Fine-tuning of dendritic cell biology by the TNF superfamily

Key Points

• TNF superfamily members have diverse functions in mammalian biology and are often expressed by cells of the immune system. TNF superfamily members are important therapeutic targets for the treatment of autoimmune disease and for encouraging tumour clearance.

• Dendritic cells (DCs) are the most potent antigen-presenting cells in the immune system. DC-intrinsic signals can dictate whether T cells respond or show tolerance to an antigen.

• The expression of TNF superfamily receptors and ligands by T cells and DCs suggests an important role for this molecular superfamily in T cell–DC crosstalk. TNF superfamily receptors on T cells can provide co-stimulation for the T cell response when they interact with their corresponding TNF superfamily ligands that have been upregulated on DCs.

• TNF superfamily receptors that are expressed directly on DCs can provide homeostatic signals to the DCs in the steady state and can function to 'license' the DCs during an immune response by virtue of T cell–DC crosstalk. Licensed DCs secrete pro-inflammatory cytokines and express CD70.

• Modulation of TNF superfamily receptor-derived signals in the gut can have a dramatic impact on mucosal inflammation and responses to mucosal pathogens. In the gut, T cell–DC crosstalk may be replaced by communication between DCs and innate lymphoid cells to provoke protective immunity and to maintain tolerance to the resident microbial communities in this location.

Abstract

Members of the tumour necrosis factor (TNF) superfamily have been implicated in a wide range of biological functions, and their expression by cells of the immune system makes them appealing targets for immunomodulation. One common theme for TNF superfamily members is their coordinated expression at the interface between antigen-specific T cells and antigen-presenting dendritic cells and, by virtue of this expression pattern, TNF superfamily members can shape T cell immune responses. Understanding how to manipulate such functions of the TNF superfamily may allow us to tip the balance between immunity and tolerance in the context of human disease.

Main

The tumour necrosis factor (TNF) superfamily is a large family of molecules that have diverse roles in mammalian biology¹. Most TNF superfamily ligands and receptors are expressed by cells of the immune system and are often induced during immune activation. As such, these molecules are appealing 'drugable targets'; indeed, inhibitors and agonists of various TNF superfamily pathways are in development or actively being prescribed in the clinic to treat autoimmune diseases and to eradicate cancers (Table 1). Since 1988, biennial meetings have brought together scientists and trainees from industry and academia to illuminate the therapeutic potential of targeting TNF superfamily members and to unravel the roles of these molecules in mammalian biology (see the 14th International TNF Conference website).

Table 1 Therapeutics targeting TNF superfamily members

As there is significant therapeutic potential in manipulating dendritic cell (DC)–T cell interactions for the treatment of complex diseases, this Review focuses on how TNF superfamily members influence various aspects of DC biology. We discuss how TNF superfamily members affect the maintenance of DCs and lymphoid tissues in the steady state, the maturation and activation of DCs during an immune response, and the function of DCs at mucosal surfaces, and we consider how TNF superfamily members can be exploited in the context of DC-targeted therapies. Furthermore, we provide an integrative model for how inputs from different TNF superfamily receptors can shape T cell responses. As therapies directed at modifying DC function gain promise in the clinic, an understanding of how TNF superfamily members influence DC biology will provide new avenues of treatment for significant health problems.

TNF superfamily members and lymphoid tissue

Members of the TNF superfamily have diverse roles in mammalian biology, including in the maintenance of leukocyte subsets and the survival of mature B cells, and this has been recently reviewed². Although a discussion on how TNF superfamily members control the survival and homeostasis of lymphocytes is beyond the scope of this Review, in this section we describe TNF superfamily receptor-mediated signals that are crucial for shaping the priming environment that sets the stage for T cell–DC crosstalk. We begin with the lymphoid tissue itself.

The TNF superfamily and lymphoid tissue development. T cell–DC crosstalk occurs within the complex microenvironment of secondary lymphoid organs. For peripheral antigens, this crosstalk occurs in lymph nodes; for orally ingested antigens, it occurs in the Peyer's patches of the gut; and for blood-borne antigens, it occurs in the spleen. The development of secondary lymphoid organs requires signalling through the lymphotoxin-β receptor (LTβR), a TNF superfamily receptor that is triggered by lymphotoxin-α (LTα)–LTβ heterotrimers (LTαβ). This concept has been recently reviewed^3,4, and it is highly relevant to the overarching role of TNF superfamily members in coordinating effective humoral and cellular immunity, especially in the generation of high-affinity antibodies that can confer long-term immunological memory against pathogens. In brief, in developing secondary lymphoid organs, populations of non-lymphoid LTβR-expressing radio-resistant cells (known as lymphoid-tissue organizer cells) interact with LTαβ-expressing radio-sensitive lymphoid-tissue inducer cells. Coordinate interactions between lymphoid-tissue organizer cells and lymphoid-tissue inducer cells result in chemokine production and the formation of an emerging lymph node anlage.

The TNF superfamily and lymphoid tissue maintenance. In adult mice, LTβR-expressing radio-resistant stromal cells also support and regulate ongoing immune responses, and directly influence T cell–DC crosstalk. The characterization of stromal cells is an emerging field, and various stromal cell subtypes perform distinct functions. For example, specialized stromal cells can be found at locations of antigen entry into lymphoid tissue; these locations include the subcapsular sinuses of lymph nodes, the subepithelial domes of Peyer's patches and the marginal zone of the spleen. In addition to supporting the development of secondary lymphoid organs in utero, constitutive LTβR signalling is required to maintain the organization of secondary lymphoid organ microenvironments and the population of these niches by antigen-presenting and antigen-transporting cells⁵. For example, subcapsular sinus macrophages, which shuttle antigens into the lymph nodes, are LT sensitive⁶, as are marginal zone macrophages of the spleen. In the domes of Peyer's patches, DCs require signalling through the TNF superfamily receptor RANK (receptor activator of NF-κB), and microfold cells, which shuttle antigens into the subepithelial lamina propria of the gut, require RANK and 4-1BB signalling^1,7. Thus, TNF superfamily members are important for maintaining the microenvironments that promote the transfer of antigens into the lymphoid tissue parenchyma for further presentation to T cells.

