Diverse developmental pathways of intestinal intraepithelial lymphocytes

Abstract

The intestinal epithelial barrier is patrolled by resident intraepithelial lymphocytes (IELs) that are involved in host defence against pathogens, wound repair and homeostatic interactions with the epithelium, microbiota and nutrients. Intestinal IELs are one of the largest populations of lymphocytes in the body and comprise several distinct subsets, the identity and lineage relationships of which have long remained elusive. Here, we review advances in unravelling the complexity of intestinal IEL populations, which comprise conventional αβ T cell receptor (TCRαβ)⁺ subsets, unconventional TCRαβ⁺ and TCRγδ⁺ subsets, group 1 innate lymphoid cells (ILC1s) and ILC1-like cells. Although these intestinal IEL lineages have partially overlapping effector programmes and recognition properties, they have strikingly different developmental pathways. We suggest that evolutionary pressure has driven the recurrent generation of cytolytic effector lymphocytes to protect the intestinal epithelial layer, but they may also precipitate intestinal inflammatory disorders, such as coeliac disease.

Intestinal intraepithelial lymphocytes (IELs) are long-lived resident effector cells that are interspersed between epithelial cells along the entire length of the intestine¹. They are mobile and constantly patrol the space between epithelial cells above the basement membrane, where they are poised for rapid activation of cytolytic and T helper 1 (T_H1) cell-type cytokine responses directed at infected or stressed epithelial cells. It is estimated that there are 25–50 million IELs in the mouse small intestine, or ~1 IEL per 10 intestinal epithelial cells (IECs)^2,3. Despite their shared properties and location, intestinal IELs encompass a surprising diversity of lineages. They are predominantly T cells, but they contain a mixture of subsets that we term conventional and unconventional IELs (BOX 1). Conventional IELs express the αβ T cell receptor (TCRαβ) together with CD4 or CD8αβ co-receptors and acquire effector properties after recognition of foreign antigens. By contrast, unconventional IELs express either TCRαβ or TCRγδ, lack expression of CD4 and CD8 or only express CD8αα homodimers and acquire effector properties after stimulation by self antigens. In addition, intestinal IELs include populations of group 1 innate lymphoid cells (ILC1s) and ILC1-like cells^4–6. The precise identities, developmental histories and modes of antigen recognition of these lineages are poorly defined, precluding an integrated understanding of their individual contributions in the intraepithelial environment.

Box 1. Nomenclature for intestinal IELs.

Historically, αβ T cell receptor (TCRαβ)⁺CD8αβ⁺ and TCRαβ⁺CD4⁺ intraepithelial lymphocytes (IELs) have been termed type A IELs, induced IELs or peripheral IELs, whereas TCRαβ⁺CD8αα⁺ and TCRγδ⁺CD8αα⁺ IELs have been termed type B IELs, natural IELs or thymic IELs on the basis of presumed similarities in development^4,24,25. However, recent reports (detailed in the main text) suggest that these assumptions were incorrect. Here, we refer to TCRαβ⁺CD8αβ⁺ IELs and TCRαβ⁺CD4⁺ IELs as conventional IELs to reflect the finding that acquisition of the IEL effector programme occurs after recognition of foreign antigens in the periphery. TCRαβ⁺CD8αα⁺ and TCRγδ⁺CD8αα⁺ IELs are termed unconventional IELs to reflect acquisition of the IEL effector programme in response to recognition of self ligands in the thymus or periphery.

Here, we focus primarily on recent advances that begin to unravel this complexity, defining different origins and developmental pathways of intraepithelial lymphoid lineages and describing underlying cellular and molecular mechanisms. Although most of the detailed knowledge is derived from mouse studies, we also consider human IELs to highlight similarities and differences with the mouse system (TABLE 1). A central emerging concept is that different developmental strategies have led to the generation of multiple lymphoid lineages that are dedicated to patrolling the epithelial layer and exerting rapid cytolytic function. It is likely that the diversity of intestinal IEL lineages represents the host response to strong evolutionary pressure from rapidly changing and evading pathogens, and such diversity may be a reason why multiple mechanisms can cause pathology in various intestinal inflammatory processes, for example, in coeliac disease.

Table 1.

Mouse and human intestinal IEL subsets

IEL subset	TCR repertoire	NK cell receptor repertoire	Ligands	Effector molecules
Mouse
Unconventional TCRαβ+ IEL	Diversea	CD94, NKG2 family and LY49 family	MHC class I, MHC class II, non-classical MHC class I, RAE1, MULT1, H60a and Qa-1	IFNγ, TNF and granzyme B
Unconventional TCRγδ+ IEL	TCRVγ7-enriched	CD94, NKG2 family and LY49 family	BTNL1, BTNL6, RAE1, MULT1, H60a and Qa-1	IFNγ, TNF and granzyme B
Conventional TCRαβ+ IEL	Diversea	None	MHC class I and MHC class II	IFNγ, TNF and granzyme B
ILC1	None	CD94, NKG2 family and LY49 family	RAE1, MULT1, H60a and Qa-1	IFNγ, TNF and granzyme B
CD8αα+ innate IEL	None	LY49E+ or LY49E−	TL	IFNγ and possibly others
Human
Unconventional TCRαβ+ IELb	ND	ND	ND	ND
Unconventional TCRγδ+ IEL	TCRVδ1-enriched	CD94, NKG2 family and KIR family	CD1, MICA, MICB, ULBP and others and possibly BTNLs	IFNγ, TNF and granzyme B
Conventional TCRαβ+ IEL	Diversea	CD94, NKG2 family and KIR family	MHC class I, MHC class II, MICA, MICB and ULBP	IFNγ, TNF and granzyme B
ILC1	None	CD94, NKG2 family and KIR family	MICA, MICB and ULBP	IFNγ, TNF and granzyme B
Intracellular CD3+ IEL	None	KIR family and others	ND	IFNγ and possibly others

BTNL, butyrophilin-like protein; H60a, histocompatibility antigen 60; IEL, intraepithelial lymphocyte; IFNγ, interferon-γ; ILC1, group 1 innate lymphoid cell; KIR, killer cell immunoglobulin-like receptor; LY49, lymphocyte antigen 49; MIC, MHC class I polypeptide-related sequence; MULT1, mouse ULBP-like transcript 1; ND, not determined; NK, natural killer; NKG2, NK cell receptor group 2; RAE1, mRNA export factor; TCR, T cell receptor; TCRV, TCR variable chain; TL, thymus leukaemia antigen; TNF, tumour necrosis factor; ULBP, UL16-binding protein.

