Lymphotoxin α3 Induces Chemokines and Adhesion Molecules: Insight into the Role of LTα in Inflammation and Lymphoid Organ Development¹

Lymphotoxin (LT)⁴ plays an important role in both inflammation and lymphoid organ development. Although these effects of LT are well documented in vivo, based most recently on work in transgenic and LT-deficient mice (1, 2), the mechanisms by which LT promotes these processes are poorly understood. LT is a member of the LT/TNF family and can be produced as a secreted LTα3 homotrimer or a cell-associated LTα1β2 heterotrimer. The homotrimer interacts with the (p55) TNFRI and (p75) TNFRII receptors, accounting for those activities which are similar to those of TNF-α; LTα1β2 binds to the LTβ receptor (LTβR) and thus carries out unique activities (reviewed in 3 . It is possible that these ligands also bind to one of the several new receptors that are being identified. In fact, information derived from studies of mice deficient in LTα or LTβ suggests that LTα might also interact with an additional, as yet undescribed, LTαR to promote lymphoid organ development (reviewed in 4 . The LT/TNF family induces lymphoid organs through combinatorial associations of the ligands and their receptors. Analysis of mice deficient in the various ligands and receptors indicates that the LTα1β2 complex contributes to the development of peripheral lymph nodes (LN) and the organization of splenic architecture and germinal centers because mice deficient in LTβ or treated with LTβR fusion protein lack these structures (5, 6). Other aspects of lymphoid development, such as Peyer’s patch (PP) formation, are also mediated as an LTα3 complex through the TNFRI, because TNFRI^−/− mice either lack (7) or have abnormal (8) PP, whereas TNF-α^−/− mice have normal PP (8, 9). Mesenteric and cervical LN development may be mediated by an LTα-specific receptor or alternatively as complementary signaling between LTα3 and LTα1β2 through TNFR1 and LTβR respectively. The mechanisms of LT-induced lymphoid organ development have not been elucidated partly because there are no developmentally appropriate in vitro systems to study these effects, and partly because recombinant murine LTα has not been available for such systems.

LT has been implicated in inflammation from the time of its initial description as an in vitro correlate of delayed type hypersensitivity (10) and by its expression by Th1, but not Th2 T cells (11, 12). LTα induces inflammation in vivo when expressed under the control of the rat insulin promoter (RIPLT mice) at the sites of transgene expression in the pancreas and kidney (13). This occurs even in islets and kidneys of RIPLT.LTβ^−/− mice (1) indicating that LTα3 alone, in the absence of LTβ, induces this inflammation. Additional data suggesting a proinflammatory role for LTα derives from studies of experimental allergic encephalomyelitis (EAE) in its correlation with encephalitogenicity of myelin basic protein-specific T cell clones (14) and the fact that LTα^−/− mice are resistant to inflammation and clinical signs of EAE (15). The observation that LTβ^−/− mice can develop EAE supports the concept that LTα3, in the absence of LTβ, plays an important proinflammatory role (15).

The studies presented here were designed to identify the mechanisms by which LT promotes inflammation and lymphoid organ development. One possible pathway is through the induction of adhesion molecules in endothelial cells. Human (h) and murine (m) TNF-α can induce adhesion molecule expression in vitro (reviewed in 16 . Sikorski et al. have demonstrated that mTNF-α induces expression of VCAM-1, ICAM, and mucosal addressin cellular adhesion molecule (MAdCAM-1) on the murine endothelial cell line bEnd.3 (17). rhLTα induces expression of ICAM and E-selectin in human endothelial cells in vitro (18). Though no studies have been conducted with rmLTα, RIPLT mice exhibit up-regulated expression of ICAM-1, VCAM-1, MAdCAM-1, and peripheral node addressin (PNAd) in the vasculature of the inflamed pancreas and kidney (19). RIPLT.RAG2^−/− mice that lack the mature T and B cell infiltrate continue to express VCAM, ICAM, and MAdCAM-1 (19), suggesting that LT in the absence of additional cytokines from mature lymphocytes induces these molecules directly. RIPLT mice crossed to mice deficient in LTβ express ICAM, VCAM, and MAdCAM-1, but fail to express PNAd.⁵ Together these data suggest that LTα3 induces VCAM, ICAM, and MAdCAM-1, while LTα1β2 may be necessary for PNAd.