In addition to maintaining stromal environments that support antigen transport, the LTαβ–LTβR pathway is important for regulating the entry of naive lymphocytes into lymph nodes. The portals of entry into lymph nodes are high endothelial venules (HEVs). The radio-resistant cells of these venules must be maintained in a 'mature' state to accommodate the entry of naive lymphocytes, and LTβR signalling is crucial for HEV maturation⁸. The expression of LTαβ has been best characterized on activated lymphocytes, where it is required for the maintenance of chemokine expression by LTβR-expressing stromal cells in lymphoid tissues. However, recombination activating gene (RAG)-deficient mice, which lack B and T cells, were found to have normal HEVs, suggesting that a non-lymphocyte source of LTαβ is required for HEV maturation. Intriguingly, a recent study has shown that expression of LTαβ by lymph node-resident DCs is important for HEV maturation, thus providing an explanation for why RAG-deficient mice have normal HEVs and also suggesting that there is direct communication between DCs and HEVs⁹.

Downstream of their entry through HEVs, lymphocytes must home to their correct niches within lymphoid tissues. The stromal cells that support T cell–DC interactions are fibroblastic reticular cells (FRCs), whereas follicular dendritic cells (FDCs) are an important scaffold for B cell follicles and germinal centres. Such stromal cells have a variety of functions in fine-tuning both immunity and tolerance, and the LT pathway is important for maintaining chemokine production by such stromal cells to orchestrate cell–cell interactions during immune responses⁵. Indeed, in the absence of LTβR signalling, segregation between T cell and B cell zones is not maintained, and the lymphoid tissue becomes highly disorganized¹⁰. Signalling via CD30 — a TNF superfamily receptor that is expressed by activated lymphocytes — is also required to maintain adequate T cell–B cell segregation in the spleen¹¹. Thus, TNF superfamily members are important for setting up the stromal cell 'space' in which T cell–DC interactions occur.

The TNF superfamily and lymph node hypertrophy. TNF superfamily members are important for supporting lymph node hypertrophy. Inflammation provokes lymph node hypertrophy, whereby lymphocyte recruitment through HEVs and DC migration into the subcapsular sinus are maximized and lymphocyte egress into the medullary cords is limited. These conditions increase the odds of cognate T cell–DC encounters. DCs can induce the proliferation of vascular endothelial cells to support draining lymph node hypertrophy¹⁰³. Vascular endothelial growth factor (VEGF) is an important mediator of inflammation-driven lymph node hypertrophy and has a key role in endothelial cell proliferation¹⁰⁴. FRCs are an important source of this growth factor during lymphangiogenesis, and the production of VEGF by FRCs depends on LTβR signalling¹³. However, VEGF can be produced by various sources and, indeed, lymphangiogenesis depends on complex interactions between different cell types, including DCs, B cells and T cells¹⁴. TNF is another important mediator of lymph node hypertrophy, and it has been suggested that tissue-resident mast cells are a major source of this TNF¹⁰⁵. It is possible that such a 'remote-control' mechanism may also be involved in modifying vascular endothelial cells and HEVs (for example, cytokines and other growth factors could be transmitted via lymph node conduits), but this remains unclear. Both TNF and LTα bind to TNF receptor 1 (TNFR1) and TNFR2, and LTα is important for lymphangiogenesis in tertiary lymphoid structures¹⁵, suggesting that TNF and LTα have complementary roles depending on the anatomical location of the lymphangiogenesis process. Finally, it was shown that lymph node hypertrophy is also dependent on the TNF superfamily member LIGHT¹², which signals through LTβR, and it has been speculated that the relevant source of LIGHT is tissue-derived Langerhans cells.

In summary, complex cellular interactions that rely on signals derived from TNF superfamily receptors must take place within secondary or tertiary lymphoid tissues to promote lymphangiogenesis. As the number of DCs that enter a draining lymph node can have a direct effect on the magnitude and quality of the immune response¹⁶, lymphangiogenesis is an important control mechanism for ensuring optimal immune responses. Furthermore, the same signals that are important for lymphangiogenesis may be required for lymphoid tissue remodelling post-infection. Indeed, this is the case for LTβR-dependent signals in the spleen¹⁷ and lymph nodes⁷ following lymphocytic choriomeningitis virus (LCMV) infection.

The TNF superfamily and DC maintenance. As mentioned above, DCs have an important role in regulating the entry of naive lymphocytes into lymph nodes. However, the main function ascribed to DCs is to capture antigens and present them to T cells (Box 1). There are several distinct DC subsets, and each subset has specialized functions and a unique anatomical location (Table 2). The TNF superfamily ligands TNF¹⁸ and CD40 ligand (CD40L)¹⁹ can promote the proliferation of bone marrow-derived progenitor cells and their differentiation into DCs. Following their egress from the bone marrow, DCs in lymphoid tissues have a short lifespan, with 50% of this population turning over in a 1–3 day period, and this turnover rate is also influenced by TNF superfamily members. For example, RANK^20,21 and CD40 (Ref. 22) promote DC survival during an active immune response. However, as neither RANK ligand (RANKL)-deficient mice²³ nor CD40-deficient mice²⁴ exhibit a deficit in peripheral DCs, the effect of these molecules may be limited to enhancing DC survival during an ongoing immune response, or perhaps these pathways are mutually redundant in the steady state. Recent data have shown that, during inflammation, 4-1BB-mediated signalling in DCs may promote DC survival by inducing the upregulation of the anti-apoptotic molecules B cell lymphoma XL (BCL-XL) and BCL-2 (Ref. 25). This is a different scenario from that found in the steady state, in which 4-1BB appears to limit DC development and restrict the number of peripheral mature DCs²⁶. As 4-1BB ligand (4-1BBL) is generally not expressed in the absence of inflammation, it is possible that 4-1BB exerts an inducible effect on DCs during an immune response.

