CELLULAR
IMMUNOLOGY
Tumor-Stimulated
140,304-3 18 ( 1992)
Release of Tumor Necrosis Factor-a by Human Monocyte-Derived Macrophages’
ROBERTDEMARCO,~JEFFREY E. ENSOR, AND JEFFREY D. HASDAY~ Division ofPulmonary and Critical Care Medicine, Department of Medicine, University of Maryland, Baltimore, Maryland 21201; and the Baltimore VA Medical Center, Baltimore, Maryland 21218 Received August 5, 1991; accepted November 20, 1991 Tumor necrosis factor-a (TNF) releaseby monocytes and macrophages may be an important determinant of the physiologic responseof the host to neoplastic disease;however, the mechanisms which regulate TNF releaseby macrophagesin hostswith neoplastic diseasesare poorly understood. The purpose of this study was to determine if cell membranes and growth medium from human leukemia cell lines and solid tumor cell lines induced TNF release by cultured human blood monocytederived macrophages.The capacity for TNF releaseand direct tumor killing was highest in monocytes cultured for 7 to 11 days. Cell membranes and culture media from K562 erythroleukemia and several small cell lung carcinoma cell lines, including H82, induced the release of up to 1500TNF units per lo6 macrophagesover 24 hr. By contrast, allogeneic peripheral blood lymphocytes, cell membranes from normal mixed donor peripheral blood leukocytes, or growth medium from normal embryonic lung fibroblasts induced the releaseof little or no TNF during culture up to 24 hr, suggestingthat this macrophage responsewas specific for tumor cells. Release of TNF by tumor-stimulated macrophages was gradual, peaking 24 hr following the addition of stimuli. Induction of macrophage TNF releasewas concentration dependent, with half-maximal TNF levels induced by 12.5 and 25 &ml cell membranes prepared from K562 and H82, respectively. Pretreatment of tumor cell membranes with polymixin B, which inhibits many of the actions of endotoxin, failed to neutralize tumor induction of TNF, suggestingthat endotoxin was not responsible for this activity. Depletion of macrophages by treatment with 3C10 monoclonal antibody and complement abrogated tumor-induced TNF release, indicating that macrophages were the source of the secreted TNF. HPLC analysis of H82 growth medium demonstrated a single peak of macrophage activating activity with approximate 40-kDa molecular weight. We have demonstrated that cell membranes and growth medium from some human leukemia and solid tumor cell lines, but not from normal human cells, induce human peripheral blood monocytes and monocytederived macrophagesto releasefunctionally active TNF. This processmay contribute to the host responseto some neoplastic diseases. 0 1992 Academic press, Inc.
INTRODUCTION Cells of monocyte/macrophage lineage possessthe capacity to play a pivotal role in the host responseto neoplastic diseases,in part through the releaseof immunomodulatory cytokines (1, 2). Following recruitment to tissue compartments, peripheral ’ This work was supported by VA Merit Review Grant No. 128444284-0002 and NIH Grant No. R29CA52741-01. ’ Supported by an American Lung Association of Maryland Research Fellowship. 3Supported by VA Career Development Award No. 1284442840003. 304 0008-8749/92 $3.00 Copyright 0 1992 by Academic Ress., Inc. All rights of reproduction in any form reserved.
MACROPHAGE
TNF-a
RELEASE
INDUCED
BY TUMOR
305
blood monocytes (PBM) differentiate into macrophages, which, in response to sequential exposure to specific signals in their microenvironment, acquire the capacity to selectively bind and kill some neoplastic cells and to maximally express several potent cytokines, including tumor necrosis factor-a (TNF) (l-3). The mechanisms which control cytokine expression by monocytes and macrophagesin tumor-burdened hostsare poorly understood. Beissertet al. (4) reported that a subsetof tumor-infiltrating macrophages present within human colorectal adenocarcinoma biopsies contain elevated levels of TNF mRNA and TNF protein, suggesting that some factors within the tumor microenvironment may induce macrophages to express TNF. We have recently demonstrated that cell membranes from several tumor cell lines directly stimulate a rapid, self-limited releaseof TNF by murine peritoneal macrophages,whereas cell membranes prepared from nontransformed cells did not induce detectable TNF release (5). Further, we found that the TNF-inducing activity could be dissociated from tumor membranes by incubation with chaotropic agents(5). Janicke and Mannel have reported that constituents of cell membranes from two human leukemia cell lines also stimulate the accumulation of TNF-specific mRNA and the releaseof functionally active TNF by freshly obtained PBM (6). Taken together, these data suggest that TNF releaseby mononuclear phagocytesmay represent a direct responseto tumor cells themselves. However, the capacity of mature human macrophages for tumorstimulated TNF releasehas not been determined. Moreover, induction by neoplastic cells of TNF releaseby human monocytes has been described only for leukemia cell lines. The capacity of human solid tumors to directly induce human mononuclear phagocytes to releaseTNF has not been previously reported. The purpose of this study was to determine whether cell membranes and culture supernatants from both leukemia and solid tumors induce expression of TNF by human monocyte-derived macrophages. Macrophages obtained from suspension cultures of mixed peripheral blood mononuclear cells (7) were characterized for expression of functions relevant to their interaction with tumor by serially evaluating their capacity to bind and kill tumor cells and to releasefunctionally active TNF during incubation with cell membrane vesicles and membrane-free growth medium from several human cell lines. The induction of TNF releasefrom optimally responsive macrophages was further characterized. METHODS Reagents. Recombinant human TNF and rabbit anti-human TNF antiserum were obtained from Genzyme (Thousand Oaks, CA). Lipopolysaccharide (LPS) prepared from Escherichia cofi 0 111:B4 was obtained from Difco (Detroit, MI). Defined newborn calf serum (NCS) was obtained from Hyclone (Logan, UT) and was heat-inactivated prior to use. Type A serum was collected by centrifugation of blood obtained from normal, consenting volunteers and allowed to coagulate at room temperature for 3 to 4 hr. Rabbit complement was obtained from Cedarlane Laboratories (Hornby, Ontario). All other reagents were obtained from Sigma Chemicals (St. Louis, MO). Cell culture. All incubations were performed in RPM1 1640 (Mediatech; Fairfax, VA) containing 50 U/ml penicillin, 50 pg/ml streptomycin, 2 mM glutamine, 1 mM sodium pyruvate (GIBCO/BRL, Grand Island, NY), 10 mM Hepes buffer, pH 7.3 (CRPMI), and, unless otherwise stated, supplemented with 10%(v/v) NCS. All media and reagentscontained lessthan 0.1 rig/ml endotoxin as determined by limulus amoebocyte lysate assay(Whittaker Bioproducts; Walkersville, MD).
