Soluble CD14 Activates Monocytic Cells Independently of Lipopolysaccharide

Abstract

The glycoprotein CD14 acts as a receptor for lipopolysaccharide (LPS), either when anchored in the myeloid cell membrane (mCD14) or as a soluble molecule (sCD14) in serum. sCD14-LPS complexes activate cells devoid of mCD14. However, the role of sCD14 independent of LPS is unknown. Therefore, the effect of sCD14 on monocyte functions was investigated in the monocytic cell lines THP1 and Mono Mac 6 and in fresh human monocytes. Under serum-free conditions, endotoxin-free human recombinant sCD14_1–348 (rsCD14_1–348) induced tumor necrosis factor alpha (TNF-α). The TNF-α effect was stronger in THP1 cells than in Mono Mac 6 cells or monocytes. It was dose dependent, with a maximum at 1 μg/ml, and time dependent, with a maximum after 2 h. sCD14 purified from urine had the same cytokine-activating capacity. In contrast, C-terminally truncated rsCD14_1–152 was inactive. The rsCD14 effect was not due to LPS contamination, since it was resistant to polymyxin and lipid IVa but sensitive to heat and trypsin. The rsCD14-induced cytokine induction was blocked by preincubation of rsCD14 with a monoclonal anti-CD14 antibody that did not recognize the LPS-binding site. Release of the TNF-α disappeared upon pretreatment of rsCD14 in 50% plasma or in complete, heat-inactivated or sCD14-depleted serum. Moreover, cytokine production was no longer observed when rsCD14 was pretreated with thrombocytes. The thrombocyte effect was dose and time dependent. In conclusion, sCD14 is able to activate myeloid cells, and the effect is prevented by the presence of plasma, serum, or thrombocytes.

In normal human blood, the soluble glycoprotein CD14 (sCD14) is present at a concentration of 2 to 3 μg/ml. It is increased in serum of patients with sepsis (20), with polytrauma (17), with atopic dermatitis (34), and with malaria (32). sCD14 has been found in urine of patients with a nephrotic syndrome (2) and in the bronchoalveolar lavage of patients with acute respiratory distress syndrome (24).

sCD14 binds not only lipopolysaccharide (LPS) (14, 26, 30) but also cell wall components of gram-positive bacteria (25). At low doses (10 to 100 ng/ml), sCD14-LPS complexes activate epithelial, endothelial, and vascular smooth muscle cells via a hitherto unknown receptor (10, 19, 23, 26). At high doses (≥10 μg/ml), preformed sCD14-LPS complexes can also activate monocytes and polymorphonuclear leukocytes (13). However, if sCD14 is added after LPS, it has an antagonistic effect on LPS-mediated activation of myeloid cells by competing with LPS for binding to membrane CD14 (14). sCD14 can bind monomeric LPS stoichiometrically at a 1:1 molar ratio (30), and it serves to shuttle LPS from its micellar LPS-binding protein (LBP)-bound form to high-density lipoprotein (12, 35). The lipid binding of sCD14 is not limited to LPS. sCD14 also binds other endogenous phospholipids in vitro; this process is catalyzed by LBP and accompanied by reciprocal transfer of LPS out of sCD14 (37). It is not known whether sCD14 physiologically interacts with endogenous lipid components in the blood or in the extracellular matrix or with other cells. Furthermore, its in vivo function is not yet clear. We therefore investigated the direct endogenous activity of sCD14 in the absence of bacterial components. We found that sCD14 was able to activate monocytic cells in the absence of plasma or serum. This activity was lost when the latter were present or when thrombocytes were added.

MATERIALS AND METHODS

Cells.