^aIndividual mice may have large expansions of individual clones, but TCR sequencing of multiple mice reveals diverse TCR usage.

^bExistence of this population remains controversial.

Conventional intestinal IELs

Development

Conventional CD4⁺ and CD8αβ⁺ effector T cell subsets represent a relative minority of the total intestinal IEL population compared with unconventional IELs, and so they are discussed only briefly to compare them with innate and innate-like IELs. Conventional CD4⁺TCRαβ⁺ and CD8αβ⁺TCRαβ⁺ IELs are derived from naive T cells that encounter antigens in the periphery, typically but not necessarily presented in the Peyer’s patches by retinoic acid-synthesizing αEβ7 integrin-positive dendritic cells. This interaction induces upregulation of expression of gut-tropic molecules such as α4β7 integrin by the effector cells, allowing them to selectively home to the intestinal epithelium after trafficking through thoracic duct lymph and blood^7,8. Most conventional TCRαβ⁺ IELs are thought to be tissue-resident effector memory T cells — a long-lived population that is poised to respond rapidly to repeat pathogenic challenge through cytolytic activity and release of type 1 cytokines, such as interferon-γ (IFNγ) and tumour necrosis factor (TNF)⁹. The antigens that drive the activation and accumulation of conventional T cells in the intestines of specific pathogen-free (SPF) mice are unknown, although they are likely to be derived from the commensal microbiota or dietary antigens, as suggested by marked reductions in this T cell population in germ-free and protein antigen-free mice¹⁰. Notably, transgenic expression of several TCRs randomly cloned from conventional CD8αβ⁺TCRαβ⁺ IELs of SPF mice led to the generation of primarily naive CD8αβ⁺TCRαβ⁺ cells in thymus and peripheral lymphoid tissue, as well as some conventional CD8αβ⁺ effector T cells in the intestinal epithelium. This finding supports the notion that, unlike unconventional IELs, conventional IELs are not developmentally programmed to home to the intestinal epithelium but arise from naive CD8⁺ T cells following exposure to antigens in the periphery¹¹.

Tissue-resident phenotype

The intestinal tissue-resident phenotype of conventional IELs — and also that of innate and innate-like lymphocytes — is defined by low-level expression of several transcription factors and receptors associated with tissue homing and recirculation, including Krüppel-like factor 2 (KLF2), L-selectin, sphingosine 1-phosphate receptor 1 (S1P1) and CC-chemokine receptor 7 (CCR7), and increased expression of CD69, which prevents surface expression of S1P1 (REF.¹²). In the presence of transforming growth factor-β (TGFβ), the expression of αEβ7 integrin is upregulated and α4β7 integrin expression is down-regulated^13–15. αEβ7 integrin binds epithelial cadherin (E-cadherin) expressed by IECs, and this interaction regulates the accumulation but not long-term retention of conventional CD8αβ⁺ IELs in the intestinal epithelium¹⁶.

Plasticity

Under certain conditions, such as in the presence of TGFβ and retinoic acid together with strong TCR signals, conventional IELs may be induced to express CD8αα^4,17,18. The converted CD4⁺CD8αα⁺TCRαβ⁺ IELs lose expression of T helper-inducing POZ/Krüppel-like factor (THPOK; also known as ZBTB7B), turn on expression of Runt-related transcription factor 3 (RUNX3) and T cell-specific T box transcription factor T-bet (also known as TBX21) and upregulate expression of IEL-associated genes such as natural killer (NK) cell receptor 2B4 and granzyme B^14,19,20. This process is mediated in part by microbial metabolites, including indole derivatives from Lactobacillus reuteri, that activate aryl hydrocarbon receptor (AHR)²¹. Intriguingly, this conversion can be observed among lamina propria regulatory T (T_reg) cells, which subsequently migrate to the intestinal epithelium, lose expression of forkhead box protein P3 (FOXP3) and THPOK and become CD4⁺CD8αα⁺ IELs in a microbiota-dependent manner²². In further support of this scenario, T cells from mice derived by nuclear transfer from a FOXP3⁺ T_reg cell from the mesenteric lymph node into an enucleated oocyte develop into either lamina propria T_reg cells or CD4⁺CD8αα⁺ IELs²³. Thus, CD4⁺ T cell responses to microbiota antigens may be characterized by plasticity and conversion between the T_reg cell and effector IEL programmes^22,23.

Unconventional intestinal IELs

A unique aspect of the intestinal intraepithelial layer compared with other organs is the predominance of unconventional, or innate-like, lymphocytes in the steady state. These lymphocytes include TCRαβ⁺ cells and TCRγδ⁺ cells, both of which can be either double negative (DN; that is, they do not express CD4 or CD8αβ co-receptors) or CD8αα positive. These populations were termed unconventional IELs on the basis of presumed peculiarities in their development^4,24,25 or in their antigen-recognition properties. As reviewed below, the major differences between unconventional and conventional IELs are the recognition of self ligands rather than foreign ligands and the acquisition of the cytolytic effector programme before rather than after exposure to infection or injury. Thus, unconventional IELs define a distinct paradigm of T cell-mediated host defence that is more akin to innate immunity than adaptive immunity. Moreover, unconventional TCRαβ⁺ and TCRγδ⁺ IEL lineages differ markedly from each other with respect to their developmental pathways and modes of antigen recognition (FIG. 1).

Fig. 1. Multiple pathways for the development of intestinal IEL lineages.

