Lymphotoxin α induces apoptosis, necroptosis and inflammatory signals with the same potency as tumour necrosis factor

Authors


Abstract

Both of the TNF superfamily ligands, TNF and LTα, can bind and signal through TNFR1 and TNFR2, yet mice mutant for each have different phenotypes. Part of this difference is because LTα but not TNF can activate Herpes Virus Entry Mediator and also heterotrimerise with LTβ to activate LTβR, which is consistent with the similar phenotypes of the LTα and LTβR deficient mice. However, it has also been reported that the LTα3 homotrimer signals differently than TNF through TNFR1, and has unique roles in initiation and exacerbation of some inflammatory diseases. Our modeling of the TNF/TNFR1 interface compared to the LTα3/TNFR1 structure revealed some differences that could affect signalling by the two ligands. To determine whether there were any functional differences in the ability of TNF and LTα3 to induce TNFR1-dependent apoptosis or necroptosis, and if there were different requirements for cIAPs and Sharpin to transmit the TNFR1 signal, we compared the ability of cells to respond to TNF and LTα3. Contrary to our hypothesis, we were unable to discover differences in signalling by TNFR1 in response to TNF and LTα3. Our results imply that the reasons for the conservation of LTα are most likely due either to differential regulation, the ability to signal through Herpes Virus Entry Mediator or the ability of LTα to form heterotrimers with LTβ.

Structured digital abstract

Abbreviations
BMDM

bone marrow-derived macrophage

cIAP

cellular inhibitor of apoptosis

ERK

extracellular signal-regulated kinase

GFP

green fluorescent protein

HVEM

herpes virus entry mediator

IκBα

inhibitor of κBα

JNK

c-Jun N-terminal kinase

LTα

lymphotoxin α

LTβ

lymphotoxin β

LTβR

lymphotoxin β receptor

LUBAC

linear ubiquitin assembly complex

MAPK

mitogen-activated protein kinase

MDF

mouse dermal fibroblast

MEF

mouse embryonic fibroblast

MLKL

mixed lineage kinase domain-like

Nec-1

necrostatin-1

NF-κB

nuclear factor-κB

PDB

Protein Data Bank

PI

propidium iodide

QVD

quinolyl-valyl-O-methylaspartyl-(2,6-difluorophenoxy)-methyl ketone

RIPK

receptor interacting protein kinase

TNFR

tumour necrosis factor receptor.

TNF

tumour necrosis factor

Introduction

The tumour necrosis factor (TNF) superfamily of cytokines regulate many physiological processes, including inflammation, proliferation, differentiation, and cell death [1]. They play pivotal roles in innate and adaptive immunity, as well as in the formation of lymphoid organs. Furthermore, excessive production of these cytokines can result in inflammatory and autoimmune diseases [2, 3], and may also contribute to tumorigenesis [4-6]. TNF (TNFSF2) is the eponymous member of the TNF superfamily of cytokines [7, 8], although its close paralogue lymphotoxin α (LTα, TNFSF1) was the first to be purified and characterized at the molecular level [9-11]. TNF receptor (TNFR)1 (TNFRSF1a) was identified as a receptor for both these ligands soon after their discovery [12].

TNF shares a common chromosomal location, 17qB1 in mice and 6p21.3 in humans, with LTα (Fig. S1A). Bony fish also have this neighbouring chromosomal organization, suggesting that a gene duplication event occurred before the divergence of fish and amphibians [13]. Despite their close kinship, the roles of TNF and LTα appear to be very different. TNF has emerged as a master inflammatory cytokine, able to regulate and amplify the production of a host of other inflammatory cytokines [14]. Surprisingly, however, given their similarity and the history of their discovery, LTα appears to have a far more restricted role.

The TNF superfamily of ligands bind as trimers to their cognate receptors; the TNF trimer binds and signals through both TNFR1 and TNFR2 (TNFRSF1b), whereas the LTα trimer (LTα3), in addition to binding these receptors, also binds and signals through herpes virus entry mediator (HVEM, TNFRSF14) [15]. LTα can also form heterotrimers with lymphotoxin β (LTβ, TNFSF3), and this LTα1LTβ2 trimer binds and signals through LTβR (TNFRSF3) [16, 17], but not through TNFR1 or TNFR2 [18]. There is no evidence for oligomerization of TNF with LTα or LTβ to signal through any known receptor.

The high degree of evolutionary conservation of the genes encoding these two proteins suggests they have nonredundant functions, and gene deletion experiments in mice have shown clear differences between the phenotypes of TNF and LTα knockout animals in the organogenesis of lymphoid tissues [19-21]. However, it is not clear whether the different phenotypes of the knockouts result from differential regulation of the two genes, from the different signalling capacities of the two homotrimeric ligands, from the singular ability of LTα to heterotrimerize with LTβ, or from a combination of these three factors. Another pertinent difference between the two ligands is that LTα3 is probably immediately processed to generate the ‘soluble’ trimer, whereas TNF is also processed but can in addition be found as a membrane-bound form. Both forms of TNF are biologically active; however, the membrane-bound form of TNF has been shown to have a higher affinity for TNFR2 [22], and TNFR1 is the main receptor for the soluble form of TNF [23].

