Introduction

Intraepithelial lymphocyte(s) (IEL) were first described by Weber in 1847 as small round cells within the epithelium of the small intestine whose primary function was in nutrient absorption.1 Later on, Fichtelius reported the presence of these cells in the epithelium of a variety of species and suggested that their primary function was related to dealing with antigens present at body surfaces.2 Guy-Grand and collaborators established in the early 1970s that IEL were mainly comprised of T lymphocytes.3 During the same period, they were reported in human4,5 and their potential role in celiac disease (CeD) was put forward by Ferguson and Holmes,6,7 and in tropical sprue by Montgomery and Shearer.8 Extensive studies of IEL in mice, especially because of the identification of naturally occurring innate-like T cell receptor (TCR)αβ+CD8αα+ IEL that are autoreactive and undergo atypical thymic development, have suggested that IEL constitute a unique subset of T lymphocytes that are unique when compared to all other lymphocytes in the body.9,10,11,12,13 While CD3−, TCRγδ+, TCRαβ+CD8αβ+, and TCRαβ+CD4+ IEL subsets are present in both human and mouse9,14,15,16 the existence in human of the mouse naturally occurring autoreactive TCRαβ+CD8αα+ IEL subset remains controversial.16

This review will focus on human tissue-resident TCRαβ+CD8αβ+ IEL. We will compare and contrast them in the context of murine naturally occurring TCRαβ+CD8αα+ and adaptively induced TCRαβ+CD8αβ+ IEL while describing their distribution along the intestine, phenotype, TCR repertoire, autoreactivity, and function in homeostasis and disease.

Categorizing the tissue-resident IEL compartment

T cells within the IEL compartment are distinct from peripheral lymphocytes17,18 and have been categorized primarily based on ontogeny into the naturally occurring (Type B) IEL and the adaptively induced (Type A) IEL.11,13 Naturally occurring IEL are composed of TCRγδ+ T cells as well as the unique mouse subset of TCRαβ+CD8αα+ T cells. Naturally occurring IEL are tissue-resident lymphocytes that seed the tissue early in life and independently of microbial colonization of the gut,19,20,21,22 whereas adaptively induced IEL are classical TCRαβ+CD8αβ+ and TCRαβ+CD4+ T cells that are generated in response to local tissue insults23 and gain residence within the IEL compartment not unlike tissue-resident memory cells.24 To that end both naturally occurring and adaptively induced IEL were shown to be stable tissue-resident populations with little to no capacity to recirculate in the periphery in parabiosis experiments.25 Finally, both human and mouse IEL express the tissue-resident hallmark marker CD10326,27 and have a cytolytic effector profile characterized by expression of the lymphocyte activation marker CD69 as well as granzyme and perforin cytolytic granules.17,18,28

Regional composition of the IEL compartment in human vs. mouse

The intestine is classically separated into two anatomically distinct regions starting proximally at the small intestine defined by the duodenum, jejunum, and ileum, followed by the large intestine defined by the colon.29 IEL can be found along the full length of the intestinal tract with the density of IEL relative to intestinal epithelial cells (IEC) being higher in the small intestine relative to the colon.30,31 These IEL are composed of T cells and innate lymphoid cells (ILCs). In an attempt to provide a resource, we present primary data side by side on the intraepithelial T cell compartments of both human and murine small intestine (duodenum) and large intestine (colon) (Fig. 1).

