Germinal center responses at barrier organ sites

Abstract

In this review, we detail the features of barrier germinal centers (GC) that form in tertiary lymphoid structures (TLSs) within non-immune organs that host a local microbiome, and posit a framework of immunity where TLS-GCs created at barrier sites are a key part of normal humoral immunity, and that they form the first layer of an intended multi-layered and increasingly defensive adaptive immune system, providing complementary tissue-specific responses, alongside the local secondary lymphoid organ network and systemic immune responses.

Introduction – An overview of TLS GCs at homeostasis and in disease

Recent evidence has identified structures resembling germinal centers (GCs) in situ at a range of barrier sites. Germinal centers are highly organized structures, typically found within secondary lymphoid organs (SLOs), which contain tightly controlled specialized zones that are maintained by well described stromal cell subsets. This exquisite structure allows for the rapid selection, expansion and mutation of B cell clones with increasing affinity to antigen presented, thus allowing for the generation of durable memory B cell and plasma cell responses, conferring immune memory and the ability to produce neutralizing antibodies (1, 2). GC reactions can also form de novo within many tissues not traditionally considered lymphoid organs. These employ a generalized mechanism of tertiary lymphoid structure aggregation, while also demonstrating tissue specific specializations dependent upon the pre-existing tissue architecture and downstream functional antibody requirements.

Autoimmune diseases of solid tissues (Hashimoto’s thyroiditis - thyroid, Sjogren’s syndrome-salivary glands, systemic lupus erythematosus – kidney, ulcerative colitis - gut), all have in common a consistent antigen depot of membrane bound self-protein, recognized as non-self (3–11). In autoimmunity, ectopic GC activity results in tissue damage and local feed-forward inflammatory loops, and auto-antibody generation often leads to morbidity (12). This antigen depot and resulting inflammation is a feature that they share with tumors, yet while autoimmune and tumor TLS likewise react to chronic complex antigen, these two classes of TLS are often diametrically opposed in their prognosis. TLS activity can act to lessen the tumor burden, by at least two mechanisms; by acting as a safe space within a tumor which otherwise causes T cells to become exhausted, and by housing B cells to create local antibodies that target neoantigens, tumor GC presence can therefore be leveraged by therapy (13–15). TLS in the context of autoimmunity have been reviewed (12, 16), as have TLS during oncological responses (17, 18), thus we will not discuss them at further detail here, instead focusing on mucosal GCs as an element of normal barrier homeostasis and tissue resident immunity, maintaining the microbiome and responding to acute pathogen encounter. (Figure 1)

Figure 1 -. An overview of barrier site germinal centers.

Mucosal sites including the lungs (A), upper airway (B), skin (C) and gut (D) all have a local microbiome, thus require immune surveillance and local response capacity. Tissue specific fibroblast architecture e.g. branchpoints of vessels in the lung, or hair follicles in the skin represent excellent areas to host a TLS. While the level of complexity for TLSs across sites can vary, they share a unifying principal – to provide a safe space to maintain B cells proximally for microbiome sampling.

The gut – mesenteric lymph nodes, Peyer’s patches, cryptopatches and isolated lymphoid follicles

The gut is well served by the secondary lymphoid organs, collectively termed the “GALT” -gut associated lymphoid tissue (Figure 2). The GALT contains mesenteric lymph nodes (mLNs) which act as any peripheral lymph node, with afferent lymphatics draining from the gut allowing for antigen presenting cells and primed B cells to enter, and efferent lymphatics allowing plasma cells to exit, ultimately seeding the intestinal lamina propria via transit through the blood (19, 20). The GALT also contains Peyer’s patches (PPs) which uniquely directly sample the intestinal lumen via microfold (M) cell uptake, followed by loading onto proximal antigen presenting cells within the subepithelial dome (SED) (21–23). PPs present a unique scenario – secondary lymphoid organs that develop in utero like lymph nodes (23), yet unlike lymph nodes, essentially do not contain primary follicles even in absence of disease, because of their continuous exposure to the diverse and everchanging gut antigen repertoire. By this logic, they are somewhat more analogous to GC-containing TLSs, with their development being dependent on primarily bacterial antigens – a hypothesis supported by the observation that germ-free mice have a reduced frequency of PPs, inactive lymph nodes and small, often indistinct GCs, and a significant reduction in intestinal IgA (24, 25). While pathogenic bacteria such as salmonella invade the gut via the PPs, commensal bacteria specifically Alcaligenes species can be found inside PPs, with their presence promoting mucosal IgA production (26, 27). In germ-free situations, any remaining germinal centers and resultant IgA PCs generated arise in response to dietary antigens, specifically from TLR contamination of food (25, 28). GCs from germ-free mice contain a markedly higher frequency of highly-selected B cells, and these successfully expanded clones are enriched for public IgH clonotypes, found across multiple individual mice, indicating that the BCR is driving selection, albeit against a much reduced breadth of antigens (29). These germ-free GCs also contain a higher frequency of IgG1+ cells, suggesting that microbial colonization ultimately results in the dominance of IgA in the gut (29). It should be noted that M cell frequency is reduced in germ-free mice, and this must have a deleterious effect on the development and activity of PPs (30). Recent work has shown that the microbiota plays a role in recruitment of naïve B cells into the PPs, not just in promotion of GC activity (31).