Studies examining double and triple knockout mice of these three ligands and comparison with TNFR1 and LTβ receptor (LTβR) knockout mice have shown that a major role of LTα is to function as a heterotrimer with LTβ to activate the LTβR signalling that is required for development of the lymphoid organs [24-27]. In contrast, the TNF knockout mouse does not show major developmental abnormalities, but shows defects in innate immune responses [28, 29]. Furthermore, although the LTβ knockout mice show a broadly similar phenotype to the LTα knockout mice, there are clear differences, as cervical and mesenteric lymph nodes are absent in LTα knockout mice but present in LTβ knockout mice [30], suggesting that LTα3 might also have a role in lymphoid organogenesis. Further evidence of a physiological role for LTα3 was demonstrated in an experimental allergic encephalomyelitis model, where LTα knockout mice showed increased resistance as compared with LTβ-deficient or wild-type littermates [31].

Although both TNF and LTα3 are able to signal via TNFR1, most TNFR1 signalling studies have focused on signalling by TNF, with LTα–TNFR1 signalling being largely neglected. In the few studies that directly compared the ability of TNF and LTα to signal via TNFR1, differences in their signal strength or capabilities were found. In particular, it was shown that LTα is less able to promote TNFR1-induced cell death and nuclear factor-κB (NF-κB) activation [32], expression of cell surface markers [33] and cytokine production [34, 35] than TNF. In one study it was observed that NF-κB activation induced by LTα occurred in a monophasic manner, whereas TNF activated NF-κB in a biphasic manner [32]. Despite data suggesting that LTα is a weaker ligand for TNFR1, other evidence shows that LTα, but not TNF and LTβ, has crucial roles in the initiation and development of some inflammatory diseases, such as rheumatoid arthritis [36, 37], graft versus host disease, and the graft versus leukaemia effect [38, 39], and also lymphangiogenesis in inflammation [40], via TNFR1 signalling.

Our understanding of TNF signalling has advanced rapidly in the last 5 years, with major breakthroughs in understanding how TNF activates NF-κB, via cellular inhibitors of apoptosis (cIAPs) and the linear ubiquitin assembly complex (LUBAC), and how TNF promotes necroptosis in a manner dependent on receptor interacting protein kinase (RIPK)1, RIPK3, and mixed lineage kinase domain-like (MLKL) [41-45]. Given the new tools that these discoveries have provided to examine TNF signalling, it is timely to re-examine and compare LTα–TNFR1 and TNF–TNFR1 signalling.

Remarkably, despite the experimental focus on TNF–TNFR1 signalling in the literature, the structure of the LTα–TNFR1 complex has been resolved [46], but that of the TNF–TNFR1 complex has not. Using the crystal structure of the LTα–TNFR1 complex [Protein Data Bank (PDB): 1TNR] and the crystal structure of TNF–TNFR2 (PDB: 3ALQ), we computationally modelled the TNF–TNFR1 interaction. Our modelling revealed differences in the ways in which LTα and TNF interact with TNFR1; however, these differences did not suggest a significant difference in affinity. Numerous biological assays were performed in wild-type and knockout cells to assess whether these observed differences impact on the ability of these two ligands to activate NF-κB and induce cell death. Despite exhaustive testing of TNFR1 signalling, we were unable to observe a difference in the signalling outcome between TNF and LTα, and conclude that there is unlikely to be a significant difference between TNF and LTα signalling from this receptor. These results therefore imply that the reason for the high degree of evolutionary conservation of LTα is most likely either differential regulation, the ability to signal through HVEM, or the ability of LTα to form heterotrimers with LTβ.

Results

The LTα–TNFR1 and the TNF–TNFR1 interfaces

TNF and LTα share 40% identical amino acid identity (Fig. S1B), and both were able to computationally dock onto TNFR1 in a similar manner, which was consistent with the crystal structures of the LTα–TNFR1 complex (PDB: 1TNR) [46] and TNF–TNFR2 complex (PDB: 3ALQ) [47] (Fig. 1A). The models predicted some significant differences between the interacting face for the LTα–TNFR1 interaction and that for the TNF–TNFR1 interaction, but the sum of these interactions did not suggest greatly different affinities for TNFR1 of TNF and LTα.

Figure 1.

Overlay of the LTα–TNFR1 (yellow–orange) complex with the TNF–TNFR1 (green–cyan) complex. (A) An alignment of the α-carbons shows high similarity between the backbones of the two complexes. Closer analysis of the side chain orientations shows some differences. Highlighted in the top insert are two significant differences between the LTα–TNFR1 (yellow–orange) complex and the TNF–TNFR1 (green–cyan) complex that neutralize each other. First, there is a hydrogen bond between Asp50 of LTα (yellow) and TNFR1 (orange) that is not present in the TNF–TNFR1 complex. However, the TNF–TNFR1 complex has an edge-face interaction between Tyr115 of TNF (green) and Trp107 of TNFR1 (cyan) that is not present in the LTα–TNFR1 complex. (B) The bottom inserts shows how the larger, charged Asp143 of TNF (green), as compared with the smaller Val158 of LTα (yellow), causes significant side chain movement of TNFR1 residues Arg77, Lys78 and Glu79, which results in the changed hydrogen bonding pattern between the two complexes.