Mono Mac 6 cells were obtained from H. Ziegler-Heitbrock (Munich, Germany), and THP1 (TIB 202), HL 60, and U937 (CRL 1593) cells were purchased from the American Type Culture Collection. All cell lines were cultured in RPMI 1640 with 1 mM sodium pyruvate, nonessential amino acids, 15 mM HEPES, 0.2% sodium bicarbonate, 15 μg of gentamicin per ml, and 10% fetal calf serum; insulin (9 μg/ml) and transferrin (1 μg/ml) were added for the Mono Mac 6 cell culture. Forty-eight hours before stimulation with sCD14, the human monocytic cell lines were pretreated with 10⁻⁸ M 1α,25-dihydroxycholecalciferol (dihydroxyvitamin D₃; kind gift of E.-M. Gutknecht, Hoffmann-La Roche Ltd., Basel, Switzerland). Monocytes were purified from heparinized blood by Ficoll density gradient centrifugation. Thrombocytes were obtained from thrombophereses in the blood donor bank. Thrombocytes were washed in phosphate-buffered saline–0.1 M EDTA (pH 7.3) and freshly used. Erythrocytes were obtained by centrifugation of heparinized blood and elimination of the buffy coat. The SW620 epithelial cell line (human adenocarcinoma CCL227; American Type Culture Collection) was stably transfected with wild-type full-length human CD14 cDNA containing the sequence encoding the glycosylphosphatidylinositol (GPI) tail or with human CD14 cDNA in which the GPI tail was replaced by a transmembrane domain and the cytoplasmic tail of tissue factor (both plasmids were a kind gift of J. D. Lee, Scripps Research Institute, La Jolla, Calif.). The two constructs were cloned into the pRc/RSV vector (Stratagene, La Jolla, Calif.). Transfected cells were cultured in RPMI 1640 with 5% fetal calf serum and 500 μg of Geneticin (Boehringer, Mannheim, Germany) per ml.

sCD14 and LPS.

Human recombinant sCD14_1–348 (rsCD14_1–348) was produced in CHO cells as described previously (20). rsCD14 was harvested from the serum-free cell supernatant, concentrated by polyethylene glycol 15000, and purified by affinity chromatography using the oxidized anti-CD14 antibody 3C10 or 63D3, coupled to Affigel-hydrazide gel, and elution in glycine (pH 2.75). The N-terminal sequence of rsCD14 was identical to that published previously (27), and the C terminus was confirmed to be STLSVGVSGTLVL by high-pressure liquid chromatography and mass spectrometry of tryptic digests (9). Urine from a patient with nephrotic syndrome was precipitated with ammonium sulfate, and then sCD14 was purified with the same method as used for the recombinant protein. rsCD14 was analyzed by silver staining and Western blotting as previously described (20). rsCD14 concentrations were determined by an enzyme-linked immunosorbent assay (ELISA) developed in our laboratory, using 63D3 (2 μg/ml) as coating and peroxidase-coupled 3C10 (500 ng/ml) as detecting antibodies (19). LPS from Salmonella enterica serovar Typhimurium was a kind gift of C. Galanos (Freiburg im Breisgau, Germany). rsCD14 and all other reagents were tested for the presence of LPS by a chromogenic Limulus assay (Chromogenix, Mölndal, Sweden) and were used only if they contained less than 10 pg of endotoxin per ml. The endotoxin-free trypsin was from Promega (Madison, Wis.). Purified rsCD14 was aliquoted, kept at −70°C, and freshly diluted before use, except when complexes between rsCD14 and LPS were formed by overnight incubation at 37°C. For testing of eventual LPS contamination, rsCD14 was preincubated with either polymyxin B or lipid IVa for 10 min at 37°C and then added to the cells.

Cytokines.

Tumor necrosis factor alpha (TNF-α) was measured with an ELISA; the reagents were kindly provided by H. Gallati (Hoffmann-La Roche). In selected assays, TNF-α was measured by bioassay using cytotoxicity of WEHI-164.13 cells (8).

Reverse transcription-PCR.

Total RNA was isolated from guanidium-isothiocyanate-acetate lysates of THP1 cells by phenol-chloroform extraction. Single-stranded cDNA was generated with oligo(dT) (Gibco BRL, Basel, Switzerland) and Moloney leukemia virus reverse transcriptase (Superscript; Gibco). PCR was performed with the following oligonucleotide primers for TNF-α: 5′ ATG AGC ACT GAA AGC ATG ATC CGG 3′ (upstream) and 5′ GCA ATG ATC CCA AAG TAG ACC TGC CC 3′ (downstream). mRNA of β-actin was amplified as a control. The primers were 5′ CAC AGA GCC TCG CCT TTG 3′ (upstream) and 5′ TGG ATA GCA ACG TAC ATG 3′ (downstream). Deoxynucleoside triphosphate was from Gibco, and Taq polymerase was from Perkin-Elmer (Norwalk, Conn.). Twenty-two cycles at 95°C, 55°C, and 92°C were performed.

