Comment

It seems that wherever one turns in mammalian biology, from conception through normal development to death, cytokines of immunological interest have become inserted as important regulators of tissue activity. Or have they? The truth is that at various stages the evolving immune system has borrowed widely to turn existing regulatory and effector molecules to new advantages.

The tumour necrosis factor (TNF) and TNF receptor (TNFR) families

The TNF and TNFR families of molecules1-3 consist of the ligands TNF, lymphotoxins alpha and beta (LT-α, LT-β), nerve growth factor (NGF), Fas ligand (FasL), CD27 ligand, CD30 ligand, CD40 ligand, OX40 (CD134) ligand, and 4–1BB ligand and their receptors (which, in most cases, are ligand specific). Of the ligands, TNF, LT-α and NGF are secreted products, which in each ease associate to form biologically active homotrimers (LT-α₃ in the case of LT-α). Both TNF and LT-α₃ are approximately equally favoured ligands for two receptors, TNFR I (55 kD) and TNFR II (75 kD). LT-β is a membrane protein that is not secreted and has not been observed as a homotrimer. Expression of LT-β on the cell surface seems to require co-expression of LT-α and their association in the heterotrimers LT-α₂β₁ (minority) or LT-α₁β₂ (majority). LT-α₂β₁ may signal via either of the TNF receptors but LT-α₁β₂ signals via a specific receptor (LTβ-R; also known as TNFR related protein, TNFR_rp).

The TNF ligand family are, in general, inducers of either cell proliferation or cell death and the receptors for some are expressed by a wide range of tissues. It would not, therefore, have been surprising if loss of function in ligand/receptor pairs in this family by natural mutations or through gene targeting had caused significant defects in general embryonic morphogenesis. In fact, this has not been the case, even when a double knockout of the genes encoding TNF-α and LT-α has prevented all signalling by these cytokines or by LT-β (table 1). Nevertheless, loss of function in genes encoding the ligand/receptor pairs FasL/Fas; TNF-α, LT-α/TNFR I and LT-α,β/LTβ-R is accompanied by morphological abnormalities that are limited to the lymphoid tissues. In the cases of TNF-α and LT, it is uncertain whether the morphological changes are primarily defects in morphogenesis or whether they are secondary to functional deficiencies in the behaviour of various differentiated cells of the lymphoid system (as seems to be the case in natural mutations of either Fas or FasL).

Organogenesis of Peyer’s patches in mice lacking TNF-α or the TNFR I

A recent publication by Pasparakis and colleagues examines the gross anatomical, histological and cellular abnormalities that follow targeted disruption of the genes encoding TNF-α and one of its receptors, TNFR I (p55TNF-R). From the foregoing discussion (table 1) it can be seen that TNF-α^−/− and TNFR I^−/− mice need not exhibit equivalent phenotype because the TNFR I also serves LT-α. Both homozygous knockout mutants (on mixed 129Sv X C57BL/6 background) had normal thymic morphology and function, grossly normal lymph nodes (LN) but had Peyer’s patches (PP) that were reduced both in size and number. Peyer’s patches were flat, they appear from the illustrations to lack subdivision into distinct follicles, while distortion of normal morphology (lack of clear follicle associated epithelium and organised patch associated villi) was more evident in the TNFR I^−/− mice. Nevertheless, the PP contained both B and T cells and there was some segregation of T cells into small presumptive interfollicular areas. Dendritic cells were present within these concentrations of T cells and also within the rudimentary dome areas, an interesting observation in light of the importance of TNF-α in the maturation of dendritic cells and their migration to lymph nodes.4 On the other hand, in both PP and LN, as well as in spleen,5 there was a failure of B lymphocytes to organise into primary follicles, a failure of germinal centre development and an absence of follicular dendritic cell (FDC) networks within areas of B cells.

