Inflammation supports immunological defence against infection, trauma, injury and cancer. The regulation of inflammation is governed by cellular communication between non-haematopoietic stromal cells, resident leukocytes and infiltrating immune cells1,2. Optimal control of this process ensures competent host defence, the elimination of the initiating antigen (or antigens) and limited tissue damage1,3. However, repeated, persistent or non-resolving episodes of inflammation lead to the inappropriate regulation of this response and promote autoimmunity, chronicity and tissue pathology1,3,4. The mechanisms that control these events are often diverse and they contribute to clinical differences in disease presentation. For example, patients who have the same underlying clinical condition invariably show differences in disease severity, rate of disease onset and response to treatment4. Although sex, age, genetics, metabolic factors and the environment are major determinants that affect the propensity to develop chronic disease, these indicators provide minimal information on the molecular or cellular pathways that drive the type of disease that is observed.

Cytokines (including interleukins, interferons and growth factors), chemokines, lipid mediators and innate-sensing mechanisms control inflammation and they have important roles in the pathology of various inflammatory diseases and cancer3,4,5, thus representing major targets for clinical intervention. For example, biological therapies that inhibit pro-inflammatory cytokines — for example, the inhibition of tumour necrosis factor (TNF) by the monoclonal antibodies infliximab, adalimumab, golimumab and certolizumab — or their receptors (for example, the inhibition of interleukin-6 (IL-6) receptor by tocilizumab) are now widely used in routine clinical practice for the treatment of rheumatoid arthritis3,4,5. These therapies display specific modes of action and have been developed on the basis of a better understanding of cytokine involvement in inflammation. Although these agents reduce the symptoms and progression of disease, not all patients respond equally to the same therapy. Indeed, 30–40% of patients with rheumatoid arthritis do not respond to any treatments and some biological therapies are efficacious in certain clinical conditions but not in others4. Such differences in therapeutic efficacy suggest that distinct inflammatory mechanisms must steer the course of disease in any one individual. Typically, the sooner after diagnosis appropriate therapy commences, the better the clinical outcome and the increased likelihood of achieving remission6. Thus, cellular events that are triggered early in the inflammatory process must influence the course of disease.

A central feature of local inflammation is the interaction between the stromal tissue compartment and immune cells1,2. Inflammatory mediators that are produced by stromal cells and tissue-resident monocytic cells control the recruitment, activation and survival of leukocytes. Cellular infiltration is traditionally viewed as the diffuse influx of immune cells, with the cells being scattered throughout the inflamed tissue. However, infiltrating leukocytes often form more organized aggregates and these promote antigen-specific adaptive immune responses that exacerbate chronic disease. Indeed, B cells, T cells, resident monocytic cells, macrophages and dendritic cells can form discrete clusters within the inflamed tissue. These regions can exist as simple lymphoid aggregates or as more sophisticated structures that histologically resemble secondary lymphoid organs (SLOs)7,8 (Table 1). The spatial organization of leukocytes into defined, compartmentalized B cell-rich and T cell-rich zones is termed lymphoid neogenesis. These structures direct various B cell and T cell responses, including the induction of effector functions, antibody generation, affinity maturation, class switching and clonal expansion. As a consequence, they are referred to as ectopic lymphoid-like structures (ELSs) or tertiary lymphoid organs (TLOs)7,8. In certain autoimmune conditions, patients who have ELSs in the inflamed tissue often respond poorly to standard biological therapy and thus remain a challenging treatment group9. However, in cancer (for example, solid tumours such as colorectal carcinoma) the presence of tumour-associated ELSs correlates with a better prognosis and they may coordinate endogenous antitumour immune responses that improve patient survival10.

Table 1 Clinical and experimental conditions that feature ELSs

In this Review, we explore the molecular and genetic basis for the formation of ELSs during infection, inflammation, autoimmune disease and cancer, and we discuss the clinical significance of ELSs in terms of prognosis and response to current biological therapies. With insight from both experimental animal models and human disease settings, we also consider the current state of ELS-directed therapies and the rationale for targeting alternative molecular pathways that are associated with ELS formation.

ELS development and function

Although ELSs display an architecture that resembles the follicular compartments that are typically seen in SLOs (Box 1), their organization ranges from simple aggregates of B cells and T cells through to highly ordered structures. Much of our understanding of ELS formation originates from findings that describe the generation of encapsulated SLOs (as reviewed in Ref. 11). Despite structural differences between SLOs and ELSs, many of the mechanisms that control the initial development, cellular composition and functional maintenance of these structures are shared. However, ELS formation is distinct from the pre-programmed ontogenic processes that are associated with secondary lymphoid organogenesis and it does not occur in all patients. Consequently, the generation of ELSs in inflamed tissues — as opposed to a more typical diffuse pattern of inflammatory infiltrate — must be governed by a defined set of inflammatory signals. The mechanisms that trigger these events are poorly defined.

Initiation of ectopic lymphoid neogenesis. Various transgenic mouse models have emphasized the role of inflammatory cytokines in lymphoid neogenesis. For example, mice that are deficient in lymphotoxin-α (LTα; which is encoded by Lta) completely lack peripheral lymphoid organs12. Conversely, tissue-specific expression of a transgene that encodes Lta in the kidney and pancreas caused severe chronic inflammation with the accompanying formation of ELSs that were capable of promoting antigen-specific responses and antibody class switching13. Moreover, the overexpression of both Lta and Ltb (which encodes LTβ) results in more prominent ELS formation compared with when only Lta is overexpressed14. Thus, in combination with LTβ, LTα enhances ectopic lymphoid neogenesis14. The activity of lymphotoxin is associated with increased expression of the homeostatic chemokines CXC-chemokine ligand 13 (CXCL13), CC-chemokine ligand 19 (CCL19) and CCL21, and the increased infiltration of T cells that express CD62L (also known as L-selectin)14. Such findings indicate that the propagation of ELSs within inflamed tissue is driven by communication between local stromal cells, tissue- specific resident mononuclear cells and infiltrating immune cells15,16. Although the actual cells that are responsible for the positioning of these structures within the inflamed tissue remain undefined, several new candidates have recently been proposed. These include lymphoid tissue inducer cells (LTi cells), IL-17-secreting CD4+ T cell populations and T follicular helper cells (TFH cells) (Fig. 1).