For example, although the backbone remains fairly stable, there are significant differences in the side chain movement of the TNFR1 residues Arg77, Lys78 and Glu79 between the LTα–TNFR1 complex and the TNF–TNFR1 complex (Fig. 1B). This is likely to be a result of differences in the way in which the side chain of Val158 in LTα and that of the equivalent residue in TNF, Asp143, interact with TNFR1. The larger, charged side chain of TNF causes the side chain of TNFR1 Arg77 to dramatically shift towards TNF, which, in turn, forces the side chain of the neighbouring TNFR1 Lys78 away from TNF. This leads to different hydrogen bond networks in the two complexes. For the LTα–TNFR1 complex, there are two hydrogen bonds present in this area: the first is between Ser82 (LTα) and Glu79 (TNFR1), and the second is between Pro155 (LTα) and Lys78 (TNFR1). For the TNF–TNFR1 complex, Lys78 (TNFR1) forms the only hydrogen bond present in this region, with Glu23 (TNF). Although there is loss of a hydrogen bond, this would not be expected to significantly change the affinity.

Another example is LTα Asp50, whose side chain forms a hydrogen bond with TNFR1 Ser72. The equivalent residue in TNF, Ala33, does not have a side chain, and therefore cannot form the hydrogen bond (Fig. 1A). However, further along the interface, TNF Tyr115 is able to form an edge-face interaction with Trp107. The equivalent LTα residue, Leu130, does not form a ring, and is thus unable to do this (Fig. 1A).

Both TNF and LTα are active as trimers. We generated Fc-tagged forms of hTNF and hLTα3 for this study, to allow easy purification, detection, and comparison. Because the Fc component promotes the formation of dimers of the trimeric form [48], we compared the oligomeric status and purity of our Fc ligands with those of commercially available trimeric counterparts, using a blue native gel followed by detection with western blotting (Fig. S1C). The commercially available trimers have a predicted molecular mass of ~ 50 kDa (LTα is slightly larger than TNF), whereas an Fc trimer has a predicted molecular mass of ~ 150 kDa. In two experiments, the commercially available trimeric forms ran anomalously at ~ 150 kDa and in a smear, which may be the result of the poor resolving capacity of the gel at lower molecular masses. The Fc-trimeric forms, on the other hand, ran at clearly defined sizes that might be consistent with trimers, dimers of trimers, and trimers of trimers. TNF and LTα3 showed similar patterns for their oligomeric state, both in their Fc-tagged and in their untagged forms.

LTα induces apoptosis and necroptosis equivalently to TNF

Our modelling suggested that TNF and LTα3 might signal differently through TNFR1, which would be consistent with earlier literature. Smac mimetics, which antagonize cIAPs, sensitize cells to TNF-induced caspase-dependent apoptosis [49] and RIPK3-dependent necroptosis [43]. To test the role of LTα in inducing apoptosis and necroptosis, we treated cells with TNF or LTα together with smac mimetics that sensitize cells to TNF-induced cell death. As previously reported [49], the smac mimetic compound A sensitized transformed mouse embryonic fibroblasts (MEFs) to TNF-induced cell death as measured by propidium iodide (PI) uptake and flow cytometry (Fig. 2A). Compound A also sensitized MEFs to LTα-induced death to a similar extent (Fig. 2A). The TNFSF member TWEAK is able to deplete cIAPs in a similar fashion to smac mimetics, and, like smac mimetics, sensitizes cells to TNF-induced cell death [49-51]. Wild-type transformed MEFs (Fig. 2A), primary mouse dermal fibroblasts (MDFs) (Fig. 2B) and the human leukaemia cell line U937 (Fig. S2A) showed comparable susceptibility to cell death induction by TNF or LTα in combination with a smac mimetic or TWEAK.

Figure 2.

TNF and LTα induce TNFR1-dependent cell death similarly. (A, B) The indicated cell types were stimulated or not stimualted with Fc–TNF (100 ng·mL−1) or Fc–LTα (100 ng·mL−1), with or without TWEAK (TW; 100 ng·mL−1) or smac mimetic (SM; 500 nm). (C) Wild-type (WT) MEFs were stimulated with various doses of either Fc–TNF or Fc–LTα with or without SM (500 nm) for 24 h, and cell death was then measured by PI staining and flow cytometry. (D) MEFs from the indicated strains were treated with 100 ng·mL−1 Fc–TNF or Fc–LTα or LTα. (E) Transformed MDFs from the indicated strains were pretreated with or without Nec-1 (50 μm) and/or QVD (10 μm) for 1 h, and then stimulated with a combination of 100 ng·mL−1 Fc–TNF or Fc–LTα and 500 nm SM. All of the stimulations were carried out for 24 h, and viability was measured by PI staining and flow cytometry: (A) N = 5; (B) N = 3 biological repeats; (C, D) N = 3; (E) N = 5 biological repeats for WT and three biological repeats for RIPK3−/− and MLKL−/−. Error bars show standard error of the mean.