Fig. 1
figure 1

A summary of the regional distribution of intraepithelial lymphocyte (IEL) subsets in human and mouse. We compare here the proportion of T cell subsets between the small intestine (duodenum) and large intestine (colon; right colon in human) while highlighting key differences between human and mouse. Human data is from healthy adults and mouse data is from 8-week-old C57BL/6 specific pathogen-free (SPF) mice from our colony at the University of Chicago. a The proportion of TCRγδ+ IEL among CD3+ IEL is summarized for human and mouse in duodenum and colon. The proportion of TCRαβ+CD8αβ+ and TCRαβ+CD8αα+ among TCRαβ+CD4− IEL and TCRαβ+CD4+ IEL among TCRαβ+ IEL is summarized for human and mouse in duodenum and colon. Ranges are based on published data in addition to our own unpublished data. b Freshly isolated IEL46 were stained with fluorescently labeled antibodies against CD45, CD3, TCRγδ, TCRαβ, CD4, CD8α, CD8β, and CD103 (human only). Top: representative flow cytometry contour plots with outliers and large dots are shown for the indicated populations (pre-gate indicated above plot) for human (left) or mouse (right) duodenum and colon. Bottom: the proportion of a given cell subset is summarized between the duodenum and colon for both human and mouse. For example, the proportion of TCRγδ+ IEL among CD3+ IEL in human is lower in the duodenum relative to the colon. The comparisons are made directly between the duodenum and colon, therefore the linear depiction of increases/decreases in proportions between the two segments are meant for simplicity and not to illustrate the progression across the other segments of the gut

Full size image

The intraepithelial T cell compartment in human is composed of TCRγδ+, TCRαβ+CD8αβ+, and TCRαβ+CD4+ T cells (Fig. 1). In addition to the three subsets above, the murine intraepithelial T cell compartment contains the naturally occurring TCRαβ+CD8αα+ IEL (Fig. 1). The proportions of these subsets and how they vary across the intestine between human and mouse is summarized in Fig. 1. The human small and large intestine is dominated by TCRαβ+ IEL. In contrast, the mouse small intestine has a more or less equal distribution of TCRγδ+ and TCRαβ+ IEL with a shift toward TCRαβ+ IEL in the colon.30,31 Interestingly, TCRγδ+ IEL increase proportionally from the small to large intestine in human while decreasing proportionally from small to large intestine in mouse (Fig. 1). Among TCRαβ+ IEL, TCRαβ+CD4+ IEL are a minor population in the small intestine but increase proportionally in the large intestine in both mouse30,31,32 and human33 (Fig. 1). Of note, the population of TCRαβ+CD4+ IEL in the human colon is primarily CD103−, a phenotype which has been described for IEL exposed to chronic antigen challenge;34 however, we cannot rule out contaminating CD103− cells from the lamina propria (Lp). A key difference is the sizeable proportion of the TCRαβ+CD8αα+ IEL in mouse, a population which is seemingly absent in human (Fig. 1). The existence of this population in human has been speculated on29 based on misinterpretation of data that failed to exclude TCRγδ+ IEL, which can express CD8αα, from the analysis.14,35 Additionally, a recent report showing a TCRαβ+CD8α+/CD8β-dim population in human intestine36 attempted to extend this observation to suggest the existence of TCRαβ+CD8αα+ IEL in human; however, the data suggests the same cell may express both CD8αα and CD8αβ dimers, a cell type that would not phenocopy the bonafide TCRαβ+CD8αα+ IEL found in mouse. Finally, the TCRαβ+CD8/CD4 double negative population which is enriched in the mouse colon30,32 and considered to be similar in nature to TCRαβ+CD8αα+ IEL37,38 as well as the TCRαβ+CD4+CD8α+ IEL39 whose inflammatory potential is modulated in the intestine via the upregulation of Runx3 and downregulation of ThPOK40,41 are both still poorly characterized in human. Single cell ex vivo transcriptional profiling of these subsets will help establish the extent to which these populations exist in human and whether or not they play a role in health or disease.

The physiological significance of the difference in distribution of various T cell subsets across the gut is still to be determined. Nonetheless, a complex array of local signals may play a role in shaping each compartment locally as can be appreciated from a study comparing the small and large intestine IEL compartments in neonatal mice vs. adult mice.21 It is well accepted that the microbial burden in the colon is higher than that in the small intestine; however, the mucus layer produced in the colon is thicker and keeps the microbiota farther from the hosts epithelial cells than in the small intestine.42,43 Furthermore, the impact of the microbiota on adaptively induced IEL is well established as these cells decrease drastically in absolute numbers in germ-free mice.19,20,22 Additionally, food is primarily absorbed in the small intestine therefore providing a unique pool of antigens that may impact IEL in the small intestine that are absent in the large intestine.

Are two cell subsets better than one?