306
DEMARCO, ENSOR, AND HASDAY
The following human tumor cell lines were used to activate macrophages: K562, an erythroleukemia, and A375 a melanoma, were obtained from the American Type Cell Collection (ATCC; Rockville, MD); the classic and variant small cell lung carcinoma lines, H 146 and H82, and the H 125 adenosquamous lung carcinoma cell line were provided by Dr. A. Doyle (Univ. of Maryland; Baltimore, MD); the small cell carcinoma cell line, H209, and two sublines, H209 myc and H209 myc/ras, which have been transfected with the oncogene c-myc alone or the combination of c-myc and Harvey-ras oncogenes, were provided by Dr. M. Mabry (8) (Johns Hopkins University; Baltimore, MD). Normal, nontransformed human embryonic lung fibroblasts (HEL), obtained from ATCC were used as a control treatment in some experiments. These cell lines were maintained in CRPMI/ 10% NCS and were routinely tested for Mycoplasma contamination by culturing in both solid and liquid media (Flow Labs., McLean, VA) using Mycoplasma arginini as a control (ATCC). The hybridoma 3C 10, which secretesa monoclonal antibody which recognizes human macrophages (9), was obtained from ATCC and maintained in Dulbecco’s modified essential medium (Mediatech) supplemented with 4 mM glutamine, 0.6 /*g/ml oxaloacetate,2 cLg/mlsodium pyruvate, 0.34 &ml bovine insulin, 100 U/ml penicillin, 100 pg/ml streptomycin, 0.2 &ml amphotericin B, 20% fetal calf serum (Hyclone), and 10% NCTC 109 medium (GIBCO/BRL). Mixed donor leukocytes (WBC), collected from plasma samples following plasmaphoresis, were provided by the University of Maryland Plasmaphoresis center. Peripheral blood lymphocytes were obtained from Ficoll-Hypaque-purified mononuclear cell suspensions, were depleted of macrophages by 2 hr adherence to plastic culture plates, and were then stimulated to proliferate by culturing for 96 hr in CRPMI/ 10% type A serum with 3 pg/ml concanavalin A. Preparation of cell membranes and cell-free culture supernatants. Membranes and supernatants were prepared from cells harvested in mid-log growth, as described (5). Briefly, cell suspensions (2 X lo* cells/ml) in ice-cold phosphate-buffered saline, pH 7.4 (PBS), containing 4 mM ethylenediaminetetraacetate, 1 mM iodoacetamide, and 1 mM phenylmethylsulfonyl fluoride were freeze-thawed twice, and lysed using a Dounce homogenizer. Nuclei were removed by centrifugation at 300g for 10 min at 4°C. The supernatants were then sonicated (Heat Systems-Ultrasonics; Plainview, NY) and the membrane vesicles were collected by centrifugation at 120,OOOgfor 60 min at 4°C washed with PBS, and the protein contents were determined by the Bradford method (10). Membranes were diluted to the desired protein concentration in CRPMI/ 10% NCS and briefly resonicated prior to use. To avoid removal of relevant membrane proteins during cell harvesting, adherent cell lines were collected by scraping without proteases.Growth medium collected from cells in late-log growth were depleted of membranes by centrifugation at 120,OOOg for 1 hr and stored at -70°C until used. None of the stimuli used in these experiments contained any detectable bioactivity in the L929 assay. Isolation and culture ofperipheral blood monocytes. Peripheral blood mononuclear cells were isolated from heparinized venous blood collected from normal, consenting volunteers by Ficoll-Hypaque gradient centrifugation as described by Boyum (11). Mononuclear cells (370 f 30 X lo6 per 200 ml blood) comprising 42 f 3% monocytes and 58 r+ 4% lymphocytes were either used immediately or suspendedto lo6 cells/ml in CRMPI supplemented with 10% type A serum and cultured in 250 ml Teflon beakers(Nalgene; Rochester, NY) as described (7). After various times in culture, cells
MACROPHAGE
TNF-a
RELEASE
INDUCED
BY TUMOR
307
were collected, counted with a hemacytometer, and the proportions of monocytes/ macrophages and lymphocytes were determined by microscopic analysis of Giemsastained cytopreparations (Leukostat; Fisher; Orangeburg, NY). Cells were allowed to adhere to the assay plates for 2 hr and nonadherent cells were removed by washing with CRPMI/lO% NCS, yielding adherent monolayers which contained at least 90% nonspecific esterasepositive cells, with viability exceeding 95%, as determined by trypan blue dye exclusion. The washed monolayers were then exposed to stimuli, and supernatants were collected at various times for measurement of TNF activity. In certain experiments, monolayers were depleted of macrophages by sequential I-hr incubations with 3ClO hybridoma-conditioned medium and an optimal dilution of rabbit complement at room temperature prior to stimulation. Tumor cell-binding assay. Binding of tumor cells to macrophageswas measured by inverted centrifugation of macrophage:tumor cultures as previously described ( 12, 13). Following 90 min incubation of “Cr-labeled H 