSF 1 28 47 + 23053+ 12 49 53 + aDifferent clones isolated on irradiated MS-5 cells were analyzed for evidence of apoptosis. bClonal extinction.

Figure 9 ELM-D cells differentiate in the presence of erythropoietin but not of GM-CSF. Cells were stained with May–Gru¨nwald Giemsa and benzidin. Original magnification ×600. (a) ELM-D cells grown in contact with stroma cells show an immature erythroblast morphology. (b) ELM-D cells cultured with rEpo and stroma cells for 25 days. Arrows show benzidine-positive mature erythroid cells. (c) and (d) No differentiation occurs in ELM-D cells grown in the presence of GM-CSF without (c) or with (d) stroma cells after 25 days of culture. Arrow depicts ELM-D cell. Ligand-induced dominant extinction of myeloid cells K Itoh et al 1760 Table 3 Induction of cell death by GM-CSF correlated with apoptosis.31,32 Factors influencing this balance in B or T cells apoptosis but not with differentiation include, among others, the differentiation state of the target cell and the presence of various cytokines.33,34 In analogy to a Condition ␣-globin /% of Tdt-positive cells results with T lymphocytes, our data suggest that higher con- (days of culture) centrations of growth factors act much faster in inducing programmed cell death of ELM-D cells. Different concentrations 14 18 21 27 of IL-3 in combination with additional factors have also been shown to influence selectively the balance between self- On irradiated MS-5 0.69/1.2 0.61/2.5 0.75/3.3 0.54/2.8 Medium 1.2/9.2 1.1/54.5 0.92/72.0 +b renewal and clonal extinction of the early myeloid progenitor 35 rGM-CSF alone 1.45/11.5 1.33/25.5 1.2/52.5 0.91/61.1 cell line, FDC-Pmix. F4-6 cellsc 5.6/− Apoptosis plays an important role in hemopoiesis by reg- ulating the rate of cell production in self-renewing progenitor 36,37 a␣-globin expression as a parameter for erythroid differentiation cells. Generally, hematopoietic growth factors are thought was calculated after being standardized by ␤-actin (␣-globin/␤- to support self-renewal of their target cells by reducing actin ratio). apoptosis and allowing their differentiation into all lineages.38 bClonal extinction. Additional pathways other than growth factor deprivation may cFriend virus-infected cells were induced with 1.2% DMSO to also regulate apoptosis during hemopoiesis. Our results dem- erythroid differentiation for positive control. onstrate that soluble growth factors can have two opposite activities in leukemic ELM-D cells, ie providing signals for growth and apoptosis. Such a mechanism might possibly be (Figure 7). Whether combinations of growth factors can substi- part of the response pattern of normal hemopoietic cells. The tute for stroma cells in maintaining permanent growth of ELM- well known observation that growth factors like GM-CSF or D is at present uncertain. The various combinations tested so G-CSF can mobilize progenitor cells could thus be attributed far, ie Epo and IL-3, Epo and GM-CSF, Epo and MS-5 CM that to the ligand-induced recruitment of progenitors from stroma contains SCF, or GM-CSF and MS-5 CM could not support that are committed to both differentiation and apoptosis. It is continued survival of ELM-D cells (unpublished data). interesting that such a flexible mechanism observed here has Importantly, we also demonstrated by serial cloning experi- also been reported in the neural system.39 ments that cell death induced by soluble growth factors is An alternative hypothesis for apoptosis induction in ELM-D dominant over the cell–cell interaction, since addition of GM- cells is that it may be related to changes in the cell cycle CSF or IL-3 induces apoptosis in the presence of MS-5 stroma during induction of differentiation. It has been shown that