Induction and recirculation of mucosal lymphocytes
Classic papers from Gowans,15–17 Cebra18 and their colleagues showed that mucosal immunoglobulin A (IgA) responses were induced in the Peyer's patches, and that IgA plasmablasts recirculate through the lymph and blood stream to home back to the intestine to mature into IgA-secreting plasma cells in the lamina propria. Similarly, a proportion of intestinal T cells (CD8αβ+ or CD4+ T-cell receptor (TCR)α/β+) are induced in the Peyer's patches and follow a similar route.19–21
These early experiments used adoptive transfer of lymphocytes to distinguish the source of lymphocyte induction.18 Induced lymphocytes were shown to recirculate by two methods: (i) by cannulating the thoracic duct to obtain samples of efferent intestinal lymph and (ii) by surgically modifying the small intestine in rats to create Thiry Vella loops. These segments of small intestine were removed from the main stream with openings onto the skin of the animal, although the vascular and lymphatic connections were not disturbed. Immunization with cholera toxin in one intestinal segment was shown to result in an IgA response in another which had not been exposed to the antigen.
The cytokine-mediated mechanisms of IgA induction were initially studied in cell culture during non-specific B-cell stimulation, in which the media was supplemented by purified cytokines, and their spontaneous effect on class switch recombination to IgA was measured.22–26 In the 1980s and 1990s the cytokine and cellular interactions required for an IgA class switch were demonstrated in vivo using murine strain combinations with spontaneous and targeted immunodeficiencies. In some cases the readout was spontaneous production of IgA, which is defective in mice deficient in transforming growth factor-β (TGF-β) signalling (TGFβRII–/–) and the tumour necrosis factor (TNF) family member A proliferation-inducing ligand (APRIL).27 In other studies a specific stimulus has been used to induce IgA: this is usually cholera toxin,28,29 which is a powerful mucosal adjuvant, and the functional outcome of mucosal immune induction can be tested by neutralization of fluid accumulation within hours of injecting a test dose of cholera toxin into a ligated intestinal segment.30–32 The cholera toxin response requires T-cell help, as it is defective in CD4–/– mice33 and animals that are major histocompatibility complex (MHC) class II deficient. Cholera toxin responses are also reduced in interleukin (IL)-4–/– mice34 as well as cytotoxic T-lymphocyte antigen (CTLA)-4-Hγ1 transgenic mice that express a CTLA-4 protein construct under the control of the immunoglobulin heavy chain promoter, which blocks CD28⇆CD80/86 costimulation signals between T cells and antigen-presenting cells.35 This led to the conclusion that the process of IgA induction was substantially T-cell dependent in vivo. However, it remained paradoxical that many of the models (CD4–/–, IL-4–/–, and CTLA-4-Hγ1 transgenics) in which cholera toxin induction of IgA was defective, nonetheless had relatively normal numbers of IgA-secreting plasma cells in the intestinal mucosa.33–35
The importance of dendritic cells for the IgA class switch in addition to interactions between B and T lymphocytes was also initially examined in ex vivo cell culture.36–39 Antigen-presenting cells have been shown to stimulate the class switch (to IgG and IgA) probably through interactions between the TNF family members B cell activating factor (BAFF) and APRIL on the antigen-presenting cells and the BAFF receptor on B cells.40,41In vivo APRIL-deficient mice have decreased spontaneous levels of IgA and reduced specific switching to T-dependent and T-independent immunization protocols.27
Induction of IgA against commensal bacteria
In contrast to toxin induction of IgA, the same process triggered by commensal bacteria is not exclusively CD4-dependent. Measurement of total IgA in mice that are deficient in T cells as a result of targeted deletions of the β and δ chains of the T-cell receptor, showed that the amount of IgA secreted was reduced to about a quarter of that in wild-type animals but there remained a T-cell independent component.42 The binding specificities to Enterobacter cloacae (a dominant aerobe of the commensal intestinal flora in the Zurich colony of specific pathogen-free mice) were identical whether studied in wild-type or T-cell deficient animals.42 In animals deficient for MHC-class II, IgA content has also been shown experimentally to be normal despite disruption of cognate interactions between antigen-presenting cells and T cells.43 T-cell independent mucosal IgA responses have also been found to confer protective immunity when C57BL/6 × 129 mice are challenged with rotavirus.44,45 Humans with defective CD40-mediated signalling have also been described with normal or high levels of serum IgA.46,47
