Bone marrow stromal cells constitute a very heterogeneous population (i.e., including fibroblasts/reticular cells, adipocytes, endothelial cells, macrophages and osteogenic cells) that is important in normal as well as leukemic hematopoiesis, including the neoplastic proliferation of immature myeloid cells in AML.1, 2, 3, 4, 5, 6, 7, 8, 9 Firstly, normal myeloid cells as well as AML blasts express adhesion molecules that bind to stromal elements. These molecules include β1 (CD29) and β2 (CD18) integrins, the immunoglobulins PECAM-1 (CD31) and LFA-3 (CD58), L-selectin and the proteoglycan CD44.6, 10–13 Secondly, stromal cells release a wide range of cytokines, including hematopoietic growth factors.5, 6, 14–16 Thirdly, stromal cells produce extracellular matrix, e.g., different collagens, proteoglycans (e.g., chondroitin sulfate, heparan sulfate, dermatan sulfate), hyaluronan and glycoproteins like fibronectin, laminin and thrombospondin.5, 10, 17 One function of these molecules is to bind growth factors and thereby create a local extracellular reservoir.17 The stromal cells together with their matrix components thereby constitute an interacting network with AML cells.11, 18
The proliferative capacity of in vitro cultured AML blasts varies between patients: spontaneous or autocrine in vitro proliferation is detected only for a subset of patients and associated with poor prognosis.19 Furthermore, only a subpopulation of the hierarchically organized AML clone appears to proliferate during in vitro culture, whereas the majority of the cells often undergo spontaneous in vitro apoptosis.20 Both the presence of hematopoietic growth factors alone as well as coculture with stromal cells (e.g., fibroblast cell lines) can then increase AML blast proliferation and inhibit apoptosis.16, 19–22 However, previous studies have not characterized in detail the effects of leukemia cells on various stromal cells. AML-induced modulation of different stromal cells may be important for the regulation of both leukemic and remaining normal hematopoiesis.23 We have therefore used standardized experimental in vitro models to compare the effects on fibroblasts and osteoblasts by native human AML cells derived from a large group of consecutive patients. Bone marrow stromal cells are heterogeneous, and even fibroblasts may differ in their functional phenotype;14 for this reason, we investigated AML effects on 2 well-characterized fibroblast lines, 3 different osteosarcoma cell lines (2 having an osteoblastic phenotype), normal osteoblasts and a heterogeneous population of normal bone marrow stromal cells.
The clinical and biologic characteristics of 55 consecutive AML patients are presented in Table I. Cells from 11 patients with ALL were also examined; 10 patients had B-cell disease and one patient, T-cell disease (Table II). 24 All patients had high peripheral blood blast counts; leukemic PBMCs were therefore isolated by density gradient separation alone (Ficoll-Hypaque, specific density 1.077; NyCoMed, Oslo, Norway). Cells were stored in liquid nitrogen.25 The percentage of leukemic blasts among PBMCs exceeded 95% for all patients as judged by light microscopy of May-Grünwald/Giemsa-stained cytospin smears.26, 27, 28
Table I. Clinical and biologic characteristics of AML patients
ALL blasts were regarded as positive for membrane molecules when >20% of cells stained positive by flow-cytometric analysis. Classification was based on the guidelines given by the European Group for the Immunological Classification, of Acute Leukemias.18 According to this classification, B-lineage ALL blasts are positive for at least 2 of the 3 markers CD19, CD22 and CD79a. Patients classified as pro-B-ALL (also referred to as B-I or null ALL) express no other B-cell differentiation antigens, common ALL (also referred to as c-ALL, pre-pre-B-ALL or B-II) express CD10, pre-B-ALL (B-III) express cytoplasmic lg and mature-B-ALL (also referred to as B-ALL or B-IV) express surface membrane Ig.
Routine screening for chromosomal abnormalities was done by analysis of cells in mitosis. The abbreviation “nt” (not tested) means that cells were not available for testing or did not proliferate in vitro; for 3 of these patients, the presence of the bcr/abl translocation (Philadelphia chromosome) was analyzed by the FISH technique.
Peripheral blood white blood cell (WBC) counts are expressed as × 109/l.
Previous chemotherapy for testicular carcinoma
Fibroblast cell lines.