Figure 2 – Germinal centers in the gut.

The gut microbiome is the most complex within the mammalian body. It therefore has a highly developed system of TLSs, which are in a chronic state of activity. Peyer’s Patches (PP) use M cells to pull antigen into the subepithelial dome (SED), where antigen presenting cells (APCs) are loaded, leading to both constitutively active GCs within the PP, as well as a steady flow of antigen to mesenteric LNs (mLNs). This provides IgA plasma cells which re-migrate and become resident within intestinal villi, as well as IgG secretion systemically. An additional mechanism of individual lymphoid follicle (ILF) expansion can be exploited if pathogenic bacteria are encountered to provide fully local rapid antibody responses.

PPs are a constitutively present, significant source of lamina propria IgA+ plasma cells, predominantly via GC-derived and to a lesser extent non-GC routes (32–34). However, there are additional structures present in the small and large intestine which can host GC B cells and plasma cells. Isolated lymphoid follicles (ILFs) have been identified in the intestines of both mice and humans, developing after birth (35–37). ILFs are closely packed clusters of lymphocytes found within the villi and are generated from cryptopatches (CPs) – small clusters of lymphocytes found at the base of intestinal crypts (38, 39). When ILFs mature, they can develop an overlying follicle-associated epithelium, which contains M cells, and host GCs, resembling a miniature PP (40). In either RORγT^−/− or Lymphotoxin Alpha (LTA)^−/− mice, which both lack PP and ILFs, reconstitution with either RORγT or LTA sufficient bone marrow allows for ILF development and restores IgA production, demonstrating that ILFs can produce IgA (41, 42).

These three functionally related structures have overlapping requirements for formation; ILFs can develop in PP deficient mice, in an IL-7R independent manner. Loss of lymphotoxin alpha and NFKB signaling (aly/aly mice) results in loss of ILFs and PPs, but not CPs. Common gamma chain mutant mice lack all 3 tissues (35, 40). ILF development is not pre-planned, and their number can increase dynamically in response to changes in microbiota. For example, in AID^−/− mice lacking high affinity IgA, anaerobic bacteria are expanded, and ILF hyperplasia occurs; antibiotic treatment of these mice to reduce anerobic bacteria burden results in hyperplasia resolution (43, 44). These ILFs can host self-sufficient GCs, being able to induce Activation-induced cytidine deaminase (Aicda) to facilitate class switch recombination (45). IgA sequencing of clones found in ILFs shows very low overlap with PP clones, further suggesting that these two systems operate independently, responding to different antigens respectively (46).

TLS in the upper airway – NALT and nasal turbinate resident cells

As the nose represents the first barrier of the respiratory system, immune cells are active both in the nose, and the post-nasal space (PNS). The PNS is surveyed by conventional lymphoid organs, termed the “nasal associated lymphoid tissue” (NALT), which in humans consists of Waldeyer’s ring of tonsils. Beyond the defined NALT structures, there is ample reporting of tertiary structures forming in the nasal subepithelium, and here we will contrast GC responses arising in these two neighboring arms of upper airway immunity. It should be noted that the adenoid and palatine tonsils themselves have somewhat different functions – each having their own microbiome (47, 48).

This nasal immune system is clinically relevant – infections such as Naegleria fowleri and viruses such as VSV (vesicular stomatitis virus) can infect the brain via the olfactory system, and mice who have circulating broadly neutralizing antibodies against VSV are not protected when VSV is introduced directly to the olfactory system – only locally embedded IgA producing plasma cells can provide protection (49, 50).