125 tumor cells (5 X lo4 per well) on quadruplicate monolayers containing 2 X lo5 macrophages in polyvinyl chloride 96-well plates (Falcon; Oxnard, CA), the wells were filled with CRPMI/lO% NCS, sealed with adhesive-backed pressure-sensitive film (Falcon), and centrifuged in an inverted position for 5 min at various speedsat ambient temperature with a Sorvall H 1OOOBrotor on a Sorvall RT6000B centrifuge. After centrifugation, the assayplates were quickly frozen at -70°C the well bottoms containing the monolayers and adherent 5’Cr-labeled target cells were collected, and the 5’Cr activity associated with the well bottoms was measured. Control wells containing only labeled tumor cells were incubated in parallel to measure the binding of tumor directly to the plastic well bottoms. The number of tumor cells bound by macrophages was calculated by the following formula: No. tumors bound =
cpm on monolayer - control cpm X No. tumors added. total cpm added
High avidity (HA) binding was defined as the number of tumor cells bound following inverted centrifugation at 13OOg( 13). Cytotoxicity assay. Macrophage-mediated cytotoxicity was measured using an 1Shr 51Crrelease assay(12). Where indicated, the assaywas performed in the presence of 10 rig/ml LPS, which did not directly alter tumor cell or macrophage viability. Ten thousand “Cr-labeled H 125 tumor cells were added to quadruplicate wells containing 2 X lo5 macrophages. Following 18 hr incubation, the plates were centrifuged (50g for 1 min), and the 5’Cr activity in the cell-free supernatantswas measured.Spontaneous release of 51Cr from tumor cells was determined from wells containing only tumor cells. Total releasable 5’Cr from labeled tumor cells was determined by solubilizing tumor cells with 10 N NaOH. The percentage of killing was calculated using the following formula: Percentage killing =
experimental release - spontaneous release x 100%. total releasable - spontaneous release
TNF bioassay. TNF activity was determined by measuring cytotoxicity for TNFsensitive L929 cells (5). L929 cells were plated into 96-well flat-bottom microtiter plates (Costar; Cambridge, MA) at 6 X lo4 cells/well in CRPMI/lO% NCS containing 1 pg/ml actinomycin D. Serial twofold dilutions of samples or recombinant TNF
308
DEMARCO, ENSOR, AND HASDAY
standard were added in triplicate to the plates. Following an 18-hr incubation at 37°C the monolayers were washed, fixed with methanol, stained with 0.1% crystal violet, and the absorbance at 540 nm was measured. One unit of TNF activity caused 50% lysis of L929 monolayers (5). In selected experiments, samples were incubated with anti-TNF antiserum or unimmunized control rabbit serum for 60 min at ambient temperature prior to their addition to L929 cells. None of the macrophage stimuli or control reagents directly affected L929 viability. Molecular weight determination of macrophageactivatingfactors. Growth medium from H82 tumor cells was concentrated 50-fold by ammonium sulfate precipitation and resolubilization in PBS. One hundred-microliter aliquots of concentrated medium were applied to a TSK 3000 molecular exclusion column and subjected to HPLC at a flow rate of 0.5 ml/min while monitoring absorbance at 280 nm. One-quartermilliliter aliquots were collected, diluted 1:1 with CRPMI/20% NCS containing 100 U/ml polymixin B, and tested for the capacity to induce 0.5 X lo6 7-day cultured monocytes to releaseTNF during a 24-hr incubation. Statistics.Data are presented as the means f standard error (SE). In figures where error bars are not evident, the error bars were smaller than the symbols. Multiple comparisons among groups were tested using the Fisher PLSD test applied to a oneway analysis of variance ( 14). RESULTS
Expression of tumor binding and killing by cultured monocytes.Peripheral blood mononuclear cells (PBMC) comprising 58 f 4% lymphocytes and 42 + 3% monocytes were cultured in suspensionfor up to 14days (Table 1). The total number of recoverable viable cells decreasedby two-thirds over 14 days; however, the relative proportion of monocytes/macrophages and lymphocytes changed little during two weeksin culture. As previously described by Loike (7), monocytes acquired morphologic characteristics of mature macrophages during culture, specifically an increase in cell size with more extensive spreading on plastic surfaces. The capacity of freshly obtained PBM and cultured macrophages to bind and kill H 125 adenosquamous lung carcinoma cells was measured (Table 2). Killing of these TABLE 1 Changes in Numbers of Viable Cells during Culture Cell density (X lo6 per ml)
Culture day
Total viable cells
Macrophages
Lymphocytes
0 1 3 7 9 11 14
1+0 0.70 k 0.05 0.62 k 0.03 0.45 + 0.04 0.44 + 0.03 0.42 + 0.04 0.33 + 0.05
0.42 f 0.24 t 0.25 + 0.19 + 0.19 + 0.17 + 0.17 +
0.58 f 0.46 + 0.37 + 0.26 r 0.25 f 0.25 k 0.16 -t
‘MeanstSE,n=
12.