Conventional αβ T cell receptor (TCRαβ)⁺CD8αβ⁺ intraepithelial lymphocytes (IELs) originate from naive CD8⁺ T cells that are primed by foreign antigens in the gut-associated lymphoid tissue (GALT) to divide, acquire a cytolytic effector programme and upregulate expression of α4β7 integrin and CC-chemokine receptor 9 (CCR9). They traffic through the lymph to reach the blood circulation and home to the gut epithelium, where on exposure to transforming growth factor-β (TGFβ), they acquire expression of αEβ7 integrin. By contrast, unconventional TCRαβ⁺ IELs acquire their effector programme and the expression of gut-homing receptors in the thymus through agonist stimulation by self antigens before reaching the gut epithelium. Unconventional TCRγδ⁺ IELs arise in the thymus (and also outside the thymus) as naive lymphocytes that already express α4β7 integrin and CCR9. They divide and acquire their effector programme in the GALT after exposure to butyrophilin-like protein (BTNL) ligands expressed by intestinal epithelial cells and then recirculate through Peyer’s patches and the lymph back to the blood and intestinal epithelium. Note that unconventional TCRαβ⁺ and TCRγδ⁺ IELs respond to self ligands rather than foreign ligands. Group 1 innate lymphoid cells (ILC1s) originate as fully differentiated effector cells from a dedicated ILC progenitor found in fetal liver and in adult bone marrow. The developmental pathway of CD4⁺CD8αα⁺ IELs is not shown here but was recently reviewed⁸⁸. AHR, aryl hydrocarbon receptor; DN, double negative; DP, double positive; IL-15Rβ, IL-15 receptor-β; NKR, natural killer receptor; PD1, programmed cell death protein 1; T-bet, T cell-specific T box transcription factor T-bet; TCRV, TCR variable chain.

Development of unconventional TCRαβ⁺ IELs

Initial studies on the development of unconventional TCRαβ⁺ IELs suggested an extrathymic origin because they could be found in nude (athymic) mice^26–28. In addition, their TCR repertoire is enriched for TCR variable β-chain (TCRVβ) families that react with mouse mammary tumour virus (MMTV)-encoded superantigen, TCR reactivities that are typically deleted during thymic negative selection^28,29. However, the fact that unconventional TCRαβ⁺ IELs are greatly reduced in number in young nude mice supports a prominent role for the thymus in their development. Some groups suggested that IELs are generated through a thymic DN pathway without transiting through the double positive (DP) stage³⁰. However, mixed bone marrow chimaera experiments using bone marrow cells with floxed Rag2 alleles and a Cre transgene driven by the Cd4 promoter (Rag2^fl/flCd4–Cre cells) together with wild-type bone marrow cells demonstrated that deletion of Rag2 at the DP stage essentially depleted the unconventional TCRαβ⁺ IEL compartment, a result that is incompatible with the proposed ‘DN pathway’ and instead supports a DP stage of development³¹. Several groups suggested that thymic IEL precursors escape thymic negative selection in a process termed agonist selection, whereby elevated TCR signals induce clonal deviation rather than clonal deletion, a process reminiscent of the development of NK T (NKT) cells and T_reg cells^29,32–35. Indeed, mice lacking store-operated calcium entry (SOCE), which are therefore unable to flux calcium following strong TCR signals, are severely deficient in unconventional TCRαβ⁺ IELs^36,37. Notably, a similar requirement for agonist signalling and SOCE has been reported for T_reg cell and NKT cell development³⁷. Additional signals are required for individual lineages. For example, co-stimulation through the CD80–CD28 interaction may have a central role in the decision between clonal deletion versus clonal diversion to the unconventional TCRαβ⁺ IEL pathway²⁹.

In support of the agonist signalling hypothesis, TCR transgenic cells developing in the presence of an agonist ligand in vivo or in vitro (for example, male mice expressing a TCR transgene specific for male antigen HY) generated cells resembling unconventional TCRαβ⁺ IELs^32,34,35. However, this particular TCR transgenic model had premature expression of the TCR, which favours the generation of DN ‘TCRγδ-like’ cells that have an inherent tendency to traffic to the intestinal epithelium, therefore confounding interpretation of these results^38–40. When expression of the HY-specific TCR transgene was delayed until the DP stage of development, the intestinal IEL fate was greatly reduced⁴¹.

The hypothesis that unconventional TCRαβ⁺ IELs are a result of agonist selection has gained renewed support from recent studies focusing on IEL TCRs randomly cloned from intestines of wild-type mice^11,42. The unconventional IEL TCR repertoire is very broad, although the diversity can be obscured by the apparently stochastic expansion of a limited number of clones^11,43. It is also largely independent of the microbiota⁴⁴. When random TCRs are cloned directly from these unconventional TCRαβ⁺ IELs and conditionally expressed using a Cd4–Cre system to exclude artefacts associated with premature TCR expression during thymic development, they drive the development of unconventional TCRαβ⁺ IELs exclusively¹¹. By contrast, TCRs cloned directly from conventional CD8αβ⁺ intestinal IELs drove the development of naive recirculating CD8αβ⁺ T cells and conventional intestinal IELs. These results demonstrated that intestinal IEL TCRs instructed the intestinal IEL lineage fate^11,42. In these experiments, unconventional but not conventional IEL TCRs induced strong signals during thymic development, as indicated by elevated levels of early growth response protein 2 (EGR2), nuclear hormone receptor NUR77 (also known as NR4A1), CD5, CD69, programmed cell death protein 1 (PD1) and the pro-apoptotic factor BCL-2-interacting mediator of cell death (BIM; also known as BCL-2L11)^11,42. Most of the thymocytes that receive a positive signal through their TCR arrest at the DP stage and die by apoptosis, but the fraction that escapes cell death downregulates CD4 and CD8αβ expression (to become DP^low cells) and selectively populates the intestinal epithelium¹¹. The DP^lowPD1^hi phenotype is associated with thymocytes undergoing negative selection at the DP stage in the thymic cortex^45–48. Indeed, transgenic expression of numerous TCRs cloned from thymic DP^lowPD1^hi cells induced signs of negative selection at the DP stage and the selective generation of unconventional TCRαβ⁺ IELs in the periphery^11,36. Furthermore, transgenic expression of more than 80 random TCRs from the preselection repertoire confirmed that the only TCRs that generated intestinal IELs were those that induced the DP^lowPD1^hi phenotype in the thymus. These studies used mixed bone marrow chimaeras in which the TCR transgenic fraction made up only a small fraction of the total T cell population to avoid the massive non-physiological expansion of T cells that occurs in lymphopenic hosts such as Rag1^−/− or Rag2^−/− mice. Therefore, the thymic DP^lowPD1^hi population is the main precursor population of unconventional TCRαβ⁺ IELs. Although there may be other thymic precursors to intestinal IELs¹², as discussed below, their frequency must be low^11,42. Characterization of thymic and peripheral stages of IEL development in this transgenic system indicated that IEL precursors downregulated CD4 and CD8αβ expression concomitantly with agonist TCR signalling. The IEL precursors also induced expression of α4β7 integrin, which is required for intestinal homing, and CD160, which interacts with herpesvirus entry mediator (HVEM; also known as TNFRSF14) on IECs. Recent thymic emigrants in the spleen show upregulated expression of SLAM family receptor NK cell receptor 2B4, whereas IELs residing in intestine upregulate expression of T-bet, RUNX3, CD8αα and αEβ7 integrin but downmodulate expression of α4β7 integrin. CCR9 is expressed by all effector T and B cells that migrate to the small intestine and is essential for intraepithelial homing of TCRVγ7⁺ IELs, whereas G protein-coupled receptor 18 (GPR18) and GPR55 are expressed by all populations of TCRγδ⁺ and TCRαβ⁺ IELs and are also important for their accumulation in the intestinal epithelial layer^3,49,50. Thus, although the IEL fate is instructed by TCR-induced signalling in the thymus, further differentiation occurs during migration and homing to the intestinal epithelium.