It was possible that the routinely used dose of 100 ng·mL−1 TNF or LTα saturated all TNFR1s, thereby preventing subtle differences in signal strength from being detected. We therefore performed a dose–response experiment in wild-type MEFs, using a constant concentration of smac mimetic that is not toxic to MEFs by itself. However, the sensitivity of MEFs to limiting doses of either LTα or TNF and smac mimetic was indistinguishable (Fig. 2C).

TNF signalling is known to be regulated by cIAPs, TRAF2, RIPK1, and sharpin, a component of LUBAC. To investigate whether these proteins were also required to regulate LTα signalling, we compared TNF–LTα signalling in MEFs with deletions or mutations of the genes for these proteins. cIAP1 knockout [49], cIAP1/2 double knockout [52, 53], Traf2−/− [54, 55] and sharpin mutant cells [56] were all sensitive to TNF-induced death, as previously reported, and were equally sensitive to LTα (Fig. 2D). Ripk1−/− MEFs are resistant to TNF-induced death [57], and were equally resistant to LTα-induced cell death (Fig. 2D). As expected, the control Tnfr1−/− MEFs were resistant to both TNF-induced and LTα-induced killing, and Tnfr2−/− MEFs behaved like wild-type cells (Fig. 2D). TNFR1 is ubiquitously expressed, whereas TNFR2 is expressed on a limited number of cell types, and has not been reported on fibroblasts. Consistent with the notion that, in fibroblasts, we are monitoring TNFR1-dependent signalling, primary TNFR1 knockout MDFs were resistant to cell death induced by TNF or LTα + smac mimetic, whereas wild-type and TNFR2-deficient MDFs were sensitive (Fig. S2B). It was possible that the higher-order oligomers of TNF and LTα increased the activity of the two ligands; however, in a direct comparison, both Fc and commercially available trimeric forms induced cell death to the same extent in both cIAP1/2 double knockout MEFs (Fig. S2C) and smac mimetic-treated MDFs (Fig. S2D). Our results therefore suggest that the life–death decision of a cell is regulated in the same way by LTα and TNF.

Treatment of cells with TNF plus smac mimetic induces apoptotic cell death, and this is therefore not blocked by the RIPK1 inhibitor necrostatin-1 (Nec-1). However, when TNF/smac mimetic-induced cell death is blocked by a caspase inhibitor such as quinolyl-valyl-O-methylaspartyl-(2,6-difluorophenoxy)-methyl ketone (QVD) [58], an alternative cell death mechanism, termed necroptosis, is activated. Necroptosis is dependent on the kinase activities of RIPK1 and RIPK3 and on the presence of the pseudokinase MLKL [43]. To determine whether TNF and LTα are able to induce necroptosis equivalently, we pretreated wild-type cells with the RIPK1 inhibitor Nec-1, prior to adding the necroptosis stimulus of TNF or LTα, smac mimetic, and QVD. As expected, wild-type MDFs were killed by TNF/QVD/smac mimetic, and the percentage of cell death could be reduced by Nec-1 pretreatment (Fig. 2E). Nec-1 protection was incomplete in two biologically independent sets of MDFs and irrespective of whether Fc or commercially available trimeric ligands were used (Figs 2E and S2D). These results show that the oligomeric state of the ligand is not a significant factor in whether cells undergo necroptosis, and that inhibition of RIPK1 by Nec-1 is not sufficient to completely block TNF-induced necroptosis in MDFs, in agreement with a recent report of RIPK1-independent form of necroptosis [59]. To test whether LTα-induced necroptosis also required RIPK3 and MLKL, we treated transformed MDFs derived from wild-type, Ripk3−/− and Mlkl−/− mice with TNF and LTα necroptotic stimuli (Fig. 2E). As previously reported, Ripk3−/− MEFs were completely resistant to the cytotoxic stimulus of TNF/QVD/Smac mimetic, and they were also not killed by LTα/QVD/smac mimetic (Fig. 2E). We have recently generated MLKL-deficient mice [60], and, consistent with reports showing that knockdown of MLKL prevents TNF-induced necroptosis [43, 61], MLKL knockout cells were also resistant to both TNF-induced and LTα-induced necroptosis (Fig. 2E).