0.03 0.03 0.03 0.03 0.02 0.02 0.03
0.04 0.05 0.03 0.03 0.01 0.03 0.03
% Viable 98.3 + 97.8 + 96.2 + 94.7 f 93.0 + 95.3 + 90.9 -t
0.7 0.6 0.8 1.3 1.7 1.0 4.4
MACROPHAGE
309
TNP-a RELEASE INDUCED BY TUMOR TABLE 2
Tumor Binding and Tumor Killing Capacity of Monocytes/Macrophages during Culture Day 0
1 3 7 9 11 14
Tumor cells bound” 16.7 f 9.3 + 9.3 + 3.7 + 4.0 + 5.0 + 2.6 +
2.3 1.4d 0.9d 0.5”’ 0.7”’ 1.4”’ 0.6”’
% “Cr releasebno LPS 2.0 f 7.5 f 3.8 + 6.5 f 10.9 f 10.4 f 5.5 f
1.4 2.0 1.1 1.4 2.4’ 2.8 I1.7
% “Cr release with LPS’ 3.3 z!z2.1 18.6 + 4.4’z 12.0 + 2.2h 15.1 + 1.9”h 20.7 + 3.5da 18.9 t 3.4d 11.4 2 2.F
a Measured by inverted centrifugation assay and expressed as number of H125 cells (X 10’) bound to 2 X lo5 macrophages with high avidity; means + SE; n = 12. b 18 hr “Cr releaseassayat effector:target ratio of 20: 1. c 10 r&ml LPS. d P < 0.01 compared with Day 0. ‘P c 0.05 compared with Days I and 3. /P < 0.05 compared with Day 0. gP < 0.05 compared with same culture day assayedwithout LPS. h P < 0.01 compared with same culture day assayedwithout LPS.
tumor cells required viable macrophages. Incubation of H125 cells for up to 72 hr with 1000 U/ml recombinant human TNF did not cause detectable cell death (data not shown). Freshly obtained PBM expressedmaximal capacity for binding tumors, which subsequently declined over 2 weeks in culture. Tumor binding by these cells shared many characteristics of selective tumor binding expressedby murine peritoneal macrophagesand murine macrophage cell lines ( 12, 13), specifically: ( 1) a requirement for divalent cations; (2) inhibition by trypsin pretreatment of macrophages; (3) inhibition by cytochalasin B, an inhibitor of microfilament function; and (4) high binding avidity (data not shown). By contrast, freshly obtained PBM did not expresscytotoxic activity for H125 tumor cells. However, the tumoricidal capacity of monocytes increasedduring culture, peaking between Culture Days 9 and 11. Release of TNF by cultured macrophages. Freshly isolated PBM released little or no detectable TNF in the absenceof exogenous stimuli (Fig. 1). Cultured macrophages spontaneously released modest levels of TNF, suggestingsome macrophage activation occurs during these culture conditions. Both PBM and cultured macrophagesreleased considerable amounts of functionally active TNF when cultured with either 2 rig/ml LPS or membrane vesicles prepared from the K562 erythroleukemia cell line. None of the stimuli used in these experiments including all preparations of tumor cell membrane vesicles and crude and purified tumor growth supernatants contained any functionally active TNF (data not shown). The capacity of macrophages to release TNF increased during culture of PBMC, peaking from Culture Days 7 to 11, and persisting through Culture Day 14. Macrophages obtained from 7- to 1l-day PBMC cultures were used for all subsequentexperiments. In all cases,cytotoxic activity of macrophage supernatants for L929 cells was neutralized by anti-human TNF antiserum (data not shown). Treatment of the adherent PBMC with 3C 10 hybridoma supematant, which contains a monoclonal antibody against human macrophages, and complement abrogated tumor-stimulated TNF release(data not shown), indicating that macrophages were the source of the secreted TNF.
310
DEMARCO, ENSOR, AND HASDAY
FIG. 1. TNF releaseby freshly obtained monocytes and cultured macrophages.Freshly isolated adherencepurified blood monocytes or monocytes cultured for up to 14 days were incubated for 6 hr (5 X 105/well) with medium atone (w), 10 rig/ml LPS (@I),or 50 &ml K562 tumor cell membrane vesicles (PB).TNF activity in the culture supematantswas then measuredusing the L929 bioassay(5). * and ** denote significant differencescompared with levels of TNF releasedby Days 0 and 1 monocytes, with P < 0.05 and P < 0.10, respectively.
We have previously demonstrated that the TNF-inducing activity of tumor cells is predominantly associatedwith the cell membrane fraction (5). To minimize variability in the activity of the stimuli, we have routinely used membrane vesicles prepared from tumor and nontransformed control cells to stimulate macrophages in these experiments. However, viable tumor cells and tumor growth media also induced macrophages to release functionally active TNF (Tables 3 and 6). TABLE 3 Induction of Macrophage TNF Releaseby Viable Tumor Cells TNF activity released(U per million macrophages) Stimulus Medium Lymphb H82 H209 H209 myc H209 my&as A315 H125
Experiment 1
Experiment 2
1 7.9 254 563 433 141 0 0
38 0 181 558 545 188 0 0
a Macrophages incubated for 6 hr; effector.stimulator cell ratio 1. bConcanavalin A-stimulated allogeneic peripheral blood lymphocytes.