A direct comparison of mouse and human NALT is complicated by the fact that mice do not have tonsils, instead having a pair of tissue aggregates at the base of the nasopharangeal duct, as well as diffuse immune cell activity along the nasal epithelium (51).

Work in humans studying upper airway immune responses is often carried out by endoscopic brushing or biopsy, however attention must be paid here, as it is not always clear whether this results in sampling of the TLS forming in the epithelium, or whether the NALT has been sampled directly – due to reaching the adenoid tonsil. This is especially true when washes and brushes which do not maintain tissue integrity are used.

A recent paper highlighted this, by purposefully sampling the midpoint of the nasal passage, and the adenoid tonsil specifically, demonstrating that GC B cells are found only within the tonsil, while memory B cells and plasma cells are enriched for IgA expression within the nasal turbinate (52). Sampling of the NALT during SARS-COV-2 infection, and in convalescence has shown that GCs form, and PCs, including IgG, IgA and IgD+ subsets line the sub-epithelial space, demonstrating that the NALT can both mature the GC B cells, and then host the resultant plasma cells to enable host defense regardless of any nasal space responses (53).

As mentioned, mice may differ from humans in this respect, as either influenza infection, or nitrophenol-conjugate based immunization, when restricted to the upper airway, can elicit the formation of TLS containing GC B cells, within the nasal turbinate, even under conditions in which migration of B cells from the NALT to the turbinate was blocked (54). This suggests that in mice, GCs can arise locally. Interestingly, this paper concludes by sampling both superior turbinate, and deep turbinate/adenoid in a human cohort, finding GC B cells in both locations – albeit with a skewing towards IgG in the adenoid, and IgA in the nasal passages (54). There is therefore some uncertainty in the literature around whether humans constitutively host GCs in the nasal mucosa or not. An older paper which sampled 150 nasal tissue blocks from children who died within 2 years of age found that disseminated immune follicles are found in 38% of the children studied (55).

While the Gailleton paper (54) demonstrates that the NALT is not dependent on adenoid seeding for its existence, this does not rule out that cells can traffic from one site to the other in normal circumstances, and work has been done to investigate such trafficking, and indeed it has been demonstrated that the NALT can seed the nasal turbinate with IgA B cells via the circulation using CCL28/PSGL-1 signaling (56).

Much like in the gut, the local nasal immune system is primarily a producer of IgA, which acts to maintain the local microbiome. Alterations to the nasal immune system – for example during chronic inflammation arising from IgE production during allergic rhinitis results in dysbiosis of the microbiome, with an increase in pathogenic firmicutes being reported (57, 58). The nasal mucosal immune system has effects which can be seen very distally; for example, chronic nasal TLS-mediated inflammation during early life causes long term microbiotal dysbiosis in the gut, and excessive IgA production by the palatine tonsil can worsen IgA nephropathy (59, 60).

Despite tremendous therapeutic potential, mucosal vaccines are hampered by the mucus found at these barrier sites – meaning that traditional vaccine formulations penetrate poorly. Recent work has sought to utilize the body’s own mechanisms of protein transport to increase vaccine uptake in barrier sites, by binding vaccines to albumin (61).

Lung – induced bronchus associated lymphoid tissue and alveolar adjacent cells

The lung has its own microbiome, the composition of which is similar to the oropharyngeal microbiome (62, 63). Normal development of the lungs is dependent on the microbiome, with germ-free mice having reduced alveolar development (64). Upon pathogenic infection, the lung draining mediastinal lymph node becomes activated, and de novo TLS structures can form, which are often termed “iBALT” (inducible bronchus associated lymphoid tissue) (65). Fine particle pollutants, such as exhaust fumes from vehicles can act as adjuvants, also promoting iBALT responses to otherwise innocuous allergens (66).

While iBALT can be found in the lungs of young children, they generally are not found at homeostasis, suggesting that ectopic GCs are not required to maintain the lung microbiome, as is seen in the gut where PPs are constitutively active (67–69).

The B cells within these iBALT structures can form ectopic GCs, and can undergo affinity maturation, therefore demonstrating a complete ability to generate local responses to pathogens, alongside the canonical draining lymph node response (70). The most advanced iBALT structures can contain follicular dendritic cells (FDCs) as well as some light zone/ dark zone polarization, albeit with less clearly defined zones than SLO GCs (70, 71). While most lung plasma cells distributed through the parenchyma are IgA+, these acute GC responses taking place in iBALT have a higher relative frequency of IgG responses (72, 73).