There seems to be a clear division of labor for mouse IEL in that naturally occurring TCRαβ+CD8αα+ IEL and induced TCRαβ+CD8αβ+ IEL have been shown to have different transcriptional and functional profiles.17,37 One critical difference is the vast array of natural killer (NK) receptors that have been shown to be enriched on TCRαβ+CD8αα+ IEL when compared to TCRαβ+CD8αβ+ IEL in mouse37,44 (Fig. 2). In contrast to TCRαβ+CD8αα+ IEL which seem to be geared toward cytolytic function and express high levels of Granzyme B, TCRαβ+CD8αβ+ IEL in mouse have a substantial capacity to produce cytokines45 such as IFN-γ when probed ex vivo (Fig. 2). Of note, TCRαβ+CD8αα+ IEL express similar levels of Granzyme B in both young and adult mice indicative of their unique developmental program whereas the capacity of TCRαβ+CD8αβ+ IEL to produce Granzyme B and IFN-γ is gained with age (Fig. 2), presumably as the gut microbiota matures leading to the generation of more adaptively induced TCRαβ+CD8αβ+ IEL. Interestingly, TCRαβ+CD8αβ+ IEL at steady state in human have dual capacity in that they express NK receptors such as NKG2D46,47 and CD94 receptors,48,49 while also carrying the capacity to produce Granzyme B and inflammatory cytokines such as IFN-γ50 (Fig. 2). However, several important differences with respect to the nature of NK receptors expressed exist between TCRαβ+CD8αβ+IEL in human and TCRαβ+CD8αα+ IEL in mouse. First, both human TCRαβ+CD8αβ+ IEL and mouse TCRαβ+CD8αα+ IEL carry NK receptors such as NKG2A/CD94 and NKG2D that can recognize non-classical MHC class I molecules;51,52 however, mouse IEL also express the Ly49 family of receptors which endows them with the capacity to recognize classical MHC class I molecules53 (Fig. 2). Second, under steady-state conditions, human TCRαβ+CD8αβ+ IEL do not express any NK receptors with immunoreceptor tyrosine-based activation motif (ITAM) adapter molecules such as DAP12,46,54 whereas mouse TCRαβ+CD8αα+ IEL can express receptors associated with DAP1244,53,54 (Fig. 2). Therefore, the only activating receptor with the capacity to induce cell proliferation and cytokine production on healthy human TCRαβ+CD8αβ+ IEL is the TCR. However, under inflammatory conditions as is associated with CeD,55 a subset of aberrant TCRαβ+CD8αβ+ IEL can gain expression of activating NK receptors that can pair with DAP12, such as NKG2C/CD9456 (Fig. 2). These cells also displayed elevated transcript levels for a variety of killer immunoglobulin-like receptors56 thus bridging the gap in functional capacity that exists between mouse and human IEL at steady state.

Fig. 2
figure 2

Here we compare the human TCRαβ+CD8αβ+ IEL subset with the TCRαβ+CD8αβ+ and TCRαβ+CD8αα+ IEL subsets found in mouse for NK receptor expression and cytokine producing capacity. a The human TCRαβ+CD8αβ+ IEL expresses NK receptors such as the activating NKG2D and inhibitory NKG2A/CD94 in the steady state, whereas the TCRαβ+CD8αα+ IEL and not the TCRαβ+CD8αβ+ IEL subset is enriched for NK receptor expression in mouse. These TCRαβ+CD8αα+ IEL express NK receptors that can be found in steady-state human IEL such as NKG2D in addition to receptors such as those of the Ly49 family and NKG2C/CD94 which can associate with the ITAM adapter molecule DAP12. Interestingly, human IEL can also gain expression of ITAM bearing NK receptors such as NKG2C/CD94 in patients with CeD. The propensity for autoreactivity of IEL TCRs is shown with more autoreactive TCRs, such as those attributed to the development of TCRαβ+CD8αα+ IEL, illustrated in red and less autoreactive TCRs illustrated in green. The threshold for TCR/non-cognate antigen interactions to result in activation can be met in the case of human TCRαβ+CD8αβ+ IEL in CeD via the costimulatory impact of inflammatory signals such as IL-15 and the engagement of activating NK receptors. b Freshly isolated IEL were treated with 50 ng/mL of PMA and 500 ng/mL of Ionomycin for 3 h to assess the expression of Granzyme B and IFN-γ at steady state for human TCRαβ+CD8αβ+ IEL from a 30-year-old individual and TCRαβ+CD8αβ+ and TCRαβ+CD8αα+ IEL from 4-week and 11-week-old C57BL/6 SPF mice, revealing human TCRαβ+CD8αβ+ IEL are potent cytokine producers and simultaneously express Granzyme B, while cytokine production is age dependent and exclusive to the mouse TCRαβ+CD8αβ+ IEL with TCRαβ+CD8αα+ IEL being more potent expressers of Granzyme B. Representative flow cytometry contour plots with outliers and large dots are shown for intracellular staining with fluorescently labeled antibody against IFN-γ and Granzyme B for human (left) and mouse (right)