LT may also contribute to inflammation and lymphoid organ development through induction of chemokines. Several chemokines are involved in inflammation (20), and Burkitt’s lymphoma receptor 1 (BLR-1), a chemokine receptor expressed on B cells, appears necessary for B cell compartmentalization and LN development (21). TNF-α induces chemokines in vitro (22, 23, 24), and an effect in vivo is suggested by a study in which Abs to TNF abrogate chemokine induction in a lung model of inflammation (25). LTα1β2 has been reported to induce RANTES and IL-8 from a melanoma-derived cell line (26). There has been one report of RANTES induction by hLTα in human mesangial cells (27), but no such analyses have been reported for mLTα.

Recently, Mackay and colleagues prepared rmLTα that was cytotoxic to the WEHI 164.13 cell line (28). mLTα was less potent than mTNF-α in this study, leading the authors to question whether LTα in the absence of LTβ has any activity and to suggest that all the proinflammatory activity of Th1 cells could be ascribed to TNF-α and the developmental roles of lymphotoxin were mediated by the LTαβ complex (28). The present studies address whether LTα exhibits independent biological activity in an in vitro target, the murine endothelial cell line bEnd.3. This model system made it possible to address several issues: 1) can in vitro studies with murine LTα provide insight into the mechanism by which LT promotes inflammation and lymphoid organ development; and 2) does LTα induce these effects in the absence of LTβ? We show here for the first time that mLTα induces expression of adhesion molecules VCAM-1, ICAM, and E-selectin in murine endothelial cells. These activities are consistent with a proinflammatory role of LTα3. LT also induced expression of MAdCAM-1, an adhesion molecule expressed on the vasculature of all developing LN (29), mature mesenteric nodes, and PP. Neither LTα nor TNF-α induced PNAd. LTα also induced expression of the chemokines RANTES, IFN-inducible protein 10 (IP-10), and monocyte chemotactic protein 1 (MCP-1). These studies, which provide potential mechanisms for the roles of LTα in development and inflammation, reaffirm the concept that LTα, in addition to its activities in the LTα1β2 complex, plays unique roles in its LTα3 form in the crucial biologic processes of inflammation and lymphoid organogenesis.

rmLTα was obtained from Dr. Jeffrey Browning (Biogen, Cambridge, MA) (28). The soluble cytokine was produced in insect cells from a cDNA provided by our laboratory (30), purified, and collected as an eluate from a TNF-R55-Fc affinity column (31). rmTNF-α, which was produced in Escherichia coli, was obtained from Dr. Regina Turetskaya (Engelhardt Institute, Moscow, Russia). rhLTα was obtained from Genentech (South San Francisco, CA).

Units of biologic activity were measured by using the sensitive nonadherent WEHI 164 fibrosarcoma cell line in the 3-(4,5-dimethylthiazol-2-yl)-2,5,diphenyl tetrazolium bromide (MTT) dye reduction assay as previously described (32). This cell line and assay were included in this laboratory’s involvement in a study defining the International Standards for LT and TNF (33). WEHI 164 cells (5 × 10³/well) were added to serial dilutions of test samples in 96-well microtiter plates in a 100-μl volume and incubated at 37°C/5% CO₂ for a total of 48 h. After 44 h, 25 μl of 5 mg/ml MTT was added and 4 h later the cells were lysed by the addition of 150 μl 0.04 M HCL in isopropanol. The amount of MTT taken up and reduced was determined 24 h later by OD reading on a Vmax plate reader (Molecular Devices, Menlo Park, CA) at 570 nm with a 650-nm reference standard. Units were calculated as the highest dilution causing 50% cytotoxicity. Percent cytotoxicity was calculated as: 100 × (1.00 − OD of sample wells)/OD of control wells.

