Autocrine Tumor Necrosis Factor (TNF) and Lymphotoxin (LT) α Differentially Modulate Cellular Sensitivity to TNF/LT-α Cytotoxicity in L929 Cells

Derivatives of the murine L929s fibrosarcoma cell line were used. L929sA, a TNF-sensitive subclone derived by limiting dilution, has a very low background of spontaneous TNF-resistant cells (<4 × 10⁻⁶ vs. 2 × 10⁻² for parental L929s cells; Vanhaesebroeck et al., 1992). The TNF-resistant L929r1.1 and L929r2 derivatives of L929s were obtained by TNF selection on L929s cells (Vanhaesebroeck et al., 1991). All L929 cell lines were cultured in DME (Life Technologies, Paisley, UK), supplemented with 5% newborn calf serum (Life Technologies), 5% FCS (Life Technologies), and antibiotics. The murine WEHI 164 cl 13 fibrosarcoma cell line (Espevik and Nissen-Meyer, 1986) was cultured in RPMI 1640, supplemented with 10% FCS and antibiotics. All cell lines were repeatedly found to be Mycoplasma-free as judged from a DNA fluorochrome assay.

Purified E. coli-derived recombinant mTNF, hTNF, and hLT-α were prepared in our laboratory and had a specific activity of 2 × 10⁸, 8.4 × 10⁷, and 3.1 × 10⁷ IU/mg protein, respectively. International standards for TNF quantification were obtained from the National Institute for Biological Standards and Control (Potters Bar, UK). Polyclonal rabbit antiserum directed against mTNF, hTNF, or hLT-α was provided by Mr. J. Van der Heyden (Roche Research, Ghent, Belgium). Polyclonal rabbit antisera directed against mTNF-R55 and mTNF-R75 were a gift of Dr. W.A. Buurman (University of Limburg, Maastricht, The Netherlands).

Supernatant of transfected cell lines was 100-fold concentrated using Centriprep-10 and Centricon-10 micro-separation devices (Amicon, Danvers, MA). TNF or LT-α were quantified in an 18-h cytotoxicity assay using WEHI 164 cl 13 cells in the presence of 1 μg/ml actinomycin D (ActD; Espevik and Nissen-Meyer, 1986). The detection limit of this assay was ∼1 pg TNF or LT-α/ml. Neutralization was performed with rabbit polyclonal antiserum specific for mTNF, hTNF, or hLT-α.

The CHL.hTNF and CHL.hLT-α fusion genes containing the membrane anchor structure of CHL (a gift from Dr. K. Drickamer, Columbia University, New York, NY), followed by mature hTNF and hLT-α, respectively, were constructed by fusion PCR (Olsen, 1992) using appropriate oligonucleotide primers. The PCR products were inserted as an EcoRI-Not fragment into an EcoRI-Not-opened pSV-SPORT1 vector (Life Technologies), which contains an E-tag COOH terminally from the inserted PCR fragments. The chimeric gene sequences were verified by DNA sequencing.

The pSV23S vector, driving the expression of a gene of interest by the constitutive SV-40 early promoter, was used in all experiments (Huylebroeck et al., 1988). Derived expression vectors were pSV23S-mTNF (Fransen et al., 1985), pSV23S-hTNF (Müller et al., 1986), and pSV23S-hLT-α (Vanhaesebroeck et al., 1992). The pSV2neo plasmid encoding the neomycin resistance (neo^r) gene under control of the SV-40 early promoter was used as a selection marker (Southern and Berg, 1982). Plasmid DNA was purified using pZ523 columns (5 Prime → 3 Prime, Boulder, CO). Stable transfection was performed by an improved DNA calcium phosphate coprecipitation protocol as described (Vanhaesebroeck et al., 1992).