Pasparakis et al believe that the primary defect in all of the secondary lymphoid tissues in their mice lies downstream of primary morphogenesis and is related more to a role of TNFα/TNFR I in the migration/organisation of B cells to form follicular structures. In particular, they favour a key role for this ligand/receptor pair in the differentiation or organisation, or both, of FDC into a network around which the follicles can develop. If correct, it is a fine point as to whether such a defect is morphogenetic (ie, failure of mesenchyme to commit to the FDC lineage during embryogenesis) or functional (failure of FDC to respond to signalling through TNFR I) and it is relevant that FDC express TNFR I and also make TNF-α.6 It is also legitimate to question whether a reduction in PP number does not, at face value, indicate at least a partial failure of morphogenesis. It is noteworthy that initial reports on PP phenotypes in TNF^−/− mice5 and TNFR I^−/−mice have been revised.7 ,8 After initial reports in which it was stated that PP were absent in both gene targeted mutants, it seems now that in each case, PP were in fact present although small and difficult to identify. This conveys a clear message that in any mutation that affects the development of PP, anything less than complete serial sections of the small intestine would fail to provide convincing evidence that small anlagen have not been missed and that PP morphogenetic foci are not present in normal numbers. It is not clear from the studies on TNF-α^−/− or TNFR I^−/− mice whether failure to signal via the TNFR I (by TNF and/or LT-α) reduces the number of anlagen (ie, is important for the initiation of the correct number of anlagen) or whether its role lies downstream and affects the subsequent development of those anlagen. Furthermore, the analysis to date is too crude to allow a conclusion that the defects in the TNF-α^−/− and TNFR I^−/− mice are equivalent and therefore, that LT-α₃ and LT-α₁β₂ produce no PP phenotype.

A recent study on an independent TNF-α^−/− knockout, this time performed in C57BL/6 mice, is informative on three counts.9 Firstly, at least in spleen, areas of B cells do contain scattered cells stained by monoclonal antibody FDC-M1, indicating that FDC indeed differentiate in TNF-α^−/−mice and that their absence is not the primary defect in PP morphogenesis in these animals (their functional state is another matter). Secondly, in this mouse strain there was only a small difference in the numbers of PP between TNF-α^−/− and wild type mice, although morphologically they resembled the structures described by Pasparakis et al. This observation suggests that TNF-α is not a primary morphogenetic stimulus for the development of PP. It raises a further question: do morphogenetic events lay down a template for coordinated epithelial differentiation (follicle associated and patch villus associated) and is this at least partially under the control of TNF-α, or does subsequent lymphoid development and function shape the epithelial architecture and is this the way that TNF-α exerts its effect on PP morphology? My own response to this question is to ask which is the most unique feature of a PP. To me, it is the follicle associated epithelium and by this reasoning, my speculation is that primary organisers in PP organogenesis will create focal fields of epithelial differentiation, which will in turn induce differentiation of underlying mesenchyme. Finally, Körner et al 9 observed that the PP “homing receptor” MadCAM was expressed by vessels in PP. This supports the contention of Pasparakis et al that failure of lymphocyte entry into PP is not responsible for the PP abnormalities in TNF-α^−/− mice. Körner et al 9 conclude that early morphogenesis of PP is controlled mainly by LT (see later), whereas the role of TNF-α is related more to the organisation of lymphocytes within the edifice so created. It would be of interest to know whether PP morphology in TNF-α^−/− mice could be restored to normal by adoptive transfer of recirculating mature lymphocytes (ie, whether the defect lies in the response of stromal components of the PP or of lymphocytes to signalling via TNF-α.)

The importance of lymphotoxins in organogenesis of Peyer’s patches

A more severe phenotype would be expected by disruption of genes which impinge more globally on the functions of the TNF/LT family of ligands (table 1). Targeting the TNFR I gene (TNFR II does not seem to be relevant in the context of this discussion) can block the actions of TNF-α, LT-α₃ and LT-α₁β₂but signalling can still occur via membrane bound LT-α₁β₂. Peyers patches in TNFR I^−/− mice seem to be reduced in number8and they have more severe disturbance of architecture than those in TNF-α^−/− mice (Pasparakis et al).8 LT-α₃ and/or LT-α₂β₁ may, therefore, have a role in the formation of normal PP, but they do not seem to be the primary organisers of PP anlagen. If all membrane bound LT activity is abolished (LT-α^−/− mice), there is a striking phenotype, with agenesis of PP and all LN, disrupted splenic white pulp architecture but normal thymic development. There seems no question but that membrane bound LT is vital in the organogenesis of secondary lymphoid organs, including PP. This conclusion is borne out (almost completely) by studies in LT-β^−/− mice, where there is selective loss of membrane associated LT. These mice lack PP and most LN but they do have mesenteric LN and variable numbers of cervical LN (albeit with abnormal structure). As might be expected, the coup-de-grâce (elimination of both TNF-α and LT-α) also leads to agenesis of PP and all LN.9 ,10