Table 2 Dendritic cell subsets

Focusing on DC survival, LT-deficient mice exhibit a marked reduction in DC numbers in the steady state²⁷. In addition to being expressed by stromal cells, LTβR is expressed by DCs, and signalling occurs following the binding of membrane-bound LTαβ or LIGHT (Box 2). Kabashima et al. showed that LTβR-deficient mice selectively lack CD11b⁺CD8⁻ DCs. Given the important function of the LT pathway in regulating the expression of chemokines and adhesion molecules²⁸, it was assumed that such a DC deficit was due to a problem with DC migration into lymphoid tissues. However, using elegant chimeric approaches and parabiotic mice, it was shown that the requirement for LTβR in maintaining CD11b⁺ DC numbers was DC intrinsic and that signalling via LTβR was important for DC proliferation²⁹. The expression of LTαβ on B cells was required for inducing homeostatic DC proliferation, which makes sense given the fact that splenic CD11b⁺ DCs are enriched in the marginal zone, where marginal zone B cells are plentiful. However, it is possible that upregulation of LTαβ on antigen-activated T cells could also provoke DC proliferation in the context of an immune response and, indeed, enforced expression of the alternative ligand LIGHT on T cells was shown to stimulate DC proliferation in vitro and in vivo³⁰. It is likely that the interaction of DCs with LIGHT- or LTαβ-expressing cells (such as B cells and T cells, or even lymphoid-tissue inducer cells in the gut) promotes DC proliferation under inflammatory conditions. Given the importance of DCs for generating an effective immune response, such redundancy in the source of ligands for LTβR would make sense.

Interestingly, a more careful examination of the CD11b⁺ DC population revealed a subpopulation of CD11b⁺ DCs that expresses high levels of endothelial cell-selective adhesion molecule (ESAM)³¹. These DCs, which turn over very rapidly, were selectively lost when either Notch or LTβR signalling was specifically ablated in DCs. Thus, it appears that this subset of rapidly proliferating CD11b⁺ DCs is specifically maintained by LTβR and Notch signalling. It is unclear whether the LTβR and Notch pathways collaborate to maintain the survival of this CD11b⁺ DC subset, or whether a more complicated form of crosstalk occurs between these pathways.

There is some information on the signals that negatively regulate DC survival. Ware and colleagues have shown that mice deficient in BTLA (B and T lymphocyte attenuator) or the TNF superfamily receptor HVEM (herpesvirus entry mediator) have increased numbers of CD8α⁻ splenic DCs compared with control mice, suggesting that the stimulation of BTLA through interaction with HVEM negatively regulates this DC subset. This was a satisfying finding because LIGHT can bind to HVEM, suggesting that BTLA–HVEM interactions serve as the negative counterpoint to LTβR signalling. Interestingly, the administration of agonistic LTβR-specific antibodies can increase DC numbers in mice, suggesting that the LTβR pathway can 'override' BTLA-mediated negative regulation of DC proliferation³². This may be important during inflammation, when the expression of LTαβ is induced in T cells³³.

In summary, TNF superfamily members have crucial roles in shaping the priming environment of the lymphoid tissue, both in the steady state and in the context of inflammation. Without these TNF superfamily receptor-dependent events, T cell–DC crosstalk would occur very inefficiently.

The TNF superfamily and the reactive lymph node

T cell priming with exogenous antigens takes place in the highly organized structures of secondary lymphoid organs and, as discussed, TNF superfamily members are important for setting up these environments. However, TNF superfamily members also have an active role in shaping the T cell response by facilitating crosstalk between T cells and DCs during T cell priming. The effects of TNF superfamily receptors are bidirectional. Some TNF superfamily receptors transmit T cell-delivered signals into DCs and promote the upregulation of cytokines and co-stimulatory molecules in a process known as DC licensing. Other TNF superfamily receptors are upregulated by activated T cells, and these T cell-intrinsic signals shape and polarize the ensuing T cell response. A complex integration of various TNF superfamily receptor-mediated signals is crucial for optimizing T cell responses (Fig. 1). In this section, we focus on those TNF superfamily receptors that are triggered specifically during T cell–DC crosstalk and their impact on the T cell response.

Figure 1: T cell–DC crosstalk is influenced by multiple TNF superfamily receptors.