TNF/LT-α sensitivity was determined as described (Vanhaesebroeck et al., 1991). In brief, cells were seeded in 96-well plates in 0.1 ml of medium at 2 × 10⁴ or 2,500 cells/well for 18- and 72-h assays, respectively. 12–18 h later, serial dilutions of TNF or LT-α, alone or in combination with either ActD (final concentration of 1 μg/ml), cycloheximide (CHX; final concentration of 25 μg/ml), or murine interferon (mIFN)-γ (200 IU/ml final concentration) were added in 0.1 ml of medium. After an 18-h (for ActD and CHX) or 72-h (for mIFN-γ) incubation period, the surviving cells were stained with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide and assayed colorimetrically (Tada et al., 1986). The percentage of surviving cells incubated in the presence of ActD, CHX, or mIFN-γ alone was never lower than 70% (vs. 100% for cells incubated with culture medium alone).

The presence of mTNF-R55 and mTNF-R75 was determined by incubation of cells with polyclonal rabbit antiserum against mTNF-R55 or mTNF-R75 (1 μg antiserum per 4 × 10⁵ cells in 200 μl), followed by incubation with biotinylated donkey anti–rabbit IgG polyclonal antiserum (Amersham Life Science, Amersham, UK), and treatment with phycoerythrin-conjugated streptavidin. All incubations were performed for 1 h at 4°C. Analysis was achieved by flow fluorocytometry using a Coulter Epics 753 equipped with an argon-ion laser (Coulter, Hialeah, FL). Acidic treatment to remove soluble TNF or LT-α from the cell surface receptors was done by incubation of the cells for 3 min at room temperature with glycine-HCl buffer (50 mM, containing 150 mM NaCl, pH 3.0).

The presence of membrane-bound hTNF or hLT-α was analyzed using monoclonal mouse antibody directed against the E-tag fused COOH terminally to mature hTNF or hLT-α (see Fig. 1; 1 μg antibody/4 × 10⁵/200 μl; Pharmacia Biotech, Uppsala, Sweden).

TNF was iodinated by using Iodo-Gen (Pierce Chemical Co., Rockford, IL) to a specific activity of 600 Ci/mmol. Binding studies were performed on L929r2 cells grown in suspension (Vanhaesebroeck et al., 1991). Cells were harvested and either untreated or incubated for 5 min at 4°C with glycine-HCl buffer to remove putative receptor-bound, endogenously produced TNF or hLT-α. Then they were incubated for 4 h at 4°C with a saturating concentration (2 nM) of ¹²⁵I-hTNF in 200 μl binding buffer (DME, supplemented with 10% FCS and 0.01% NaN₃). Cells were pelleted through a naphthalate solution containing a mixture of 1/3 dinonyl phthalate and 2/3 dibutyl phthalate; the retained radioactivity was determined in a γ-counter. Specific binding of ¹²⁵I-hTNF was determined by subtracting the amount of radioactivity bound to the cells in the presence of a 500-fold excess of unlabeled TNF from the radioactivity in the presence of ¹²⁵I-hTNF alone. Aspecific binding of ¹²⁵I-hTNF was about 15% of the total binding.

Previously, we demonstrated that in contrast to TNF, transfection of LT-α in TNF/LT-α–sensitive L929sA cells does not result in stable LT-α–producing cells (Vanhaesebroeck et al., 1992). However, transfection of genomic hLT-α/hTNF in L929s cells led to the production of both hTNF and hLT-α, excluding the possibility that this cell type is intrinsically unable to produce functional LT-α (Vanhaesebroeck et al., 1992). To determine whether coexpression of TNF or, more likely, TNF/LT-α resistance is necessary to efficiently express LT-α, we transfected hLT-α cDNA in TNF-resistant L929s derivatives. L929r1 cells constitutively produce low levels of TNF and are completely TNF-resistant, whereas non-TNF-producing L929r2 cells can be sensitized to TNF cytotoxicity by inhibition of transcription or translation with ActD or CHX, respectively (Vanhaesebroeck et al., 1991). After transfection of hLT-α cDNA in combination with the neo^r gene in L929r1 and in L929r2 cells, G418-resistant colonies were obtained (Table I), which were tested for the production of mTNF and hLT-α. In L929r1 cells, the cytotoxic activity in the supernatants of cells transfected with neo^r alone could be completely neutralized by antiserum against mTNF (Fig. 2,A), whereas antiserum against hLT-α was also required to completely block the cytotoxic activity of supernatants of clones obtained after cotransfection with the hLT-α gene (Fig. 2,B). The cytotoxic activity in supernatants of L929r2 transfectants was completely neutralized by antiserum directed against hLT-α (Fig. 2 C). These results indicate that both types of TNF-resistant cells can produce hLT-α. Clearly, efficient hLT-α production does not require TNF gene expression; a preexisting TNF/LT-α resistance is sufficient for L929 cells to become LT-α producers after transfection of the LT-α gene.