HFL1 cells (ATCC, Manassas, VA; ATCC number CCL-153), derived from the lungs of a fetus (16–18 weeks old) have a diploid karyotype; the recommended medium is Ham's F12K with 10% FCS. Hs27 cells (ATCC number CRL-1634), derived from the normal foreskin of a newborn, have a normal male karyotype; the recommended culture medium is DMEM with 10% FCS. Both lines were expanded in the recommended media and stored in frozen aliquots.25
Human osteosarcoma cell lines.
Cal72 cells (Deutsche Sammlung von Zellkulturen und Mikroorganismen, Braunschweig, Germany) have been characterized previously and have a phenotype close to normal osteoblasts with an adherent growth pattern and a broad cytokine release profile.29, 30 Pilot experiments demonstrated that this cell line released high levels of IL-6, VEGF, HGF and MCP-1. The SJSA-1 cell line (ATCC number CRL-2098) has a similar growth pattern and cytokine profile, whereas the lines Saos-2 (ATCC number HTB-85), SK-ES-1 (ATTC number HTB-86), U2OS (ATCC number HTB-96), 143.98.2 (ATCC number CCL-11226) and KHOS-32IH (ATCC number CRL-1546) showed either an epithelial growth pattern or a different cytokine release profile.
Normal human osteoblasts and bone marrow stromal cells.
These cells were delivered in frozen vials (Clonetics-BioWhittaker, Walkersville, MA), stored in liquid nitrogen until thawed and used directly in the coculture assay. Both cell populations showed a purity of 95% and tested negative for mycoplasma, HIV-1, hepatitis B and hepatitis C (PCR). The culture medium in these experiments was osteoblast growth medium containing 10% FCS (Clonetics-BioWhittaker).
Normal human osteoblasts were derived from a healthy 16-year-old male Caucasian. Cells were derived by explant, and characterization after in vitro redifferentiation showed positivity after staining for alkaline phosphatase and bone mineralization (von Kossa stain). Normal bone marrow stromal cells were derived from a healthy 20-year-old female Caucasian. Bone marrow mononuclear cells were then derived by gradient separation (specific density 1.077) and cultured in Myelocult growth medium (Clonetics-BioWhittaker) for 4 weeks. Stromal cells represent the adherent cell population of these in vitro cultured cells and are a heterogeneous population of fibroblasts, reticular cells, endothelial cells, macrophages and fat cells.
The following culture media were used: Ham's F12K medium, DMEM, McCoy's medium (all from ATCC), RPMI-1640 (GIBCO, Paisley, UK), Stem Span SFEM (Stem Span Technologies, Vancouver, Canada), X-vivo 10, X-vivo 15 and X-vivo 20 (all from Clonetics-BioWhittaker). X-vivo and Stem Span media can be used for serum-free cell culture; the other media were supplemented with 10% heat-inactivated FCS (Clonetics-BioWhittaker). Unless otherwise stated, we used Stem Span medium supplemented with 10% FCS and gentamicin 100 μg/ml because (i) pilot experiments demonstrated that FCS was required for optimal growth of fibroblasts and osteosarcoma cell lines, (ii) all sarcoma cell lines and HFL1 fibroblasts showed strong proliferation in this medium (Hs27 fibroblasts showed 50% reduced proliferation compared to the recommended medium) and (iii) the medium should be regarded as suitable for culture of native human AML cells.31 Proliferation showed wide variation between cell lines even when using this medium, the highest proliferation being observed for the 143.98.2 sarcoma cell line.
The following soluble mediators were used at 50 ng/ml (unless otherwise stated, cytokines were from Preprotech, Rocky Hill, NJ): angiogenin (R&D Systems, Abingdon, UK), angiopoietin-2 (R&D Systems), angiostatin (Calbiochem, Darmstadt, Germany), bFGF, endostatin (Calbiochem), Flt3 ligand (Flt3-L), G-CSF (Roche, Basel, Switzerland), GM-CSF (Sandoz, Basel, Switzerland), IFN-γ, IL-1α, IL-1β, IL-2, IL-3, IL-4, IL-9, IL-10, IL-13, IL-15, IL-16, IL-17, M-CSF, SCF, TNF-α, VEGF, IL-1RA (R&D Systems).