Multiple tumour necrosis factor (TNF) superfamily ligands and receptors sequentially participate in T cell–dendritic cell (DC) crosstalk. The events that occur following the activation of antigen-bearing DCs are outlined in the figure. DCs are activated by pathogen-associated molecular patterns (PAMPs) or damage-associated molecular patterns (DAMPs) and consequently upregulate their expression of CD40, MHC molecules and the co-stimulatory molecules CD80 and CD86. The presentation of peptide antigens in the context of MHC class II molecules to antigen-specific CD4⁺ T cells, together with co-stimulatory signals (from CD80 and/or CD86), results in the activation of CD4⁺ T cells and the upregulation of the DC licensing factors CD40 ligand (CD40L) and lymphotoxin-α₁β₂ (LTα₁β₂). Expression of CD40L and LTα₁β₂ on activated antigen-specific CD4⁺ T cells induces signalling through CD40 and the LTβ receptor (LTβR), and this licenses DCs. CD40 signalling results in the production of interleukin-12 (IL-12) and the upregulation of CD70, CD86, 4-1BB ligand (4-1BBL), OX40 ligand (OX40L) and GITR ligand (GITRL), whereas LTβR signalling leads to the production of type I interferons (IFNs). PAMPs and DAMPs also contribute to these events. Priming of CD8⁺ T cells by MHC class I-restricted peptides results in the upregulation of CD27, 4-1BB, OX40 and GITR (glucocorticoid-induced TNFR-related protein). Stimulation of these receptors on CD8⁺ T cells by their cognate TNF superfamily ligands, in combination with IL-12 and type I IFNs, results in robust CD8⁺ T cell activation, proliferation and effector function, as well as the formation and maintenance of CD8⁺ T cell memory. PRR, pattern-recognition receptor; TCR, T cell receptor.

Inducible co-stimulatory TNF superfamily receptors. Given that TNF superfamily members are often upregulated by antigen-activated lymphocytes during priming, it is not surprising that these molecules have an important role in T cell–DC interactions. The experimental systems used to determine the contribution of signalling mediated by different TNF superfamily members during T cell responses are quite diverse and can often lead to different results. This is probably due to the relative levels of antigen and/or the extent of inflammation in each setting (Table 3). TNF superfamily members that co-stimulate a T cell response are defined here as mediators of T cell-intrinsic signals that augment T cell activation during T cell receptor (TCR) ligation. As the expression of these TNF superfamily receptors on T cells is induced following antigen-mediated activation, these receptors are termed inducible co-stimulatory TNF superfamily receptors.

Table 3 Effects of inducible co-stimulatory TNF superfamily receptors on the T cell response

Inducible co-stimulatory TNF superfamily receptors can make distinct and complimentary contributions to the shaping of an effective T cell response, with some being required for optimal priming and others for memory T cell formation and/or maintenance. A common theme for such interactions is that expression of the cognate TNF superfamily ligand on DCs often requires pre-stimulation of the DCs through CD40 or through pattern-recognition receptors (PRRs), suggesting that the expression of these TNF superfamily ligands requires regulation via a priori activation signals in DCs (Fig. 1). This makes sense given that many of these ligands have been implicated in tissue pathology during autoimmune and allergic responses. We have focused here on co-stimulatory TNF superfamily receptors that are activated by membrane-bound ligands displayed by DCs (for brevity, we have omitted some TNF superfamily receptor–ligand interactions that occur between other immune cells, such as B cell–T cell interactions). Listed in Table 3 are examples of T cell-expressed co-stimulatory TNF superfamily receptors — namely, OX40, 4-1BB, CD27 and GITR (glucocorticoid-induced TNFR-related protein). When engaged by their ligands expressed by activated DCs, these receptors can shift the balance between robust immunity and an abortive or even a tolerogenic immune response to antigens.

The TNF superfamily and DC licensing. DC licensing is a process that occurs early in an immune response and is achieved through cognate interactions between antigen-specific CD4⁺ T cells and antigen-bearing DCs (Fig. 1). DC licensing can be supported by various signals derived from T helper (T_H) cells, such as interferon-γ (IFNγ). TNF superfamily ligands expressed by antigen-specific CD4⁺ T cells can also influence the phenotype and the function of antigen-bearing DCs, thereby licensing DCs to cross-prime CD8⁺ T cell responses. For example, these ligands can alter the expression of MHC molecules and co-stimulatory molecules (such as CD80 and CD86) and the production of cytokines such as interleukin-12 (IL-12) by DCs. CD40L and LTαβ are both upregulated on antigen-specific CD4⁺ T cells within 24 hours of infection or immunization³³, and these TNF superfamily ligands can 'license' DCs by signalling through CD40, which is upregulated on activated DCs, and through LTβR, which is constitutively expressed on DCs. In the context of immunization, these two pathways are not always mutually compensatory during immunization regimes that do not promote substantial inflammation (that is, regimes involving cell-associated or protein antigens that are processed by DCs in the absence of adjuvant, or involving antigens that are not derived from replicating pathogens). In this section, we compare the effects of CD40 versus LTβR signalling in DCs to gain clues as to why more than one TNF superfamily member may be required for DC licensing.

Mechanism for CD40-mediated licensing. CD40 has a very important role in licensing DCs to prime CD8⁺ T cell responses. Blocking CD40-mediated signalling impairs both primary and memory T cell responses against a broad range of immunogens³⁴. Moreover, CD40 ligation can render an otherwise tolerogenic response immunogenic³⁵. Expression of CD40 on DCs is crucial for cytotoxic T lymphocyte (CTL) function³⁶ and for CD8⁺ T cell memory³⁷, and the ligation of CD40 on DCs results in the upregulation of co-stimulatory molecules (including CD80, CD86 and CD70) as well as of other TNF superfamily members, such as 4-1BBL, OX40 ligand (OX40L) and GITR ligand (GITRL). The upregulation of CD70 expression by DCs is a crucial component of CD40-mediated DC licensing, as CD70 blockade abrogates the pro-inflammatory effects of CD40 stimulation in vivo³⁸. CD40 also induces the expression of pro-inflammatory cytokines, such as IL-1β, IL-6 and IL-12, by DCs. Of these cytokines, IL-12 is most restricted to licensed DCs and is clearly important for CD8⁺ T cell responses, as defective or absent IL-12 signalling impairs CD8⁺ T cell clonal expansion, function and memory³⁹.