We further analyzed the effect of autocrine production of TNF and/or LT-α on TNF/LT-α sensitivity. As shown in Table I, transfection of TNF or hLT-α cDNA did not result in counterselection, neither in L929r1 nor in L929r2 cells. The TNF/LT-α–resistant phenotype of the L929r1 cell line was not influenced by autocrine hLT-α production (data not shown). Expression of mTNF levels (even minimal ones) in L929r2 cells conferred complete resistance to TNF/LT-α cytotoxicity in the presence of ActD, CHX, or mIFN-γ (Table II; data for CHX and mIFN-γ are not shown). Expression of hTNF induced only partial TNF/LT-α resistance (Table II), with a level of resistance proportional to the amount of hTNF produced in the supernatant, a phenomenon also observed in hTNF L929sA transfectants (Vanhaesebroeck et al., 1992). In contrast, no alteration in TNF/LT-α susceptibility was observed after LT-α production, not even by high LT-α production levels (Table II). So it is clear that the absence of TNF/LT-α resistance in LT-α–producing cells is not due to lower cytokine expression levels as compared with TNF-producing cells. This different effect of autocrine TNF vs. LT-α cannot be explained by a differential sensitivity of production of these cytokines to a treatment with metabolic blockers (Table III). ActD or CHX treatment for 18 h was found to have a similar effect on LT-α and TNF secretion in L929r2 transfectants. ActD treatment (at a concentration used in standard cytotoxicity assays) resulted in a twofold drop in cytokine expression levels, whereas treatment with CHX reduced the cytokine production levels by a factor of 20– 50 (Table III). These observations show that autocrine LT-α expression, in contrast to TNF, is not capable of modulating the TNF-resistant/susceptible phenotype of L929r2 cells.

Previously, we demonstrated a correlation between the induction of TNF resistance and the downmodulation of TNF receptors (Vanhaesebroeck et al., 1992). TNF-R55 and TNF-R75 expression in the different L929r2 transfectants was first measured by flow fluorocytometric analysis. In the fully resistant mTNF-producing L929r2 transfectants, no TNF-R55 or TNF-R75 expression could be detected (Fig. 3,B). On hTNF L929r2 transfectants, a downmodulation of TNF-R55 but not of TNF-R75 was observed (Fig. 3, C–D). This downmodulation was proportional to the amount of hTNF production (Fig. 3,C, low producer; D, high producer). In contrast, hLT-α–producing L929r2 cells showed no downmodulation of TNF-R55 or TNF-R75 (Fig. 3,E). Despite the apparent absence of detectable TNF-R55 on some hTNF transfectants, TNF could still induce cytotoxicity, which is mediated via TNF-R55 in L929r2 cells (Vercammen et al., 1995). This suggests that low numbers of functional TNF-R55 must be present on the hTNF-producing cells. To detect these and to confirm the data obtained with flow cytometric analysis on the expression levels of TNF-R55, we examined the presence of specific binding sites for ¹²⁵I-hTNF (TNF-R55) on mTNF-, hTNF-, or hLT-α–producing L929r2 transfectants. Specific hTNF binding could be detected on neo^r transfectants. This binding disappeared after treatment with glycine-HCl at pH 3 (Fig. 4), indicating that a low pH leads to the dissociation of TNF from its receptor. Pretreatment with glycine-HCl had no effect on subsequent hTNF binding, demonstrating that acid pretreatment did not remove the TNF-R55 molecules from the plasma membrane or had any adverse side effects. No specific ¹²⁵I-hTNF binding could be detected on mTNF transfectants, even not after pretreatment with glycine-HCl (Fig. 4). This is in agreement with the absence of receptor molecules found on mTNF-producing L929sA cells (Vanhaesebroeck et al., 1992). L929r2 transfectants producing high amounts of hTNF showed considerably reduced hTNF binding, which completely disappeared by glycine-HCl treatment (Fig. 4). Pretreatment of hTNF transfectants with glycine-HCl could increase ¹²⁵I-hTNF binding (Fig. 4), indicating that a fraction of the TNF-R55 molecules on the cell surface was occupied by secreted hTNF. It may be noted that ¹²⁵I-hTNF binding on hLT-α transfectants was completely similar to that of neo^r transfectants (Fig. 4), indicating that autocrine hLT-α production did not downmodulate TNF-R55 on the plasma membrane of L929r2 cells. In conclusion, induction of unresponsiveness to TNF/LT-α–mediated cytotoxicity is correlated with downmodulation of the TNF/LT-α receptor on the cell surface. Moreover, in contrast to mTNF and hTNF, hLT-α is not able to mediate such TNF-R55 downmodulation.