Following infection with viruses such as influenza, in addition to large iBALT structures, small numbers of B cells become resident in the lung parenchyma, and during re-infection, these resident memory B cells will become activated, quickly localizing to areas of infection, and differentiating into plasma cells, which can localize to alveoli (73–75). This tissue resident response allows for faster responses to previously encountered pathogens, and can occur in a T cell independent manner, unlike the PCs generated and responding from local iBALT tissues (73, 74). These lung IgG and IgA plasma cells, from either PC clusters or iBALT, are capable of long-lived survival, and expand following subsequent infection (76, 77). In models of interstitial lung inflammation, naïve B cells within the lung can become activated following interaction with CXCR5⁻ Tph (T peripheral helper) cells and macrophages, and adopt a GC-like phenotype, then class switching to IgA and undergoing somatic hypermutation, developing a unique repertoire autonomous to that of the mediastinal lymph node (78).

Additional research has shown that influenza-specific lung memory B cells are often broadly reactive, being capable of binding the post-fusion form of HA (hemagglutinin) which is only present during active viral replication and thus is not an antigen which is seen by the mediastinal lymph node (79, 80). This observation draws parallels to that of the PP and ILF of the gut – while being relatively proximal, they host unique clones of B cells, resulting in a broader pool of antigens being surveilled.

Skin – Dermal tertiary lymphoid organ structures and B cell clusters

While the skin is not a wet mucosal surface it is an important large barrier site. Plasma cells are present within the skin at homeostasis, primarily producing natural IgM (81). B cells have also been recently identified within the dermis of the skin and can form GCs not only during inflammation, but also during health. These observations were made in the context of healthy volunteers as a control for melanoma patients, and present as small dense clusters of cells, including IgG switched B cells (82, 83).

B cells can become increasingly activated during disease settings, during Hidradenitis Suppurativa (HS), germinal center derived B cells produce IgG and IgA autoantibodies against metabolic byproducts (84, 85), while during atopic dermatitis (AD), B cell infiltration into the skin increases, with B cells having a more activated phenotype (CD27+) while producing more IgE, suggesting an allergic phenotype, even if local GCs have not yet been described in this disease (86, 87).

The introduction of a new commensal bacteria onto the skin of SPF mice generates an IgM and IgG response, with the draining lymph nodes becoming activated and forming GCs. As well as LN GCs becoming active, novel tertiary structures form, and these can produce IgG2 antibodies independently of the draining LN (88). This concept has been taken even further – by introducing a new bacterium which expresses tetanus toxin proteins, mice can become systemically immune to the toxin, and repeated administrations further generated a mucosal response, including IgA responses in the nasal space (89). Immunoglobulins detected in normal skin are clonally restricted, demonstrating that a small pool of resident B cells are present (90). These publications demonstrate that the skin has a two-part autonomous humoral immune system, with both trafficking and skin-resident B cells forming local GCs, which then feed a distinct local PC population. This system can maintain barrier homeostasis, via natural IgM and IgA, and where appropriate generate high affinity IgG in response to encounters with novel pathogens.

Barrier germinal centers have universal commonalities and tissue specific adaptations

A shared feature of all barrier sites discussed here is that immune surveillance is continuously upheld, to maintain homeostasis of the microbiome. This continual monitoring is carried out by tissue specialized macrophages and DCs, such as dermal dendritic cells in the skin, alveolar macrophages in the lung, and where necessary, cross-barrier sampling can be first performed by M cells in the gut, prior to APC processing. This allows for sensing of alterations to homeostasis, resulting in recruitment of lymphocytes to local areas, and the trafficking of antigens to more distal immune hubs, such as the relevant draining lymph nodes. During such times of homeostatic disruption, for example colonization of the space with a novel pathogen, pre-existing immune tissues - draining lymph nodes, tonsils and Peyer’s patches physically expand, as they generate new germinal centers, and see infiltration of pre-existing germinal centers with new antigen and corresponding B cells (91). Concurrently, across all relevant sites, local structures form de novo, with features specialized to the organ in which they form.