Studies of IgA sequences also suggest indirectly that the response to commensal bacteria does not depend on conventional germinal centre reactions in which the affinity of the antibodies is improved by sequential accumulation of somatic hypermutations.48 This is unlikely to merely reflect excess antigen binding to B-cell receptors, since germinal centres form selectively in Peyer's patches and mesenteric lymph nodes in mice in which the B-cell receptor (BCR) has been deleted, but a low level antigen-independent constitutive signal is delivered by B-cell expression of the Epstein–Barr virus protein LMP2A containing an immunoreceptor tyrosine-based activation motif.49 Experiments with antibiotics in BCR-deficient LMP2A mice suggest that BCR-independent signals from the intestinal flora are sufficient to drive germinal centre formation in the mucosal lymphoid system, although the details are unknown.49 In fact, even germinal centre formation is not obligatory for IgA induction, which occurs efficiently in the TNF receptor I-deficient strain.42 Sequence analysis of the alpha heavy chain and spectratyping of the CDR3 region length also shows that the repertoire of the (VHα) variable region in Peyer's patch or lamina propria tissues of mouse and man is surprisingly restricted given the diversity of the commensal flora.48,50 Somatic mutation of intestinal VH genes increases with age in humans51 although we do not know whether this has occurred by classical affinity maturation of the BCR or alternative signals from intestinal bacteria. Overall, the observations suggest that induction of IgA by commensal bacteria is rather a primitive system in which the production of large amounts of antibody against bacterial surface molecules with relatively low affinity, yet broad specificity, is useful to limit their local colonization or penetration through the epithelial layer.
In adult mice there are two sources of B-cell precursors.52 The bone marrow has stem cells that give rise to the conventional lineage of (B2) B cells. There are also precursors in the pleuroperitoneal cavities for a different (B1) lineage which are distinguished from B2 cells by higher levels of B1 staining for surface IgM, Mac-1 and CD5, and weaker staining with antibodies against B220 and IgD. Actually, the independence of these lineages is controversial, because experiments in mice with B cell receptor signalling abnormalities or a fixed antigen-binding specificity show that whether B cells exhibit the B1 or B2 phenotype is dependent on the specficity and strength of signalling from the B cell receptor.49,53 IgM antibodies derived from B1 cells are reactive with polysaccharide microbial antigens (induced in a T-independent fashion), and are encoded by unmutated VH genes.52 The contribution of B1 cells to intestinal IgA has been estimated by reconstitution of lethally irradiated animals with sources of B1 and B2 cells where there are distinctive allotypic differences in secreted immunoglobulins. These experiments give estimates of about half the intestinal IgA and most of the T-independent IgA being B1 derived.42,54,55 In a different experimental system, germ-free allotype chimeric mice were generated by repetitive antibody depletion of endogenous neonatal B cells followed by transfer of peritoneal cells.56 The final chimeras still had considerable numbers of recipient B1 cells in the peritoneal compartment (15–39%), so the system would underestimate the B1 contribution to secreted antibodies on the basis of their allotypic differences. However, 56–70 days after recolonization with bacteria the donor allotype contribution to intestinal IgA was less than 15%. In a third experimental system the intestinal IgA in MHC class II-deficient animals was studied. Here, the levels of intestinal IgA were relatively normal, despite the T-cell deficiency and disruption of cognate B–T interactions, but intestinal IgA became very reduced when the animals also carried the xid mutation resulting in deficient B1 cells.43 These inconsistent results leave open the exact contribution of B1 cells to intestinal IgA in mice. In man, CD5+ B cells producing polyspecific antibodies form 15–20% of the adult B-cell repertoire and constitute most neonatal B cells, although there is no significant pleuroperitoneal B-cell precursor population as in mice.57–59 It is not possible to assess the contribution of these B1-like cells to intestinal IgA directly in man.