Figure 1: Mechanisms that control the induction and maintenance of ectopic lymphoneogenesis.
figure 1

a | Various cell types have been implicated as potential initiators of ectopic lymphoid-like structure (ELS) formation. Although the precise mechanisms require further clarification, the cell types that are associated with this process may include interleukin-17 (IL-17)-secreting CD4+ T helper (TH) cells (not shown) and CD4+ lymphoid tissue inducer (LTi) cells. b | These cell populations are attracted to inflammatory sites by certain chemokine signals, including CXC-chemokine ligand 13 (CXCL13) and CC-chemokine ligand 21 (CCL21). Within these inflamed lesions, resident stromal cells contribute to the cellular organization of lymphoid aggregates. Pro-inflammatory mediators — including IL-7 and lymphotoxin α1β2 (LTα1β2) — regulate processes that affect the chemokine expression profile that is necessary for further recruitment of B cells and T cells, the spatial arrangement of these cells into defined clusters and the control of angiogenesis. c | Although persistent antigen presentation by follicular dendritic cells (FDCs) and B cells supports the long-term maintenance of these structures, cell types such as CD4+ T follicular helper (TFH) cells are essential for relaying immunological instructions to the B cells, which, in turn, ensure the continued action of TFH cells. Of potential relevance to the development of ELSs in inflamed tissues is the ability of defined CD4+ TH effector subsets to acquire TFH cell-like characteristics. Commitment of cells towards a TFH cell-like phenotype in inflamed tissues may aid the development or the activities that are associated with ELSs. BCL6, B cell lymphoma 6; HEV, high endothelial venule; ICAM1, intercellular adhesion molecule 1; ICOS, inducible T cell co-stimulator; IL-7R, IL-7 receptor; LTβR, lymphotoxin-β receptor; LTo cell, lymphoid tissue organizer cell; PD1, programmed cell death protein 1; RANK, receptor activator of NF-κB; RANKL, RANK ligand; RORγt, retinoic acid receptor-related orphan receptor-γt; VCAM1, vascular cell adhesion molecule 1.

PowerPoint slide

The accumulation of CD4+CD3CD45+ LTi cells at sites of lymph node development is an early event in secondary lymphoid organogenesis17,18,19. This is coordinated by the expression of CXCL13, receptor activator of NF-κB ligand (RANKL; also known as TNFSF11) and IL-7, which control the recruitment, survival and activation of LTi cells that express CXCR5 and IL-7 receptor (IL-7R; also known as CD127)20. Central to this process is the interplay between LTi cells and stromal mesenchymal cells — known as lymphoid tissue organizer cells — which act as the anchor for lymph node development11. The release of IL-7 and RANKL by activated lymphoid tissue organizer cells promotes the expression of LTα1β2 by LTi cells, which, in turn, engages the LTβ receptor (LTβR) on lymphoid tissue organizer cells that also express vascular cell adhesion molecule 1 (VCAM1) and intercellular adhesion molecule 1 (ICAM1). This promotes homeostatic chemokine release and vascularization by high endothelial venules (HEVs), which also contribute to lymphoid neogenesis11,15,20,21.

Although stromal lymphoid tissue organizer cells and LTi cells are important in secondary lymphoid organogenesis, their involvement in ELS formation is less clear. However, stromal tissue cells may acquire lymphoid tissue organizer cell-like properties in ELSs21,22,23. For example, synovial fibroblasts from the joints of patients with rheumatoid arthritis release homeostatic chemokines and cytokines that may contribute to ectopic lymphoid neogenesis within the inflamed synovium2,24,25,26. Moreover, gene expression profiling of synovial tissue from patients with rheumatoid arthritis has identified an IL7 signature in ELS-associated synovitis26.

The recent discovery of innate adult LTi cells that are part of the innate lymphoid cell (ILC) family18,19 raises the possibility that these cells also contribute to inflammation-associated ELS development. For example, the adoptive transfer of adult LTi cells into Cxcr5−/− mice induces the formation of intestinal lymphoid tissues19. Ectopic lymphoid tissues that develop in response to transgenic overexpression of the Il7 gene also require LTi cells17. Adult LTi cells express the transcriptional regulator retinoic acid receptor-related orphan receptor-γt (RORγt) and can produce IL-17, which are characteristics of group 3 ILCs18,19. This phenotype suggests an ancestral relationship between adult LTi cells and IL-17-producing CD4+ T helper (TH17) cells18,27. Furthermore, various studies have now linked IL-17 and TH17 cells with ectopic lymphoid neogenesis, where they have a role in chronic allograft rejection, experimental autoimmune encephalomyelitis (EAE) and the development of inducible bronchus-associated lymphoid tissue (iBALT)28,29,30. For example, ELS formation in the central nervous system is associated with TH17 cells that express the lymphoid tissue-associated glycoprotein podoplanin (also known as gp38 in mice)29. Podoplanin-deficient mice display defective development of lymph nodes and germinal centre structures, and podoplanin expression by fibroblastic reticular cells is associated with the development of the micro-architecture of the T cell zones29,31. Thus, podoplanin expression is a defining feature of tissues that display active lymphoid neogenesis. Although the precise function of podoplanin is unclear, it may support the retention of TH17 cells within these sites29,30. In addition to these studies, various reports have also noted roles for B cells and TNF-secreting F4/80+ myeloid cells in ELS generation32,33. The formation of ELSs in different tissues and in response to different forms of immunological activation suggests that the mechanisms driving ELS expansion are complex. Inherent similarities in effector functions or cellular plasticity may render various cell types able to promote tissue-specific ELS development.