The following cytokine-specific neutralizing antibodies (R&D Systems, neutralization data reported by distributor) were used: (i) antihuman GM-CSF: mouse IgG1 MAb 3209.01, 0.3–0.5 μg/ml of this antibody neutralizing 50% of the biologic activity of GM-CSF 0.5 ng/ml; (ii) antihuman HGF: mouse IgG1 MAb 24612.111, 0.1–0.3 μg/ml neutralizing 50% of the biologic activity of HGF 100 ng/ml; (iii) antihuman VEGF: mouse IgG2b MAb 26503, 0.04–0.08 μg/ml neutralizing 50% of the biologic activity of VEGF 10 ng/ml; (iv) antihuman SCF: polyclonal goat antiserum, 0.05–0.1 μg/ml of this preparation neutralizing 50% of the biologic activity of SCF 10 ng/ml. Anti-GM-CSF and anti-HGF were used at 10 μg/ml; anti-SCF at 2 μg/ml and anti-VEGF were used at 1 μg/ml. Control cultures were prepared using corresponding normal mouse IgG1 (anti-GM-CSF, anti-HGF) and IgG2b (anti-VEGF) antibodies or normal goat antiserum (anti-SCF) (all supplied by R&D Systems) at the same concentrations as the specific antibodies.
Preparation of AML supernatants.
AML blasts (2 × 106 cells in 2 ml Stem Span/well) were cultured in 24-well tissue culture plates (Costar 3524; Costar, Cambridge, MA) for 48 hr before supernatants were harvested
In vitro culture of nonleukemic cells alone.
Frozen nonleukemic cells were precultured in the recommended medium in a 75 cm2 culture flask until they were confluent, then trypsinized and distributed into flat-bottomed microtiter wells (104 cells/well cultured in 200 μl medium; Costar 3596 96-well culture plates). After 3 days 37 kBq of 3H-thymidine (TRA 310; Amersham, Aylesbury, UK) were added in 20 μl saline, and nuclear radioactivity was determined 18 hr later by liquid scintillation counting.
Coculture of AML and nonleukemic cells separated by a semipermeable membrane.
Cultures were prepared in Transwell culture plates (Costar, Transwell 3401), where cells in the lower, larger compartment were separated from cells in the upper, smaller well by a semipermeable membrane with pore size 0.4 μm. Nonleukemic cells were initially cultured for 3 days in the lower chamber (104 cells in 1 ml medium) before 106 native human AML blasts were added in 0.5 ml medium to the upper well. Cocultures were incubated for 6 days before 3H-thymidine (280 kBq/well) was added; cultures were thereafter incubated for an additional 18 hr before AML cells were removed. All cultures were observed under a microscope before harvesting to ensure that cells/osteoblasts/fibroblasts did not form a confluent layer. Adherent cells were then washed in isotonic saline before 300 μl/well of trypsin-EDTA solution (Stem Cell Technologies) were added and nuclear radioactivity was assayed in 70 μl aliquots.
Coculture of AML cells in direct contact with fibroblasts.
HFL1 and Hs27 fibroblasts were precultured for 3 days (104 cells/well in 1 ml medium, Costar 3524 culture plates) before AML cells were added (2 × 106 leukemia cells in 0.5 ml medium). Cocultures were incubated for 6 days before 3H-thymidine (280 kBq/well) was added. Nonadherent cells were removed 18 hr later; light microscopy then demonstrated that the remaining cells formed a uniform layer with a morphologic appearance consistent with fibroblasts and that they did not form a confluent layer in any culture. Adherent cells were trypsinized and nuclear radioactivity was assayed as described above.
Analysis of GM-CSF, IL-1β and IL-1RA levels.
GM-CSF, IL-1β and IL-1RA levels were determined by ELISA (Quantikine ELISA kit, R&D Systems) performed strictly according to the manufacturer's instruction. Standard curves were prepared as recommended, and differences between duplicates were generally <10% of the mean. Minimal detectable levels were as follows: GM-CSF, 3 pg/ml; IL-1β, 2.8 pg/ml; and IL-1RA, 14 pg/ml.
Presentation of data
The mean response of triplicate determinations was used in all calculations and statistical comparisons. Significant proliferation was defined as 3H-thymidine incorporation exceeding 1,000 cpm and exceeding the mean counts per minute of the negative control by at least 3 SD. Incremental responses (stimulated response in cpm – negative control) were used for all calculations. A significant alteration of proliferation was defined as (i) a difference exceeding 2,000 cpm and (ii) this difference being >20% of the control.
GM-CSF and IL-1RA supernatant levels.
A significant alteration in GM-CSF and IL-1RA supernatant levels was defined as a difference exceeding 20% of the control when paired samples were compared. Incremental GM-CSF levels were defined as the levels in cocultures minus the summarized levels for control cultures containing corresponding AML cells and nonleukemic cells alone.