Full size image

TCRαβ+CD8αα+/TCRαβ+CD8αβ+ IEL TCR repertoire

Analysis of the TCR repertoire based on measurement of the length of the hypervariable - complementary determining region 3 (CDR3), using littermate controls and genetically identical mice, concluded the TCR repertoires of both naturally occurring TCRαβ+CD8αα+ and adaptively induced TCRαβ+CD8αβ+ IEL are highly restricted and non- overlapping, indicating these two subsets are not clonally related.57 Furthermore, there was no evidence for shared/public T cell clones even in mice born from the same mother and living in the same cages. Finally, the degree of oligoclonality was significantly higher in IEL than peripheral lymph node T cells. All together, these results suggest IEL undergo considerable expansion and/or are selectively accumulated, the antigens driving their expansion are diverse, and there is a stochastic component to the selection of particular TCRs undergoing oligoclonal expansion in a given mouse. Interestingly, an elegant study comparing the TCR repertoire of adaptively induced TCRαβ+CD8αβ+T cells in the thoracic duct and the epithelial compartment of the same mouse showed T cells sharing the same TCR are polyclonal in the thoracic duct and oligoclonal in the epithelium.58 Given that naïve TCRαβ+CD8αβ+T cells primed in the Peyer’s patches and mesenteric lymph nodes transit through the thoracic duct before returning through the blood stream in the intestine, these results suggest that the major clonal expansion takes place within the intestine. Furthermore, identification of similar T cell clones across the length of the small intestine suggests that T cells primed at one location of the gut seed the whole intestine after they have circulated through the thoracic duct.57 More recent studies show indeed that TCRαβ+CD8αα+ IEL subsets can be selected by a diverse set of MHC and MHC-like molecules.38,59 Furthermore, there is evidence for antigen-driven expansion of naturally occurring TCRαβ+CD8αα+ and adaptively induced TCRαβ+CD8αβ+ IEL. In H-Y TCR transgenic mice, TCRαβ+CD8αα+ IEL are expanded in the gut of male but not female mice,60 and the TCR repertoire of naturally occurring and adaptively induced IEL is less oligoclonal in the absence of microbiota.61 In germ-free mice, this polyclonality is accompanied by a significant decrease in the number of IEL, in particular of adaptively induced TCRαβ+CD8αβ+ and TCRαβ+CD4+ IELs, while the number of naturally occurring IEL is less decreased,20 with TCRγδ+ IEL being preserved in particular.19,22 These results suggest the highly oligoclonal nature of the IEL TCR repertoire results from antigen-driven selection. Whether oligoclonality of IEL, in particular of naturally occurring TCRαβ+CD8αα+ IEL, is antigen driven remains to be determined. Non-antigen-driven expansions can be potentially explained by epithelial factors such as IL-15.62,63 Dietary factors that are required for the generation of ligands for the aryl hydrocarbon receptor64 have also been shown to play an important role in the expansion and maintenance of naturally occurring IEL.65 The more significant decrease of TCRαβ+CD8αα+ IEL in food antigen-free mice as compared to germ-free mice22 may point to the critical role dietary factors play in their TCR-independent expansion.