Organization of cells within ELSs. Inflamed tissues displaying ELS-associated pathology have increased expression of homeostatic chemokines that govern the cellular composition and functional properties of ELSs34,35,36,37. These include CXCL12, CXCL13, CCL19 and CCL21. In such tissues, cellular communication between local stromal cells, tissue-specific resident mononuclear cells and infiltrating immune cells propagates the development of ELSs and enables them to function in a similar manner to germinal centres (Box 1). Such interactions affect leukocyte trafficking and angiogenesis but are also instrumental in governing the organization of cells within these structures. For example, CXCL13 and CCL21 control the segregation of B cells and T cells in ELSs34,35. By contrast, CCL19 and CXCL12 promote lymphocyte infiltration and the positioning of dendritic cells, B cells and plasma cells within these aggregates34 (Fig. 2). These specialized properties probably reflect differences in the ability of individual homeostatic chemokines to promote the expression of LTα and LTβ by B cells and T cells34. For example, transgenic expression of Cxcl13 in pancreatic tissue promotes the LTα1β2-mediated formation of ELSs that display defined lymphoid zones, HEVs and stromal reticulum35. Thus, homeostatic chemokines affect the size, cellular composition and organization of ELSs. Such features affect the functional properties of ELSs and their impact on pathology.

Figure 2: Cytokines and chemokine regulate ELS formation and function.
figure 2

The inducible formation of ectopic lymphoid structures (ELSs) mimics the ontogenic process of secondary lymphoid organ (SLO) development, whereby the cytokines interleukin-7 (IL-7), lymphotoxin-α (LTα), LTβ and receptor activator of NF-κB ligand (RANKL) have a direct role in initiating chemokine-directed lymphoid organogenesis. It is becoming clear that, during ELS formation, effector cytokines that are produced in response to chronic inflammation, infection, autoimmunity and cancer are indirect regulators that substitute for the cytokines that are involved in SLO formation. Recent studies have highlighted novel roles for immune cell subsets in ectopic lymphoid neogenesis. For example, T helper 17 (TH17) cells and T follicular helper (TFH)-like cells are associated with ELS development in the central nervous system and the lungs29,30,38. Various TH cell subsets can also acquire TFH cell-like characteristics and effector functions29,43,44,45,46,47, and these cells may contribute to ELS development in autoimmune and infectious diseases. The discovery of a role for these novel cell types brings with it a host of regulators that may positively and negatively regulate ELS formation and function. These include factors that determine the commitment or effector characteristics of defined T cell populations — for example, the control of TH17 cells and TFH-like cells by IL-2, IL-6, IL-21, IL-27 and type I interferons (IFNs) — and mediators that have altered expression in ELSs within defined anatomical locations (for example, IL-4, IL-5, IL-17, IL-21 and IL-27). Potentially, these may serve as alternative therapeutic targets or agents for ELS-associated pathology. CCL, CC-chemokine ligand; CXCL, CXC-chemokine ligand; FDC, follicular dendritic cell; LTi cell, lymphoid tissue inducer cell; LTo cell, lymphoid tissue organizer cell; TNF, tumour necrosis factor.

PowerPoint slide

However, the contribution of individual homeostatic chemokines to ELS formation may also depend on the anatomical site or ongoing disease processes. Although transgenic expression of Ccl21 in the pancreas drives pancreatic lymphoid neogenesis, the ectopic expression of Ccl21 in the skin does not36,37. Whether these differences reflect a hierarchy of chemokine-mediated outcomes that affects the degree of architectural organization displayed by ELSs or whether they reflect specific characteristics that are associated with 'permissive' versus 'non-permissive' tissues remains unknown.

Maintenance of ELSs within tissues. Considerable attention has been given to the development of lymphoid structures, however, the presence of ELSs is often a transient feature of inflamed tissue. Thus, the mechanisms that control the maintenance of these aggregates during disease may have greater clinical significance. Recent studies have linked TFH cells — or markers of TFH cell activity — to the formation, maintenance and function of ELSs29,38,39,40.

TFH cells promote B cell activities and high-affinity antibody generation in germinal centres41,42. Recent studies of T cell plasticity in mice suggest that TH1, TH2 and TH17 cells may also acquire TFH cell-like characteristics and effector functions during their differentiation29,43,44,45,46,47. For example, TH17 cells can display TFH cell-like characteristics during their differentiation, including the secretion of IL-21 and the expression of TFH cell-associated molecules, such as signal transducer and activator of transcription 3 (STAT3), interferon-regulatory factor 4 (IRF4), MAF and inducible T cell co-stimulator (ICOS)48,49,50,51,52,53. These observations may help to explain the ability of TH17 cells to promote ectopic lymphoid neogenesis28,29,30 and may also explain the regulation of ELS activity in diseases that are associated with TH1, TH2 or TH17 cell responses. Memory and effector TH cells that lack typical TFH cell-like features are, nevertheless, also capable of providing B cell help as they express CXCL13, IL-4, IL-21 and CD40L54,55,56. Consequently, both positive and negative regulators of effector T cell responses may influence the maintenance of ELS activity (Fig. 2).