Mechanism for LTβR-mediated licensing. T cell help is not mediated exclusively through CD40. The discovery of a help-dependent CD8⁺ T cell response that is independent of CD40 and CD40L⁴⁰ supports the existence of additional licensing pathways. An absence of LTβR signalling during T cell–DC interactions renders DCs functionally defective ex vivo³³, suppresses CD8⁺ T cell-mediated graft rejection⁴¹, protects the host from CD8⁺ T cell-driven systemic shock following LCMV infection⁴² and impairs antiviral CTL responses^43,44. Our laboratory has examined how DC-intrinsic LTβR-dependent signalling may modulate CD8⁺ T cell responses in vitro and in vivo. Although IL-12 expression was not readily induced by LTβR signalling in vitro, we found that LTβR signalling (either alone or in synergy with lipopolysaccharide) induces the production of IFNα and IFNβ by DCs³⁶. This is consistent with the finding that LTβR drives type I IFN expression in stromal cells in response to viral infection^45,46. Although LTβR-deficient DCs have defects in their ability to prime CD8⁺ T cell responses, our studies found that exogenous IFNα could rescue normal CD8⁺ T cell population expansion in vitro³⁶. This suggests that LTβR-induced production of IFNα and IFNβ by DCs is an important requirement for CD8⁺ T cell priming in certain settings. IFNα/β receptor (IFNAR) signalling has been shown to be involved in priming antiviral CD8⁺ T cell responses, in particular by promoting the expansion of antigen-specific CD8⁺ T cell populations⁴⁷. Although CD8⁺ T cell-intrinsic IFNAR signalling has been implicated in this process^47,48, type I IFNs produced by DCs downstream of LTβR signalling may also function in an autocrine manner, as IFNAR-deficient DCs may have a defective stimulatory function⁴⁹. Therefore, it is reasonable to conclude that an important outcome of LTβR signalling in DCs is the induction of type I IFN production for optimal CD8⁺ T cell priming.

Are CD40 and LTβR signals in DCs redundant or cooperative? T_H cell-expressed LTαβ and CD40L are both required to license DCs for CTL priming. Therefore, the CD40 and LTβR signalling pathways are likely to collaborate to promote DC maturation. Among other effects, CD40 and LTβR induce, respectively, the expression of IL-12 and the expression of type I IFNs. These cytokines have been implicated in priming CD8⁺ T cell responses and in programming functional CD8⁺ T cell memory. IL-12- or IFNα-mediated signalling in antigen-activated CD8⁺ T cells was shown to induce chromatin remodelling that favours the acquisition of CD8⁺ T cell effector and memory functions, possibly providing an explanation for how qualitative differences in T cell priming produce lasting effects in memory cells¹⁰⁶. It is unclear how CD40-induced IL-12 and LTβR-induced type I IFNs can distinctly and complementarily programme short- and long-term gene expression in CD8⁺ T cells, and many questions concerning CD40- and LTβR-mediated DC licensing remain. The ligands for these receptors are upregulated in concert, and both CD40 and LTβR share many of the same signalling pathways, with some exceptions (Box 2); thus, the unique signals that induce IL-12 versus IFNα and IFNβ production remain to be determined. Furthermore, it is unclear which aspects of the CD8⁺ T cell response depend on the production of IL-12 and type I IFNs by DCs, and which aspects require LTβR- and CD40-induced expression of other soluble or cell-bound immune mediators.

The TNF superfamily in peripheral inflammation

As reviewed above, the activation of TNF superfamily receptors can directly co-stimulate T cell responses and can licence DCs so that they become more immunostimulatory. The contributions of both T cell-associated and DC-associated TNF superfamily receptors shape the nature of the T cell response to foreign and self antigens in the periphery, thereby tipping the balance towards immunity rather than tolerance. DCs in mucosal tissues are distinct from other DC populations and have the challenging task of maintaining tolerance to a large variety of normally harmless antigens, including antigens from host-resident commensal microorganisms, innocuous inhaled protein antigens and ingested food antigens (Table 2). As exciting new data implicate TNF superfamily members in the regulation of immune responses in mucosal environments, we focus here on TNF superfamily receptors in the gut and in the airways. An important role for TNF superfamily receptors in other, non-mucosal, peripheral organs⁵⁰ and in the skin⁵¹ is also emerging, but a discussion of these roles is beyond the scope of this Review.

The TNF superfamily and intestinal inflammation. The maintenance of tolerance in mucosal tissues is complicated by an ever-present and diverse host-resident microbial community. At such sites, DCs must maintain ongoing tolerance to prevent inflammation and tissue damage. This is a delicate balance, and the nature and maturation status of the DCs in these locations are an important key to understanding how tolerance to the microbiota is maintained. Indeed, it is in these locations that the activation of TNF superfamily receptors on DCs may profoundly affect immune homeostasis.