LTα activates NF-κB by assembling a similar signalling complex with TNFR1 as that formed by TNF

Whereas our data indicate that TNF and LTα induce TNFR1-dependent apoptosis and necroptosis equivalently, the usual outcome of TNF signalling is inflammatory cytokine production, in large part driven by activation of the transcription factor NF-κB. Activation and nuclear translocation of the heterodimeric p50/p65 NF-κB occurs following TNF-induced degradation of the inhibitor of κBα (IκBα) protein, which normally retains NF-κB in the cytoplasm [62]. TNF-induced activation of NF-κB occurs in a phasic manner [63], because IκBα is one of the main targets of activated NF-κB, and rapidly shuts down NF-κB transcription. Provided that TNF is continuously present, this feature of the signalling pathway results in waves of IκBα degradation and NF-κB activation, with the first two degradations being the most clearly distinguished. It has been reported that LTα, on the other hand, induces only a single wave of IκBα degradation and activation of NF-κB [32], and we wished to determine why this was so. We therefore examined IκBα degradation in transformed MEFs (Fig. 3A), primary MDFs (Fig. 3B) and in the U937 cancer cell line (Fig. S3A) over a 4–8-h period following addition of either TNF or LTα. However, in these experiments, over this time course in these cells we were unable to observe any difference between the biphasic degradation of IκBα induced by LTα or TNF. Nuclear translocation of NF-κB subunits p65, Rel-B and c-Rel following IκBα degradation also appeared to be induced similarly by TNF and LTα (Fig. 3C). We did not observe increased nuclear translocation of p52 after either TNF or LTα stimulation, consistent with the notion that TNFR1 is the main receptor stimulated by these ligands.

Figure 3.

TNF and LTα activate canonical NF-κB signalling in the same manner. (A, B) Transformed wild-type (WT) MEFs and primary WT MDFs were stimulated with 100 ng·mL−1 Fc–TNF or Fc–LTα for the indicated times. Cell lysates were generated, separated on SDS/PAGE gels, and analysed by western blotting (WB) for IκBα degradation. (C) WT MEFs were treated for the indicated times with 100 ng·mL−1 Fc–TNF or Fc–LTα. Then, cell lysates from cytoplasmic (Cyto.) and nuclear (Nuc.) fractions were analysed by western blotting for translocation of the indicated NF-κB subunits. (D) U937 cells were treated with either 2 × Flag–Fc–TNF or 2 × Flag–Fc–LTα for the indicated times. TNFR1 was immunoprecipitated from cell lysates by tandem affinity purification, and analysed by western blotting. (E) The indicated MEFs were stimulated with 100 ng·mL−1 of either Fc–TNF or Fc–LTα for the indicated times, and cell lysates were analysed for IκBα degradation. *Nonspecific band.

We then tested the composition of the TNFR1 signalling complexes following TNF or LTα stimulation by immunoprecipitation with 2 × Flag–Fc-tagged ligands and tandem affinity purification (Fig. 3D). It has been shown that, upon TNF stimulation, RIPK1 is recruited to the TNFR1 signalling complex [64], where it is ubiquitylated by cIAPs [56, 65-67]. Consistent with previous reports of TNF signalling, we observed ubiquitylated RIPK1 within 5 min after TNF treatment, and this also occurred following LTα stimulation. We also detected cIAP1 and phosphorylated IKK1/2 in both TNF-induced and LTα-induced complexes.

Our observations strongly suggested that there is no significant difference in signalling between TNF and LTα. We reasoned, however, that subtle differences might be revealed in different knockout MEFs that are more or less competent in promoting IκBα degradation following TNF–TNFR1 signalling. However, as in the death assays, there was no difference between IκBα degradation induced by TNF or and that induced by LTα (Fig. 3E). Of particular note, degradation of IκBα was completely blocked in Tnfr1/ MEFs, but not in Tnfr2/ MEFs, again consistent with the fact that, in MEFs, the TNF/LTα signalling mainly occurs via TNFR1.

LTα induces inflammatory cytokine production by macrophages to the same extent as TNF

Macrophages constitute a major physiological source of TNF, and respond to TNF by inducing inflammatory signals and the production of cytokines [34]. Some studies have suggested that LTα also stimulates macrophages to express inflammatory cytokines [68, 69], but TNF and LTα signalling have not been directly compared in these cells. We therefore examined signal transduction and cytokine production in bone marrow-derived macrophages (BMDMs) stimulated by Fc–TNF or Fc–LTα.

As in the other cell types previously examined, TNF and LTα induced rapid degradation of IκBα, even with a 10-fold lower concentration (10 ng·mL−1) of the ligands (Fig. 4A). TNF and LTα also induced rapid phosphorylation of the mitogen-activated protein kinases (MAPKs) p38, extracellular signal-regulated kinase (ERK) and c-Jun N-terminal kinase (JNK) with similar kinetics and intensity. To examine the role of the two TNF receptors, TNFR1 and TNFR2, we used BMDMs from the respective knockout mice. Whereas wild-type and Tnfr2−/− cells rapidly and potently induced IκBα degradation and MAPK phosphorylation, Tnfr1−/− cells were unresponsive (Fig. S4). This indicates that the main receptor for TNF and LTα in BMDMs is TNFR1.