The murine brain endothelial cell line, bEnd.3 (34) was obtained from W. Risau (Max Planck Institute, Bad Neuheim, Germany). Cells were grown to confluence in RPMI 1640 (Life Technologies, Grand Island, NY) containing 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM l-glutamine, and 10% FCS (Life Technologies, Gaithersburg, MD). Cells were incubated with or without cytokines for various times, as indicated. For flow cytometry, cells were harvested with 0.25% trypsin and 1 mM EDTA, suspended in FACS buffer (PBS containing 1% BSA, 0.1% sodium azide, and 1% heat-inactivated rabbit serum), and 1 × 10⁶ cells were delivered per tube for immunofluorescence staining. Cells were incubated on ice for 15–30 min with primary Abs. Abs used were: murine anti-ICAM YN 1/1.7 (35) (hybridoma was purchased from American Type Culture Collection, Manassas, VA, and the supernatant was used at a dilution of 1:10), anti-murine VCAM (PharMingen, San Diego, CA; 0.5 μg), anti-murine MAdCAM-1 (MECA 367 diluted 1:5–1:10; generously provided by E. Butcher) (36); anti-murine PNAd (MECA 79, diluted 1:5; provided by E. Butcher) (37), and anti-murine E-selectin (PharMingen; 1 μg). In all experiments, rat IgG was used as a negative control (Jackson Immunoresearch, West Grove, PA; 1 μg). After incubation with primary Ab, cells were washed and incubated with goat anti-rat IgG conjugated with phycoerythrin (PharMingen, 0.5–1 μg) as secondary Ab. Samples were analyzed with a FACScan flow cytometer (Becton Dickinson, Mountain View, CA). Data were analyzed using CellQuest software (Becton Dickinson). The specific mean fluorescence for each sample was calculated by subtracting background (IgG).

bEnd.3 cells were treated with 100 U/ml of rmLTα for 4, 8, 16, or 24 h. Total RNA was isolated using Trizol (Life Technologies) according to manufacturer’s recommendations. Chemokine mRNA levels were determined using multiprobe RNase protection assay system (PharMingen). This system detects mRNA for lymphotactin, RANTES, eotaxin, macrophage inflammatory protein (MIP) -1β, MIP-1α, MIP-2, IP-10, MCP-1, and TCA-3 as well as the housekeeping genes L32 and GAPDH. ³²P-labeled riboprobes were synthesized from the plasmid template set using T7 polymerase after which the DNA template was digested with DNase. Five micrograms of total endothelial cell RNA were hybridized with the riboprobes in a 10-μl volume overnight at 56°C. RNA samples were digested and electrophoresed on a 5% polyacrylamide gel according to manufacturer’s recommendations. The level of mRNA expression was quantified by densitometric analysis using Imagequant software (Molecular Dynamics, Sunnyvale, CA).

Supernatants from LT-treated bEnd.3 cells were analyzed by ELISA for chemokine proteins. RANTES was detected with a Quantikine ELISA kit purchased from R & D Systems (Minneapolis, MN). MCP-1 was determined by sandwich ELISA. Nunc maxisorp ELISA plates (Fisher, Pittsburgh, PA) were coated with 4 μg/ml of hamster polyclonal anti-mouse MCP-1 Ab in 0.1 M NaHCO₃ buffer, pH 8.2, overnight at 4°C. Plates were washed three times with PBS and blocked with 2% BSA/PBS/0.05% Tween for 2 h at room temperature. Then, 100 μl of rmMCP-1 standard (PharMingen) or supernatant (either undiluted or diluted 1:2 in 2% BSA/PBS) were added to the plate and incubated overnight at 4°C. Plates were washed four times with PBS/Tween and biotinylated anti-murine MCP-1-detecting Ab (PharMingen) was added at 1 μg/ml and incubated for 45 min at room temperature. Plates were washed six times with PBS/Tween. Alkaline phosphatase-conjugated avidin (Zymed, South San Francisco, CA) was added at 1:1000 and incubated for 30 min at room temperature. Plates were washed eight times with PBS/Tween and developed with p-nitrophenyl phosphate (Pierce, Rockford, IL; 10 μg/ml) in diethanolamine buffer. Absorbance was quantified on a Vmax plate reader at 405 nm. Samples were analyzed in duplicate. IP-10 was not analyzed, as matching Ab pairs were not available.