Immunofluorescence.

THP1 and Mono Mac 6 cells were stained according to a previously described method (18). Briefly, they were incubated for Fc receptor blockade with 5 μl of normal rabbit serum and then for 30 min at 4°C with the anti-CD14 antibody My4-fluorescein isothiocyanate (Becton Dickinson, Mountain View, Calif.) or an unrelated fluorescein isothiocyanate-conjugated immunoglobulin (Ig) as a control. Fluorescence and viability, assessed with propidium iodide, were measured in a FACScan (Becton Dickinson).

RESULTS

Kinetics of the CD14 effect on monocytic cells.

First, we performed activation experiments with rsCD14 and LPS under serum-free conditions. When we tested each component separately, we observed a TNF-α response with rsCD14 but not with LPS. This rsCD14 effect was time dependent and transient. The TNF-α release was maximal after 2 h in THP1 and Mono Mac 6 cells. After 24 h, TNF-α release was back to baseline (Fig. 1A). LPS at 1 ng/ml did not induce TNF-α under the same serum-free conditions (Fig. 1B). In purified human monocytes, TNF-α production increased to 1,527 ± 696 (mean ± standard error of the mean [SEM]), 2,091 ± 914, and 2,523 ± 1,281 pg/ml after 4, 6, and 16 h of stimulation with 1 μg of rsCD14 per ml. The increase in TNF-α protein was preceded by the appearance of TNF-α mRNA 30 min after addition of CD14 to THP1 cells. At this time, the TNF-α mRNA signals in unstimulated and LPS-stimulated cells were similar and barely detectable. After 2 h, the CD14-induced TNF-α mRNA was weaker and the LPS signal was unchanged (Fig. 1A, inset).

FIG. 1.

Kinetics of rsCD14- and LPS-induced TNF-α in human monocytic cells. THP1 cells (10⁵/well) or Mono Mac 6 cells were incubated for 0.5 to 24 h in serum-free medium (A) with rsCD14 (1 μg/ml) (A) or with LPS from S. enterica serovar Typhimurium (1 ng/ml) (B). Thereafter, TNF-α was determined in the supernatant. Means ± SEM of four experiments are shown. (Inset) Reverse transcription-PCR of TNF-α mRNA 30 min and 2 h after culture of THP1 cells in serum-free medium alone (lane 1), with rsCD14 (1 μg/ml) (lane 3), or with LPS (1 ng/ml) (lane 5). Lanes 2, 4, and 6 show β-actin as loading controls.

Dose-response curve of the sCD14 effect on monocytic cells.

rsCD14 caused a dose-dependent increase of TNF-α release in monocytic cells, with a plateau at 0.5 to 10 μg of rsCD14 per ml. The response was six times higher in THP1 cells than in Mono Mac 6 cells (Fig. 2A). Under the same serum-free conditions, even high concentrations of LPS did not induce any TNF-α release from THP1 cells and had only a weak stimulatory effect in Mono Mac 6 cells (Fig. 2B). In the presence of 5% serum, LPS caused the expected, dose-dependent TNFα response in both THP1 and Mono Mac 6 cells. Maximal TNF-α values of 1,500 and 3,000 pg/ml at 10 and 100 ng, respectively, of LPS per ml were obtained.

FIG. 2.

Dose-response curve of rsCD14- and LPS-induced TNF-α in human monocytic cells. THP1 cells (10⁵/well) or Mono Mac 6 cells were incubated for 4 h in serum-free medium with rsCD14 (0.1 to 20 μg/ml) (A) or with S. enterica LPS (0.1 to 1,000 ng/ml) (B). Thereafter, TNF-α was determined in the supernatant. Means ± SEM of three experiments are shown.

To determine whether the rsCD14 effect was modulated by LPS, rsCD14 was preincubated with increasing doses of LPS for either 10 min at room temperature or overnight at 37°C before addition to the THP1 cells. In the 10-min rsCD14-LPS mixtures, the TNF-α response remained the same as with rsCD14 alone (Fig. 3A). This finding indicates that functionally active rsCD14-LPS complexes did not form at the concentrations used within 15 min. When rsCD14 was preincubated with LPS overnight to allow complex formation (13), there was a slight enhancement of the CD14 effect by LPS, yet the maximum response was similar to that with 10-min rsCD14-LPS mixtures (Fig. 3B).