dif- (Figure 6). Any direct interaction between GM-CSF and ferent aspects of the process of erythroid differentiation can stroma, that could be responsible for this, can be excluded be uncoupled, one part leading to functional differentiation, because MS-5 cells do not express the GM-CSF receptor the other to changes in cell cycle and cell death.40 Uncoup- (unpublished data). Detailed analysis of serial cloning experi- ling of differentiation may thus have occurred in ELM-D. GM- ments show that the final process of clonal extinction by sol- CSF or IL-3 may then activate only that part of the differen- uble growth factors is not a stochastic event but reflects an tiation pathway that is responsible for cessation of cycling and endogenous programming or ‘precommitment’ to clonal apoptosis, but is deficient in leukemic cells and does not deletion. This process is not an ‘all or nothing decision’ but induce functional differentiation. However, reduced cell size, requires at least one cell duplication to be fixed. In this phase typical for normal erythropoiesis, is not found in ELM-D cells induction of apoptosis can be reversed by renewed stroma cell induced to apoptosis. contact (Figure 7). Thereafter apoptosis is irreversibly induced. The presence of DNA breaks demonstrates that cell death is due to apoptosis (Figure 8) rather than differentiation as Acknowledgements judged also by morphological examination (Figure 9) and lack ␣ of upregulation of -globin message (Table 3). The This work was supported by grants of Deutsche Forschungsge- mechanism(s) by which the induction of apoptosis in ELM-D meinschaft to WO (Os 31/14). The Heinrich-Pette Institute is is prevented by re-contact to stromal cell is unclear. Several financially supported by Freie und Hansestadt Hamburg and reports demonstrate the importance of stroma in suppressing Bundesministerium fuer Gesundheit, PRH and RB by grants apoptosis.25,26 Activation of bcl-2 which is known to attenuate 27,28 from the Cancer Research Campaign and the Medical apoptosis might be one possibility. In fact, we observed Research Council. KI was a fellow of the Alexander von that bcl-2 message is upregulated by the contact of MS-5 Humboldt Foundation. stroma in ELM-D (unpublished observation) and in another stroma-dependent cell line Myl-D7.29 The mechanisms by which factors such as GM-CSF or IL-3 References can induce apoptosis in stroma-dependent ELM-D cells are unknown. Recently, involvement of adhesion molecules in 1 Dexter TM. Regulation of hemopoietic cell growth and develop- induction of apoptosis in hemopoietic cells has been sug- ment: experimental and clinical studies. Leukemia 1989; 3: gested since stimulation of CD43 can induce apoptosis in 469–474. erythroid progenitors by abrogating their factor-dependent 2 Dorshkind K. Regulation of hemopoiesis by bone marrow stromal proliferation. After cross-linking of CD43 adhesion molecules, cells and their products. Ann Rev Immunol 1990; 8: 111–137. apoptosis was initiated in the presence of cytokines.30 3 Kishimoto TK, Larson RS, Corbi AL, Dustin ML, Staunton DE, Whether GM-CSF in our coculture system activate CD43- Springer TA. The leukocyte integrins. Adv Immunol 1989; 46: 149–182. mediated apoptosis in ELM-D cells remains unclear. 