Cryptopatches sprout mature ILFs in the gut, which then play host to a distinct pool of PCs. In the lung, local GCs can form, and PC clusters can independently mass within the parenchyma, proximal to airways. PCs can embed within the nasal epithelium, and clusters of lymphocytes within the skin can increase in size and complexity. All of these local structures can support IgA class-switching, therefore, a universal general mechanism is present, in which local IgA responses act to curtail changes to tissue microbiome homeostasis, and failure to do so allows for further, more distal responses, and the takeover of GCs in tissue-draining lymph nodes, often characterized by IgG biased responses, which are more extensively selected, higher affinity, and more potently neutralizing. While effective, these can be a source of tissue damage, especially in the context of autoimmunity, when overriding factors mean that the response is abnormally strong, or cannot be adequately reversed, when the initial pathogen has been addressed. Hence, a multilayered system would primarily facilitate the local regulation of mutualistic commensals by non-inflammatory IgA produced via barrier-GCs, whilst simultaneously establishing GC ‘hubs’ in nearby LNs should more extensive antibody evolution be required for responses geared towards pathogen clearance, systemic protection, or generation of memory responses.

Table 1:

Overview of site-specific TLS responses

TLS type	Antigen source	Structure and Presence	Cellular composition	Antibody output	Functional output/prognosis
Lung	Commensal bacteria, pathogens (virus/bacteria), environmental pollutants (66, 74).	Bronchial walls: iBALT with GCs form after infection, poorly organized LZ/DZ (70). Perivascular: PC/memory clusters. Absent in healthy adults, more frequent in pediatrics (68, 69).	Varies from simple B cell clusters in early iBALT, to full GCs, including Tfh, follicular dendritic cells (FDCs), NK cells. Memory B cells and PCs often in separate clusters.(70, 73, 74, 78).	Primarily IgG within iBALT, IgA within parenchyma (72, 73).	Protective during acute infection. (74–76, 80). Pathological during COPD, Rheumatoid Arthritis-associated lung disease, hypersensitivity pneumonitis (92).
Gut (small intestine)	Commensal bacteria, dietary antigens (25, 28, 91).	Intestinal submucosa: PPs constitutively present, highly organized (23). Lamina propria: ILFs dynamically present, less organized than PP (41, 42).	PPs and mature ILFs: All GC-related cells; APCs, GC B cells, Tfh cells, FDCs, macrophages, M cells within follicle associated epithelium (23, 41, 43, 46).	IgA, some IgG during disease (9, 25, 28, 33).	Protective - Microbiome homeostasis maintenance. Can be pathological during IBD (9–11).
Upper airway	Commensal bacteria, inhaled pathogens, environmental antigens	Nasopharynx: GCs present. Turbinate: PC and memory B cells (53–56).	GC B cells, Tfh cells. M cells, DCs, FDCs and macrophages in NALT GCs. Memory B cells and plasma cells in nasal turbinates (53–56).	Mix of IgA and IgG (49, 52, 53, 56).	Protective. Hyperactivated during allergic rhinitis (57, 58).
Skin	Commensal derived antigens. (85, 88, 89).	Dermis: GCs form after colonization/immunization (88, 89).	Varies from small B cell only clusters to GCs, with GC B, T cell zones and some FDCs. (81, 82, 88, 89).	Diffuse IgM PCs, TLS primarily IgG (often IgG2) (81, 82, 88, 90).	Protective. Pathological during HS and AD (84–87).
Tumor	Tumor neoantigens / tumor associated antigens (13, 14).	Within or adjacent to tumor. Variable organization depending on tumor type and maturity (17, 18).	Varies from small B cell only clusters to well-developed GCs, with GC B, PCs, Tfh, CD8s, cDC1s, Tregs (13, 14, 17, 18).	Primarily IgG (13, 14, 17, 18)	Mostly protective, correlates with maturity of GCs. (13, 14, 17, 18)
Autoimmune	Self-antigens (3, 5, 6, 12).	Location is disease dependent, GCs can be present, although dysregulated.	Mature TLS can have GCs with GC B cells, memory B cells, PCs, Tfh cells, FDCs, macrophages (7, 8, 12).	IgG/IgA – site dependent (6, 7).	Pathological (3, 5–8, 12).
Peripheral LN	Infection, vaccines (1, 2).	Highly organized, textbook GC structure, developmentally fixed location (1, 2).	Full complement of spatially organized GC cell types, including specialized stroma such as fibroblastic reticular cells (FRCs) (1, 2).	Primarily IgG. Subclass dependent on type of infection (1, 2).	Protective (1, 2).