It is interesting that lymph nodes vary in their requirements for LT-α₁β₂ to induce organogenesis. A fascinating addition to this story is the use of maternally administered LTβ-R-Ig fusion protein to block membrane bound LT at various stages of gestation within the fetus.11Administration of fusion protein as early as nine days’ gestation did not inhibit morphogenesis of mesenteric LN, whereas development of other LN was blocked even if fusion protein treatment was delayed until day 12 (brachial LN) through to day 16 (popliteal LN). In contrast, morphogenesis of PP was prevented even when treatment was given on day 18 of gestation. These results correlate with the late maturation of the gut in rodents and the appearance of PP anlagen in the late prenatal period.12 More needs to be done to examine the precise time at which genes that encode TNF-α, LT-α, LT-β, and their receptors are transcribed in the local mesenchyme that gives rise to LN and PP.

IgA production in gene targeted mice

Mice with selectively targeted genes encoding LT-α13 ,14 or LT-β14 or with combined knockouts of the genes encoding TNF-α and LT-α9 ,10provide a unique opportunity to assess the importance of PP for IgA production (although other sites, such as bronchus associated lymphoid tissue, could also be affected). LT-α^−/− and LT-β^−/− mice have very low concentrations of IgA in serum and in faeces, although serum concentrations of IgM and IgG are relatively normal. Mesenteric LN, present in LT-β^−/−mice but not in LT-α^−/− mice, do not seem to confer a greater capacity to produce IgA. The presence of normal concentrations of serum IgG in these mice suggests that the defect in IgA production results from agenesis of PP rather than to a more general failure to support isotype switching. In TNF-α/LT-α^−/− double knockout mice, serum IgA and mucosal IgA plasma cells are even more profoundly depressed and there is also reduction in serum IgG concentrations.9 ,10 These findings may indicate that some switching to the IgA isotype can occur at other sites (eg, the spleen) in LT-α^−/− and LT-β^−/− mice.

In contrast, the structural disorganisation in PP of TNF-α^−/− mice does not lead to reduction in serum IgA concentrations.9 The same seems to be true for TNFR I^−/− mice, which have been reported to lack PP8 but seem now to have small and disorganised PP (Pasparakis et al).

Implications for mucosal immunologists and gastroenterologists

What lessons can be learned from gene targeted mice with abnormal or absent PP? Firstly, the mutants in which PP are affected (relatively) selectively offer opportunities for exploring the roles of these organs in responses to immunising and tolerising antigens delivered orally. Secondly, the mutants with absolute agenesis of PP offer an interesting model in which to explore the permeability of the villous mucosa to antigens and macromolecules, independent of PP (and, presumably, isolated lymphoid follicles). Thirdly, how do the results in rodents relate to humans, in whom it is reported that numbers of PP increase until puberty?15 Does this represent delayed morphogenesis de novo, or does it represent the activation of latent anlagen (perhaps by the action of cytokines such as TNF-α)? Do PP that involute in later life disappear, or can they be reactivated? Our own recent studies suggest that a range of isolated lymphoid aggregations that are present in normal human small intestine can be activated by microbial colonisation. Finally, how do these findings on early morphogenesis relate to the “lymphoid neogenesis” observed in transgenic mice when LT-α is expressed ectopically under a tissue specific promoter?16 Is follicular hyperplasia in the small intestine, follicle formation in the stomach in atrophic gastritis or the mononuclear infiltrate in Crohn’s disease a recapitulation of morphogenetic events that occur normally under control of cytokines during early development? Love them or hate them, cytokines rule.