The CCR7-dependent recruitment of dendritic cells to developing ELSs represents an important homeostatic event that drives the initial propagation of adaptive immune responses within these sites. For example, infiltrating CD4+ T cells form tight clusters with dendritic cells and this promotes T cell proliferation57. As ELSs mature, dendritic cells within these structures continue to support the efficient priming of T cell responses through antigen presentation and they contribute to class-switch recombination, antibody generation and the formation of new lymphatic vessels58,59,60. The importance of dendritic cells in ELSs is best illustrated by studies of mucosal immunity following viral infection. The depletion of dendritic cells during viral lung infection leads to impaired germinal centre reactions and a disruption of ELS architecture59,60. The sustained activation of dendritic cells within ELSs is therefore required for both their formation and their functional maintenance.

The function of ELSs as germinal centres. There is conclusive evidence that ELSs not only recapitulate the cellular, molecular and structural organization of SLOs but that they can also support the function of germinal centres. In the germinal centres of SLOs, B cells undergo affinity maturation and differentiation to memory B cells and plasma cells via antigen-driven selection61 (Box 1). This process includes antibody fine-tuning through somatic hypermutation and class switching, both of which affect antigen recognition and the effector capacity of the antibody62,63,64,65. Consistent with the role of ELSs as functional ectopic germinal centres, activation-induced cytidine deaminase (AID) is expressed at the mRNA and protein level within these structures, and studies in patients and animal models support the involvement of AID in autoimmunity66,67,68,69, infection70 and allograft rejection28. For example, AID controls local antibody affinity maturation (including somatic hypermutation), as evidenced by a restricted profile of variable (V)-gene repertoire usage, highly mutated V regions and the oligoclonal diversification of infiltrating B cells and plasma cells68,71,72,73. Active class switching is also confirmed by the detection of Iγ–Cμ and Iα–Cμ circular transcripts within ELSs, which marks ongoing class-switch recombination from IgM to IgG and IgA, respectively67,68,74. Thus, ELSs in inflamed tissues retain the necessary molecular machinery to support in situ antibody diversification, isotype switching, B cell differentiation and oligoclonal expansion. However, as reviewed below, although the final outcome of lymphoid neogenesis is generally protective in the context of infection and cancer, it is often deleterious in the setting of autoimmunity and graft rejection.

ELSs in protective immunity and disease

ELSs as sites of anti-pathogen immune responses. Infection with bacteria (such as Mycobacterium tuberculosis and Helicobacter pylori) and viruses — such as influenza virus, hepatitis C virus (HCV) and several sialotropic viruses — has been associated with the formation of ELSs in animal models39,59,74,75 and in humans76,77,78. These ELSs resemble highly organized SLOs. Most remarkably, in mice, infection-triggered lymphoid neogenesis seems to be preferentially induced at permissive mucosal sites, and results in the de novo formation of iBALT and of lymphoid tissues associated with the salivary glands79,80.

Consistent with their immunological functions, infection-associated mucosal ELSs can mount protective immune responses in situ. Examples of this include influenza virus-induced and M. tuberculosis-induced lymphoid neogenesis in the lungs of rodents. Following influenza virus infection of the upper respiratory tract, mice develop functional ELSs79 that are maintained by lymphoid tissue-associated chemokines (for example, CXCL13 and CCL21) and LTβ, which are produced by infiltrating CD11b+CD11c+ dendritic cells59. These ELSs promote the in situ differentiation of antiviral plasma cells that are specific for the nucleoprotein of influenza virus59. Notably, in LTα-deficient mice — in which SLOs do not develop — reconstitution with Lta+/+ bone marrow allows for ELS development in response to influenza virus infection81. These bronchial ELSs can maintain antiviral immunity and memory recall responses in the absence of SLOs81. A role of ELSs in antimicrobial immunity is further illustrated by their involvement in M. tuberculosis infection, where they promote granuloma formation and prevent the dissemination of infection38,75. Although these studies demonstrate a non-redundant role for ELSs in certain infections, they also indicate that pathogen recognition by innate-sensing mechanisms may contribute to the formation of these structures.

Overall, these data suggest an evolutionary role of ELSs at sites of infection where they coordinate protective immunity in addition to, and often independently from, SLOs. However, a failure to eradicate a pathogen has considerable clinical implications and can lead to the development of autoimmunity (for example, in chronic HCV infection) and lymphoma (for example, in H. pylori-associated chronic gastritis). Thus, ectopic lymphoid neogenesis in response to microbial challenge influences the natural history and clinical course of chronic infection and associated complications.

ELSs in autoimmunity and the emerging role of Epstein–Barr virus. The presence of ELSs with the appearance of fully functional ectopic germinal centres has long been described in the inflamed target organs or tissues of patients who are affected by autoimmune diseases, including the synovial tissue in rheumatoid arthritis (as reviewed in Ref. 82), the meninges in multiple sclerosis69, the salivary glands in Sjögren's syndrome (as reviewed in Ref. 83), the thymus in myasthenia gravis (as reviewed in Ref. 84) and the thyroid gland in Hashimoto's thyroiditis85. In these clinical disorders, ELSs develop in response to disease-specific autoantigens, which also ensure the long-term maintenance of these structures within the inflamed tissue.

The presence of ELSs in autoimmune conditions perpetuates autoimmunity towards disease-specific antigens. In this respect, many of the regulatory mechanisms that govern tolerance within SLOs are not seen in autoimmune disease-associated ELSs. For example, in SLOs, autoantigen-binding B cells may be excluded from entering germinal centres and they lack responsiveness to CXCL13 owing to a downregulation of CXCR5 (Ref. 86). However, autoimmune disease-associated ELSs permit the entry of autoreactive B cells87. This allows for the differentiation of the B cells into high-affinity autoreactive plasma cells that release disease-specific autoantibodies, such as anti-citrullinated protein antibodies (ACPAs) in rheumatoid arthritis67, and antibodies against the ribonucleoproteins Ro and La (also known as Sjögren's syndrome antigens A and B) in Sjögren's syndrome88, against thyroglobulin and thyroperoxidase in Hashimoto's thyroiditis85, and against the nicotinic acetylcholine receptor in patients with myasthenia gravis84.