A fundamental biologic effect of LT is cytotoxicity, so the activity of the rmLTα was analyzed first. The WEHI 164 assay was used to compare the cytotoxic activity of three preparations of recombinant-derived cytokines, mLTα, hLTα, and mTNF-α. The results (Fig. 1) indicate that all three cytokines induced complete killing of WEHI 164 cells. The preparations differed in potency in that 1.9 ng of mLTα was needed for 50% cytotoxicity, whereas comparable killing required only 1.2 pg hLTα or 0.19 pg mTNF-α. Treatment with mLTα revealed a comparable cytotoxicity in four separate experiments. Nevertheless, LTα induced cytotoxicity with starting material diluted as much as 1:20,000, indicating that this preparation of mLTα was very potent. This demonstration of mLTα3 cytotoxic activity validates the use of this material for evaluating new biologic activities of LT.

The ability of LT to modify adhesion molecule expression in endothelial cells was investigated to obtain insight into direct effects of LTα in promoting inflammation and lymphoid organogenesis. To establish sensitivity of murine endothelial cells to LTα, the ability of hLTα to induce expression of adhesion molecules was first examined using the murine endothelial cell line bEnd.3 (Fig. 2). hLTα (100 U/ml) stimulated expression of VCAM, ICAM, and MAdCAM-1. VCAM expression increased 6.7-fold above baseline levels within 4 h and persisted for 24 h. Similarly, ICAM expression increased 3.4-fold within 4 h and decreased slightly over 24 h. The most dramatic change in adhesion molecule expression was evident with MAdCAM-1. MAdCAM-1 expression increased 3.1-fold within 4 h and continued to increase over the 24-h assay period to a final increase of 15.5 times baseline levels. PNAd was not induced by hLTα at any time point. These observations establish that LTα can induce adhesion molecule expression in murine endothelial cells. The kinetics of induction are consistent with previous observations of dermal microvascular endothelial cells that demonstrate early induction of VCAM and persistent expression of E-selectin (38). The 18-h time point was chosen for subsequent experiments.

The ability of mLTα3 to modify adhesion molecule expression in endothelial cells was of particular interest to test whether this was one mechanism whereby mLTα promotes inflammation and lymphoid organogenesis. mLTα was evaluated for its ability to induce ICAM-1, VCAM-1, E-selectin, MAdCAM-1, and PNAd. mLTα (100 U/ml) induced VCAM, ICAM, and MAdCAM-1 expression; however, 500 U/ml was necessary for induction of E-selectin. A representative experiment with 500 U/ml mLTα is shown in Fig. 3. Although bEnd.3 cells constitutively express relatively high levels of ICAM-1, this level was increased 2.3-fold following mLTα treatment. Baseline expression of VCAM-1 was lower and was increased 25-fold following treatment with mLTα. E-selectin was induced 3-fold above baseline by mLT. MAdCAM-1 expression was dramatically induced by treatment with mLTα. At 100 U/ml, MAdCAM-1 was induced 14-fold over baseline. The induction was even more striking when cells were treated with 500 U/ml. This treatment induced MAdCAM-1 expression 26-fold above baseline. mLTα did not induce expression of PNAd at any of the doses tested (up to 1000 U/ml).

The relative ability of mLTα, hLTα, and TNF-α to induce adhesion molecules was then tested. All three cytokines induced VCAM, ICAM, E-selectin, and MAdCAM-1 but not PNAd (Fig. 4). The individual cytokines did differ in their effectiveness at inducing different adhesion molecules. The murine cytokines, mLTα and mTNF-α, were much more effective at inducing VCAM expression than the hLTα3. In contrast, all three cytokines were comparable in their ability to induce ICAM expression (2- to 2.5-fold induction). mTNF-α was slightly more effective at inducing E-selectin than either mLT or hLT. Importantly, this comparative study demonstrated that mLTα was the most potent inducer of MAdCAM-1 expression. In this experiment, mLTα induced MAdCAM-1 expression to a level three times higher than that induced by hLTα or mTNF-α. This experiment was conducted three times with treatments of 100 or 500 U, and mLTα was always more effective at inducing MAdCAM-1 than the other two cytokines. None of the three cytokines induced PNAd.