MACROPHAGE
311
TNF-cu RELEASE INDUCED BY TUMOR
TNF releaseby cultured macrophages was stimulated during 6 hr incubation with low concentrations of LPS, with half-maximal and maximal TNF releaseinduced by 1 and 10 rig/ml LPS, respectively (Fig. 2A). Induction of macrophage TNF releaseby cell membranes prepared from K562 tumor cells and H82, a small cell lung carcinoma cell line, demonstrated similar concentration dependence with peak TNF release induced by 12.5 pg/ml K562 or 25 pg/ml H82 (Fig. 2B). Somewhat lower levels of TNF were releasedduring incubation with higher concentrations of tumor cell membranes. The kinetics of TNF release induced by 2 rig/ml LPS or tumor membrane are shown in Fig. 3. Compared with LPS-induced TNF release,which peaked after 6 hr of stimulation, TNF releaseinduced by tumor cell membranes was more gradual and
A 0
0
20
40 Concentration
80 of LP!3 (nglml)
40 80 120 Concentration of Tumor Membrane
80
160 @g/ml)
100
200
FIG. 2. Concentration dependence of TNF release induced by LPS and tumor cell membrane vesicles. Peripheral blood monocytes cultured for 9 to 11 days were purified by adherence to plastic and incubated for 6 hr with various concentrations of (A) LPS or (B) cell membrane vesicles prepared from K562 (w) or H82 (0) tumor cell lines. TNF activity present in the culture supematants was measured using the L929 cytotoxicity. assay (5). * and ** denote significant differences compared with levels of TNF released by monocytes incubated in parallel without added stimuli, with P < 0.01 and P < 0.05, respectively.
312
DEMARCO, ENSOR, AND HASDAY
0
4
12 0 16 Incubation time (hours)
20
24
FIG. 3. Kinetics of TNF release by monocytes stimulated with LPS and tumor cell membrane vesicles. Peripheral blood monocytes cultured for 9 to 11 days were purified by adherence to plastic and incubated for up to 24 hr with medium alone (m), 10 @ml LPS (Cl), or 50 &ml cell membrane vesicles prepared from normal pooled donor peripheral blood leukocytes (WBC) (0) or K562 (A) or H82 (+) tumor cells. TNF activity present in the culture supernatants was measured using the L929 cytotoxicity assay (5) and presented as a percentage of each individual’s maximal TNF release over a 24-hr incubation. * and ** denote differenceswith P < 0.0 I and P < 0.05, respectively, compared with TNF activity present in culture medium at time 0.
sustained, continuing to increase over the 24 hr studied. Macrophages incubated in medium alone or with control membranes prepared from normal pooled mixed donor leukocytes released little or no functionally active TNF during 24 hr of incubation. Viable allogeneic peripheral blood lymphocytes (Lymph), stimulated to proliferate with 3 pg/ml concanavalin A for 96 hr, also failed to induce macrophages to release TNF (Table 3), suggesting that target cell proliferation was not sufficient to induce macrophages to releaseTNF. We have previously excluded a contribution by endotoxin contamination of tumor membranes in stimulating TNF release by murine peritoneal macrophages (5). The inability of membrane vesicles prepared from normal cells in parallel with tumor cell membrane vesicles to induce macrophage TNF release suggestedthat induction of TNF release by human macrophages was an intrinsic property of the tumor cell. However, we further evaluated the possible contribution of endotoxin contaminant to the observed induction of TNF by pretreating tumor membrane with polymixin B, which neutralizes many of the biologic actions of endotoxin ( 15) (Table 4). Whereas polymixin B inhibited LPS-stimulated TNF release by 80 to 90%, similar treatment of K562 and H82 membrane vesicles did not alter their ability to induce macrophages to release TNF. As we have previously demonstrated with murine peritoneal macrophages(5), tumor cell lines which grow in vitro in suspension stimulated TNF release by human macrophages, whereas tumor ceil lines which grow as adherent monolayers stimulated releaseof little or no TNF (Tables 3 and 5). Three small cell lung carcinoma cell lines, H82, a variant, and H146 and H209, two classic small cell carcinomas and a subline of H209 transfected with the c-myc oncogene, stimulated releaseof considerable levels
313
MACROPHAGE TNF-ol RELEASE INDUCED BY TUMOR TABLE 4 Polymixin B Neutralization of Macrophage Activation by LPS and Tumor Membrane No polymixin B
Stimulus” LPS 2 rig/ml LPS 5 rig/ml LPS 10 rig/ml K562 H82 WBC’
168 + 261 + 340 f 75.2 + 292 + 0
19 69.9 75 41.9 47.6
Polymixin B* 23.8 f 18.2 f 39.1 f 15.7 f 318 + 10 f
14d 12.7d 30.8d 30.6 68.4 8.2
a Incubated 6 hr with LPS or 50 &ml cell membrane from indicated cell; means ? SE; n = 3. * Stimulus preincubated with 100 U/ml polymixin B for 1 hr at room temperature. c Mixed donor peripheral blood leukocytes. d P < 0.05 compared with same stimulus preincubated without polymixin B.