FIG. 3.

Effect of rsCD14 and LPS in combination. rsCD14 (20 μg/ml) was mixed with S. enterica LPS (20 ng/ml to 2 μg/ml) and preincubated either for 10 min at room temperature (A) or overnight at 37°C (B) before addition to THP1 cells (10⁵/well) at the final concentrations indicated. Cells were incubated for 4 h in serum-free medium. Thereafter, TNF-α was determined in the supernatant. Means ± SEM of three experiments are shown.

In purified human monocytes, the maximal TNF-α response with 1 μg of rsCD14 per ml was 1,542 ± 38 pg/ml; under the same serum-free conditions, LPS did not induce release of TNF-α.

rsCD14-induced monocytic cell activation was not limited to TNF-α; in both cell lines and monocytes, interleukin-6 (IL-6) was also released in a time- and dose-dependent way (data not shown).

Specificity of the CD14 effect.

Several experiments were performed to exclude LPS contamination of rsCD14 as a source of stimulation. First polymyxin B, which is known to bind LPS (7), was added to rsCD14. It did not modify the effect of rsCD14 in either THP1 or Mono Mac 6 cells (Table 1). Second, lipid IVa, which is known to be a strong LPS antagonist in human cells (11), was preincubated with rsCD14. It was unable to modify the effect of rsCD14 on TNF-α (Table 1). We confirmed, however, that lipid IVa was functional as an LPS antagonist in our system, since LPS-induced TNF-α in the presence of serum (TNF-α at 339 pg/ml) was totally inhibited by preincubation of LPS with lipid IVa (no TNF-α).

TABLE 1.

Effects of polymyxin B and lipid IVa on rsCD14-induced TNF-α production in human monocytic cells^a

Treatment	Dose	TNF-α (pg/ml)
		THP1	Mono Mac 6
Medium		0	0
rsCD14	1 μg/ml	1,699 ± 403	351 ± 34
Polymyxin B	10 μg/ml	0	9 ± 4
rsCD14 + polymyxin B	1 μg/ml + 10 μg/ml	2,317 ± 1,186	426 ± 56
Lipid IVab	1 μg/ml	0	ND
rsCD14 + lipid IVa	1 μg/ml + 1 μg/ml	2,002	ND
LPS	1 ng/ml	16 ± 16	24 ± 12
LPS + polymyxin B	1 ng/ml + 10 μg/ml	9 ± 9	12 ± 6

^aTHP1 or Mono Mac 6 cells were cultured for 4 h at 37°C without or with rsCD14 that had been preincubated with polymyxin B or lipid IVa for 10 min at 37°C. Thereafter, the level of TNF-α in the supernatant was determined. Results are means ± SEM of at least three experiments, except for treatment with rsCD14 plus lipid IVa, in which case two experiments were performed. ND, not determined. 

^bKind gift of H. D. Flad (Borstel, Germany). 

Third, rsCD14 was pretreated with a protease or heat. Both procedures are expected to destroy protein but not LPS. Preincubation of rsCD14 for 3 h at 37°C with 1 or 5 μg of trypsin abolished 18 or 81%, respectively, of its activity (Fig. 4A). rsCD14 was resistant to heating at 56°C but displayed only 25% of its activity after 30 min of heating at 75°C (Fig. 4B). Since LPS alone was not active in serum-free conditions, it could not be influenced by heat or trypsin. Trypsin alone did not activate THP1 cells. Silver stains of rsCD14 without and with trypsin treatment showed that the band of pure rsCD14 disappeared after addition of trypsin (data not shown).

FIG. 4.

Specificity of the effect of CD14. rsCD14 (1 μg/ml) or S. enterica LPS (1 ng/ml) was preincubated for 3 h at 37°C with 1 or 5 μg of trypsin (A) or for 30 min at 37, 56, or 75°C (B). Then the mixture was added to THP1 cells (10⁵) and incubated for 4 h at 37°C. TNF-α was determined in the supernatant. Trypsin alone did not induce TNF-α, nor did it modify the TNF-α standard curve. Means of two experiments (A) or means ± SEM of three experiments (B) are shown.