Our knowledge of how the TNF superfamily affects the homing, homeostasis and function of lamina propria-resident DCs is limited, but some key concepts are emerging (Fig. 2). Recent work by Fu and colleagues has revealed a DC-intrinsic role for LTβR signalling in orchestrating immunity to Citrobacter rodentium infection. However, unlike the classical T cell–DC crosstalk described for the LT system in peripheral lymphoid organs^33,36, DCs in the gut interact with innate lymphoid cells in colonic lymphoid follicles⁵². Therefore, expression of LTαβ on innate lymphoid cells is important not only for orchestrating the formation of innate lymphoid follicles and colonic lymphoid follicles (via crosstalk with stromal cells expressing LTβR)⁵³, but also for 'conditioning' LTβR-expressing DCs in the gut. In the gut environment, LTβR signalling appears to be essential for the production of IL-23, which in turn provokes the production of IL-22 by innate lymphoid cells⁵⁴. These studies provide a putative mechanism for the C. rodentium-induced colitis that is attenuated in LT-deficient mice⁵⁵, and perhaps for other models of colitis in which LT inhibition has been beneficial⁵⁶. Interestingly, a similar paradigm has been reported in colitis triggered by CD40 ligation. Powrie and colleagues showed that agonistic CD40-specific antibodies could induce colitis in the absence of T cells, and the CD40-specific treatment induced IL-23 production by DCs⁵⁷. In the absence of T cells, the authors showed that IL-17 was produced in the colitic gut tissue, and it is tempting to speculate that similar crosstalk occurs between CD40-expressing DCs and CD40L-expressing innate lymphoid cells. Nevertheless, the link to IL-23 in these studies provides an explanation for the efficacy of CD40L-specific blocking antibodies in the treatment of colitis in mouse models^58,59.

Figure 2: TNF superfamily members and immune responses in the intestine.

In the small intestine, dendritic cells (DCs) are present in the organized Peyer's patches (not shown), in isolated lymphoid follicles and in the lamina propria. DCs express different cell-surface receptors depending on their location in the small intestine. DCs can interact with innate lymphoid cells (ILCs) to provoke ILC activation. DC-expressed members of the tumour necrosis factor (TNF) superfamily may be important for promoting the expression of interleukin-23 (IL-23), which provokes the production of IL-22 by ILCs to protect the host from infection. This form of crosstalk is reminiscent of the crosstalk that occurs in the periphery during DC licensing. CD40L, CD40 ligand; CX₃CR1, CX₃C-chemokine receptor 1; LTα₁β₂, lymphotoxin-α₁β₂; LTβR, LTβ receptor.

Interestingly, other TNF superfamily receptors appear to have an anti-inflammatory role in the gut. In particular, HVEM–BTLA signalling attenuates intestinal inflammation⁶⁰, suggesting a possible counterpoint to the LTβR pathway in the regulation of mucosal DCs. This is not unlike the negative regulatory role described for HVEM–BTLA signalling in curtailing DC survival in the periphery³². Furthermore, RANK signalling in gut-associated DCs appears to provoke IL-10 secretion, and the administration of RANKL promotes oral tolerance⁶¹. The anti-inflammatory effects of RANK stimulation in DCs are associated with increased numbers of regulatory T (T_Reg) cells in the gut, and the anti-inflammatory function of T_Reg cells during colitis is abrogated by RANK blockade⁶². Thus, the integration of TNF superfamily receptor signals in DCs may dictate the delicate balance between tolerance and immunity in the complex environment of the gut. How such signals are affected by the composition of the local microbiota, and vice versa, remains unexplored.

The TNF superfamily and airway inflammation. Although it is clear that TNF superfamily members are important for the clearance of lung-resident pathogens (Table 3), a role for TNF superfamily receptors in chronic inflammation in the lungs is emerging. Not unlike the gut mucosal tissue, the lungs are a site where coordinate interactions can occur between infiltrating leukocytes, airway-resident DCs (see Table 2) and the epithelial cells of the tissue itself. For example, OX40–OX40L interactions can drive inflammation and promote eosinophilia in the lungs following exposure to allergens, and OX40 signalling in memory CD4⁺ T cells appears to be important for the recall response of these cells to antigens in the lungs⁶³. However, signalling by OX40 in airway epithelial cells was also shown to be important for the production of thymic stromal lymphopoietin (TSLP), which drives an inflammatory T_H2-type immune response in the lungs⁵¹. This could feasibly fuel a 'vicious circle' involving the repeated priming of memory T_H2 cells by allergen-presenting OX40L⁺ lung DCs, and the continual polarization to a T_H2 phenotype because of crosstalk with lung epithelial cells. However, it is important to keep in mind that during chronic allergen exposure, lung remodelling is also quite deleterious to the host, and the notion that TNF superfamily receptors simply control pro-inflammatory T cell responses in the context of mucosal tissues is likely to be over-simplistic. For example, during chronic allergen exposure (such as that induced in the chronic house dust mite model), tissue fibrosis and lung remodelling are driven by LIGHT–LTβR and LIGHT–HVEM interactions, and the deleterious effect of LTβR signalling in the lungs was associated with the production of transforming growth factor-β (TGFβ), an important pro-fibrotic mediator⁶⁴. Thus, a complex interplay between leukocytes and the lung microenvironment is important for the polarization of allergic responses as well as for the subsequent fibrotic response that impairs lung function.

Chronic inflammation in the lungs can also induce the formation of ectopic lymphoid structures known as inducible bronchus-associated lymphoid tissue (iBALT). These structures, which are observed in chronic obstructive pulmonary disease (COPD) as well as during and after influenza virus infection, support local immune responses such as plasma cell generation. Although lymphoid-tissue inducer-like cells are important for the formation of isolated lymphoid follicles in the gut, other types of cells can initiate iBALT formation in the airways. DCs have been shown to be important for iBALT formation and maintenance⁶⁵, and, interestingly, these DCs were an important source of LTαβ, a result echoed by the more recent finding that LTαβ expression by DCs is required for HEV maturation⁹. Whether LTαβ on DCs stimulates DC-expressed LTβR in an autocrine manner remains unclear; however, it is likely that this reservoir of LTαβ does communicate with the underlying stroma to induce the production of lymphocyte-attracting chemokines (such as CXCL13) that serve to recruit other leukocytes into the iBALT. More recently, T_H17 cells were also shown to be important in the formation of iBALT, whereas LTαβ–LTβR interactions were needed to sustain these structures⁶⁶. Thus, not unlike the scenario of chronic allergen exposure, during which TNF superfamily interactions between leukocytes and the lung microenvironment are important for lung remodelling, the orchestration of iBALT formation may also rely on cognate interactions between multiple cell types, including DCs, stromal cells and iBALT-infiltrating lymphocytes. As iBALT formation could contribute to pathogen clearance, further study of how TNF superfamily members affect these cellular interactions may inform vaccine design.