Another unresolved issue is where class switch recombination might occur for B1 cells.60 Flow cytometry of the characteristic B2/B1 markers shows that effectively all Peyer's patch B cells have the B2 phenotype and B1 cells in the peritoneum are IgM+. In animals that are deficient for the TGF-β receptor (TGFβRII–/–), B1 lymphocytes do appear in the Peyer's patches, although whether this is prolonging transitory presence in normal circumstances is unknown.61 Class switch recombination has been described in the intestine outside the Peyer's patches62 although isolated lymphoid follicles may be contributing to this process.63 It is has also been suggested that B1 class switch recombination may take place in the mesenteric lymph nodes.48,64
We found that significant intestinal IgA only occurred in strains with some B-cell structures in the intestine, although these could be very disorganized, for example without follicular dendritic cells and germinal centres in mice deficient for the TNF receptor I.42,65 Regardless of relative B1/B2 contributions, the IgA response has a restricted VH repertoire of intestinal immunoglobulin α heavy chains in mouse48 and man.50 This supports the primitive nature of the overall IgA response, mostly induced in the presence of commensal bacteria, as a high capacity, broad specificity, low affinity system, not dependent on conventional B–T interactions for affinity maturation. The system is eminently suitable for producing antibodies that will bind to a very diverse bacterial flora with multiple redundant surface epitopes.42,66
Intestinal IgA can be experimentally induced in C57BL/6 wild-type mice by repeated administration of live bacteria into the intestine.6 Unlike the T-dependent cholera toxin response, which induces serum IgG and IgE in addition to IgA67 the commensal conditioning method of mucosal stimulation is entirely specific for IgA. Conditioning doses of bacteria result in loading of live bacteria in Peyer's patch dendritic cells (DC) – when these DC are isolated and cultured with naive mesenteric B ± T cells, both surface expression and secretion of IgA by the B cells is induced. Neither the in vivo conditioning response nor the ex vivo IgA induction will work when the animals are treated with heat-killed bacteria.6
The mechanisms for the specificity of IgA induction, as opposed to class switch recombination to other isotypes are not clearly understood. TGF-β signalling from diverse mucosal cell types is clearly important from models both in vitro22 and in vivo61 and there may also be direct interactions between antigen-presenting cells and B cells enhanced by the TNF family members BAFF and APRIL.27,40 These unconventional mechanisms of class switch recombination may also be able to occur without prior B-cell receptor engagement.65,68
Despite the enormous amount of IgA that is secreted daily across the intestinal epithelium, there are very few studies that address its function in relationship to commensal intestinal bacteria. Mice that are genetically deficient in the polymeric immunoglobulin receptor (pIgR) that transports IgA and IgM across the epithelial cell layer69–71 have a protein-losing enteropathy in which serum proteins are lost into the intestinal lumen as a result of damage to the paracellular permeability barrier.72 Two functional mechanisms of mucosal IgA secretion have been described. First, IgA antibody-coated commensal bacteria are excluded from penetrating the intestinal epithelium. This observation came from experiments in which the intact mucosa was challenged either by recolonizing germ-free animals or by delivering an experimental dose of intestinal bacteria to animals that already had an established commensal flora.6,9 Overall therefore IgA is protective against penetration of luminal bacteria, presumably by limiting their motility or access to the epithelial surface, but it is possible that IgA receptors of M cells facilitate sampling of live bacteria in the Peyer's patches and isolated lymphoid follicles.73–75 Second, in the absence of IgA, luminal densities of the commensal organisms are not properly controlled.63 The evidence for this comes from activation-induced cytidine deaminase deficient (AID–/–) mice, which have an anaerobic overgrowth in the lower intestine. AID–/– mice are deficient in IgA (and other class-switched isotypes) and affinity matured IgM as a result of defective class switch recombination and somatic hypermutation.
These mechanisms of IgA induction against commensals are only a component of the overall way in which the immune system adapts to the presence of such a large load of intestinal bacteria. In the following section we will discuss host–commensal bacterial mutualism in its wider immune context.