4 Butcher EC. Leukocyte-endothelial cell recognition: three (or Crosslinking of T or B cell antigen receptors also induces more) steps to specificity and diversity. Cell 1991; 67: 1033–1036. either proliferation and differentiation or, alternatively, 5 Pardi R, Inverardi L, Bender JR. Regulatory mechanisms in leuko- Ligand-induced dominant extinction of myeloid cells K Itoh et al 1761 cyte adhesion: flexible receptors for sophisticated travelers. Immu- 22 Zsebo KM, Wypych J, McNiece IK, Lu HS, Smith KA, Karkare SB, nol Today 1992; 13: 224–230. Sachdev RK, Yuschenkoff, VN, Birkett NC, Williams LR, Satyagal 6 Chabot B, Stephenson DA, Chapman VM, Besmer P, Bernstein VN, Tung W, Bosselman RA, Mendiaz EA, Langley KE. Identifi- A. The proto-oncogene c-kit encoding a transmembrane tyrosine cation, purification, and biological characterization of hematopoi- kinase receptor maps to the mouse W locus. Nature 1988; 335: etic stem cell factor from buffalo rat-conditioned medium. Cell 88–89. 1990; 63: 195–201. 7 Zsebo KM, Williams DA, Geissler EN, Broudy VC, Martin FH, 23 Flanagan JG, Chan DC, Leder P. Transmembrane form of the kit Atkins HL, Hsu R-Y, Birkett NC, Okino KH, Murdock DC, ligand growth factor is determined by alternative splicing and is Jacobsen FW, Langley KE, Smith KA, Takeishi T, Cattanach BM, missing in the Sld mutant. Cell 1991; 64: 1025–1035. Galli SJ, Suggs SV. Stem cell factor is encoded at the Sl locus of 24 Friel J, Itoh K, Heberlein C, Stocking C, Bergholz U, Harrison P, the mouse and is the ligand for the c-kit tyrosine kinase receptor. Ostertag W. Antagonism of soluble and membrane inserted stem Cell 1990; 63: 213–224. cell factor (SCF) in long-term maintenance of stroma dependent 8 McNiece IK, Langley KE, Zsebo KM. Recombinant human stem hematopoietic cells. Curr Biol 1997 (submitted). cell factor synergises with GM-CSF, G-CSF, IL-3 and Epo to stimu- 25 Manabe A, Coustan-Smith E, Behm FG, Raimondi SC, Campana late human progenitor cells of myeloid and erythroid lineages. Exp D. Bone marrow-derived stromal cells prevent apoptotic cell Hematol 1991; 19: 226–231. death in B-lineage acute lymphoblastic leukemia. Blood 1992; 79: 9 Ben-David Y, Bernstein A. Friend virus-induced erythroleukemia 2370–2377. and the multistage nature of cancer. Cell 1991; 66: 831–834. 26 Spooncer E, Heyworth CM, Dunn A, Dexter TM. Self-renewal and 10 Itoh K, Sasaki R, Ono K, Tezuka H, Sakoda H, Sawada H, Hitomi differentiation of interleukin-3-dependent multipotent stem cells K, Nakane H, Uchiyama T, Uchino H, Mori KJ. Stromal cell are modulated by stromal cells and serum factors. Differentiation dependent growth of leukemic cells from murine erythroblastic 1986; 31: 111–118. leukemia. Jpn J Cancer Res (Gann) 1988; 79: 931–937. 27 Hockenberry D, Nunez G, Milliman C, Schreiber RD, Korsmeyer 11 Nibbs RJB, Itoh K, Ostertag W, Harrison PR. Differentiation arrest SJ. Bcl-2 is an inner mitochondrial membrane protein that blocks and stromal cell-independent growth of murine erythroleukemia programmed cell death. Nature 1990; 348: 334–336. cells are associated with elevated expression of ets-related genes, 28 Fairbairn LJ, Cowling GJ, Reipert BM, Dexter TM. Suppression of but not mutation of p53. Mol Cell Biol 1993; 13: 5582–5592. apoptosis allows differentiation and development of a multipotent 12 Itoh K, Ono K, Sawada H, Tezuka H, Sakoda H, Nakane H, Uchi- hemopoietic cell line in the absence of added growth factors. Cell yama T, Uchino H, Mori KJ. Establishment and characterization 1993: 74: 823–832. of a transplantable erythroblastic leukemia in C3H mice. 29 Itoh K, Friel J, Kluge N, Kina T, Kondo-Takaori A, Kawamata S, Leukemia Res 1988; 12: 471–478. Uchiyama T, Ostertag W. A novel hematopoietic multilineage 13 Laker C, Stocking C, Bergholz U, Hess N, DeLamarter JF, Ostertag clone, Myl-D-7, is stromal dependent and supported by an W. Autocrine stimulation after transfer of the granulocyte macro- alternative mechanism(s) independent of stem cell factor/c-kit phage colony-stimulating factor gene and autonomous growth are interaction. Blood 1996; 87: 3218–3228. distinct but interdependent steps in the oncogenic pathway. Proc 30 Bazil V, Brandt J, Tsukamoto A, Hoffman R. Apoptosis of human Natl Acad Sci USA 1987; 84: 8458–8462. hematopoietic progenitor cells induced by cross-linking of surface 14 Sibley CH, Tomkins GM. Isolation of lymphocyte cell variants. CD43, the major sialoglycoprotein of leucocytes. Blood 1995; 86: Cell 1974; 2: 218–222. 