The TNF superfamily in human disease

The first therapeutics to target the TNF superfamily were drugs that neutralize the activity of TNF. These drugs are particularly useful for the treatment of inflammatory bowel disease (IBD) and rheumatoid arthritis, and this makes sense because TNF has pleiotropic effects on the inflammatory cascade, including (but not limited to) neutrophil recruitment⁶⁷. Although TNF blockers have had a tremendous impact on these chronic diseases, other TNF superfamily members clearly have an important role in the aetiology of autoimmune diseases. Despite this, there are only two other TNF superfamily pathways targeted by approved therapeutics (see Table 1 for current TNF superfamily-directed therapies that have been approved by the US Food and Drug Administration, that are in clinical trials or that have been proved efficacious in preclinical animal models). Although TNF blockers have potent anti-inflammatory properties, many of the pathways listed in Table 1 also have an important role in DC biology. As attenuating DC function can either prevent further T cell activation or induce tolerance, DC-directed therapies may have a profound impact on a wide variety of diseases, including multiple sclerosis, graft-versus-host disease, diabetes and IBD. As such, when aberrant DC activity is at the heart of disease aetiology, these therapies merit consideration. In this section, we focus on the promise of targeting TNF superfamily members for the eradication of cancers, and we explore the unmined potential of targeting TNF superfamily receptor-mediated DC licensing as a means to reprogramme immunogenic DCs so that they acquire tolerogenic (or non-stimulatory) properties.

TNF superfamily receptor co-stimulation and tumour immunity. As the expression of TNF superfamily ligands by DCs is generally induced by PRR stimulation, these molecules may be limiting during immune responses that do not provoke substantial inflammation. Immune responses to tumours can often be augmented by the administration of agonist antibodies specific for TNF superfamily receptors. However, agonist antibody therapies can have risks for the host⁶⁸. Enforced expression of TNF superfamily ligands by DCs, which are the natural cellular source of these molecules, can augment the tumoricidal activity of T cells in a more defined manner. Such experiments have revealed that T cell–DC interactions via TNF superfamily members are normally limiting but have significant potential for promoting tumour clearance. Just as these pathways have a diverse role in shaping the T cell response to foreign antigens, the activation of distinct TNF superfamily receptors on T cells can have unique effects on the nature of the T cell response to tumours. For example, the enforced expression of TNF superfamily ligands by DCs can break tolerance to tumours (in the case of CD70)⁶⁹, relieve suppression by T_Reg cells (in the case of OX40L)⁷⁰, provide direct co-stimulation to tumour-specific T cells (in the case of GITRL or OX40L)^71,72, polarize T_H cells towards a T_H1 phenotype (in the case of OX40L)⁷³ and enhance antitumour memory T cell responses (in the case of 4-1BBL)⁷⁴. Given the diverse functions of these TNF superfamily receptors in eliciting antitumour responses, combinatorial therapies in which multiple TNF superfamily ligands are delivered to DCs may make sense.

Therapeutic manipulation of TNF superfamily receptor-dependent DC licensing. DCs are crucial drivers of T cell-mediated autoimmune disease in rodent models, and TNF superfamily receptors that are involved in promoting DC maturation have been implicated in pathological inflammation. Inhibition of CD40 signalling with a blocking antibody specific for CD40L protects from disease in animals models of multiple sclerosis, type 1 diabetes, collagen-induced arthritis and colitis. Inhibition of LTβR signalling also protects from these T cell-driven diseases, although, similarly to the effects of CD40 signalling inhibition, protection usually involves reduced severity or delayed progression of the disease rather than full tolerance induction. Blocking CD40 or LTβR signalling can prolong allograft survival, but lasting tolerance is only achieved by combination therapies that couple CD40 or LTβR signalling inhibition with co-stimulatory molecule blockade or immunosuppressants. Importantly, dual inhibition of CD40 and LTβR signalling can also result in allospecific tolerance induction¹⁰⁷ and may provide a means of altering DC licensing by promoting the apoptosis of alloreactive T cells and/or by encouraging the differentiation of T_Reg cells. These strategies may be the key to establishing long-lasting tolerance without the undesirable effects of long-term treatment with broad-acting immunosuppressants.

Conclusion and perspectives

Even before an immune response occurs, TNF superfamily members exert influence on lymphoid tissues in a way that maximizes T cell–DC encounters and also promote the homeostasis of the DCs themselves. One can therefore imagine that, within ectopic tertiary lymphoid structures, the blockade of TNF superfamily signals may compromise pathogen clearance (as in the case of iBALT structures) or may attenuate chronic inflammation (as in the case of ectopic tertiary lymphoid structures found at sites of autoimmune attack). This could be quite relevant for diseases such as secondary progressive multiple sclerosis, in which immunomodulation in the peripheral compartment appears to have no clinical benefit. Beyond these homeostatic functions, however, TNF superfamily receptor-orchestrated crosstalk between T cells and DCs can lead to enhanced T cell proliferation, cytokine secretion and memory T cell maintenance in the context of an immune response. In the periphery, amplifying TNF superfamily receptor signals can potentially break tolerance to tumours or to self antigens. In the mucosal tissues, where DCs must constantly process harmless antigens and maintain tolerance, the dysregulation of TNF superfamily receptor activation and/or expression may have dire consequences. Taking into consideration the complimentary functions of TNF superfamily members in T cell–DC crosstalk, one can imagine selecting combinations of TNF superfamily receptor-targeted therapies to handle distinct disease aetiologies. Thus, there exists considerable untapped potential in targeting TNF superfamily members, either alone or in combination, for the treatment of significant health problems that exert a tremendous burden on individuals, their families and society.