Since membrane retention of TNF is crucial for downmodulation of TNF receptors and induction of TNF/LT-α resistance (Decoster et al., 1998), the differential effect of autocrine-produced TNF as compared with LT-α can be explained by a different pathway of producing mature TNF and LT-α. Hence, it is possible that if LT-α would be produced as a membrane-bound form like TNF, it would also confer receptor downmodulation and TNF/LT-α resistance. To test this hypothesis, we exchanged the classical signal sequence of LT-α for the membrane anchor structure of CHL (Mellow et al., 1988); by choosing the latter instead of the TNF presequence we avoided any possible specific functional contribution of the TNF presequence. CHL is a trimeric, type II transmembrane liver glycoprotein (Chiacchia et al., 1984; Steer et al., 1990), which contains a membrane anchor allowing trimerization of the extracellular domain. Transfection of L929r2 cells with the neo^r gene, combined with the CHL.hLT-α chimeric gene (Fig. 1) or with the CHL.hTNF chimeric gene (used as a control), yielded normal numbers of G418-resistant colonies (data not shown). L929r2 cells transfected with CHL.hLT-α secreted between 1.25 and 30 IU hLT-α/ml; in the culture supernatant of CHL.hTNF transfectants between 1.5 and 14 IU/ml hTNF were detected. Expression of membrane-bound fusion protein could be revealed via flow fluorocytometric analysis in CHL.hLT-α as well as in CHL.hTNF transfectants (Fig. 5). These results clearly demonstrate that expression of either the CHL.hTNF or the CHL.hLT-α chimeric gene gives rise to production of a membrane-bound and secreted form comparable to the biosynthesis of wild-type TNF. We further analyzed the effect of autocrine-produced CHL.hLT-α or CHL.hTNF on TNF/LT-α sensitivity. As shown in Table IV, expression of these chimeric proteins induced unresponsiveness to the cytotoxic effect of TNF or hLT-α. This is in contrast to the effect of autocrine LT-α production and is comparable with the effect of autocrine-produced hTNF (Table II). Since induction of TNF/LT-α resistance is correlated with downmodulation of TNF receptors on the cell surface, we examined the expression level of receptors on the plasma membrane of L929r2 transfectants expressing CHL.hTNF or CHL.hLT-α fusion protein. These transfectants showed considerably reduced binding of ¹²⁵I-hTNF (TNF-R55) compared with neo^r transfectants (Fig. 6). Pretreatment with glycine-HCl buffer could increase ¹²⁵I-hTNF binding (Fig. 6), indicating that a fraction of the TNF-R55 molecules on the cell surface was occupied by secreted hTNF or hLT-α. This phenomenon was also observed in hTNF-producing L929r2 cells (Fig. 4). As expected, no downregulation of TNF-R75, determined via flow cytometric analysis, could be revealed on CHL.hTNF or CHL.hLT-α transfectants (data not shown). Taken together, we showed that membrane retention of hLT-α is crucial and sufficient for the induction of TNF/LT-α resistance and downmodulation of TNF-R55 on the cell surface. Furthermore, these results strongly indicate that the differential effect of autocrine-produced TNF and LT-α on TNF/LT-α sensitivity and downmodulation of TNF/LT-α receptors can be explained by their different way of processing for secretion.