In human, the oligoclonal nature of the TCRαβ+CD8αβ+ IEL TCR repertoire has been demonstrated in the small intestine66 and colon.67 As for mice, no public TCRs could be identified. Importantly, concomitant analysis of Lp lymphocytes showed a significantly more polyclonal TCR repertoire.67 Frequently, a unique CDR3 amino acid sequence accounted for more than 50% of all IEL sequences for a given TCR Vβ-chain.67 These observations pose the question of whether the highly oligoclonal nature of the TCR repertoire of human TCRαβ+CD8αβ+ IEL is dependent on antigen. The antigen-driven nature of a T cell expansion can be shown in different ways such as presence of conserved amino acids in the CDR3 that are encoded by distinct nucleotide sequences, presence of a conserved amino acid motif flanked by different amino acid sequences, or association of the same Vβ-chain with different Vα-chains. Using such criteria, existence of unequivocal antigen drive was not only shown in human TCRαβ+CD8αβ+ IEL but also linked to the expression of activating vs. inhibitory NK receptors.49 The combination of major clonal expansion with evidence for antigen-driven selection suggests that IEL respond to specific antigens that drive their expansion. However, many questions remain unanswered in both human and mouse. How can one explain the extremely clonal repertoire of IEL that contrasts not only with the repertoire of peripheral T cells but also with Lp T cells? Such a clonal repertoire was described during the Listeria recall response but not in the primary effector and memory TCRαβ+CD8αβ+T cell response.68 This suggests that TCRαβ+CD8αβ+ IEL may undergo chronic recall responses. In both human and mouse, it is extremely rare to identify IEL in mitosis. Human TCRαβ+CD8αβ+ IEL are KI67− under steady-state conditions.69 In mice, 0.2–3% of IEL are in mitosis.70,71 Using electron microscopy, it was estimated that 5% of IEL were immunoblast.72 Furthermore, an elegant study suggested that while adaptive TCRαβ+CD8αβ+ IEL divide in the GALT and the thoracic duct lymph, naturally occurring TCRαβ+CD8αα+ IEL divide after they have entered the epithelium.71 Where these recall responses take place and in response to what antigens remain to be determined as does the nature and location of the memory T cells that feed the response.

TCRαβ+CD8αα+/TCRαβ+CD8αβ+ IEL and autoreactivity

Studies in mouse suggest that autoreactivity is primarily a characteristic associated with the naturally occurring TCRαβ+CD8αα+ IEL subset. This IEL subset was shown to be selected by self-antigens restricted by non-classical and classical MHC class I and II molecules during thymic development.38,59,73,74,75 The current line of thought is self-reactive T cells that failed to undergo negative selection are destined to preferentially migrate and expand in the intestine,76,77 where they acquire CD8αα and granzyme.71 In addition to having an autoreactive TCR, these naturally occurring innate-like lymphocytes express activating NK receptors37,44,53 that enable them to recognize self-antigens induced under conditions of stress and inflammation78 (Fig. 2). This latter autoreactivity is destined to recognize modifications of self that signal the presence of pathogens and transformed cells. Interestingly, TCRαβ+CD8αα+ IEL are poorly reactive through their TCR,37 which is in line with the proposed role for the CD8αα homodimer as a corepressor,79 suggesting they respond mainly to innate signals.