Fourth, rsCD14 was pretreated with anti-CD14 antibodies (big2, big6, big7, and big16) directed against epitopes unrelated to LPS (28). One of the antibodies (big16) prevented rsCD14-induced TNF-α release in THP1 (Fig. 5A) and Mono Mac 6 cells (data not shown). An isotype control antibody (HLA-DR, IgG2b) did not blunt the response to rsCD14. Pretreatment of the cells with the same big16 (IgG2b) anti-CD14 antibody did not alter rsCD14-induced TNF-α production, provided that the antibody was washed off before stimulation (Fig. 5B). This finding indicates that membrane CD14 was not involved in the action of sCD14. Similarly, TNF-α production after LPS stimulation in serum was not affected by pretreatment of the cells with either the big16 antibody or the isotype control (Fig. 5C). This result confirms that the big16 antibody was indeed not directed against the LPS-binding site of CD14. The big16 antibody not only abolished the effect of rsCD14_1–348 but also reduced the TNF-α response to truncated rsCD14_1–323 by 89% ± 3%. This result indicates that big16 binds CD14 proximal to amino acid 323.

FIG. 5.

Effects of antibodies on rsCD14-induced TNF-α release. (A) rsCD14 (1 μg/ml) was preincubated for 15 min at 37°C with the Ig2b anti-CD14 antibody big16 (kind gift of C. Schütt, Greifswald, Germany) or with an isotype control antibody and then added for 4 h at 37°C to THP1 cells (10⁵). Thereafter, TNF-α was determined in the supernatant. (B) The Ig2b anti-CD14 antibody big16 or an isotype control antibody was preincubated for 15 min at 37°C with THP1 cells and then washed off before addition of rsCD14 (1 μg/ml) in serum-free medium. (C) Identical to panel B but with addition of S. enterica LPS (1 ng/ml) in medium with 5% serum. Cells from panels B and C were also cultured for 4 h at 37°C before TNF-α was determined in the supernatants.

Effects of different sCD14 preparations.

To ascertain that the activating capacity was not a special feature of the purified rsCD14_1–348, the serum-free supernatant of CHO cells transfected with either CD14_1–348 or vector was directly applied to THP1 cells. The CD14-containing supernatant induced a strong response of TNF-α (Table 2); the supernatant harvested from the vector-transfected cells was inactive. In addition, we examined whether natural CD14 was also able to activate THP1 cells. To this end, CD14 purified from urine was tested for its monocyte-activating capacity. Urinary sCD14 caused a similar TNF-α induction in the THP1 and Mono Mac 6 cell line as rsCD14_1–348. To define the functionally active moiety of CD14, we compared the activity of our rsCD14_1–348 with those of two different rsCD14 preparations with a shortened C terminus. rsCD14_1–323 lacks the peptides which most probably contain the GPI-anchoring sequence (3). This protein caused only a sevenfold-lower TNF-α release than rsCD14_1–348. The half molecule rsCD14_1–152 containing the N-terminal half responsible for LPS binding (15, 31) had no stimulatory activity (Table 2). Thus, the C-terminal part of sCD14 was responsible for the release of TNF-α.

TABLE 2.

TNF-α liberation upon stimulation of monocytic cells with different sCD14 proteins^a

Treatment	Dose (μg/ml)	TNF-α (pg/ml)
		THP1	Mono Mac 6
Unstimulated		0	0
Supernatant of CD141–348-transfected CHO cellsb		2,689	ND
Urinary CD14	1	3,695 ± 980	391 ± 53
Purified rsCD141–348	1	4,215 ± 1,120	340 ± 115
rsCD141–323c	1	628 ± 80	118 ± 39
rsCD141–152c	1	19 ± 11	0

^aTHP1 cells were incubated for 4 h at 37°C with urine-derived purified sCD14 or the different molecular forms of rsCD14, and then the level of TNF-α in the supernatant was determined. ND, not determined. 

^bSupernatant of CHO cells transfected with vector alone did not stimulate THP1 cells. 

^cKind gift of R. Tapping, Scripps Research Institute. 

Natural inhibitors of the sCD14 effect on monocytic cells.