Although the mechanisms that are responsible for the preferential accumulation of autoreactive B cells in ELSs are not fully understood, a direct role has been proposed for Epstein–Barr virus (EBV) in the development of autoimmunity. EBV is a γ-herpesvirus that is retained throughout the life of an infected individual and that promotes the survival and proliferation of B cells89,90. Emerging evidence suggests that ectopic follicles in the target organs of patients with multiple sclerosis, myasthenia gravis, rheumatoid arthritis and Sjögren's syndrome frequently harbour latent EBV infection69,91,92,93. In these sites, EBV-transformed B cells and plasma cells display autoreactivity to disease-specific autoantigens, such as citrullinated fibrinogen in rheumatoid arthritis92 and the ribonucleoprotein Ro in Sjögren's syndrome93. Autoreactive EBV-infected B cells are therefore predicted to transit from the peripheral compartment into ectopic germinal centres where they undergo differentiation to high-affinity autoreactive plasma cells. Thus, niches of EBV colonization within ELSs might be considered as a hallmark of organ-specific autoimmunity.

EBV infection may also explain another remarkable feature of ELSs, which is their ability to persist in an activated state for several weeks in the absence of recirculating immune cells from the periphery. This is best demonstrated in chimeric severe combined immunodeficiency mice that are engrafted with synovial or thymic tissue from patients with rheumatoid arthritis or myasthenia gravis, respectively. ELSs within the transplanted tissue maintain their follicular organization and autoantibody generation despite both the absence of 'new' cells infiltrating the grafts and the impaired host immune response67,94. This is of crucial importance, as ELS neutralization is probably fundamental to avoiding the re-emergence of autoreactive clones that are capable of driving disease relapse or resistance to therapy.

ELSs in transplantation. ELSs have been observed in almost all types of human grafts that have been removed owing to chronic rejection, including kidneys, lungs and hearts95. Transplant-associated ELS formation seems to recapitulate the molecular pathways that are triggered during the ontogeny of SLOs. These ELSs harbour ectopic germinal centres in which naive B cells differentiate into anti-HLA-producing plasma cells and memory B cells96. However, ELS formation in allografts differs from that of canonical SLOs in a number of ways: first, the considerable amounts of TH17 cell-associated cytokines and growth factors — such as B cell-activating factor (BAFF; also known as TNFSF13B) — that facilitate autoreactive B cell survival; second, the continuous release of alloantigens from the injured tissue that are trapped locally by defective lymphatic drainage; and third, the apparent defect of local immunoregulatory mechanisms. As a consequence, there is an excessive immune response within the ELSs that contributes to allograft rejection95. However, two recent publications in distinct experimental models (renal and cardiac allografts) have reported that the presence of ELSs promotes allograft tolerance and graft function97,98. On this basis, it is postulated that under certain conditions, ELSs may amplify beneficial immune responses — that are associated with B cells and T cells displaying a regulatory profile — to promote graft tolerance.

ELSs in cancer. In cancer, ELSs have been documented in many tumours, including those that are associated with lung99,100, colon10,101, breast40,102,103 and germ cell cancers104,105. However, not every patient with a specific type of cancer develops ELSs and when they do occur, their contribution to disease varies considerably. Some tumour types are more likely to induce ELS formation than others, which indicates that certain tumours provide a microenvironment that is conducive to lymphoid neogenesis. Given the stark difference between the immunosuppressive nature of the tumour microenvironment compared with chronic inflammatory foci, it is perhaps surprising that ELSs develop at all in tumours104. Although CXCL13, CCL19 and CCL21 have a crucial role in the formation of cancer-associated ELSs, LTi cells seem to be dispensable as ELSs can form in their absence100,104,105. Permissive tumours probably induce the production of lymphoid chemokines not only through the expression of LTα1β2 but also through the expression of pro-inflammatory cytokines — such as TNF, IL-1β and IL-6 — which contribute to the transcriptional regulation of CXCL13, CCL19 and CCL21. However, the precise stimuli that initiate the development of cancer-associated ELSs within the immunosuppressive tumour microenvironment require further investigation.

Clinical implications

As discussed above, ELSs are found in patients with diverse medical conditions but invariably, in each condition, ELSs form in some patients but not in others (Table 1). Therefore, important questions relating to this phenomenon remain to be answered. For instance, what determines the formation of ELSs in some individuals but not in others? Is ELS development typical of different disease subtypes (ab initio) or is ELS formation the inevitable result of persistent inflammation — that is, are ELSs the cause or the consequence of chronicity? To what extent do ELSs contribute to ongoing inflammation and tissue damage (disease prognosis)? In this section, we address these questions in turn by providing recent data from the literature, as well as from our own personal experience.

What governs ELS formation in some individuals but not in others? This question still remains mostly unanswered but it is likely that genetic and/or environmental factors shape the inflammatory response to give rise to diverse structural outcomes within individual tissue microenvironments.

Considering the genetic influence first, evidence is beginning to emerge from various genetic studies that have found associations between specific pathways that are involved in ELS formation and/or function and susceptibility to autoimmune diseases. For example, four single nucleotide polymorphisms (SNPs) have been identified in the IL2–IL21 locus at chromosome region 4q27 that are associated with several autoimmune diseases106,107,108 and that can influence circulating levels of IL-21 (Ref. 109). Additionally, SNPs in loci that are near to the CXCR5 gene have been linked with susceptibility to primary biliary cirrhosis110, Sjögren's syndrome111, systemic lupus erythematosus (SLE)112 and multiple sclerosis113.