Other thymic precursors for unconventional IELs have been suggested, mostly on the basis of transfer experiments into lymphodepleted recipients^12,51–53. For example, a population of CD4⁺CD8αβ⁺CD8αα⁺TCRαβ⁻ thymocytes selectively generated unconventional TCRαβ⁺ IELs upon intrathymic transfer into irradiated hosts⁵³. These triple-positive cells belonged to the cycling DP blast stage, implying that the putative IEL precursors could be identified before expression of their (autoreactive) TCR. Another candidate thymic precursor for unconventional IELs was recently reported on the basis of the observation that a mature (medullary) T-bet⁺CD161⁺ DN thymic subset lacking expression of a TCR recognizing αGalCer–CD1d tetramers (and therefore distinct from TCRVα14⁺ NKT cells) could give rise to unconventional IELs after intravenous transfer into V(D)J recombination activating protein (RAG)-deficient hosts¹¹. In the same study, the transfer of DP^lowPD1^hi thymocytes gave rise to larger numbers of unconventional IELs, indicating that the αGalCer–CD1d tetramer-negative T-bet⁺CD161⁺ DN thymocytes were a minor precursor in the transfer assay. The above studies involving transfers of minor cell populations into lymphopenic hosts are confounded by several factors, including aberrant differentiation in the ‘empty’ host and the outgrowth of potential contaminating cells over several weeks after the initial injection. Furthermore, although they suggest that a particular cell type can generate unconventional IELs in an empty host, they do not establish the real contribution of that particular precursor cell in physiological competitive conditions. Further studies of TCR transgene expression are required to clarify these issues.

MHC ligands for unconventional TCRαβ⁺ IELs

Early work indicated that the presence of unconventional TCRαβ IELs in the intestine required β₂-microglobulin (β₂m) but not classical MHC class I molecules H2-K^b or H2-D^b, suggesting that these IELs recognize non-classical MHC class I molecules^54–56. However, an analysis of individual IEL-derived TCRs transgenically expressed in mice lacking I–A^b MHC class II molecules or β₂m suggested the recognition of both MHC class II and MHC class I ligands, including non-classical MHC class I ligands^11,42. In fact, in a majority of cases, IEL TCRs were found to be inherently broadly MHC cross-reactive, recognizing many MHC class I and MHC class II molecules across multiple MHC haplotypes³⁶. This was demonstrated by in vitro experiments in which IEL TCR-expressing DP thymocytes tended to be strongly reactive against a panel of fibroblasts and splenocytes expressing different MHC haplotypes and by in vivo experiments in which IEL TCR-expressing DP thymocytes received strong signals in mice with different MHC haplotypes.

This property of cross reactivity is very rarely observed among TCRs driving conventional CD4⁺ and CD8⁺ T cell development. Notably, agonist signalling by cross-reactive TCRs depends on the engagement of the CD4 or CD8αβ co-receptor¹¹. Therefore, the downregulation of both CD4 and CD8αβ, a characteristic feature of the thymic differentiation of intestinal IEL precursors, might dampen their TCR reactivity to self-MHC ligands and allow them to escape negative selection in the thymus and avoid overt reactivity in the intestine. Such cross-reactive TCRs arise at a frequency of ~5% in the randomly generated preselection TCR repertoire and are expressed by a majority of the thymocytes that undergo negative selection in the thymic cortex³⁶. They probably overlap with cross-reactive TCRs that were previously observed in mice engineered to express a single peptide–MHC class II complex, in which negative selection was limited^57,58. The cross-reactive nature of the TCR repertoire of DP^lowPD1^hi thymocytes explains why they are largely preserved in mice that lack either MHC class I or MHC class II molecules¹². However, it does not explain why mature unconventional IELs are nearly absent in the intestinal epithelium of β₂m-deficient mice but largely preserved in MHC class II-deficient mice. It is possible that MHC class I reactivity is more important than MHC class II reactivity for the maintenance of the unconventional IEL pool in the intestinal epithelial layer. In that regard, by binding the non-classical mouse MHC class Ib molecule thymus leukaemia antigen (TL; also known as H2-T3), CD8αα may provide crucial signals to IELs in the intestinal epithelial layer, although TL-deficient mice did not have an obvious defect in the unconventional IEL populations^59,60. Although more studies are required to better define the relevant MHC ligands of IELs in the thymus and intestinal environments, the expression of crossreactive TCRs indicates that IELs might be responding to broad changes in classical and non-classical MHC levels rather than to the presentation of particular peptides, a scenario reminiscent of signalling for NK cells. In agreement with this notion, the differentiation of unconventional IELs was unaltered in germ-free mice, although changes in frequency were suggested in some reports^61–64.

Maturation and maintenance of unconventional TCRαβ⁺ IELs

Unlike other agonist-selected T cell lineages, unconventional TCRαβ⁺ IELs lack a lineage-defining transcription factor and are instead currently characterized by expression of a set of key transcription factors, including T-bet and RUNX3, that drive cytolytic differentiation and are uniformly expressed and required by all other intestinal IEL lineages^19,20,65,66.