To gain further insight into the mechanisms by which LT promotes inflammation and lymphoid organogenesis, we analyzed the ability of rmLTα3 to induce chemokine mRNA and protein expression. bEnd.3 cells were treated with 100 U/ml rmLTα, and the chemokine expression pattern was analyzed. RNase protection analysis of bEnd.3 cells revealed a low constitutive accumulation of RANTES, IP-10, and MCP-1 mRNA (Fig. 5,A). These increased dramatically within 4 h of treatment with mLTα. These mRNA levels decreased through the observation period, but still remained elevated above untreated levels at 24 h. A similar time course and chemokine profile was seen in two separate experiments. Densitometric analysis was performed on a shorter exposure of the blot depicted in Fig. 5,A by normalizing the level of chemokine mRNA expression to the level of GAPDH housekeeping gene. This analysis confirmed LT-induced RANTES, IP-10, and MCP-1 mRNA expression peaked at 4 h, then decreased at 16h and 24 h (Fig. 5,C). Although RANTES, IP-10, and MCP-1 are the predominant chemokines induced by LT, upon overexposure of the autoradiogram, MIP-1β and MIP-2 mRNAs were also detected (Fig. 5,B, lane 7). This expression profile is distinct from that observed in LN. RNase protection analysis of that tissue revealed expression of a different chemokine array, namely lymphotactin, RANTES, eotaxin, MIP-1α, IP-10, and TCA-3 (Fig. 5 B, lane 8).

Chemokine protein was analyzed by evaluating the supernatants from LT-treated cells by ELISA. bEnd.3 cells produced a low constitutive level of RANTES and MCP-1 protein that increased within 8 h of treatment with rmLTα (Fig. 6) and continued to increase through the 24-h evaluation period.

The data presented here indicate that rmLTα3 has significant biologic effects in vitro. mLTα kills WEHI 164 cells and induces expression of adhesion molecules associated with inflammation (VCAM, ICAM, and E-selectin) and lymphoid organs (MAdCAM-1) and chemokine mRNA and protein in a murine endothelial cell line, bEnd.3. This is the first report that rmLTα has any effects other than killing. The studies have profound implications with regard to determining the mechanisms by which LT induces inflammation and lymphoid organ development and are supportive of recent work with LTα transgenic and knockout mice.

Though LTα was one of the first cytokines to be discovered (39, 40, 41), and certainly the first of the still expanding TNF family, analysis of its biologic activity has been hampered by the absence of adequate quantities of recombinant murine material that might allow absolutely unambiguous distinction between LTα and TNF-α. hLTα has been available, and, in fact, the first crystal structure analysis of a TNF family member was based on the interaction of hLTα (TNFβ) with TNFRI (42). Induction of cytotoxicity and adhesion molecule expression by rhLTα (18) and cytotoxicity with supernatants from murine Ag-activated CD4TH1 clones that accumulate LTα and TNF-α mRNA (3, 11, 32) are consistent with the conclusion that mLTα and mTNF-α carry out similar activities through TNFR1. This is also supported by the observations that the inflammation induced by LTα is qualitatively similar to that induced by TNF-α in transgenic mice (43) and that, at least in the case of LT, this occurs through TNFRI (1). In the current studies, we document the biologic activities of rmLTα, namely the induction of adhesion molecules and chemokines.

The data presented here clearly demonstrate that LTα3 alone is biologically active and induces inflammatory mediators independent of LTβ. This is consistent with previous publications from this laboratory indicating that inflammation in RIPLT.LTβ^−/− mice is quantitatively similar to that observed in RIPLT.wt mice (1). In addition, RIPLT.LTβ^−/− mice expressed the same chemokines expressed in the RIPLT.wt mice, namely RANTES, IP-10, and MCP-1, at the sites of transgene expression. These data indicate the the LTαβ complex is not required for LT-induced expression of these chemokines.⁵ The current study indicates that LTα is less potent than mTNF-α in signaling through TNFR1 as more LTα is necessary to induce 50% cytotoxicity. A quantitative, but not qualitative, difference between these cytokines is also observed in vivo by comparing the infiltrates of the islet in RIPLT and RIPTNF transgenic mice (43). The cellular composition of the inflammation is identical, but the RIPTNF islets contain more infiltrating cells. One important difference between these cytokines is that while TNF can either suppress or exacerbate the progression of nonobese diabetic insulin-dependent diabetes mellitus (44, 45, 46), LTα in our hands does not have this effect (A. Kratz et al., manuscript in preparation). This may be because LT has a lower sp. act., and in most situations LTα reaches biologic levels that are proinflammatory and prolymphoid organogenic, but not immunosuppressive.