Although no sub-analysis with stratification of patients for the presence or absence of ELSs has been carried out in the above studies, the increased genetic susceptibility to autoimmunity that is associated with loci near to both the IL21 and CXCR5 genes corresponds with the crucial role of IL-21-producing CXCR5+ TFH cells in the formation and function of ELSs. This is also potentially of crucial relevance as novel strategies to specifically target TFH cells are being developed for therapeutic use in autoimmune conditions (as discussed below). It is likely that the influence of genetic factors will become clearer as larger scale analyses of stratified disease subsets are carried out according to pathobiology (for example, the presence or absence of ELSs).

With regard to the environmental factors that influence ELS formation, diverse host immune responses to infection lead to different outcomes. Thus, it is possible that in 'permissive' patients, an active EBV infection is established in the diseased tissue and may drive ELS formation through its unique ability to infect and activate B cells114,115. This hypothesis was recently supported by our own work, in which we found a high frequency of EBV-infected B cells and plasma cells in ELS-containing synovia from patients with rheumatoid arthritis92. Notably, EBV was not detected in control synovia from patients with osteoarthritis or in patients with rheumatoid arthritis whose synovia showed diffuse infiltration of immune cells but lacked ELSs92. Similar results were seen in the salivary glands of patients with Sjögren's syndrome93.

Is ELS formation pre-determined or a result of persistent inflammation? The frequency at which ELSs occur in patients with chronic inflammation typically varies from 20–40%7,82. However, it is unknown whether ELSs are a typical feature of different disease subtypes from the beginning (ab initio) or whether they form as a result of persistent inflammation. The reason this this remains unknown is twofold: first, in humans, clinical presentation is often delayed and it is thus impossible to precisely determine the early events in the pathological process; and second, there are an insufficient number of studies (and an insufficient number of large cohort studies) in which serial biopsies have been taken to establish the 'pre-determined' versus the 'evolutionary' nature of ELSs. Nonetheless, data is beginning to emerge from a cohort of 300 patients with rheumatoid arthritis who had shown symptoms for less than 12 months; biopsies were taken at the time of disease presentation and 70% of individuals were also biopsied 6 months post-treatment116. Approximately 40% of patients displayed histological evidence of ELSs before receiving any disease-modifying therapy. Thus, ELSs seem to be present ab initio and to define a specific disease subset from the beginning. However, in another early arthritis cohort of 93 patients with only 17 biopsy repeats at 6 months, ELSs have been reported as being related to the degree of inflammation and not related to specific disease subtypes117. No serial biopsy data is available from other diseases and, therefore, further work is required in this area to establish the precise ontogeny of ELSs in the context of chronic inflammation.

To what extent do ELSs control disease progression and outcome? From the above discussion, it is clear that ELSs are dynamic structures, the main function of which is to potentiate the immune response at sites of disease. The clinical presentation of ELSs may therefore influence disease severity, the rate of disease progression and clinical outcome. Ultimately, this will depend on the ability of the host to clear triggering antigens, as well as the stage and the nature of the disease in question. In the subsequent paragraphs, we provide some examples of the influence of ELSs on the progression and outcome of infection, autoimmune conditions, transplantation and cancer.

On the basis of the evidence from animal models showing a protective role of mucosal ELSs in pathogen control, it is anticipated that the presence of ELSs in infected humans could alter the natural history and clinical manifestations of chronic infections. Accordingly, in M. tuberculosis infection, ELS formation in the peripheral rim of lung granulomas is typically associated with the latent, asymptomatic form of tuberculosis118. Conversely, sites of active infection that have cavitary tuberculous lesions lack B cell follicles118. Thus, ELSs coordinate the local host response to pulmonary tuberculosis118. ELSs probably exert similar mechanisms of pathogen containment in other chronic infections, such as H. pylori and HCV76,77. However, it is now clear that in the absence of pathogen eradication, ELSs also act as a double-edged sword by contributing to autoimmunity and lymphomagenesis. Although tuberculosis has long been associated with the development of various autoimmune phenomena119, the association with ELSs has not been investigated. Conversely, in chronic HCV-related hepatitis, elevated CXCL13 levels in liver biopsies identify patients that develop mixed cryoglobulinaemia120. Together with the clonal analysis of intrahepatic B cells in portal tracts121, these data indicate that intraportal lymphoid follicles contribute to the evolution from polyclonal to oligoclonal to monoclonal autoreactive B cell activation that occurs as a result of chronic HCV infection. Pathogen-driven chronic B cell activation within ELSs is also associated with lymphoproliferative disorders, as evident in HCV-associated B cell lymphoproliferation122 and H. pylori-driven gastric mucosa-associated lymphoid tissue (MALT) B cell lymphomas76. The development of gastric (and salivary gland) MALT lymphomas has been linked to the genetic instability that is associated with the process of aberrant somatic hypermutation, whereby dysregulated expression of AID leads to mutations in the regulatory and coding sequences of genes that regulate B cell survival and proliferation (such as PAX5 and MYC)123. Interestingly, AID expression in gastric and salivary gland MALT lymphomas is confined to ELSs and is not found in malignant marginal zone B cells66,123, which suggests that lymphomagenesis is dependent on sustained antigen stimulation within ELSs. Accordingly, gastric MALT lymphomas are crucially dependent on H. pylori-specific T cells124 and the eradication of H. pylori results in both ELS resolution and tumour regression125. Although in this context lymphomagenesis is ultimately driven by chronic infection, a striking and unique feature of MALT lymphomas is that malignant B cells frequently originate from precursors that express an autoreactive B cell receptor with homology to rheumatoid factor126. Thus, autoimmunity and MALT lymphomagenesis can represent a continuum of the same antigen-driven response in which ELSs have a central role.