Wilcoxon's test for paired samples and the Mann-Whitney U-test/Wilcoxon test for 2 independent samples were used for comparison of continuous data. Spearman's ρ test was used for correlation analysis. Differences were regarded as statistically significant at p < 0.05.
Native human AML cells inhibit proliferation of HFL1, but not Hs27, fibroblasts
AML blasts were cultured together with HFL1 fibroblasts (Table I, patients 1–44) in transwell cultures prepared in Stem Span medium with FCS. HFL1 fibroblasts cultured alone showed relatively strong proliferation, which was decreased during coculture with AML blasts (Fig. 1, n = 44, p = 0.007); this effect was reproduced in repeated experiments for 11 patients. The number of HFL1 fibroblasts was also counted after coculture with AML blasts derived from 16 consecutive patients (Table I, patients 1, 3, 7, 9, 10, 17–24, 26, 47, 54), and the reduced numbers of HFL1 fibroblasts after cocultures confirmed that the decreased 3H-thymidine incorporation during coculture reflected a true reduction in the number of fibroblasts (mean decrease 38%, Wilcoxon's test for paired samples p = 0.048). Decreased HFL1 proliferation was also detected in the presence of exogenous hematopoietic growth factors (transwell cultures with GM-CSF + SCF + Flt3-L, 23 patients examined; data not shown) and when AML cells and fibroblasts were cultured in direct contact (Table I, patients 1–33; Wilcoxon's test for paired samples p = 0.01).
HFL1 fibroblasts were also cultured in the recommended FCS-containing medium supplemented with 50 μl/well (final dilution 1:4) of AML cell supernatants (Table I, patients; 1, 3–5, 7, 10, 14–16, 19–25, 27–29). These selected patients represent a consecutive group. Supernatants increased fibroblast proliferation significantly (Fig. 2, p < 0.001). This enhancement was maintained in the presence of exogenous angiogenin, angiopoietin-2, angiostatin, SCF and VEGF, whereas the enhancement was not detected in the presence of exogenous IL-1β, bFGF and endostatin (data not shown). Thus, the AML-associated inhibition of HFL1 proliferation during in vitro culture depends on the crosstalk between the 2 cell populations.
The effects of AML blasts on Hs27 fibroblasts differed from those on HFL1 fibroblasts: (i) Hs27 cells showed relatively low in vitro proliferation; (ii) proliferation was increased during coculture with AML blasts in transwell cultures (Fig. 1, n = 33, p = 0.001); and (iii) proliferation of Hs27 cells was not altered during coculture in direct contact with AML blasts (data not shown).
The effects of AML blasts on HFL1 and Hs27 fibroblast proliferation showed no correlation with peripheral blood counts (platelets, hemoglobin levels) at the time of diagnosis, signs of AML cell differentiation (FAB classification and CD34 expression) or cytogenetic or genetic Flt3 abnormalities of the AML cells (data not shown).
Native human AML blasts inhibit proliferation of osteoblastic sarcoma cells
Native AML blasts were cultured in transwell cultures together with Cal72 (Table I, patients 1–10, 12, 13, 16–23, 26, 28, 29, 31, 34–36, 38, 42, 43, 45–47, 49, 50–54), SJSA-1 and 143.98.2 osteosarcoma cells (patients 1–10, 12–14, 16–26, 28–36, 38, 40, 42, 43, 45–54). Proliferation of osteoblastic Cal72 (Fig. 1; Wilcoxon's test for paired samples n = 39, p < 0.001) and SJSA-1 (Fig. 1; n = 47, p = 0.01) sarcoma cells was significantly decreased during coculture. The effects on Cal72 proliferation were reproduced in repeated experiments for 11 patients, and the growth inhibition of osteoblastic sarcoma cells showed no association with karyotype or Flt3 abnormalities in AML cells (data not shown). Numbers of Cal72 and SJSA-1 cells were also compared for AML-free cultures and cocultures with AML blasts derived from a consecutive subset of 15 patients (Table I, patients 1, 3, 7, 9, 10, 17–23, 26, 47, 54), and decreased numbers of sarcoma cells were then detected after coculture both for Cal72 (median decrease 47%, Wilcoxon's test for paired samples p = 0.048) and SJSA-1 (median decrease 59%, p < 0.0005) cells.
In contrast to the effects on Cal72 and SJSA-1 cells, AML blasts had only minor and divergent effects on the proliferation of the epithelioid cell line 143.98.2 (data not shown) and of the 4 other osteosarcoma cell lines (11 patients examined, data not shown).