Thymic IEL precursors express high levels of CD122 (also known as IL-2Rβ; a shared subunit of receptors for IL-2 and IL-15). However, nude mice engrafted with an IL-15-deficient or IL-15 receptor-α (IL-15Rα)-deficient neonatal thymus generated a normal intestinal IEL compartment, suggesting that IL-15-induced signalling is not required at the thymic stage⁶⁷. By contrast, IL-15-induced signalling was crucial for the maturation and/or survival of unconventional TCRαβ⁺ IELs in the periphery. T-bet-deficient mice also lack intestinal IELs, which is consistent with an important role for the IL-15–T-bet–RUNX3 axis in their development^63,65.

Human unconventional TCRαβ⁺ IELs

Until recently, humans were thought to lack the equivalent of mouse unconventional CD8αα⁺TCRαβ⁺ intestinal IELs. However, a recent report provided some evidence that these cells avoided detection because they express both CD8αα and CD8αβ co-receptors⁶⁸. The development and function of human CD8αα⁺TCRαβ⁺ IELs and their relationship to mouse unconventional TCRαβ⁺ IELs need further study. Some differences are already apparent, as several human intestinal IEL clones were found to react with CD1 family molecules in vitro⁶⁹. By contrast, unconventional TCRαβ⁺ IEL populations are unaltered in CD1d-deficient mice⁵⁶.

Development of unconventional TCRγδ⁺ IELs

TCRγδ⁺ IELs constitute the largest fraction of intestinal IELs in mice. Although their general phenotype and transcriptional programme closely resemble those of unconventional TCRαβ⁺ IELs, their development is distinct, at least with regard to the well-defined TCRVγ7⁺ subset, which predominates among TCRγδ⁺ IELs. TCRVγ7⁺ IELs are the descendants of a perinatal wave of γδ T cells generated in the mouse thymus. But the thymus may not be the only site of TCRVγ7⁺ IEL generation because about half of the population could be found in nude mice¹⁰. Using mice in which cells that have recently undergone TCR gene rearrangement are fluorescently marked, recent thymic emigrants expressing TCRVγ7 could be detected among intestinal IELs³. This TCRVγ7⁺ population seemed to be selectively activated in gut-associated lymphoid tissues (GALT), as indicated by the induction of expression of CD69, cell cycle proteins and α4β7 integrin in the Peyer’s patches and by the presence of activated cycling blasts in the thoracic duct lymph³. It has been reported that neither lymph nodes nor Peyer’s patches are absolutely required for the generation of a normal TCRVγ7⁺ IEL compartment in the intestine, suggesting alternative locations for priming¹⁰. In addition, this study suggested that activation and cell cycle progression were initiated in the gut epithelium rather than the Peyer’s patches. Thus, these studies suggest that, unlike unconventional TCRαβ⁺ IELs, TCRVγ7⁺ IELs exit the thymus as naive cells and undergo activation in the gut or GALT before gaining residence in the intestinal epithelium.

Ligands of unconventional TCRVγ7⁺ IELs

TCRVγ7⁺ IELs develop normally in mice lacking MHC class II molecules and β₂m, and their activated effector programme is unperturbed in germ-free and antigen-free mice, suggesting that both their development and activation rely on yet unknown endogenous ligands^10,55,56. A remarkable series of studies uncovered the role of butyrophilin and butyrophilin-like gene family members, which are structurally related to the B7 family molecules and are expressed on the surface of various epithelial cells, as potential ligands of various populations of TCRγδ⁺ T cells⁷⁰. Butyrophilin-like protein 1 (BTNL1), BTNL4 and BTNL6 are expressed in the intestinal epithelium, and mice lacking BTNL1 expression in the intestinal epithelium had a selective loss of TCRVγ7⁺ intestinal IELs, with residual cells displaying a naive CD122^lowCD8α⁻THY1^hiCD5^hi phenotype. In vitro experiments further suggested that BTNL1 cooperates with BTNL6 for activation of TCRVγ7⁺ IELs, as indicated by CD69 induction¹⁰. Thus, in contrast to unconventional TCRαβ⁺ IELs, whose effector programme is induced in the thymus, activation of the TCRVγ7⁺ IEL effector programme takes place in the periphery and is driven by self ligands such as BTNL molecules expressed in the intestinal epithelium. Whether the expression of these BTNL molecules can be modulated has not been reported, leaving open the possibility that binding of co-ligands is necessary for full TCRVγ7 stimulation, similar to the activation of human TCRVγ9⁺ T cells by butyrophilin and phosphoantigens⁷¹. Importantly, extending the mouse observations, BTNL3 and BTNL8 seem to interact with TCRVγ4⁺ cells in the human colon¹⁰, although other TCRVδ1⁺ cells² seem to recognize CD1 molecules^72,73. Thus, although the discovery of crucial interactions between TCRγδ and BTNL surface ligands in both mice and humans constitutes a long-awaited breakthrough in the field, the details of these interactions need to be further defined.

Maturation and maintenance of TCRγδ⁺ IELs

TCRVγ7⁺ IELs and unconventional TCRαβ⁺ IELs have a very similar phenotypic, functional and transcriptional cytolytic profile^74,75. Like unconventional TCRαβ⁺ IELs, TCRVγ7⁺ IELs require IL-15-induced signalling and T-bet and AHR expression and they do not depend on either the microbiota or dietary antigens for development or maintenance^10,63,76,77. In contrast to unconventional TCRαβ⁺ IELs, however, neither TGFβ nor αEβ7 integrin seem to be required for the development or maintenance of unconventional TCRγδ⁺ IELs, although these studies did not distinguish the TCRVγ7⁺ population from other TCRγδ⁺ IELs^14,15.

Intestinal innate lymphoid cells

Development of ILC1s

ILC1s are a TCR-negative population of lymphocytes that have a type 1 functional programme, similar to other IEL populations, but represent a minority of intestinal IELs (<1%). They originate from a fetal liver or bone marrow lineage (LIN)⁻ IL-7Rα⁺ and α4β7 integrin-positive precursor termed the common ILC progenitor, which transiently expresses the transcription factor promyelocytic leukaemia zinc-finger protein (PLZF; also known as ZBTB16) and gives rise to all categories of ILC1s, ILC2s and ILC3s in various tissues⁷⁸.