In chronic inflammatory and autoimmune conditions, the B cell-rich inflammatory infiltrate that is typically seen in ELSs is associated with a strong potential for local immune activation and cellular differentiation. This promotes autoantibody production, complement activation and pro-inflammatory cytokine release, which drives the inflammatory cascade, autoimmunity and tissue destruction. For example, studies in rheumatoid arthritis have reported worse disease outcomes in patients with ELSs116. In these cases, increased joint destruction is associated with the production of pro-inflammatory cytokines127, as well as osteoclast-activating molecules such as RANKL128 and the RANKL-inducing molecule TNF-related weak inducer of apoptosis (TWEAK; also known as TNFSF12)129. In autoimmune thyroiditis, most ELS B cells recognize the autoantigens thyroglobulin and thyroperoxidase, and this leads to glandular destruction85. In SLE, a restricted repertoire of isotype-switched antibodies are deposited in the glomerular basement membrane of the kidneys with consequent organ damage130. Finally, the presence of ELSs in the inflamed meninges of patients with progressive multiple sclerosis can be determined through analyses of B cell markers, and the occurrence of ELSs is associated with a gradient of cortical neurodegeneration and a more aggressive course of disease131.

ELSs are thought to drive chronic transplant rejection by enhancing intra-graft allogeneic responses, as grafts in which the ELSs are most functional have a shorter life expectancy28,96. However, the validity of this conclusion has been questioned by the fact that, by definition, only de-transplanted, failed grafts were included in the analysis. In addition, as discussed above, there is evidence from experimental transplant models that regulatory B cells and ELSs may govern graft tolerance. Further work is therefore required to clarify the contribution of ELSs to graft rejection versus survival.

As discussed above, different cancers are more or less immunogenic and hence, are more or less prone to induce ELSs. Analyses of the cytokine and chemokine gene signatures, the number of tumour-infiltrating lymphocytes and the level of ELS organization have been used to stratify patients and types of cancer as 'immune response positive' and 'immune response negative'. Although these classification criteria are not purely based on ELSs, there is a general consensus that patients who have immune-response-positive tumours (for example, breast cancer tumours) have a better prognosis than those with immune-response-negative tumours102,132,133,134. Similar results have been described in colon cancer, where immune response positivity was associated with the increased survival of patients independently of tumour stage, previous treatment or microsatellite instability10. As expected, CXCL13 and CCL19 expression was also associated with a good prognosis10. Finally, gene expression profiling of immune-response-positive and immune-response-negative human melanomas identified CXCL13 and CXCL8 as components of a discrete group of 12 genes that were found to be diagnostic of immune response status135. In another study, the mean survival of patients with melanoma was 55 weeks in the immune-response-positive group compared with 18 weeks in the immune-response-negative group136.

In summary, ELS formation in peripheral tissues seems to follow normal pathophysiological programmes in response to the need to increase local immune responses and clear exogenous antigens, alloantigens and/or autoantigens. When such effector programmes are successful, they lead to improvements in disease pathology and a better prognosis, whereas their failure or ineffectiveness leads to persistent inflammation, tissue damage and worse outcomes. This bivalent functionality has substantial clinical implications when it comes to targeting ELSs for therapy.

ELSs as therapeutic targets

The modulation of ELSs for therapeutic purposes has attracted marked interest both from the scientific community and from industry. On the basis of the functional and clinical data discussed above, it would be desirable to potentiate the activity of ELSs in infection and cancer but to inhibit their activity in the context of transplantation and chronic inflammatory conditions. Current cancer trials are testing whether ELS-related pathways can be exploited to enhance antitumour immunity. Although cytokine and chemokine administration, and tumour antigen vaccination represent promising therapeutic approaches to target ELSs, the long-term efficacy of these strategies remains debatable137,138. Similar vaccination approaches are also being considered as adjunct therapies that promote local antimicrobial immunity in the lungs139.

By contrast, there is a strong rationale for therapeutic inhibition of ELSs in graft rejection, and in chronic inflammatory and autoimmune conditions. In a study of patients showing chronic renal allograft rejection, treatment with anti-CD20 (rituximab) failed to cause regression of ELSs despite the successful depletion of B cells from the peripheral blood140. This suggests that the local inflammatory microenvironment in ELSs may facilitate B cell survival and allow evasion of rituximab-mediated depletion140. In the context of autoimmunity, biological interventions that promote the clearance or inhibition of autoreactive lymphocytes have the potential to disrupt ELS involvement and restrict deleterious adaptive immune responses. In this regard, understanding the impact of TNF inhibitors on synovial ELSs is of major importance. In a recent study, patients with rheumatoid arthritis who had synovial ELSs displayed an inferior response to TNF inhibitors compared with patients who did not have synovial ELSs9. In addition, the presence of ELSs in synovial tissue before treatment was reported to be an independent negative predictor of the response to TNF inhibitors9. Furthermore, high protein-level expression of IL-7R, which is normally associated with ELSs, was reported to be associated with a poor response to TNF inhibitors141. Therefore, the development of selective ELS-targeted therapies has attracted marked interest both from the scientific community and from industry.

Currently licensed therapies for inflammatory diseases, including CD20-specific (for example, rituximab) and IL-6R-specific (for example, tocilizumab) monoclonal antibodies, oral Janus kinase inhibitors (for example, tofacitinib) and T cell activation blockers (for example, abatacept) probably target processes that are associated with ELS activities4. However, these drugs also possess broader modes of action3,4,5. Given the prominent role of lymphotoxin and homeostatic chemokines in ELS formation, initial approaches targeted this axis. Pharmacological inhibition of LTβ using an LTβR–immunoglobulin fusion protein resulted in disease amelioration in animal models of arthritis142, autoimmune sialoadenitis143 and type 1 diabetes144. Although, baminercept (which is a humanised LTβR–IgG1 fusion protein) failed to show clinical efficacy in a Phase II clinical trial in patients with rheumatoid arthritis, its utility in Sjögren's syndrome is currently under investigation (Table 2). Recently, a new LTα-specific monoclonal antibody has completed a Phase I clinical trial in patients with rheumatoid arthritis145 but the results of the Phase II study have not yet been published. It is noteworthy that none of the above studies have stratified patients for the presence of ELSs in the disease-affected tissue, which would probably influence the reported extent of clinical responses. However, the baminercept trial in Sjögren's syndrome includes a salivary gland biopsy sub-study, which will inform on ELS modulation after treatment.