The effects of AML blasts on osteoblastic sarcoma cells showed no association with signs of AML cell differentiation (FAB classification and CD34 expression), cytogenetic abnormalities or genetic Flt3 abnormalities of AML cells (data not shown).
GM-CSF is involved in crosstalk between fibroblasts and native human AML cells
To further characterize the molecular mechanisms involved in crosstalk between AML blasts and nonleukemic cells, we prepared transwell cultures with and without cytokine-inhibitory mediators. AML blasts derived from 12 randomly selected patients (patients 7, 9, 10, 17, 21, 23, 29, 31, 42, 51, 54, 55) were cultured with HFL1 fibroblasts and Cal72 osteoblastic sarcoma cells in transwell cultures in the presence of neutralizing anti-GM-CSF, anti-HGF, anti-VEGF, anti-SCF or corresponding control antibodies. Pilot experiments demonstrated that these mediators did not alter the proliferation of fibroblasts cultured alone (data not shown). Fibroblast proliferation was increased by anti-GM-CSF for a majority of patients (10 of 12, Wilcoxon's test for paired samples p < 0.005), though exogenous GM-CSF alone did not alter proliferation of fibroblasts cultured alone (data not shown). Furthermore, an increase corresponding to >2,000 cpm and exceeding the control by at least 20% was also observed for certain patients in the presence of anti-HGF (2/12) and anti-SCF (2/12). Proliferation of Cal72 osteoblastic sarcoma cells was altered (i.e., the difference being >2,000 cpm and >20% of control) only for a minority of patients by the neutralizing antibodies (anti-GM-CSF 1/12, anti-HGF 0/12, anti-SCF 1/12, anti-VEGF 0/12). Furthermore, leukemia cells derived from 11 other randomly selected patients (Table I, patients 1, 9, 10, 17–21, 23, 25, 26) were cultured with Cal72 and HFL1 cells in medium alone and in the presence of IL-1RA 50 ng/ml. Exogenous IL-1RA increased nonleukemic cell proliferation (i.e., a difference corresponding to >2,000 cpm and >20% of controls) for only a minority of patients when testing HFL1 (2/11) and Cal72 cells (3/11). We conclude that (i) GM-CSF is usually involved in crosstalk between fibroblasts and native human AML cells but (ii) additional mechanisms that differ between patients and may involve HGF, SCF and IL-1β also appear to be involved in the AML-associated inhibition of fibroblasts and osteoblasts.
GM-CSF levels can be increased during coculture of AML blasts with HFL1 fibroblasts and osteoblastic sarcoma cells
Constitutive GM-CSF secretion by native AML blasts was investigated for patients 1–54 (Table I), and detectable release was defined as detectable supernatant levels in at least 2 of 3 experiments when AML blasts were cultured alone for 7 days. Detectable GM-CSF release was then observed for 40 patients (median level 26.1 pg/ml, variation range 2.3–1,260 pg/ml). Constitutive GM-CSF release was also detected for fibroblast and sarcoma cell lines, the levels in control cultures of each coculture experiment being for HFL1 fibroblasts always <41.0 pg/ml, for Hs27 fibroblasts <35.0 pg/ml, for SJSA-1 sarcoma cells <32.0 pg/ml, for Cal72 sarcoma cells <5.6 pg/ml and for 143.98.2 sarcoma cells >1,000 pg/ml.
Native human AML blasts were cocultured in transwell cultures with HFL1 and Hs27 fibroblasts (Table I, patients 1–34) and with the osteoblastic sarcoma cell lines Cal72, SJSA-1 and 143.98.2 (Table I, patients 1–10, 12, 13, 16–23, 26, 28, 29, 31, 34–36, 38, 42, 43, 45–47, 49, 50–54). Increased GM-CSF levels were detected when AML blasts with constitutive GM-CSF secretion (20 of the 34 patients) were cocultured with HFL1 fibroblasts both in transwell cultures (Fig. 3, Wilcoxon's test for paired samples p = 0.004) and when cells were cultured in direct contact (p = 0.008, data not shown). Increased levels were also observed during coculture with Cal72 and SJSA-1 cells (Fig. 3, p < 0.001 for both) when testing AML blasts with constitutive GM-CSF release (31 of the 39 patients). GM-CSF levels for AML cells cultured alone showed strong correlations with the levels in cocultures both when testing HFL1, Cal72 and SJSA-1 nonleukemic cells (Spearman's ρ test, correlation coefficients >0.89 and p < 0.001 for all 3 cell lines). In contrast, an increase was not observed in experiments including (i) AML blasts with low/undetectable GM-CSF release (Fig. 3); (ii) coculture of AML blasts with Hs27 fibroblasts, these levels being significantly lower than for HFL1 cells (Wilcoxon's test for paired samples p < 0.001); and (iii) cocultures with the epithelial sarcoma cell line 143.98.2, decreased GM-CSF levels being detected (p = 0.002, data not shown).