of TNF by macrophages, whereas the non-small cell lung carcinoma cell line, H125, did not induce TNF release. Interestingly, a subline of H209 transfected with both cmyc and H-ras oncogenes and which no longer expressessmall cell markers (8), and grows in vitro as semiadherent clusters of cells, consistently induced releaseof lessthan one-third as much TNF as the nonadherent H209 parent cell line (Tables 3 and 5). Induction of TNF release by growth medium from tumor cells. Cell-free growth medium from tumor cell lines whose cell membranes stimulated macrophages to releaseTNF also induced TNF releaseby macrophages (Table 6). Moreover, growth medium from tumor cell lines whose cell membranes did not induce macrophages to release TNF and growth medium from nontransformed human embryonic lung fibroblasts stimulated release of little or no TNF by cultured macrophages. Induction of TNF release by tumor cell growth medium was concentration dependent with as TABLE 5 Induction of Macrophage TNF Releaseby Tumor Cell Membranes Stimulus”
n
Growth characteristics*
TNF activity released (U/ lo6 macrophages)
Medium WBC K562 H82 H146 H125 A315 H209 H209 myc H209 mycfras
36 11 21 7 8 17 4 2 2 2
Fresh Suspension Suspension Suspension Adherent Ahderent Suspension Suspension Semiadherent
25.6 k 7.9 14.6 + 6.8 505 f 1Old 345 Y!I99d 569 f 221d 12.8 + 5.9 6.2 + 6.0 560 f 3 489 f 56 148 + 8
0 Macrophages incubated for 6 hr with 50 &ml cell membranes from indicated cells. * Normal in vitro growth pattern. c Means -CSE. dP i 0.05 compared with medium, WBC, HI25, and A375.
314
DEMARCO, ENSOR, AND HASDAY TABLE 6 Effects of Growth Medium from Tumor Cells and Nontransformed Cells on TNF Releaseby Macrophages Stimulus”
n
Medium H82 KS62 H125 A375 Spleen’ HELd
9 14 14 8 3 8 3
TNF activity release (U/ 1O6macrophages)b 0 92lk
282'
1510 f 459e 9.7 f 2.9 0 45 *20 18.8 f 0.9'
a Cell-free growth medium collected from late-log cultures of indicated cells. b Macrophages incubated 6 hr with 1:4 dilution of growth media. ’ Concanavalin A-stimulated splenic lymphocytes from C3H/HeN mice. d Human embronic lung fibrobiasts. eP < 0.01 compared with TNF releasedby macrophages cultured in fresh medium.
little as a 1:32 dilution of K562 growth medium inducing 1O6macrophages to release over 600 TNF U during a 24-hr incubation (data not shown). The kinetics of TNF releaseinduced by tumor growth medium were identical to the kinetics of TNF release stimulated by intact tumor cells and tumor cell membrane vesicles (data not shown). Molecular weight determination of H82 tumor-associated macrophage activating factor. Growth medium from H82 tumor cells was concentrated 50-fold by ammonium sulfate precipitation and analyzed by molecular exclusion HPLC (Fig. 4). Recovery of activity following ammonium sulfate precipitation was virtually complete. HPLC demonstrated a single peak of TNF-inducing activity associatedwith a minor protein peak of approximately 40-kDa molecular weight, and accounting for 66% of the total activity applied to the column. DISCUSSION We have shown that human peripheral blood monocytes and macrophagesobtained from suspension culture of peripheral blood mononuclear cells release functionally active TNF when incubated with growth medium and cell membranes prepared from some leukemia and solid tumor cell lines, but not during culture with membranes or growth medium from nontransformed cells (Tables 3, 5, and 6; Figs. l-3). These results confirm our previous report of specific induction by some tumor cell lines of TNF releaseby murine peritoneal macrophages (5). These findings are also consistent with and extend the reported observations of Janicke and Mannel who described the induction by two human leukemia cell lines of TNF-specific mRNA accumulation and TNF releaseby freshly obtained human peripheral blood monocytes (6). In this study, we have further characterized the induction of TNF release by both human leukemia and solid tumor cell lines. We have shown that viable tumor cells, crude preparations of their cell membranes, and their growth medium all induce macrophages to releaseTNF with similar kinetics and in a similar concentration dependent fashion. We have previously demonstrated that TNF-inducing activity for
MACROPHAGE TNF-a RELEASE INDUCED BY TUMOR
100
6OC
‘I
z E “0 -c 4rm b
‘I 80
500
’ I
E
’ I A
0
5
10
315
’
I
’
I
I
k
i-5 20 25 Fraction Number
/
30
FIG. 4. Molecular exclusion HPLC analysis of concentrated H82 tumor cell growth medium. Growth medium from H82 late-log cultures of H82 tumor cells was concentrated 50-fold by ammonium sulfate precipitation, resuspended in PBS, and 100 ~1 aliquots were applied to a TSK 3000 HPLC column and eluted with PBS at 0.5 ml per min. Absorbance at 280 nm, was monitored, 0.25-ml aliquots were collected, and were added to 0.5 X lo6 Day-7 cultured monocytes in 0.25 ml CRPMI/20% NCS containing 100 U/ ml polymixin B. Following 24 hr incubation, TNF activity of cell-free supematants was measured using the L929 cytotoxicity assay (5). Molecular weights (kilodaltons) of protein standards are indicated by arrows.