Table 2 Novel therapies targeting ELSs in ELS-positive autoimmune diseases

Blockade of CXCL13, CCL19 and CCL21 or their receptors CXCR5 and CCR7 is another potential therapeutic strategy that has shown some promise in animal models. CXCL13 blockade ameliorated collagen-induced arthritis146 and autoimmune sialoadenitis147 but not autoimmune diabetes148. Of note, although CXCL13 inhibition altered the structure of ELSs, it did not influence their function or the incidence of diabetes148. To date, no developmental compound targeting this pathway has been tested in human clinical trials. However, there is evidence that blocking immune cell recirculation to ELSs in the context of autoimmune disease is a promising therapeutic strategy.

Fingolimod (also known as FTY720) — which blocks the egress of lymphocytes from SLOs via functional antagonism of the sphingosine-1-phosphate receptor — has shown favourable efficacy in relapsing–remitting multiple sclerosis149 and has received approval from the US Food and Drug Administration (FDA) for this indication. Fingolimod selectively reduces the frequency of circulating CCR7+CD4+ T cells (that is, naive and central memory T cells) and traps this subset of T cells in SLOs, which prevents their migration into the central nervous system150.

Given the role of TFH cells in ELSs, there is also growing interest in blocking key TFH cell-related signatures — for example, the co-stimulatory molecules ICOS and ICOS ligand (ICOSL) or the key TFH cell-related cytokine IL-21. Blockade of IL-21–IL-21R signalling ameliorated disease in animal models of arthritis151 and SLE152, and an anti-human IL-21 monoclonal antibody (NNC0114-0006) has recently completed Phase II clinical trials in rheumatoid arthritis and a Phase I clinical trial in SLE, but no published data are currently available (Table 2). Pharmacological ICOS blockade has also shown benefit in experimental arthritis153 and a mouse model of lupus154 via inhibition of TFH cell responses. Similarly, the concomitant inhibition of ICOS and CD40L co-stimulation protects non-obese diabetic mice (NOD mice) from diabetes155. As a consequence, an anti-ICOSL monoclonal antibody (AMG557; also known as mAb-3B3) has recently entered Phase I clinical trials in SLE (Table 2). The co-stimulatory signalling that is mediated by ICOS–ICOSL and CD40–CD40L is also required for TH17 cell survival48,153 and, as such, may contribute to the formation or maintenance of ELSs within inflamed tissue. Thus, it will be extremely interesting to investigate whether directly targeting the IL-17 pathway with novel therapeutics (such as secukinumab, ixekizumab and brodalumab) that are currently in late-stage clinical development (Table 2) will modulate ELSs in the context of autoimmune disease. Finally, it remains to be established whether therapeutics that target factors that are associated with B cell survival and proliferation disrupt ELSs or impair their role as functional niches for autoimmune B cell activation. In this regard, data are eagerly anticipated from a biopsy-based Phase II clinical trial in Sjögren's syndrome with belimumab, which is an anti-BAFF monoclonal antibody that has received FDA approval for SLE (Table 2). Such studies should elucidate whether BAFF inhibition is sufficient to disrupt ELS functionality in the salivary glands.

Concluding remarks and future directions

An increasing number of clinical investigations emphasize that ectopic lymphoid neogenesis is a common occurrence at sites of inflammation, whereby ELSs form an integral part of the immune response to infections, tumours and autoantigens. Although experimental animal models have defined certain mechanistic aspects relating to ELS development, function and maintenance, parallel investigations in human conditions remain relatively sparse by comparison. It is important to determine whether the clinical consequences of ELS activities are beneficial (for example, immunity to infections or cancer) or detrimental (for example, the promotion of autoimmunity or graft rejection). Clinical trials in conditions that are typically associated with ectopic lymphoid neogenesis may broaden our understanding of ELS involvement in these diseases. An increasing number of novel biologics that target mediators that are central to ELS biology are now entering the clinical arena. However, the key goal is to identify specific molecular signatures that will predict at an early stage of the disease process whether ELSs will form and that can be used to inform decisions regarding the most appropriate therapy for individual patients. For example, there may be little point in targeting pathways that promote ELS development if these structures have already become a prominent feature of the underlying pathology by the time a patient is diagnosed. Instead, the use of agents that block the long-term maintenance of ELSs within these inflamed tissues may prove to be a more suitable approach in this scenario. To realize this ambition, clinicians need to consider more precise diagnostic approaches in their patient management protocols. The use of ultrasound-directed biopsy techniques has revolutionized oncology treatment and the introduction of molecular pathology in autoimmune conditions is now being used to enhance the mechanistic understanding of the crucial pathways that are driving disease. For example, ultrasound-guided synovial biopsy studies in patients with rheumatoid arthritis have highlighted the clinical heterogeneity of synovitis within the disease-affected tissue116. The tremendous progress in miniaturized technologies and high-throughput 'multi-omic' approaches will enable clinicians to move away from defining disease mainly on the basis of symptoms and signs, and towards a new taxonomy that integrates molecular signatures into systematic algorithms to map the observed pathology onto existing disease classifications. This information will lead to improved patient stratification, better and more appropriate clinical management and an increased likelihood of remission, and will eventually fulfil the 'promise of personalized medicine'.