The strong TNF-α production by rsCD14 in vitro contrasts with the lack of TNF-α liberation by whole blood containing sCD14. Therefore, we looked for natural inhibitors, either cells or soluble factors. Based on the well-known interaction of monocytes with platelets (21), we tested the effect of platelets on the activity of rsCD14. After preincubation of platelets with 1 μg of rsCD14, the TNF-α release from added THP1 cells was reduced in a dose-dependent way (Fig. 6A). Platelets rapidly interacted with rsCD14, with a half-maximal inhibition at 12 min (Fig. 6A, inset). Thrombocytes did not bind but only inactivated rsCD14, because sCD14 was immunochemically measurable in thrombocyte supernatants but could no longer stimulate THP1 cells. We measured 586 and 943 ng of sCD14 per ml after a 2-h preincubation without and with thrombocytes, respectively. To investigate whether the inhibition required the presence of thrombocytes during THP1 cell stimulation, alternatively the preincubation was performed in rotating tubes, and the thrombocytes were pelleted before addition of the supernatant to THP1 cells. Under these conditions, thrombocytes were also able to inhibit the rsCD14-induced TNF-α. However, this inhibition was less complete; 82% of the rsCD14 activity still remained with 3 × 10⁷ thrombocytes, but it also dropped to 20% with 3 × 10⁸ thrombocytes. The reaction was slower, with 89, 83, 55, and 18% remaining after 10, 30, 60, and 120 min, respectively. As a control for the specificity of the thrombocyte effect, rsCD14 was also preincubated with erythrocytes; this treatment did not lead to any dose- or time (not shown)-dependent abolishment of the rsCD14-mediated activation of THP1 cells (Fig. 6B).

FIG. 6.

Thrombocytes as natural inhibitors of the effect of rsCD14 on monocytic cells. rsCD14 was preincubated for 2 h at 37°C with 3 × 10⁶ to 3 × 10⁸ thrombocytes (A) or with 10⁷ to 10⁹ erythrocytes (B). Thereafter, the mixture was added to THP1 cells (10⁵) and incubated for 4 h at 37°C. TNF-α was determined in the supernatant by bioassay; 100% TNF-α corresponds to 4,288 pg/ml. (Inset) Time curve of inhibition by thrombocytes (3 × 10⁸).

In addition, we tested serum and plasma as inhibitors. The cytokine-inducing capacity disappeared completely after a 2-h preincubation of rsCD14 in 25% plasma; it disappeared nearly totally after a 2-h preincubation of rsCD14 in 50% complete, heat-inactivated or sCD14-depleted serum (Fig. 7). These data indicate that the inhibitor is heat resistant and not consumed during coagulation. Moreover, it is not eliminated with sCD14 during the depletion procedure.

FIG. 7.

Plasma and serum as natural inhibitors of the effect of rsCD14 on monocytic cells. rsCD14 was preincubated for 2 h at 37°C in serum-free medium with increasing concentrations of plasma, complete serum, heat-inactivated serum, or sCD14-depleted serum; thereafter, the mixture was added for 4 h at 37°C to THP1 cells (10⁵). TNF-α was determined in the supernatant by bioassay; 100% TNF-α corresponds to 1,310 ± 37 pg/ml.

Effect of cell type and mCD14.

rsCD14 induced TNF-α in purified monocytes, in THP1 cells, and in Mono Mac 6 cells. In contrast, HL 60 and U937 cells did not respond, even after dihydroxyvitamin D₃ treatment (data not shown). The role of membrane CD14 (mCD14) was evaluated by three approaches. First, as shown in Fig. 5B, an antibody which could block rsCD14 in solution did not prevent rsCD14 activity when added to the cells. Second, mCD14 expression did not correlate with the response to rsCD14 (Table 3). Without dihydroxyvitamin D₃, mCD14 was nearly undetectable by fluorescence-activated cell sorting in THP1 cells, whereas Mono Mac 6 cells expressed it weakly (Table 3). After addition of dihydroxyvitamin D₃, the level of mCD14 was increased slightly in THP1 cells and strongly in Mono Mac 6 cells, yet the TNF-α response was much stronger in THP1 cells than in Mono Mac 6 cells (Table 3). Third, rsCD14 was applied to SW620 epithelial cells which had been transfected with either GPI-linked or transmembrane CD14 or with vector alone. IL-8 was measured after stimulation of the three cell types with rsCD14. None of them responded to rsCD14 alone (IL-8, less than 40 pg/ml). As a control, IL-8 production by LPS-rsCD14 complexes (100 and 1 μg/ml) was measured and found to be 323 ± 27 pg/ml in vector-transfected SW620 cells, 1,134 ± 172 pg/ml in SW620 cells with transmembrane CD14, and 871 ± 325 pg/ml in SW620 cells with GPI-anchored CD14.