murine peritoneal macrophages could be dissociated from tumor cell membranes by treatment with chaotropic agents (5). The demonstration of TNF-inducing activity in growth medium from some of the tumor cell lines described in this report suggests that some tumor cells may secrete or shed soluble factors which can induce macrophages to release TNF. There is presently insufficient data to determine whether the TNF inducing activity associated with tumor cell membranes results from directed insertion of these factors into the cell membrane (16) or the secondary binding of secretedTNF-inducing factors to the surface of the tumor cell. The macrophagesused in theseexperiments were obtained from 7- to 1l-day cultures of mixed peripheral blood mononuclear cells (7). Macrophages cultured under these conditions have been reported to acquire morphologic and biochemical characteristics of mature macrophages. We have shown that during culture these macrophages also acquire tumoricidal capability and enhanced capacity to releaseTNF spontaneously and in response to stimulation with either LPS or tumor cell membranes (Table 2; Fig. 1). In fact, the observed induction of TNF releaseby tumor may result from an interaction between tumor factors and activating factors generated during culture of the mononuclear cells. The time course of differentiation of human macrophages under these culture conditions differed from that of in vivo activated murine peritoneal macrophages (13, 17) in that the capacity for high avidity tumor binding by human cultured monocytes waned prior to expression of maximal tumoricidal capacity. Nonetheless, human macrophages cultured under these conditions for 7 to 11 days appeared to be optimally activated for tumoricidal activity and TNF release,both of which subsequently decreasedwith further culture. The abrogation of tumor-stimulated TNF releaseby pretreating human mononuclear cell monolayers with anti-macrophage
316
DEMARCO,
ENSOR,
AND
HASDAY
antibody (3C 10 hybridoma) and complement and our previously reported findings of tumor-stimulated TNF release by murine peritoneal macrophages (5) and murine macrophage cell lines (18) establish the macrophage as the source of the TNF. We have previously demonstrated that tumor-induced TNF releaseby murine peritoneal macrophages is rapid, peaking after 2 to 3 hr, and is self-limited with nearly undetectable levels by 6 hr after addition of tumor cell membranes (5). By contrast, TNF release by human macrophages was more gradual, progressively rising during 24 hr incubation with cell membranes from either H82 or IS562 tumor cells (Fig. 2). TNF releaseinduced by another small cell carcinoma cell line, H 146, followed a time course similar to that of H82- and K562-induced TNF release (data not shown). By comparison, LPS, a potent activator of macrophages, stimulated a more rapid release of TNF by human macrophages(Fig. 3), which is consistent with the rapid appearance of circulating TNF observed in patients with sepsisor in human volunteers who have received a single administration of bacterial endotoxin (19, 20). Macrophages obtained from 7- to 11day peripheral blood mononuclear cell cultures were sensitive to stimulation by low concentrations of LPS and tumor, with maximal TNF releaseinduced by 2 to 10 rig/ml LPS or as little as 12.5 pg/ml tumor membrane (Fig. 2). We had previously excluded endotoxin contamination of tumor membrane vesicle preparations as the actual stimulus of TNF releaseby murine peritoneal macrophages (5). Several lines of evidence suggestthat endotoxin was not responsible for the induction by tumor of human macrophage TNF releasereported here, including ( 1) low (co.1 rig/ml) concentrations of detectable endotoxin in tumor preparations, (2) absenceof TNF-inducing activity in nontransformed cell membranes and growth medium prepared in parallel with tumor cells, and (3) the failure of polymixin B to neutralize the TNF-inducing activity of tumor preparations. Whereas Janicke and Mannel (6) reported that monocytes stimulated with viable K562 erythroleukemia cells accumulated TNF-specific mRNA, but did not release detectable TNF, we found considerable TNF activity in supernatants from human macrophages cultured with membrane vesicles and growth medium from K562 cells (Figs. 1 and 2B; Tables 5 and 6). Janicke and Mannel presented evidence suggesting that viable K562 cells may obscure monocyte TNF releaseby removing the cytokine from the culture medium via TNF receptors on their surface (6). In this study, we stimulated macrophageswith isolated K562 cell membranes, which would be incapable of internalizing and recycling TNF:receptor complexes, thus limiting their ability to remove TNF from culture medium and masking induction of TNF release.In further support of this contention, Janicke and Mannel reported that a protein fraction partially purified from K562 cell membranes stimulated monocytes to releasedetectable TNF protein (6). As we have previously reported about tumor-induced TNF releaseby murine peritoneal macrophages, some tumor cell lines did not induce release of TNF by human macrophages (Tables 3, 5, and 6). Interestingly, all tumor cell lines which induced human macrophages to release TNF, including an erythroleukemia and three small cell lung carcinoma cell lines, grew as nonadherent suspension in vitro, whereas all tumor cell lines tested, which grew as adherent monolayers in vitro, failed to induce macrophagesto releaseTNF. Moreover, transfection of the H209 small cell carcinoma cell line with c-myc and H-ras oncogenes caused both a change in in vitro growth to a semiadherent pattern and a partial loss in TNF-inducing capacity (Tables 3 and 5). This association between in vitro growth patterns and macrophage-inducing activity