In human, as previously discussed, TCRαβ+CD8αα+ IEL are, if not non-existent, extremely rare. Furthermore, reactivity to self-antigens such as CD1 was not detected in humans ex vivo, but rather seen in IE-CTL lines generated after multiple in vitro stimulations with PHA and peripheral blood monocytes.66 In contrast, TCRαβ+CD8αβ+ IEL show signs of antigen-driven expansion,49 suggesting that likewise to TCRαβ+CD8αβ+ IEL in mice, they are adaptively induced in response to exogenous antigens. However, unlike murine TCRαβ+CD8αβ+ IEL, all human TCRαβ+CD8αβ+ IEL express the activating NK receptor NKG2D,46,47 in addition to other activating NK receptors such as CD94 and NKR-P1A,48,49 indicating that they are poised to recognize modified self and respond to stress and inflammatory signals. Importantly, it was shown that NKG2D and CD94 receptors have the ability to kill targets through engagement of these receptors and in absence of TCR engagement.46,80 Finally, because both NKG2D46,47,80 and CD9449 can act as costimulatory molecules for the TCR they can significantly reduce the TCR activation threshold, hence potentially enabling the TCR to recognize non-cognate, low-affinity antigens in vivo. Conversely, induction of the CD8αα homodimer on TCRαβ+CD8αβ+ IEL in mouse is thought to increase the TCR activation threshold.79 Such a function for the CD8αα homodimer based on the observation of a TCRαβ+CD8α+/CD8β-dim population in human intestine36 is possible but yet to be demonstrated. The ability to recognize self-antigens is further enhanced when IL-15,81 a cytokine induced under conditions of stress, inflammation and infection, is upregulated.80,82,83 Of note, IL-1546,48 and NKG2D46 signaling in TCRαβ+CD8αβ+ IEL induces effector programs that are not observed in memory TCRαβ+CD8αβ+T cells. These observations led us to propose that activation of human TCRαβ+CD8αβ+ IEL is driven by recognition of non-classical MHC class I molecules by NK receptors and IL-15, signifying the presence of ongoing tissue distress.80,84 Concomitantly, a role for NKG2D in the rejection of skin tumors expressing non-classical MHC class I molecules Rae-1 and H60 by skin epithelial TCRγδ+ T cells was reported.85

In summary, both human and mouse IEL display autoreactive properties. However, whereas in mice this property is innately displayed by TCRαβ+CD8αα+ IEL that have been selected by self-antigens, in human adaptively induced TCRαβ+CD8αβ+ IEL can display these properties via the activating NK receptors they express and their ability to respond to IL-15. Because mouse IEL have been mainly studied in specific pathogen-free conditions, it remains possible that mouse TCRαβ+CD8αβ+ IEL could acquire expression of NK receptors under infectious and inflammatory conditions and gain the capacity to respond to stress signals.

IEL function during homeostasis

It is important to preserve the integrity of the intestinal epithelium given the burden of foreign antigens in the lumen so it is no surprise that tissue-resident lymphocytes are enriched in the epithelium. Given this proximity to the barrier, IEL are thought to participate in tissue surveillance and maintenance of barrier function. To accomplish this task, the IEL compartment is endowed with the capacity to respond to tissue stress via NK receptor/non-classical MHC-like molecule interactions as well as pathogen invasion via classical antigen-specific TCR/MHC interactions (Fig. 2). These functions can be accomplished by a singular cell in human in the TCRαβ+CD8αβ+ IEL but are differentially distributed in mouse between naturally occurring and adaptively induced IEL as can be appreciated by the predominant expression of NK receptors on TCRαβ+CD8αα+ IEL (Fig. 2). One of the primary functions of IEL is cytotoxicity37,50,86,87,88 and can be inferred from their potent expression of granzymes,17,18,71 which endow them with the capacity to lyse infected or aberrant cells.88,89 However, establishing requirement for a given function of IEL has been challenging due to the lack of models where IEL can be selectively ablated in addition to the redundancy in functional capacity between different IEL subsets. For instance, TCRαβ+ IEL are the primary producers of IFN-γ in response to Listeria monocytogenes infection, but in β2-Microglobulin-deficient mice, which lack adaptively induced TCRαβ+CD8αβ+ IEL, this response is compensated for by the β2-Microglobulin-independent TCRγδ and TCRαβ+CD8αα+ IEL subsets.87

Naturally occurring IELs have also been suggested to exert an immuno-regulatory role under homeostatic conditions. Interestingly, transfer of naturally occuring TCRαβ+CD8αα+ IEL but not adaptively induced TCRαβ+CD8αβ+ IEL is protective against the development of colitis in a CD4+CD45RBhi T cell transfer colitis model.90 In line with this observation, TCRδ−/− mice, which lack naturally occurring TCRγδ+ IEL, infected with Eimeria vermiformis show increased epithelial cell damage,91 an observation that fits with the proposed role for TCRγδ+ IEL in tissue repair mediated via secretion of keratinocyte growth factor.92 Furthermore, naturally occurring TCRγδ+ IEL actively survey epithelial cells at homeostasis and respond to Salmonella infection by increasing their movement speed and localization while enhancing expression of antimicrobial genes to participate in the early phase of the immune response.93