Maturation and maintenance of ILC1s

ILC1s have a CD161⁺NKp46⁺CD122⁺CD160⁺CD94⁺NKG2D⁺ phenotype, can produce type 1 cytokines and become cytolytic upon stimulation by cytokines such as IL-12, IL-15 and IL-18 (REFS^5,79). Their transcriptional profile is similar to those of other IELs. However, surprisingly, mouse intraepithelial ILC1s do not seem to express the canonical intestinal IEL integrin αEβ7 integrin⁸⁰. Intestinal ILC1s are not affected by TGFβ deficiency, perhaps owing to compensation by other TGFβ family cytokines such as inhibins and bone morphogenetic proteins that are expressed at high levels in the intestine⁸⁰. Importantly, ILC1s require DNA-binding protein inhibitor ID2 for their development⁷⁸. IL-15 minimally affects ILC1s in the intraepithelial compartment but is required for ILC1s in the lamina propria^5,81. Further work is needed to dissect the relationship between ILC1s in these adjacent tissues and their seemingly disparate cytokine requirements. Like other intestinal IELs, ILC1s require T-bet but not eomesodermin for development, and neither AHR nor the microbiota is required for their development^5,82. ILC1s are also present among human intestinal IELs and are phenotypically and functionally similar to their mouse counterparts, with the exception that they express different families of NK cell receptors and express αEβ7 integrin⁵.

Other ILC1-like populations

In addition to ILC1s, several populations of ILC1-like cells are present in the intestinal epithelium and may be involved in regulating intestinal immune responses^6,83,84. Notably, CD8αα⁺ innate IELs are an ILC1-like, non-T cell IEL population that is developmentally distinct from unconventional TCRαβ⁺ and TCRγδ⁺ IELs; they do not require ID2 for their development but express αEβ7 integrin and a single type of NK cell receptor, lymphocyte antigen 49E (LY49E; also known as KLRA5)^6,83. Strikingly, there is a reduced frequency of CD8αα⁺ innate IELs in TL-deficient mice, suggesting an interaction between CD8αα and TL in these cells⁶.

Similarly, human IELs contain innate lymphocytes that express intracellular CD3 and bear a striking resemblance to the CD8αα⁺ innate IELs in mice⁸⁵. In fact, similar to mouse CD8αα⁺ innate IELs, human intracellular CD3⁺ innate lymphocytes develop independently of ID2 but require IL-15 (REF.⁸⁶). Notch signals in precursors have been proposed to initiate T cell lineage commitment but are terminated by granzyme B-mediated cleavage of Notch⁸⁶. Notably, intracellular CD3⁺ lymphocytes may be the precursors to frequent lymphomas associated with refractory sprue or uncontrolled coeliac disease⁸⁶.

Intestinal IEL functions

Lineage redundancy

All intestinal IEL lineages share stereotypic functional properties, including patrolling behaviour within the epithelial layer and potent cytolytic and T_H1 cell-type activities, which has made it difficult to assign crucial functional roles to individual IEL lineages (FIG. 2). The cytokine and cytolytic functions of most IELs can be stimulated either by the cytokines IL-12, IL-15 and IL-18, which are produced by IECs and lamina propria antigen-presenting cells, or by engaging activating NK cell receptors that are expressed by all IEL lineages (with the exception of conventional CD8αβ⁺TCRαβ⁺ IELs)^5,87,88. Thus, infectious and inflammatory conditions that induce these cytokines and receptors trigger the recruitment of IEL lineages in a promiscuous manner, potentially accounting for the substantial degree of redundancy observed in the intestinal IEL response. Redundancy of ILCs was observed in patients with severe combined immunodeficiency (SCID) who received allogeneic bone marrow transplants. These individuals show successful T and B cell reconstitution, but not ILC reconstitution, yet they do not have increased susceptibility to any particular infections, even after nearly 40 years⁸⁹, suggesting that ILCs are dispensable for protection against infection in humans.

Fig. 2. Stereotypic functions of intestinal IEL lineages.

Intestinal intraepithelial lymphocyte (IEL) lineages engage in multiple interactions with epithelial cells, dietary components and the microbiota. They have common properties, including expression of type 1 cytokines and cytolytic properties mediated by granzyme B upon exposure to IL-12, IL-18 and IL-15 produced in the local environment. They can directly kill stressed epithelial cells that express ligands for activating natural killer (NK) cell receptors such as NK cell receptor group 2, member D (NKG2D). In addition to these shared properties, individual lineages respond to specific ligands, including butyrophilin-like proteins (BTNLs) for T cell receptor variable γ-chain 7 (TCRVγ7)⁺ IELs and MHC ligands for TCRαβ⁺ IELs. Note that the TCRαβ of unconventional TCRαβ⁺ IELs is cross-reactive, such that it can recognize multiple MHC ligands. AHR, aryl hydrocarbon receptor; HVEM, herpesvirus entry mediator; IL-12R, IL-12 receptor; IL-15R, IL-15 receptor; IL-18R, IL-18 receptor; ILC1, group 1 innate lymphoid cell; T-bet, T cell-specific T box transcription factor T-bet; TL, thymus leukaemia antigen.

However, intestinal IEL lineages have distinct recognition properties that may contribute to their selective recruitment. For example, TCRVγ7⁺ IEL activation might be regulated by epithelial expression of BTNL ligands, unconventional TCRαβ⁺ IEL activation may be controlled by the levels of expression of classical and non-classical MHC molecules, and conventional TCRαβ⁺ IELs can be directly activated in situ by their cognate peptide ligand. These distinct means of activation support the concept that protection may be achieved through multiple layers of innate, innate-like and adaptive lymphocytes that reside at the same intestinal mucosal barrier and are partially redundant.

At present, there are few well-documented reports demonstrating the specific contribution of a single IEL lineage in response to an injury or infectious challenge. For example, mice lacking TCRγδ⁺ IELs exhibited a transient increase in intestinal barrier penetration and splenic dissemination of Salmonella enterica subsp. enterica serovar Typhimurium after oral inoculation in germ-free mice, but the defect could only be detected transiently in the first few hours after infection, suggesting that other mechanisms were involved as well⁹⁰. How TCRγδ⁺ IELs could limit S. Typhimurium penetration after oral inoculation was in part based on myeloid differentiation primary response protein 88 (MYD88) signalling by epithelial cells and production of antimicrobial peptides such as regenerating islet-derived protein 3γ (REG3γ) by TCRγδ⁺ IELs⁹⁰. Such minor and transient defects would likely be difficult to detect in most experimental situations, illustrating the challenge in assigning important protective functions to intestinal IELs.