GM-CSF levels in cultures containing AML cells alone and AML cells cocultured with any of the nonleukemic cells showed no correlation with differentiation (FAB classification or CD34 expression), AML cell karyotype or genetic Flt3 abnormalities (data not shown).
IL-1β levels (but not levels of IL-1RA) are increased during coculture of AML blasts with HFL1 fibroblasts and osteoblastic sarcoma cells
Both normal as well as leukemic hematopoiesis can be stimulated by IL-1, and native human AML blasts can release both IL-1 and IL-1RA.32 The balance between IL-1 and IL-1RA release will probably determine the effect on neighboring normal hematopoiesis. IL-1β is usually released at higher levels than IL-1α by native human AML blasts,26 and we therefore compared levels of IL-1β in cultures containing AML blasts alone, osteoblasts/fibroblasts alone and cocultures of AML blasts with HFL1 fibroblasts (Table I, patients 1–33) with the osteoblastic sarcoma cell lines Cal72 and SJSA-1 (Table I, patients 1–10, 12, 13, 16–23, 26, 28, 29, 31, 34–36, 38, 42, 43, 45–47, 49, 50–54). Constitutive IL-1β release was detected only for a subset of patients (median level 21.6 pg/ml, range 4.8–416 pg/ml). Increased IL-1β levels were observed when AML blasts with constitutive IL-1β release were cocultured with HFL1 fibroblasts and Cal72 and SJSA-1 osteoblastic sarcoma cells (Fig. 4, Wilcoxon's test for paired samples p < 0.001 for all 3 cell lines); this increase was also maintained when AML cells and HFL1 fibroblast were cultured in direct contact. IL-1β levels for AML cells cultured alone showed strong correlations with the levels in cocultures when testing HFL1, Cal72 and SJSA-1 nonleukemic cells (Spearman's ρ test correlation coefficients >0.90 and p < 0.001 for all 3 cell lines). In contrast, IL-1β was not detected for culture containing only nonleukemic cells and for a small minority of cocultures containing AML cells without constitutive IL-1β secretion (undetectable levels <2.8 pg/ml).
We analyzed IL-1RA levels after in vitro culture of AML blasts derived from 50 patients (Table I, patients 1–16, 18–36, 38–43, 46–54). IL-1RA was detected in the supernatants for all patients (median level 1,100 pg/ml, range 20–2,800 pg/ml), and patients with AML-M4/M5 had increased levels (Mann-Whitney U-test p = 0.02). AML blasts from 26 randomly selected patients were available for additional studies of IL-1RA levels after coculture with HFL1 fibroblasts (patients 1–7, 9–13, 18–30, 33). Detectable IL-1RA levels were then observed for 25 patients (median 305 pg/ml, range <14.0–2,700 pg/ml). IL-1RA levels for AML cultures and corresponding cocultures showed a significant correlation (Fig. 5, Spearman's ρ test, n = 26, correlation coefficient 0.53, p = 0.005), though levels were lower in cocultures (Wilcoxon's test for paired samples p = 0.049). Furthermore, a significant increase in the IL-1β/IL-1RA ratio was observed for HFL1 transwell cultures (median 0.065, range 0.001–4.762) compared to corresponding control cultures containing AML blasts alone (median 0.020, range 0.002–0.151, Wilcoxon's test for paired samples p = 0.002).
Even though IL-1RA levels showed no correlation with AML blast karyotype for cells cultured alone, patients with normal karyotype (n = 10, median 2,075 pg/ml, range 27–2,700 pg/ml) showed higher IL-1RA levels in HFL1 cocultures than patients with abnormal karyotypes (n = 10, median 182 pg/ml, range <14–1,690 pg/ml, Mann-Whitney U-test p = 0.023).