TABLE 3.

Effect of dihydroxyvitamin D₃ on the TNF-α response to rsCD14 in human monocytic cells^a

Cell type	Dihydroxy- vitamin D3 (M)	mCD14 mean fluores- cence intensity	TNF-α (pg/ml)
THP1	0	4	69 ± 57
	10−8	42	1,077 ± 155
Mono Mac 6	0	29	229 ± 44
	10−8	185	431 ± 18

^aTHP1 or Mono Mac 6 cells were cultured for 2 days without or with 10⁻⁸ M dihydroxyvitamin D₃ and then seeded at 10⁵ cells/well and stimulated with rsCD14 (1 μg/ml). After 4 h at 37°C, the supernatant was harvested for TNF-α measurement. The results are mean values from three experiments. The cells were stained with CD14 antibodies by indirect immunofluorescence and analyzed in a flow cytometer. The mean fluorescence intensity acquired from the viable cells is indicated. 

DISCUSSION

The results of this investigation show that natural and recombinant sCD14 are able to directly stimulate monocytic cells in vitro and that this effect is blocked by incubation of sCD14 with thrombocytes or plasma. The activation of monocytic cells by sCD14 was clearly independent of LPS, as shown by multiple control experiments.

Previously, sCD14 was always studied in the context of LPS and viewed as the soluble receptor for bacterial glycolipids (25, 33). We obtained evidence that sCD14 activates monocytes and macrophages independently of LPS. The arguments outlined below include biochemical properties of sCD14, kinetics of the response, structure-function relationship, and blocking conditions.

Prior to the study, all solutions used in the production of sCD14 and sCD14 itself were found endotoxin free in the Limulus test. This result was also true for supernatants of CHO cells transfected with either vector or CD14-containing plasmids. However, THP1 cells were activated with the supernatant from the CD14-positive CHO cells but not from the vector-transfected CHO cells (Table 2). We also treated rsCD14 with polymyxin B and lipid IVa, two substances known to inhibit the action of LPS (7, 22). The resistance of the rsCD14 response to polymyxin B does not completely exclude LPS contamination, since polymyxin B does not inhibit the responses to all LPS preparations equally and also has other actions on cells independent of LPS such as the inhibition of protein kinase C (1, 4). In contrast, the persistence of the rsCD14 activation in the presence of lipid IVa is a strong argument that sCD14 was free of LPS. Lipid IVa was used as a competitive antagonist of LPS effects in many different human cells and shown to inhibit LPS-mediated signals at low concentrations without interfering with LPS-CD14 binding (5, 11, 16). In contrast, measures which inactivate proteins (protease and heat) but not LPS abolished the sCD14 effect (Fig. 4).

We performed our experiments with rsCD14 under serum-free conditions. This precludes an activation by LPS (Fig. 2), since LBP, catalyzing LPS transfer to sCD14, was lacking. In its absence, LPS-sCD14 complex formation can take place only slowly, i.e., within 1.5 h (12); the response to sCD14 occurred within 30 min at the mRNA level, and within 1 h TNF-α protein was strongly expressed (Fig. 1). In addition, the ratios of LPS and sCD14 used to obtain functionally active complexes were between 1:6 and 1:100 (35, 36). If we had indeed functional LPS in the active sCD14 preparations (0.1 to 10 μg/ml), it would have been easily detectable (1 ng/ml) in the Limulus test (sensitivity of 3 pg/ml). Also, mixtures of rsCD14 with LPS at up to 100 ng/ml did not induce a stronger TNF-α production than rsCD14 alone. Even after formation of complexes between rsCD14 and LPS by overnight preincubation, the TNF-α response was only slightly higher than with rsCD14 alone (Fig. 3), which indicates that rsCD14 did not activate cells via LPS-rsCD14 complexes. Our observation is in contrast to investigations by Hailman et al. (13), who found an IL-6 response in macrophages upon activation with rsCD14-LPS complexes but not with freshly mixed rsCD14 and LPS. Since CD14 binds many bacterial components beyond LPS (25), another contaminant may have mediated its effect. Hence, urinary and CHO cell-derived CD14 were equally active; it is unlikely that these two fluids contained the same contaminant.