Why then have multiple lineages evolved if they carry out largely redundant functions? It is possible that individual lineages contribute apparently minor advantages against only certain pathogens that had an evolutionary impact. It is also possible that pathogens have evolved multiple layers of subversion that disable entire aspects of innate immunity, for example, the disruption of type I interferon signalling by rotavirus^91,92. In that context, new lineages that can escape these evasion mechanisms might be advantageous, even though they might also be subverted later in evolution.

Effector mechanisms

Regardless of which IEL lineage is most important in response to a given challenge, there are several key signalling pathways that encompass the canonical IEL effector programme and highlight the importance of IEL–IEC interactions not only in host defence but also in regulating the homeostatic crosstalk between the microbiota, epithelial cells and IELs. Cytokines including IL-15 and IL-18 are produced by IECs either in the steady state or in response to MYD88 signals, respectively, and lead to the maintenance, recruitment and activation of IELs^93,94. MYD88 signalling in epithelial cells is required for basal expression of REG3γ by unconventional TCRγδ⁺ IELs⁹⁰. Unconventional TCRαβ⁺ and TCRγδ⁺ IELs can kill target cells independently of TCR activation by triggering of NK cell receptors in a response that is enhanced by cytokines such as IL-12 and IL-18 (REF.⁸⁷). Furthermore, both unconventional TCRαβ⁺ and TCRγδ⁺ IELs have decreased TCR sensitivity compared with conventional TCRαβ⁺ IELs, suggesting that TCR signalling alone is insufficient to drive IEL effector responses in vivo⁹⁵. Patrolling behaviour along the intestinal villi is common among IELs (it has not been formally demonstrated for ILC1s, but is likely), and as recently demonstrated for TCRγδ⁺ IELs, this behaviour is dynamically regulated by MYD88 signalling in IECs^49,96,97. Intravital microscopy showed that in mice infected with S. Typhimurium or Toxoplasma gondii, TCRγδ⁺ IELs survey a smaller area of the intestinal epithelium but more frequently reach the intercellular space between IECs (a process termed ‘flossing’)⁹⁶. These recent findings converge to highlight the dynamic behaviour of all IELs and their potential to receive signals close to the luminal space.

Most IELs express CD160, and its interaction with HVEM, a TNF superfamily member expressed by IECs, drives antimicrobial peptide release from IECs⁹⁸. Inhibition of the CD160–HVEM interaction by administration of CD160-specific antibody to wild-type mice reduced the production of antimicrobial peptides and led to increased death following infection with Citrobacter rodentium⁹⁸. This interaction triggers bidirectional signalling, as ligation of CD160 by HVEM augments effector cytokine production by NK cells and enhances their cytolytic capabilities⁹⁹.

AHR expression by IELs links IEL development and maintenance to environmental cues derived from the intestinal lumen. AHR-deficient mice or mice lacking dietary AHR ligands derived from cruciferous vegetables have fewer IELs and show decreased IEC turnover, which is consistent with a previously demonstrated role for TCRγδ⁺ IELs in controlling IEC proliferation^76,100,101. AHR-deficient mice develop more severe colitis following dextran sodium sulfate (DSS) treatment than wild-type mice and can be rescued by transfer of TCRγδ⁺ IELs before initiation of DSS challenge⁷⁶. Similarly, in a colitis model induced by CD45RB^hi T cell transfer into SCID mice, prior transfer of unconventional TCRαβ⁺ IELs, but not TCRγδ⁺ IELs, lessened disease severity¹⁰². By contrast, CD8αα⁺ innate IELs seem to have a pro-inflammatory role in colitis induced by administration of CD40-specific antibody, which suggests that different IEL lineages are recruited by different stimuli¹⁰³.

Human diseases

Inflammatory bowel disease

The potential involvement of IELs as innate producers of pro-inflammatory cytokines in inflammatory bowel disease (IBD) development and progression is under active investigation. There is evidence in both humans and mice that under inflammatory conditions, ILC3s can lose expression of retinoid-related orphan receptor-γt (RORγt; also known as RORC) and upregulate T-bet expression to become ILC1-like cells that express NK cell receptors and CD122 and have a pathogenic role in IBD^104–107. Indeed, ILC1s accumulate in the intestinal epithelium and lamina propria of patients with active Crohn’s disease lesions and express high levels of IFNγ, suggesting a role for ILC1s in the initiation, maintenance or control of IBD^5,105.

Coeliac disease

Coeliac disease is an inflammatory disorder in which a subset of genetically susceptible individuals who express HLA-DQ2 or HLA-DQ8 develop T_H1 cell-type inflammation in the small intestine, leading to villous atrophy and subsequent nutrient malabsorption¹⁰⁸. Massive expansion of intestinal IEL populations is a hallmark of the disease. Gluten-reactive CD4⁺TCRαβ⁺ T cells are characteristically present in the intestines of patients with coeliac disease. However, these cells are restricted to the lamina propria and do not seem to directly cause disease pathology¹⁰⁹. By contrast, conventional CD8αβ⁺TCRαβ⁺ IEL populations are markedly expanded and activated but do not seem to be gluten specific¹⁰⁸. In healthy individuals, conventional CD8αβ⁺TCRαβ⁺ IELs express inhibitory NK cell receptors including CD94–NKG2A^110,111, whereas in patients with coeliac disease, increased levels of IL-15 upregulate the expression of activating NK cell receptors including NKG2C and NKG2D. These activating receptors licence CD8αβ⁺TCRαβ⁺ IELs to kill neighbouring epithelial cells that express stress-induced ligands such as MHC class I polypeptide-related sequence A (MICA)^112–114.

Although they do not seem to be gluten reactive, the expanded CD8αβ⁺TCRαβ⁺ IEL populations are oligo-clonal and show evidence of antigen-driven selection on the basis of TCR complementarity-determining region 3 (CDR3) sequence analysis, suggesting that the TCR has a role in CD8αβ⁺TCRαβ⁺ IEL-mediated coeliac disease pathogenesis¹¹¹. Indeed, the expression of activating versus inhibitory NK cell receptors seems to be controlled in part by TCR specificity¹¹¹. One hypothesis is that activating NK cell receptors and IL-15-mediated stimulation lower the TCR signalling threshold required for activation, allowing for TCR-mediated killing of stressed epithelial cells in an MHC-dependent manner but without the need for cognate peptide ligands¹¹⁵.