Figure 4.

BMDMs respond similarly to LTα and TNF stimulation by expressing inflammatory cytokines. (A) BMDMs were stimulated with 10 ng·mL−1 of either Fc–TNF or Fc–LTα, and cell lysates were analysed for degradation of IκBα and phosphorylation of p38, ERK1/2 and JNK by western blotting (WB). (B) Supernatants from BMDMs stimulated with 100 ng·mL−1 Fc–TNF or Fc–LTα for 24 h were collected and analysed for the expression of the indicated cytokines by Bioplex. Lines represent means of at least four biological repeats. Undetectable was considered as zero. Student's t-tests were performed, comparing TNF with LTα: interleukin (IL)-1-α, P = 0.12; IL-9, P = 0.31; IL-12 (p40), P = 0.89; IL-12 (p70), P = 0.30; IL-13, P = 0.15; eotaxin, P = 0.71; KC, P = 0.84; monocyte chemoattractant protein-1 (MCP-1), P = 0.06; MIP-1α, P = 0.02; macrophage inflammatory protein (MIP)-1β, P = 0.11; RANTES, P = 0.24. Asterisks represent significant differences, P < 0.05; NS, not significant.

Cytokine production induced by TNF and LTα over a 24-h stimulation period was measured by collecting supernatants and measuring cytokines with a Bioplex assay. Of the 23 cytokines and chemokines tested, 11 were found to be induced by TNF or LTα (Fig. 4B). For 10 of these, we observed no statistical difference between TNF-induced and LTα-induced expression (Fig. 4B).

Discussion

In this study, we constructed models of the TNF–TNFR1 and LTα–TNFR1 interfaces. Although these models showed structural differences between the interfaces, the sum of these interactions suggested little difference in affinity. However, affinity is not the only predictor of signalling, and the different binding modalities could affect signalling from TNFR1. Indeed, earlier reports have suggested that TNF and LTα signal differently through TNFR1, either in intensity or in duration of the signal. However, we found that TNF and LTα induced cell death, MAPK activation and NF-κB activation with the same pattern in a range of different wild-type cells.

We reasoned that subtle differences in the manner in which the two ligands signal might be more easily detected in knockout cells that are hypersensitive or defective in TNF–TNFR1 signalling. However, loss of cIAPs and their adaptor, TRAF2, sensitize cells to both TNF-induced and LTα-induced cell death [49, 70]. Likewise, cpdm cells that are deficient in the protein sharpin, a component of LUBAC that is essential for full-strength TNF signalling [56], are equally sensitive to TNF-induced and LTα-induced cell death. Smac mimetic and TWEAK stimulation deplete cIAPs, and are both known to sensitize cells to TNF-induced cell death. Consistent with the cIAP knockout phenotype, these treatments also sensitize cells equally to TNF-induced or LTα-induced death.

Loss of cIAPs sensitizes cells to TNF-induced cell death, and, if caspase activity is simultaneously blocked, this induces necroptosis. Although this form of programmed cell death has not been completely defined, it depends on the kinase activity of the kinases RIPK1 and RIPK3 and on the presence of the pseudokinase MLKL [43, 61]. We have shown that LTα can induce necroptosis with the same potency as TNF, and that, like TNF, it requires RIPK3 and MLKL to cause this cell death.

Chaturvedi et al. showed, in the leukaemic cell line ML-1a, that, with low concentrations of ligands, TNF-induced NF-κB activation occurred in a biphasic manner, but LTα activated NF-κB degradation with a monophasic pattern [32]. In our hands, however, transformed MEFs and primary MDFs degraded IκBα in a biphasic manner following either TNF or LTα stimulation. Although we also tested low concentrations of each ligand, we were unable to detect any IκBα degradation with the dose of 1 pm that they used. It is possible that the differences in the responses of the leukaemic cell line are attributable to signalling through a different receptor from TNFR1, e.g. TNFR2 [71] or HVEM.

Consistent with the results of our death assays, the ability of TNF and LTα to induce NF-κB activation depended on cIAP1 and the LUBAC component sharpin. Although TRAF2 deficiency is not always sufficient to reduce TNF-induced NF-κB activation [72], in several independently derived TRAF2 knockout MEF lines we have observed a reduction in TNF induced NF-κB activation [70] as compared with wild-type MEFs. Again, and consistent with the death assays, TRAF2 knockout MEFs that were deficient in both TNF and LTα induced NF-κB activation. The role of RIPK1 in TNF-induced NF-κB activation is controversial. Although many studies have shown that RIPK1-deficient Jurkat T cells are defective in TNF induced NF-κB activation, and it is known that RIPK1 ubiquitylation can provide a platform for the recruitment and activation of the IKK kinases [42, 73], we and others have shown that RIPK1 is not absolutely required for TNF-induced NF-κB activation [45, 57, 74]. Consistent with our earlier findings, both TNF and LTα are able to activate IκBα degradation within 15 min of their addition to Ripk1−/− MEFs.