Another argument against the contamination of sCD14 with LPS was the transient character of the TNF-α response. LPS-induced TNF-α in the presence of serum is detectable for more than 24 h (6), yet the rsCD14 effect disappeared after 16 h. The short-lasting effect after sCD14 application could be explained by its internalization or degradation and could indicate that the proinflammatory effect of this endogenous protein is tightly regulated.

Our structure-function analyses revealed that the C-terminal part strongly contributed to the activity of sCD14; namely, shortening of only 25 amino acids, from sCD14_1–348 to sCD14_1–323, caused an 85% loss of activity (Table 2). sCD14 binds to LPS with amino acids 39 to 44; amino acids 57 to 64 are near the binding site (15, 28). Truncation of sCD14 to the N-terminal moiety (amino acids 1 to 152), which still includes the LPS-binding site, led to disappearance of the effect (Table 2). Thus, sCD14-mediated monocyte activation did not involve the part of the molecule responsible for LPS binding.

The results of our blocking results are in line with the foregoing observation. We used four different antibodies mapping to epitopes beyond amino acid 152; one of them was able to neutralize rsCD14. This antibody acted only when it was pretreated with rsCD14, not when added to the cells prior to stimulation (Fig. 5). The latter observation indicates that blockade of mCD14 could not prevent the effect of rsCD14 and leads to the question of the receptor for sCD14. The results of the present study allow us to attribute several characteristics to this putative receptor. One is its tissue distribution. Only myeloid cells, not an epitheloid cell line like SW620, were responsive. Among the myeloid cells, only monocytes and two dihydroxyvitamin D₃-differentiated lines, THP1 and Mono Mac 6, not HL 60 or U937 cells, reacted after differentiation. The expression of mCD14 per se did not influence the response, since CD14-transfected SW620 cells could not be stimulated with rsCD14. Also, THP1 cells reacted much more strongly than Mono Mac 6 cells yet exhibited four times less mCD14.

Finally, there remains the question of the physiological relevance of our observation. The rsCD14 effect was found with recombinant material in serum-free medium. Under physiological conditions, CD14 is abundant in serum, yet no TNF-α is measurable. We therefore postulated a cellular or soluble inhibitor to explain the absence of stimulation in normal blood. Our data show that serum and plasma were inhibitory. This factor was characterized by the following properties: it was not consumed during coagulation, since it was also present in serum; it was heat resistant, since it withstood 56°C (as does CD14 itself [Fig. 4]); and it was not eliminated with CD14 during its depletion from serum by immunoaffinity chromatography, since CD14-depleted serum also inhibited rsCD14 activity. Serum was inhibitory; in contrast, urine apparently did not contain the inhibitor or lost it during the CD14 purification procedure, since urine-derived CD14 was a potent TNF-α inducer. This observation is in agreement with the report on a urinary factor copurified with sCD14 with cytokine-inducing capacity (29). It is unknown whether this soluble factor is identical with the inhibitory activity which we found associated with platelets. We found that preincubation of rsCD14 with thrombocytes precluded its effect on THP1 cells. We have shown that sCD14 was not absorbed by the platelets but that it was inactivated by a platelet component. Since sCD14 binds to phospholipids (37), possible inhibitory candidates are phospholipids. Finally, the structure-function relationship between the length of rsCD14 and its activity on THP1 cells can be related to the observation that rsCD14 is inactive in the presence of serum or platelets. This finding suggests that the latter components may inactivate CD14 by C-terminal degradation.

In conclusion, we found that sCD14 directly activates myeloid cells. Under physiological conditions, this activation is precluded by a platelet-derived, possibly lipidic inhibitor. The nature of this inhibitor and the pathological circumstances under which sCD14 could be liberated from the inhibitor and deploy its activity are now being investigated.