Notably, TCRγδ⁺ IEL populations are also markedly expanded in coeliac disease^116,117. Indeed, gluten challenge of patients with coeliac disease allows for the detection of gut-tropic TCRγδ⁺ cells even in the peripheral blood¹¹⁸. However, in contrast to their TCRαβ⁺ IEL counterparts, the population of TCRγδ⁺ IELs minimally contracts after removal of gluten from the diet, suggesting that extensive remodelling of the TCRγδ⁺ IEL niche occurs^119,120. There is evidence to suggest that TCRVδ1⁺ clones remain predominant in the intestinal epithelium after gluten challenge, but whether more subtle changes in the repertoire have occurred, for example, owing to antigen-driven selection and regulation by butyrophilin-like molecules, remains an open question¹¹⁸.

Conclusions

Intestinal IELs typically express T_H1 cell-type cytokines and have cytolytic properties, yet they comprise multiple different lineages that follow distinct developmental pathways to acquire these similar effector properties. We speculate that an evolutionary arms race between pathogens and the intestinal immune system has resulted in the recurrent emergence of lineages that patrol the intestinal barrier for host protection against infection and injury. Redundancy of lymphoid lineages is also observed at other sites, including the liver, lung and skin (BOX 2), which is likely to reflect selection pressures to create a multi-layered protection system at mucosal barriers. The multiplicity of these cytolytic lineages may account for the growing evidence that they are involved in the pathogenesis of inflammatory diseases such as coeliac disease and IBD. Further studies should clarify the relative importance of the different IEL subsets in the context of different categories of pathogens and modes of injury. Ultimately, an improved understanding of early immune responses at mucosal barriers and their dysregulation in inflammatory disorders will lead the way to novel diagnostic and therapeutic approaches that improve human health.

Box 2. Tissue-resident lymphocytes: a rapid, multi-layered protection system.

The presence of multiple functionally similar, but developmentally distinct, cell populations is well established in the intestine and also in other organs. In the intestinal epithelium, conventional αβ T cell receptor (TCRαβ)⁺ intraepithelial lymphocytes (IELs), unconventional TCRαβ⁺ IELs, unconventional TCRγδ⁺ IELs and group 1 innate lymphoid cells (ILC1s) all produce interferon-γ (IFNγ) and granzyme B in response to IL-12 and IL-18 and, with the exception of conventional TCRαβ⁺ IELs, in response to stimulation of their natural killer (NK) cell receptors by stress-induced ligands (see the figure, part a).

In the liver, NK T (NKT) cells, γδ T cells, ILC1s and mucosa-associated invariant T (MAIT) cells reside within the lumen of the sinusoids, interact with sinusoidal endothelial cells via integrins and contribute to host defence against infectious agents presented by Kupffer cells^121,122 (see the figure, part b). Constitutive interactions between leukocyte function-associated molecule 1 (LFA1) on NKT cells and intercellular adhesion molecule 1 (ICAM1) on endothelial cells are essential for their resident phenotype. The CD1d molecules expressed by Kupffer cells allow NKT cells to sample microbial lipids present in the portal system. All of these cell populations rapidly secrete type 1 cytokines and exhibit cytolytic properties in response to IL-12 and IL-18 secreted in their environment. These functions are also triggered by recognition of lipid ligands by CD1d-restricted NKT cells, microbial metabolites by MHC class I-related gene protein (MR1)-restricted MAIT cells and stress ligands by NK cell receptors (such as NK cell receptor group 2, member D (NKG2D)) expressed by all lineages.

In the lungs, NKT cells and ILC2s contribute to both the host response against bacterial infections and the initiation of allergic responses^123,124 (not shown).

In the dermis, there is functional overlap between IL-17-producing γδ T cells, NKT cells and ILC3s. Specifically, application of an imiquimod-containing cream to skin causes a psoriasis-like reaction that is dependent on IL-17 family cytokines and is ameliorated by depletion of γδ T cells or ILC3s¹²⁵. Conventional T helper 17 (T_H17) cells may have a pathogenic role in psoriasis as well. Likewise, in the skin-draining lymph nodes, IL-17-producing NKT cells, ILC3s and IL-17-producing γδ T cells are colocalized beneath the subcapsular sinus (SCS) macrophages and respond to IL-1β and IL-23, thus providing a multi-layered type 17 defence^126,127 (see the figure, part c).

Glossary

CD8αα

A homodimer of CD8α molecules (in contrast to the CD8αβ heterodimer found on conventional CD8⁺ T cells) that is expressed by most intestinal intraepithelial lymphocytes; its function is not known, although it is known to bind the MHC class Ib molecule thymus leukaemia antigen (TL)

Innate lymphoid cells (ILCs)

Cells that develop from a common lymphoid progenitor but do not express lineage markers associated with other lymphocytes, such as recombined antigen receptors. These cells rapidly secrete effector cytokines in response to activation and have been subdivided into three main groups on the basis of whether they produce T helper 1 (T_H1) cell-type, T_H2 cell-type or T_H17 cell-type cytokines

Peyer’s patches

Highly specialized lymph nodelike structures found on the anti-mesenteric side of the small intestine. They orchestrate immune responses against luminal antigens taken up by M cells

Aryl hydrocarbon receptor (AHR)

A ligand-dependent transcription factor that is activated by dietary and microbial metabolites and that regulates the development of most intestinal intraepithelial lymphocytes

αGalCer–CD1d tetramers

Tetrameric forms of CD1d molecules bound to α-galactosylceramide (αGalCer) that have sufficient affinity for the T cell receptor of natural killer T (NKT) cells to allow the detection of NKT cells by flow cytometry

Refractory sprue

Pathological changes of the small intestine, including crypt hyperplasia and villous atrophy, that do not resolve in patients with coeliac disease on a gluten-free diet

Mucosa-associated invariant T (MAIT) cells

A population of T cells with a semi-invariant αβ T cell receptor that recognizes riboflavin metabolites presented by the non-classical MHC class I molecule MHC class I-related gene protein (MR1)