MACROPHAGE
TNF-a RELEASE INDUCED BY TUMOR
317
is similar to our previously reported observations of murine peritoneal macrophages (5) and suggeststhat a common characteristic of tumor cell membranes may stimulate macrophages and prevent adherence to plastic during growth. In further support of the inability of adherent tumor cell lines to directly induce macrophage TNF release, Webb and Gerrard (21) recently reported that the adherent A375 human melanoma cell line alone did not induce human monocytes to release TNF, but A375 tumor enhanced the induction of TNF releaseby interferon-y. Several cytokines have been reported to either directly stimulate macrophages to releaseTNF or prime macrophagesfor TNF release,including interleukin-3 (22), GMCSF (22), and interferon-y (22, 23). However, as we have previously reported (5), TNF-inducing activity of tumor membranes and supernatants is resistant to heat. We have further characterized the factor expressing TNF-inducing activity in H82 growth medium as a 40-kDa protein. The observed molecular weight and heat stability of this tumor-associated macrophage activator suggeststhat it may not be identical with any cytokines known to induce macrophage TNF release;however, these data do not exclude the possibility that tumors may induce macrophage release by secreting cytokines. Whereas TNF is cytotoxic for a narrow spectrum of tumor cells, this cytokine expressesa broad range of immunologic and physiologic modulating effects (24, 25), which can contribute to regulation of tumor immune defenseand affect the well-being of the host (26). In these experiments, H 125 adenosquamous lung cancer cells, which are resistant to direct killing by recombinant TNF, but which are efficiently killed by macrophages, failed to stimulate TNF release. Small cell carcinoma cell lines, which induce releaseof large amounts of TNF by macrophageswere not susceptible to macrophage-mediated killing (data not shown). This dissociation of TNF-inducing activity and sensitivity to macrophage-mediated killing supports the contention that TNF is primarily an important regulator of immunologic and physiologic responseto tumor, rather than an effector with direct, broad tumoricidal activity (27). We have demonstrated that human macrophages obtained from mixed peripheral blood mononuclear cell cultures selectively released TNF during culture with cell membranes or growth medium from some leukemia and solid tumor cells, but not from nontransformed cells. This process may play an important role in the immunologic and physiologic response of patients to neoplastic diseases. REFERENCES 1. Fidler, I. J., and Schroit, A. J., Biochim. Biophys. Ada 948, 151, 1988. 2. Nathan, C. F., J. Clin. Invest. 79, 3 19, 1987. 3. Chensue, S. W., Remick, D. G., Shmyr-Forsch, C., Beals, T. F., and Kunkel, S. L., Am. J. Puthol. 133, 564, 1988. 4. Beissert, S., Bergholz, M., Waase, I., Lepsien, G., Schauer. A., Pfizenmaier, K., and Kronke, M., Proc.
Natl. Acad. Sci. USA 86, 5064, 1989. Hasday, J. D., Shah, E. M., and Lieberman, A. P., J. Immunol. 145, 371, 1990. Janicke, R., and Mannel, D. N., J. Immunol. 144, 1144, 1990. Loike, J. D., Kozler, V. F., and Silverstein, S. C., J. Exp. Med. 159, 746, 1984. Mabry, M., Barges, M., Falco, J. P., Casero, P. A., Nelkin, B. D., Jasti, R., Baylin, S. B., Br. Cancer J., in press. 9. VanVoorhis, W. C., Steinman, R. M., Hair, L. S., Luban, J., Witmer, M. D., Koide, S., and Cohn, Z. A., J. Exp. Med. 158, 126, 1983. 10. Bradford, M., Anal. Biochem. 72, 248, 1976. 11. Boyum, A., &and. J. Clin. Lab. Invest. (Suppl.) 21, 1, 1968. 5. 6. 7. 8.
318
DEMARCO, ENSOR, AND HASDAY
12. Hasday, J. D., and Crawford, E. K., J. Immunol. Methods 114,243, 1988. 13. Somers, S. D., Whisnant, C. C., and Adams, D. O., J. Immunol. 136, 1490, 1986. 14. Winer, B. J., “Statistical Principles in Experimental Design,” 2nd ed. pp. 199-200. McGraw-Hill, New York, 1971. 15. Beutler, B., Krochin, N., Milsark, I. W., Luedke, C., and Cerami, A., Science 323,977, 1986. 16. Walter, P., and Lingappa, V. R. Annu. Rev. Cell Biol. 2,499, 1986. 17. Adams, D. O., and Hamilton, T. A., Annu. Rev. Immunol. 2,284, 1984. 18. Hasday, J. D., Shah, E. M., Harris, A. M., Walker, W. S., Roberson, S. M., and Giannini, S. H., Cell. Immunol. 136, 349, 1991. 19. Waage, A., Brandtzaeg, P., Halstensen, A., Kierulf, P., and Espevik, T., J. Exp. Med. 169, 333, 1989. 20. Michie, H. R., Manogue, K. R., Spriggs, D. R., Revhaug, A., O’Dwyer, S., Dinarello, C. A., Cerami, A., Wolff, S. M., and Wilmore, D. W., N. Engl. J. Med. 318, 1481, 1988. 21. Webb, D. S., and Gerrard, T. L., J. Immunol. 144, 3643, 1990. 22. Hart, D. H., Whitty, G. A., Burgess,D. R., and Hamilton, J. A., Immunology 71, 76, 1990. 23. Luedke, C. E., and Cerami, A., J. Clin. Invest. 86, 1234, 1990. 24. Talmadge, J. E., Phillips, H., Schneider, M., Rowe, T., Pennington, R., Bowersox, O., and Lenz, B., Cancer Res. 48, 544, 1988. 25. Tracey, K. J., Lowry, S. F., and Cerami, A., J. Infectious Dis. 157, 41, 1988. 26. Balkwill, F., Burke, F., Talbot, D., Tavemier, J., Osborne, R., Naylor, S., Durbin, H., and Fiers, W., Lancet 1, 1229, 1987. 27. Havell, E. A., Fiers, W., and North, R. J., J. Exp. Med. 167, 1067, 1988.