The function of adaptively induced IEL has been primarily studied in the context of infection, and although evidence for antigen-specific responses in naturally occurring IEL are still lacking, antigen-specific adaptively induced TCRαβ+CD8αβ+ IEL can be quantified by MHC-tetramer staining post infection with either vesicular stomatitis virus or Listeria monocytogenes.94,95 In accordance, the protective potential of adaptively induced TCRαβ+CD8αβ+ IEL is best demonstrated in models of adoptive transfer of antigen-specific IEL in various mouse infection models.88,96

The role for human IEL under homeostatic conditions is still poorly understood. What can be said for certain is human IEL have potent cytolytic capacity which can be mediated via engagement of the TCR or NK receptors such as NKG2D.46 Additionally, the potent cytokine production by TCRαβ+CD8αβ+ IEL in healthy individuals (Fig. 2) strongly suggests they are responding to stimuli in vivo; however, the specificity of these responses and the degree to which functional redundancy between various IEL subsets exists is yet to be determined.

IEL function during pathology

A role for IEL in pathology was implied primarily because of their increase in enteropathies such as CeD, tropical sprue, and parasite infections.4 IEL were also reported to be increased in graft vs. host disease, allograft rejection, autoimmune enteropathies, and inflammatory bowel disease; however, their increase in these disorders is less significant and occurs late in the disease process.4,97,98 Because of the technical limitations that prevent the selective elimination of IEL in mouse models of disease, their requirement in immunopathology could not be established. In a mouse model expressing the antigen ovalbumin in the epithelium, TCRαβ+CD8αβ+T cells specific for ovalbumin (OT-1 TCR Transgenic T cells) preferentially migrated to the inductive and effector sites of the intestinal mucosa without causing epithelial cell destruction. However, once mice were infected with a virus expressing ovalbumin, OT-1 T cells induced epithelial cell destruction and villous atrophy.99 Using human fetal small intestinal explants, it was shown that activation of T cells could induce villous atrophy,100 but the respective role of IEL and Lp lymphocytes could not be delineated, neither in the mouse nor the human experimental model.

In human, the role of TCRαβ+CD8αβ+ IEL in disease is best established in CeD, a T cell-mediated small intestinal enteropathy induced by dietary gluten in genetically susceptible HLA-DQ2 or HLA-DQ8 individuals.55,84,98,101 Intraepithelial lymphocytosis is a hallmark of CeD and used in clinic for the diagnosis of CeD.102 However, the inability to identify gluten-specific TCRαβ+CD8αβ+ IEL initially led scientists in the field to propose that the increase in IEL was secondary to the activation of dietary gluten-specific TCRαβ+CD4+T helper-1 (TH1) cells and did not play a role in the pathogenesis of CeD.103 However, similarly to patients with latent autoimmune diabetes of adults who preserve a functional pancreas despite the presence of an adaptive immune response against beta-islet antigens,104,105 potential CeD patients conserve a normal intestinal architecture despite having developed an adaptive immune response against gluten,106,107 suggesting that gluten-specific TH1 cells are not the effector cells mediating tissue destruction. The discovery that activating NK receptors and their ligands are upregulated in TCRαβ+CD8αβ+ IEL and epithelial cells in active CeD but not in patients on a gluten-free diet,46,47,56 provided a mechanism through which TCRαβ+CD8αβ+ IEL can destroy epithelial cells despite not being gluten-specific.84 The observation that activating NK receptors and their ligands are not upregulated in potential CeD108 further supports that TCRαβ+CD8αβ+ IEL are the key effector T cell subset mediating epithelial cell destruction and villous atrophy in CeD. In agreement with observations in human, mice in which gluten-specific TH1 IFN-γ-producing T cells were induced in absence of TCRαβ+CD8αβ+ IEL activation failed to develop villous atrophy,109,110,111 in contrast to mice in which TCRαβ+CD8αβ+ IEL acquired lymphokine killer-like activity.112 In addition to TCRαβ+CD8αβ+ IEL, innate-like IEL lacking surface TCR expression were involved in the development of villous atrophy in patients with refractory CeD,113,114,115 an indolent or cryptic innate intraepithelial lymphoma that rarely complicates CeD. In both active and refractory CeD, upregulation of IL-15 in the epithelium is thought to play a critical role in the activation of IEL and epithelial cell destruction.81,116,117 In line with a role for IL-15 in promoting tissue destruction, a study showed that when mice overexpressing IL-15 in the epithelium were crossed to ovalbumin-specific CD4 TCR transgenic mice, they developed villous atrophy when ingesting ovalbumin.118 Altogether, the observations in CeD point toward TCRαβ+CD8αβ+ IEL being a key effector cell able to mediate epithelial cell destruction based on the recognition of stress and inflammatory signals. Intriguingly, TCRαβ+CD8αβ+ IEL in active CeD share numerous functional features with TCRαβ+CD8αα+ IEL in that they can exert NK-like properties46 and even express NK receptors associated with ITAM-bearing adapter molecules56 (Fig. 2a).