Previously, we have demonstrated that after transfection in TNF/LT-α–sensitive L929s cells, TNF production induces resistance to the cytotoxic effect of both autocrine and exogenous TNF (Vanhaesebroeck et al., 1992). In contrast, transfection with the hLT-α cDNA gene did not allow to isolate clones that produced hLT-α or had become TNF/LT-α resistant. Normally, LT-α expression is restricted to lymphoid cell types (Paul and Ruddle, 1988). In the case of human tumor cell lines only those of lymphoid origin have been demonstrated to express and secrete LT-α after stimulation, whereas TNF can be produced by tumor cell lines of both lymphoid and nonlymphoid origin (Krönke et al., 1987). In this report, we demonstrate that nonlymphoid L929 murine fibrosarcoma cells can produce LT-α, but only when these cells are resistant to TNF/LT-α cytotoxicity. Remarkable differences were found between the effects of autocrine TNF and LT-α in TNF-resistant L929r2 cells. The latter cells can be sensitized to the cytotoxic action of TNF/LT-α by addition of ActD, CHX or mIFN-γ. But when tested under these conditions, autocrine mTNF production prevented the L929r2 cells to become sensitive. Autocrine hTNF production resulted in a partial resistance to TNF and LT-α under these conditions. In contrast, even high levels of autocrine hLT-α production did not alter the TNF/LT-α–resistant/ sensitive phenotype after addition of ActD, CHX or mIFN-γ. Since we demonstrated a correlation between induction of resistance and downmodulation of cell surface receptors after TNF production in L929s cells (Vanhaesebroeck et al., 1992; Decoster et al., 1998), we investigated whether TNF and LT-α differ in their capacity to downmodulate these receptors on the cell surface. As documented previously for L929s cells, both receptor types were completely downmodulated on L929r2 cells after autocrine mTNF production. hTNF and hLT-α L929r2 transfectants showed an unaltered expression level of TNF-R75 on their cell surface. This is in agreement with data indicating that hTNF and hLT-α only interact with mTNF-R55 (Lewis et al., 1991). Whereas hTNF transfectants showed considerable downmodulation of TNF-R55, autocrine hLT-α production had no effect on TNF-R55 expression levels. A possible lower binding affinity of hLT-α for mTNF-R55 might explain the inability of hLT-α to downregulate TNF-R55. However, this is very unlikely, since hLT-α seems to compete better than hTNF with ¹²⁵I-hTNF for binding on L929 cells (Browning and Ribolini, 1989). Moreover, hLT-α shows even a higher binding affinity than hTNF for binding on L929 cells (Hass et al., 1985; Vanhaesebroeck et al., 1992). A more likely hypothesis is that the different biosynthesis pathways of TNF and LT-α are at the basis of our observation. TNF is initially synthesized as a transmembrane TNF proform, which is then proteolytically cleaved at the cell surface, releasing the secreted, trimeric TNF form. In contrast, LT-α polypeptides are secreted in the lumen of the endoplasmic reticulum concomitantly with cleavage of the signal sequence. Membrane retention of TNF was shown to be crucial for the induction of TNF/ LT-α resistance and downmodulation of the cell-surface TNF receptor (Decoster et al., 1998). Moreover, we had already demonstrated previously that a functional interaction between TNF and TNF receptor is required to that end; expression of a biological inactive mutein with strongly reduced TNF receptor-binding capacity could indeed not induce TNF receptor downmodulation nor TNF resistance (Vanhaesebroeck et al., 1992).