Box 1 | Priming of T cell responses by DCs

In the absence of infection, dendritic cells (DCs) participate in central and peripheral T cell tolerance by presenting and cross-presenting self antigens or innocuous environmental antigens to induce T cell death, anergy or regulatory function. Semi-mature, self-antigen-bearing DCs migrate to the draining lymph nodes to tolerize self-reactive T cells by presenting peptide–MHC complexes in the absence of co-stimulatory signals (see the figure, left panel). During an infection, the movement of DCs from the tissue via afferent lymphatics into lymph nodes increases dramatically (see the figure, right panel). Inflammatory signals induce the upregulation of co-stimulatory molecules and chemokine receptors that encourage DC homing to lymphoid tissues⁷⁵. Lymphoid tissue-resident DCs are also induced to mature by pro-inflammatory signals, and this is accompanied by changes in their localization within the lymphoid tissue to optimize T cell encounter and priming⁷⁶. Innate stimuli also promote encounters between antigen-bearing DCs and rare, antigen-specific naive T cells by inducing the production of T cell chemoattractants by DCs⁷⁷, thus encouraging contact between DCs and T cells. During these protracted interactions, DCs present MHC–peptide complexes to T cells to induce T cell receptor (TCR) activation, and they also provide various co-stimulatory signals (for example, through CD80 and CD86). In response to these activating signals, T cells upregulate their expression of tumour necrosis factor (TNF) superfamily ligands (see Fig. 1). The production of pro-inflammatory cytokines by DCs is also required for optimal T cell priming, and cytokine expression can be triggered by the activation of TNF superfamily receptors on DCs.

PAMP, pathogen-associated molecular pattern.

Box 2 | Signalling pathways initiated by CD40 and LTβR

Displayed in the figure are the various receptors and ligands from the tumour necrosis factor (TNF) superfamily that are involved in T cell–dendritic cell (DC) crosstalk. CD40 and the lymphotoxin-β receptor (LTβR) are the best-studied examples of TNF superfamily receptors that are relevant to DC licensing.

Ligand binding to CD40 or LTβR induces receptor trimerization and the recruitment of multiple TNF receptor-associated factors (TRAFs), which are adaptor molecules that couple TNF superfamily receptors to downstream mediators. CD40 molecules may oligomerize further, perhaps recruiting an altered spectrum of adaptors and kinases to activate different pathways. CD40 signalling involves TRAF1, TRAF2, TRAF3, TRAF5 and TRAF6, whereas LTβR uses TRAF2, TRAF3 and TRAF5.

Some studies have teased apart roles for individual TRAFs in the activation of specific signalling cascades downstream of receptor activation; however, many of these interactions remain to be confirmed in DCs⁷⁸. TRAF-mediated signalling downstream of CD40 induces the activation of p38, AKT, protein kinase C (PKC) and JUN N-terminal kinase (JNK) as well as both the classical and alternative nuclear factor-κB (NF-κB) pathways⁷⁹. In non-stimulated cells, alternative NF-κB signalling is held in check through the association of NF-κB-inducing kinase (NIK), TRAF2 and TRAF3 with inhibitor of apoptosis protein 1 (IAP1) and IAP2, allowing for IAP1–IAP2-mediated ubiquitylation of NIK and its subsequent degradation. CD40 ligation activates TRAF2 ubiquitin ligase activity, which induces K63-linked ubiquitylation of IAP1–IAP2. This results in the K48-linked ubiquitylation and subsequent degradation of TRAF3, the accumulation of NIK and the activation of downstream signalling.

LTβR also activates JNK and the classical and alternative NF-κB pathways, although the mechanisms are perhaps distinct. For example, although TRAF6 is crucial for CD40 signalling, LTβR signalling does not involve TRAF6, and this fundamental difference could explain the differences in the cytokines produced by these two pathways (interleukin-12 versus type I interferon production). TRAF3 degradation and NIK release downstream of LTβR stimulation may involve the competitive displacement of NIK by LTβR and TRAF2-mediated ubiquitylation of TRAF3 (Ref. 80). Moreover, whereas CD40 signalling first results in canonical NF-κB activation and then activates the alternative NF-κB pathway, LTβR signalling can mediate exclusive and rapid activation of alternative NF-κB signalling⁸¹.

LTβR can also induce the nuclear localization of less-well-characterized NF-κB dimers, including p52–p65 and p50–RELB⁸². Finally, CD40 and LTβR signalling occur in the same DC simultaneously, together with other stimulatory signals (such as pattern-recognition receptor activation) that use common adaptors and that promote the activation of the same transcription factors. It is not clear how common signalling molecules are triaged or sequestered in the context of multi-receptor ligation.

4-1BBL, 4-1BB ligand; CD40L, CD40 ligand; GITR, glucocorticoid-induced TNFR-related protein; GITRL, GITR ligand; HVEM, herpesvirus entry mediator; LTα₁β₂, lymphotoxin-α₁β₂; OX40L, OX40 ligand; RANK, receptor activator of NF-κB; RANKL, RANK ligand.