Conclusion

The tissue-resident intraepithelial T cell compartment is shaped by the local environment (oral antigen, microbial signals, region-specific IEL–IEC interactions) as can be appreciated by the different proportion of various cell subsets between the small and large intestine in both human and mouse. Therefore, studies involving IEL should consider the distribution of IEL subsets across the gut; especially those involving the design of mouse models to investigate human relevant questions as there are also differences in the distribution of subsets between the two species. One such difference is the relative absence of the naturally occurring mouse TCRαβ+CD8αα+ IEL in human. A close comparison between the human and mouse IEL subsets highlights a critical aspect of IEL function which is the dual capacity to recognize both modified self and non-self. Whereas in mouse, recognition of modified self via NK receptors is restricted to the naturally occurring TCRαβ+CD8αα+ IEL and recognition of non-self is best described for adaptively induced TCRαβ+CD8αβ+ IEL, in human a singular TCRαβ+CD8αβ+ IEL subset has the capacity to do both given the dual expression of NK receptors and the TCR. The pathogenic role and propensity for autoreactivity of the human TCRαβ+CD8αβ+ IEL subset has been best characterized in CeD where inflammatory signals such as IL-15 can combine with NK receptor/stress ligand engagement to result in non-cognate antigen activation of IEL TCRs, ultimately resulting in tissue destruction.

The role for the local environment in shaping the IEL compartment is further highlighted by the observation that both naturally occurring TCRαβ+CD8αα+ and adaptively induced TCRαβ+CD8αβ+ IEL have private oligoclonal TCR repertoires in the steady state. To that point, it has been challenging to uncover the intimate specificities of IEL. This is most likely due to the rich source of antigen they encounter, be it microbial, dietary, or epithelial antigens and thus studies geared toward removal of one source of antigen may simply result in the expansion of IEL specific to the remaining source of antigen. The reduction in IEL numbers observed in both food antigen-free and germ-free mice suggests IEL may be reactive to dietary and microbial antigens; however, there has been no identification of dietary or commensal-specific IEL.

The reason for the absence of the TCRαβ+CD8αα+ IEL in human is still unclear. Given the observations that TCRαβ+CD8αβ+ IEL in mice increase with age and that germ-free mice have a significant reduction of TCRαβ+CD8αβ+ IEL, it remains possible that wild mice and mice under chronic inflammatory conditions would lose the TCRαβ+CD8αα+ IEL subset at the expense of the TCRαβ+CD8αβ+ IEL that may acquire expression of NK receptors akin to those in human. Interestingly, TCRγδ+ IEL, which are also considered naturally occurring in mouse, are also significantly underrepresented in the human intestine.

The IEL compartment as a whole is comprised of many cell subsets, including non-T cell subsets of ILCs, and these subsets are influenced by their environment under steady (small intestine vs. colon) and pathogenic conditions. What remains to be elucidated is whether there is a clear division of labor between IEL subsets and if not to what extent does redundancy exist in their functions. Finally, whether the tissue-resident IEL compartment is stable and the extent to which it can be reshaped as a result of chronic insults and inflammatory conditions is a question with important physio-pathological consequences.