To confirm the hypothesis that membrane retention of hLT-α is sufficient to downmodulate TNF/LT-α receptor expression and to induce resistance to the cytotoxic effect of exogenous TNF or LT-α, we analyzed the effect of expression of a CHL.hLT-α fusion gene. Like TNF, CHL is a type-II, trimeric transmembrane protein and the CHL membrane anchor allows trimerization of the extracellular domains. Expression of this CHL.hLT-α gene gave rise to both a membrane-bound and a secreted form of hLT-α, a feature characteristic of TNF synthesis. CHL.hLT-α transfectants exhibited TNF/LT-α resistance and showed downmodulation of TNF-R55, but not of TNF-R75. This is consistent with the fact that hLT-α can only interact with mTNF-R55 and not with mTNF-R75 (Lewis et al., 1991). The data obtained with CHL.hLT-α are in contrast with the results for hLT-α, expressed as a secreted protein, and are analogous to the results obtained with hTNF. Thus the inability of LT-α to induce TNF/LT-α resistance and downmodulation of the TNF/LT-α receptor is unambiguously explained by the absence of membrane retention of LT-α during processing for secretion.

It is remarkable that the mammalian genome codes for two closely related cytokines, i.e., TNF and LT-α, which have different biological functions and which are synthesized and processed by a different mechanism. The results here described illuminate some implications of this difference. First, our data suggest a possible dual biological role of membrane-bound TNF. The latter, produced by different cell types, such as macrophages, some T cells and some tumor cells, leads to primarily local inflammatory and/or immune reactions; but it is also involved in downmodulation of the TNF receptor, hence avoiding that TNF-producing cells become a target of their own product (Fiers, 1993). Second, the different effect observed for autocrine-produced TNF and LT-α might possibly explain why some cell types, such as HL-60, produce both TNF and LT-α; indeed, autocrine TNF downregulates the TNF/LT-α receptors and avoids counterselection. Third, tumor cells expressing TNF may have been selected to do so in order to avoid negative effects by TNF (or LT-α) released as an inflammatory/immune response of the host. Fourth, our observations might also partially explain why LT-α is mainly produced by cells of lymphoid origin, since such cells either also express TNF or are devoid of TNF-R55 receptors (Fiers et al., 1986; Beyaert and Fiers, 1998). Finally, it is quite possible that the production of LT-α during development corresponds mainly to formation of membrane-bound LT-α/LT-β complexes.

In summary, the observation that in contrast to LT-α, TNF expression protects L929 against TNF/LT-α–mediated cytotoxicity, is explained by the different pathway of processing of TNF and LT-α for secretion. Production of TNF or, as shown in this paper, of hLT-α as a membrane-bound form allows the producer cells to downmodulate TNF-R55 (and TNF-R75, depending on the ligand), such that they become unresponsive to their own product. This essential role of membrane anchoring is further supported by the fact that production of secreted TNF after substitution of the presequence for the signal sequence of IL-6 did not induce TNF resistance and downmodulation of the TNF receptors on the cell surface (Decoster et al., 1998), while membrane anchoring of LT-α in a similar construct as TNF did result in receptor downmodulation and resistance.

The authors thank A. Raeymaekers and Dr. P. Ameloot for cytokine preparations, J. Van der Heyden and Dr. W.A. Buurman for antisera, and Dr. K. Drickamer for CHL cDNA. W. Burm, D. Ginneberge, A. Meeus and M. Van den Hemel are acknowledged for technical assistance.

ActD

actinomycin D

CHL

chicken hepatic lectin

CHX

cycloheximide

hLT

human lymphotoxin

hTNF

human TNF

IL

interleukin

LT

lymphotoxin

mIFN

murine interferon

mTNF

murine TNF

neo^r

neomycin-resistant

TNF

tumor necrosis factor

TNF-R55

55-kD TNF receptor

TNF-R75

75-kD TNF receptor