Decreased osteoblast proliferation and increased GM-CSF/IL-1β levels during coculture of native human AML blasts and normal osteoblasts/stromal cells
Normal stromal cells and osteoblasts were incubated in osteoblast growth medium in transwell cultures together with AML blasts derived from 15 consecutive patients (Table I, patients 1, 3, 7, 9, 10, 17–23, 26, 47, 54). Light microscopy verified that viable, adherent stromal cells were present in all cultures but 3H-thymidine incorporation did not exceed 1,000 cpm for any culture. In contrast, proliferation of normal osteoblasts alone corresponded to 2,230 ± 433 cpm; proliferation was slightly increased for 3 patients and decreased to undetectable levels for 7 patients (Fig. 6). Furthermore, GM-CSF and IL-1β levels were also determined in culture supernatants. Osteoblast growth medium was suboptimal for AML cells, and constitutive GM-CSF/IL-1β release was detected only for 3 and 2 patients, respectively, in these experiments. Increased GM-CSF/IL-1β levels were then observed for all these patients during coculture with normal osteoblasts as well as normal bone marrow stromal cells (data not shown).
ALL blasts have only minor effects on proliferation of fibroblast and osteoblastic sarcoma cells
ALL blasts were cultured in transwell cultures together with HFL1 fibroblasts (Table II, patients 1–11), Hs27 fibroblasts (patients 1–11), Cal72 osteoblastic sarcoma cells (patients 1–7), SJSA-1 osteoblastic sarcoma cells (patients 1–7) and 143.98.2 osteoblastic epithelial sarcoma cells (patients 1–7). ALL blasts had only minor and divergent effects on the proliferation of nonleukemic cells (data not shown).
Bone marrow stromal cells constitute a very heterogeneous population that forms an interacting network together with matrix components and leukemia cells in AML patients.3–13, 15–17 The molecular bidirectional crosstalk with stromal cells appears to support AML blast proliferation,3–9, 16, 20, 22 and our own prior studies demonstrated that both fibroblast cell lines HFL1 and Hs27, the osteoblastic sarcoma cell lines Cal72 and SJSA-1 and normal osteoblasts33, 34 can increase AML blast proliferation as well as release of proangiogenic IL-8 by native human AML cells. Fibroblasts also have an additional antiapoptotic effect,22 and this effect is observed both with Hs27 and HFL1 fibroblasts (Ryningen, unpublished). Thus, the fibroblast and osteoblast cell lines included in our present study support leukemic hematopoiesis through several mechanisms. However, at the same time, our present results show that native human AML cells release soluble mediators that inhibit stromal cell proliferation. AML cells may thereby indirectly affect remaining normal hematopoiesis and contribute to bone marrow failure together with direct inhibitory effects like AML cell release of IL-1RA.
The major part of our experiments was performed with well-characterized cell lines; this approach allows comparison of AML-induced effects between patients in standardized experimental models. Both HFL1 and Hs27 fibroblasts support leukemic hematopoiesis.33, 35 The sarcoma cell line Cal72 has an osteoblastic phenotype and supports normal hematopoiesis,29, 30 and SJSA-1 cells had a similar phenotype. To further evaluate the biologic relevance of our experimental results, we also included studies of normal osteoblasts and a heterogeneous population of normal bone marrow stromal cells.
Osteoblasts and fibroblasts are closely related,36 and the effects of native human AML cells on HFL1 fibroblasts, osteoblastic sarcoma cells (Cal72 and SJSA-1) and normal osteoblasts showed several similarities. Firstly, AML cells inhibited proliferation of all these cells. This growth inhibition was detected by a 3H-thymidine incorporation assay. An alternative explanation for this decreased 3H-thymidine incorporation after 7 days of in vitro coculture could be increased proliferation with formation of confluent layers before day 7 and thereby decreased 3H-thymidine at the end of the culture period. However, several observations make this last explanation unlikely: (i) cell counting confirmed that cultures contained a decreased number of nonleukemic cells; (ii) all cultures were observed under a microscope before harvesting, and adherent cells were not confluent in either control cultures or AML cocultures; (iii) 143.98.2 cells showed the strongest in vitro proliferation when cultured alone or in the presence of native human AML cells, but for this cell line no AML-associated growth inhibition was detected. We therefore conclude that the decreased 3H-thymidine incorporation by HFL1, Cal72 and SJSA-1 cells represents a true AML-associated growth inhibition. Secondly, the presence of fibroblasts, osteoblastic cells and normal osteoblasts could increase GM-CSF and IL-1β levels when these cells are cocultured with constitutive GM-CSF/IL-1β-secreting AML cells. These increased cytokine levels are probably caused by increased release from AML cells because the levels for AML cells cultured alone showed a wide variation and significant correlations with the corresponding cocultures. Thirdly, these effects showed no associations with FAB classification, cytogenetic abnormalities or genetic Flt3 abnormalities, thereby suggesting that the growth inhibition and increased GM-CSF/IL-1β levels are common functional characteristics of otherwise heterogeneous AML cell populations.