The effect of LTα on primary cells has not been extensively examined. Broudy et al. showed that TNF is more potent than LTα in inducing expression of haematopoetic growth factor and adhesion molecules on human endothelial cells [35]. However, they did not quantify adhesion molecules or growth factors directly. Another study by Figari et al. examined the effects of TNF and LTα on neutrophil migration and superoxide production, and showed that both ligands behave similarly, although, at lower concentrations, TNF was more potent [75]. Oster et al. reported that TNF, but not LTα, could induce the production of colony-stimulating factor 1 on human monocytes [34]. We examined the effects of TNF and LTα on BMDMs, and showed that activation of MAPKs, as judged by their phosphorylation, and activation of NF-κB were indistinguishable over a time course of 2 h. Furthermore, cytokine production measured 24 h poststimulation was not significantly different between the two ligands.

Our data suggest that most of the differences that have been reported for TNF and LTα knockouts, and which are mostly related to lymphoid organogenesis [24-27], are unlikely to depend on differential signalling through TNFR1. Furthermore, there were only minor differences between the signalling capacity of the oligomeric Fc forms of the ligands and those of commercially available trimeric forms, making it unlikely that the form of the ligand dramatically alters signalling. It remains possible, although beyond the scope of this study, that the physiologically relevant soluble LTα trimer signals differently from TNF through TNFR2, or that specific cell types respond differently.

Bacterial and viral pathogens use different strategies to block innate immune signalling. One common strategy is to block TNF signalling, and poxviruses encode soluble decoy TNFRs to block TNF. Remarkably, CrmE from vaccinia virus [76], CrmC from cowpox virus [77] and 2L from Tanapox virus [78] appear to bind and block TNF but not LTα. This shows that LTα and TNF binding are separable, and that LTα does not exert a strong evolutionary pressure on these viruses.

Although LTα was cloned before TNF, it is clear that our understanding of its biology has lagged far behind. It is clear from studies of knockout mice that one of the major functions of LTα is to heterotrimerize with LTβ and signal through LTβR. Nevertheless, the LTα trimer exists in vivo, and can exacerbate inflammatory diseases such as graft versus host disease [38], arguing for the importance of investigating signalling by this relatively understudied TNFSF ligand. Whether LTα3 exacerbates disease simply by augmenting existing TNF–TNFR1 signalling, or signalling through HVEM, or finally by differential signalling through TNFR2 remains open to investigation.

Experimental procedures

Modelling of the LTα–TNFR1 and TNF–TNFR1 interfaces

The crystal structures of TNF in complex with TNFR2 (PDB code: 3ALQ [47]) and of LTα in complex with TNFR1 (PDB code: 1TNR [46]) were downloaded from the PDB (http://pdb.org/) [79]. Chain A (TNF) and chain R (TNFR2) of 3ALQ and chain A (LTα) and chain R (TNFR1) of 1TNR were extracted and saved separately in sybylx1.2 (http://tripos.com/). TNF was docked onto TNFR1 with the zdock server (http://zdock.bu.edu/). The top 50 results were visually analysed with the objective of finding a solution with a crude overall 3D structure as the known crystal structure of the LTα–TNFR1 complex [46]. Thus, the second result was submitted to the rosettadock server (http://rosettadock.graylab.jhu.edu/) [80]. The top result from the server was then downloaded and used for further analysis. As a control, LTα was redocked onto its receptor, with the same method described above for the TNF docking, and compared with the crystal structure of the same complex. In this case, the first result from zdock proved suitable for submission to rosettadock. All visual analyses were performed and all images were created with pymol (http://pymol.org/).

Cell culture and reagents

All cells were maintained in DMEM/10% fetal bovine serum. MEFs were generated with a previously described protocol [49], and transformed with an SV40 large T antigen expressing lentivirus. Stable NF-κB–green fluorescent protein (GFP) reporter MEFs were generated with the NF-κB lentiviral reporter vector pTRH1–mCMV–NF-κB–dscGFP from System Biosciences. MDFs were isolated from the dermis of adult mice 48 h after incubation with collagenase from Sigma. To generate BMDMs, cell suspensions from bone marrow were cultured for 6 days in DMEM/10% fetal bovine serum and 20% L929 conditioned media (containing Macrophage colony-stimulating factor (M-CSF) in Petri dishes, and the attached BMDMs were then seeded and assayed on the following day. All of the tissue harvesting procedures were performed according to the guidelines of the Animal Ethics Committee of the Walter and Eliza Hall Institute of Medical Research (WEHI).

Ligands and compounds

293T cells maintained in a defined synthetic medium were transiently transfected with expression constructs for Fc–TNF, Fc–LTα, Fc–TWEAK, 2 × Flag–Fc–TNF and 2 × Flag–Fc–LTα, and the ligands were purified from the supernatant with protein A beads (GE Healthcare, Rydalmere, NSW, Australia). Soluble LTα was prepared by treating the Fc-fusion protein with prescission protease (GE Healthcare) and using protein A beads to remove the severed Fc. Recombinant TNF and LTα were purchased from Life Technologies. The smac mimetic, compound A, has been previously described [49], and was synthesized by TetraLogic Pharmaceuticals. The pan-caspase inhibitor QVD (MP Biomedicals, Seven Hills, NSW, Australia) and the RIPK1 inhibitor Nec-1 (Biomol, Farmingdale, NY, USA) were added 1 h prior to TNF and smac mimetic treatment.