The observations reported here confirm and extend in vivo studies on the role of LTα in inflammation (3, 19).⁵ They indicate that LTα alone, in the absence of an LTαβ complex, effectively induces expression of adhesion molecules VCAM, ICAM-1, E-selectin, and MAdCAM-1 and the chemokines RANTES, IP-10, and MCP-1. We have previously demonstrated that VCAM, ICAM, MAdCAM-1, and PNAd are induced in vivo in the RIPLT mouse model (19). The induction of adhesion molecules is consistent with the marked inflammation induced in the RIPLT mouse and other in vivo studies that strongly implicate LTα in the inflammation observed in EAE (14, 15). We have also recently demonstrated that the RIPLT transgene induces the same array of chemokines in vivo as the rmLTα3 induced in these studies.⁵ The array of chemokines induced by LTα is consistent with the type of cells that accumulate in RIPLT-induced lesions, namely mononuclear cells. All three of these chemokines attract T cells (47, 48, 49). In addition, MCP-1, and to a lesser extent RANTES and IP-10, also attract monocytes (47, 50, 51, 52).

The in vitro data reported here provide insight into the mechanism by which LT promotes lymphoid organ development through induction of adhesion molecules expressed on high endothelial venules. One of the most dramatic effects of LT in this study was its induction of MAdCAM-1, a marker of all developing LN (53), mature mesenteric LN, and PP (36), and also some instances of chronic inflammation (19, 54). The fact that mLTα does this more effectively than equal units of TNF-α is consistent with its crucial role in the development of LN and PP. The data are consistent with our in vivo observation that LTα induces MAdCAM-1 through TNFRI in the absence of LTβ,⁵ but may also act through another as yet undescribed receptor. This also suggests that LTα may use more than one receptor to induce MAdCAM-1 expression. It is already apparent through studies of knockout mice that LTα induces PP both through an interaction with TNFRI and the LTβR (5, 7, 8). Thus induction of MAdCAM-1 by LTα could contribute to these effects. TNF-α does not appear to contribute in any way to the development of LN or PP (8, 9) although in vitro it can induce MAdCAM-1 (Ref. 17, and data provided here). Perhaps it does not participate in LN development because it is not expressed at the appropriate time and place in embryogenesis. None of the preparations used here induced expression of PNAd from the bEnd.3 cell line. This is consistent with in vivo data indicating that RIPLT mice need LTβ to induce expression of that adhesion molecule and consistent with the fact that LTβ^−/− mice lack all peripheral LN (5), though they have mesenteric and cervical LN. A test of the hypothesis that LTαβ complex is necessary for peripheral LN in part through induction of PNAd would be to determine whether LTα1β2 induces PNAd in vitro.

LT may also contribute to lymphoid organ development through its induction of chemokines. Thus far only one chemokine, through its receptor, BLR-1, has been shown to play a role in lymphoid organ development, but it is likely that others will be described. Mice deficient in the gene that codes for blr1 have profound defects in B cell trafficking and lack inguinal LN and have missing or defective PP (21). It will be interesting to determine whether LT can induce BCA-1/BLC, the ligand for this receptor (55, 56), or other chemokines involved in lymphoid trafficking in development. The in vitro studies initiated here will continue to provide insight into the role of LT in inflammation and development by allowing the analysis of the molecular basis of its effects in these crucial and mechanistically related processes.

We thank Jeffrey Browning for rmLTα3 and Regina Turetskaya for rmTNF-α. We also thank Matthew Hanson for helpful advice with flow cytometry, Margot Iverson for technical assistance, Eugene Butcher for his generous gift of Abs, and Walter Risau for the gift of the bEnd.3 cell line.