The AML-associated growth inhibitory and GM-CSF/IL-1β-increasing effects were detected only for HFL1 fibroblasts, osteoblastic sarcoma cells and normal osteoblasts/stromal cells, whereas these effects were not observed when testing other osteosarcoma cells (the epithelial line 143.98.2 was included in all experiments together with Cal72 and SJSA-1) and the Hs27 fibroblast cell line. Taken together, these observations suggest that the AML-associated effects on sarcoma cells depend on the osteoblastic phenotype rather than on osteosarcoma-associated characteristics. Fibroblasts may also show different phenotypic characteristics,14 and our results suggest that the AML-associated effects are observed only for certain fibroblast subsets. However, our in vitro culture conditions appeared to be suboptimal for Hs27 cells, and this may also contribute to the differences between HFL1 and Hs27 fibroblasts.
In a separate series of experiments, we compared the effects on HFL1 fibroblasts cocultured with native human AML cells and in the presence of culture supernatants harvested from AML cells that were cultured alone. The divergence between supernatant (increased HFL1 proliferation) and coculture (decreased proliferation) experiments clearly demonstrated that the AML-associated growth inhibition is dependent on bidirectional crosstalk between the cells.
GM-CSF was released by both nonleukemic cells as well as most AML blast populations, and GM-CSF was detected in most cocultures with fibroblasts and osteoblastic sarcoma cells. Experiments with GM-CSF neutralizing antibodies demonstrated that GM-CSF was important for the crosstalk between AML cells and fibroblasts, but exogenous GM-CSF alone did not alter fibroblast proliferation. Our hypothesis is therefore that the increased GM-CSF levels mainly have effects on AML cells during coculture; AML blasts are usually GM-CSF-responsive, and GM-CSF can be an autocrine growth factor for native human AML cells.26, 27, 28 GM-CSF effects on AML blasts may then affect the cytokine release profile and thereby contribute indirectly to the effects on fibroblasts. Furthermore, AML blasts derived from different patients vary in cytokine release profile,26, 27, 37 and other mediators may also contribute to the growth inhibition in certain patients, as suggested by our cytokine neutralization studies.
Our present results confirmed that IL-1RA was released by all AML cell populations, especially by AML-M4/M5 cells,38 whereas IL-1β levels were generally lower and detectable levels were observed for only a subset of patients. Increased IL-1β levels were detected during coculture of AML blasts with HFL1 fibroblasts and osteoblastic sarcoma cells but only for the subset of patients with constitutive IL-1β release. Although release of IL-1RA was decreased during coculture, the remaining levels were still relatively high. IL-1 is important for normal hematopoiesis,32 and the persistent local release of IL-1RA by AML cells may thereby contribute to decreased normal hematopoiesis and bone marrow failure in AML patients, at least for the large subset of patients with low/undetectable IL-1 release. IL-1β and IL-1RA levels for AML blasts cultured alone were significantly correlated to the corresponding levels in cocultures, suggesting that AML blasts are the major source of both mediators during coculture.
Osteoblasts are a part of the stem cell niche, i.e., the bone marrow microcompartment where hematopoietic stem cells are located.39, 40 Previous studies have described reduced numbers of colony-forming fibroblasts in AML bone marrow,6 and our present results support this conclusion and suggest that the number of other important stromal cells may also be reduced in AML. Although the local levels of certain AML-derived cytokines are increased during coculture, the molecular mechanisms behind the AML-induced antiproliferative effect on nonleukemic cells are probably heterogeneous and not caused by a single mediator. A growth-inhibitory effect on osteoblasts mediated via local cytokine networks may thereby represent a mechanism for AML-induced inhibition of the remaining normal hematopoiesis. This inhibition may be further increased by similar inhibitory effects on fibroblast subsets and the release of direct inhibitory mediators like IL-1RA (IL-1β/IL-1RA ratio <1.00 is usually maintained even in fibroblast/AML cell cocultures). Taken together with previous results (Ryningen, unpublished),5, 7, 22, 34, 37 we therefore suggest that nonleukemic stromal cells may be important in disease development in AML patients through several mechanisms, including enhancement of leukemic hematopoiesis, stimulation of bone marrow angiogenesis and reduced stimulation of remaining normal hematopoiesis.