502–511. 15 Nishikawa S, Kusakabe M, Yoshinaga K, Ogawa M, Hayashi SI, 31 Benhamou LE, Cazenave P-A, Sarthou P. Anti-immunoglobulins Kunisada T, Era T, Sakakura T, Nishikawa SI. In utero manipu- induce death by apoptosis in WEHI-231 B lymphoma cells. Eur J lation of coat color formation by a monoclonal anti-c-kit antibody: Immunol 1990; 20: 1403–1407. two distinct waves of c-kit-dependency during melanocyte devel- 32 Smith CA, Williams GT, Kingston R, Jenkinson ER, Owen JT. Anti- opment. EMBO J 1991; 10: 2111–2118. bodies to CD3/T-cell receptor complex induce death by apoptosis 16 Heberlein C, Fischer KD, Stoffel M, Nowock J, Ford A, Tessmer in immature T cells in thymic cultures. Nature 1989; 337: 181– U, Stocking C. The gene for erythropoietin receptor is expressed 184. in multipotential hematopoietic and embryonal stem cells: evi- 33 Ucker DS, Myers J, Obermiller PS. Activation driven T cell death: dence for differentiation stage-specific regulation. Mol Cell Biol II. Quantitative differences alone distinguish stimuli triggering 1992; 12: 1815–1826. nontransformed T cell proliferation or death. J Immunol 1992; 17 Anderson DM, Lyman SD, Baird A, Wignall JM, Eisenman J, Rauch 149: 1583–1592. C, March CJ, Boswell HS, Gimpel SD, Cosman D, Williams DE. 34 Lenardo MJ. Interleukin-2 programs mouse ␣␤ T lymphocytes for Molecular cloning of mast cell growth factor, a hematopoietin that apoptosis. Nature 1991; 353: 858–861. is active in both membrane bound and soluble forms. Cell 1990; 35 Heyworth CM, Dexter TM, Kan O, Whetton AD. The role of 63: 235–243. 18 Just U, Friel J, Heberlein C, Tamura T, Baccarini M, Tessmer U, hemopoietic growth factors in self-renewal and differentiation of Klinger K, Ostertag W. Upregulation of lineage specific receptors IL-3 dependent multi-potential stem cells. Growth Factors 1990; and ligands in multipotential progenitor cells is part of an 2: 197–211. endogenous program of differentiation. Growth Factors 1993; 9: 36 Koury MJ. Programmed cell death (apoptosis) in hematopoiesis. 291–300. Exp Hematol 1992; 20: 391–394. 19 Gorczyca W, Bruno S, Darzynkiewicz RJ, Gong J, Darzynkiewicz 37 Sachs L, Lotem J. Control of programmed cell death in normal Z. DNA strand breaks occurring during apoptosis: their early in and leukemic cells: new implications for therapy. Blood 1993; 82: situ detection by the terminaldeoxynucleotidyl transferase and 15–21. nick translation assays and prevention by serine protease inhibi- 38 Williams GT, Smith CA, Spooncer E, Dexter TM, Taylor DR. tors. Int J Oncol 1992; 1: 639–648. Haematopoietic colony stimulating factors promote cell survival 20 Itoh K, Tezuka H, Sakoda H, Konno M, Nagata K, Uchiyama T, by suppressing apoptosis. Nature 1990; 343: 76–79. Uchino H, Mori KJ. Reproducible establishment of hemopoietic 39 Rabizadeh S, Oh J, Zhong L, Yang J, Bitler CM, Butcher LL, Bred- supportive stromal cell lines from murine bone marrow. Exp esen DE. Induction of apoptosis by the low-affinity NGF receptor. Hematol 1989; 17: 145–153. Science 1993; 261: 345–348. 21 Williams DE, Eisenmann J, Baird A, Rauch C, van Ness K, March 40 Heyman MJ, Meyer S, Martin F, Steinlein P, Beug H. Self-renewal CJ, Park LS, Martin U, Mochizuki DY, Boswell HS, Burgess GS, and differentiation of normal avian erythroid progenitor cells: Cosman D, Lyman SD. Identification of a ligand of the c-kit proto- regulatory roles of the TGF␣/c-ErbB and SCF/c-kit receptors. Cell oncogene. Cell 1990; 63: 167–174. 1993; 74: 157–169.