Death assay and flow cytometry

Cells cultured in 12-well or 24-well tissue culture plates were harvested 24 h after stimulation with TNF/smac mimetic/QVD/Nec-1, cell death was measured by PI staining, flow cytometry was performed on a FacsCalibur flow cytometer (BD Biosciences, North Ryde, NSW, Australia), and data processing was performed with weasel (WEHI, Parkville, VIC, Australia).

Immunoprecipitation, nuclear fractionation, western blotting, and antibodies

Immunoprecipitation was performed with the cell lysates of U937 cells treated for 5 min with either 2 × Flag–Fc–TNF or 2 × Flag-Fc-LTα by tandem affinity purification; details have been described previously [45]. Nuclear and cytoplasmic fractionation were performed with the NE-PER Nuclear and Cytoplasmic Extraction Kit (Pierce, Rockford, IL, USA), following the manufacturer's instructions. Cell lysates were prepared with DISC buffer (1% NP-40, 10% glycerol, 150 mm NaCl, 20 mm Tris, pH 7.5, 2 mm EDTA, Roche complete protease inhibitor cocktail, 2 mm sodium orthovanadate, 10 mm sodium fluoride, β-glycerophosphate, N2O2PO7). Cell lysates were loaded in NuPAGE Bis-Tris gels (Life Technologies, Mulgrave, VIC, Australia), and transferred onto Immobilon-P poly(vinylidene difluoride) membranes (Millipore, Billerica, MA, USA) or Hybond-C Extra (GE Healthcare). Membranes were blocked, and antibodies were diluted in 5% skimmed milk powder or BSA in 0.1% Tween-20/NaCl/Pi or NaCl/Tris. Antibodies against phospho-p65, p100/p52, phospho-ERK1/2, phospho-JNK1/2, phospho-p38, p38, ERK1/2, JNK1/2, IκBα, phospho-IKK1/2 (Cell Signaling, Danvers, MA, USA), β-actin (Sigma Aldrich, St. Louis, MO, USA), RIPK1 (BD Biosciences), lamin A/C, HSP60, c-Rel, RelB, p65 (Santa Cruz, Dallas, TX, USA), TNF and LTα (Abcam, Cambridge, UK) were used for western blotting. Signals were detected by chemiluminescence (Millipore) after incubation with secondary antibodies conjugated to horseradish peroxidase.

Blue native PAGE

Blue native PAGE was performed essentially as previously described [81]. Blue native PAGE loading dye [5% Coomassie Blue R-250 (Bio-Rad, Hercules, CA, USA) in 500 mm 6-aminohexanoic acid, 100 mm Bis-tris, pH 7.0] was added to each sample. Gels were electrophoresed in anode buffer (50 mm Bis-tris, pH 7.0) and blue cathode buffer (50 mm Tricine, 15 mm Bis-tris unbuffered containing 0.02% Coomassie Blue G-250). Blue cathode buffer was replaced with clear buffer (without Coomassie Blue) when the dye front was one-third of the way through the resolving gel. Gels were transferred to polyvinylidene difluorid (PVDF) in Tris-glycine transfer buffer containing 20% methanol and 0.037% SDS. Prior to immunoblotting, blots were destained in 50% methanol and 25% acetic acid, and washed with NaCl/Tris.

Cytokine measurement

Supernatants were collected from BMDMs cultured in 24-well tissue culture plates, and treated with Fc–TNF or Fc–LTα for 24 h. Twenty-three-plex mouse cytokine assays (Bio-Rad) were performed following the manufacturer's instructions, and cytokines were measured with a Bio-Rad Bioplex machine.

Acknowledgements

We would like to thank P. Schneider for the initial LTα, TNF and TWEAK constructs that were adapted by us, and D. Vaux for comments on the manuscript. This work was supported by National Health and Medical Research Council of Australia (NHMRC) project grants 541901, 541902, 602516, 637342, and 1025594, program grant 1016647 and an Independent Research Institutes Infrastructure Support Scheme Grant (361646) from the Australian National Health and Medical Research Council, the Australian Cancer Research Fund and Victorian State Government Operational Infrastructure Support. J. M. Murphy is supported by the Australian Research Council (Future fellowship FT100100100). W. S. Alexander is supported by the NHMRC (Research Fellowship 575501). M. W. Parker is an NHMRC Research Fellow. J. Silke is a member of the Scientific Advisory Board of TetraLogic Pharmaceuticals. U. Nachbur is supported by the Swiss National Science Foundation (SNSF, fellowship PA00P3_126249).

Ancillary