Cutting Edge: Eomesodermin Is Sufficient To Direct Type 1 Innate Lymphocyte Development into the Conventional NK Lineage Free

Innate lymphoid cells are enriched at mucosal barriers in close proximity to epithelial surfaces, where they exert crucial functions during infection, tissue injury, and inflammation. Type 1 innate lymphocytes (ILCs) express the NK1.1 and NKp46 lineage Ags as well as the transcription factor T-bet. Type 1 ILCs can be further subdivided into discrete populations based on expression of Eomesodermin (Eomes), a close homolog of T-bet (1, 2). Eomes⁺ conventional NK cells (cNKs) express integrin CD49b (DX5) and MHC class I–specific Ly49 receptors and circulate through lymphoid and nonlymphoid organs. In contrast, Eomes⁻ type 1 helper innate lymphoid cells (hILC1s) express integrin CD49a and are anchored to their tissues and mucosal sites of origin (1–7). hILC1s and cNKs are thought to arise from separate lymphoid progenitors (6, 8). Nonetheless, type 1 ILC lineages are closely related transcriptionally and temporal deletion of Eomes from cNKs unveils latent hILC1-like characteristics (1, 9). We used genetic approaches to test the necessity and sufficiency of Eomes to determine the outcome of type 1 ILC maturation. Our results lead us to the conclusion that Eomes expression is a key determinant of cNK versus hILC1 maturation following lineage specification.

Mice were housed in specific pathogen-free conditions and used in accordance with the Columbia University Institutional Animal Care and Use Committee's guidelines. C57BL/6 Eomes^flox/flox, Tbx21^flox/flox, Tbx21^−/− (10, 11), NKp46-Cre⁺ (12), and Id2^hCD5/hCD5 reporter (13) mice were generated as previously described. They were bred locally to obtain the compound genotypes Eomes^flox/floxNKp46-Cre⁺, Tbx21^flox/floxNKp46-Cre⁺, and Eomes^flox/floxNKp46-Cre⁺Tbx21^−/−.

Detailed characterization of the Tbx21-Eomes transgenic (Tg) mouse strain will appear elsewhere. Briefly, Eomes cDNA followed by an SV40-PolyA site was inserted into the ATG translational start site of T-bet in the Tbx21 bacterial artificial chromosome clone RP23-237M14 (CHORI) by Red/ET recombination technology (Gene Bridges). This strategy enabled T-bet regulatory elements to drive Eomes expression instead of T-bet expression from the Tg allele. The bacterial artificial chromosome construct was used for pronuclear injection of C57BL/6 zygotes. For some experiments, Tg mice were crossed to Id2^hCD5/hCD5 reporter mice that express human CD5 from an internal ribosome entry site–driven cassette knocked in to the 3′-untranslated region of Id2 (13). Furthermore, Tg mice were crossed to Tbx21^−/− mice for some analyses. Aged Tg mice had no evidence of skin lesions, intestinal inflammation, or hepatosplenomegaly.

Single-cell suspensions were prepared from liver, spleen, bone marrow, uterus, salivary gland, thymus, and lung tissues. Liver lymphocytes were isolated at the interphase of a 40/60% Percoll gradient (1). Uterus and salivary gland were digested in the presence of DNase I and collagenase D or liberase TL (5, 7), whereas lung digestion additionally incorporated trypsin inhibitor. Cells were stained with a LIVE/DEAD fixable dead cell stain kit (Invitrogen) and blocked with Abs against CD16/CD32 (BD Biosciences) prior to staining with Abs for CD3, CD11b, CD27, CD43, CD122, human CD5, integrin α_v, killer cell lectin–like receptor subfamily G, member 1 (KLRG1), Ly49G2, Ly49H, NKp46, and TRAIL (eBioscience); and CD19, CD49a, CD49b, CD127, Ly49D, Ly-6G/Ly-6C (Gr-1), NK1.1, NKG2A/C/E, and TER-119 (BD Biosciences). Cells were treated with Cytofix/Cytoperm kits prior to staining with Abs for the transcription factors T-bet (eBioscience) and Eomes (eBioscience) or the cytokine TNF-α (BD Biosciences).

Statistics were calculated using Prism (GraphPad Software). Differences between groups were quantified using the unpaired t test. Statistical significance was reached at p < 0.05.

Within various lymphoid and nonlymphoid tissues examined, we could identify hILC1s and cNKs among lineage-negative, NK1.1⁺NKp46⁺ cells based on reciprocal expression of integrins CD49a and CD49b (DX5), respectively (Fig. 1A). We previously examined cNK development in mice with pan-hematopoietic deficiency of Eomes (1). To determine if Eomes expression at the onset of ILC development is necessary for cNK differentiation, we intercrossed Eomes^flox/flox and NKp46-Cre⁺ mice (12). Eomes^flox/floxNKp46-Cre⁺ mice had a substantial loss of CD49a⁻DX5⁺ cNKs (Fig. 1A), which supports the prior suggestion that Eomes is genetically upstream of CD49b induction (1). Remnant NK1.1⁺NKp46⁺ cells in Eomes conditional knockout mice were CD49a⁺DX5⁻, CD49a⁺DX5⁺, or CD49a⁻DX5⁺ cells (Fig. 1A), and the persistence of these cells did not appear to result from inefficient deletion of Eomes (Supplemental Fig. 1A). These findings are consistent with a role for Eomes upstream of CD49a repression, which is supported by prior evidence of direct binding of Eomes to the locus encoding CD49a (14). Examination of TRAIL, CD127, and integrin α_v confirmed that deletion of Eomes shifts the balance of type 1 ILCs from cNKs to helper-like ILC1s (Supplemental Fig. 1B).

To test the contribution of T-bet to residual type 1 ILC development in Eomes conditional knockout mice, we examined Eomes^flox/floxNKp46-Cre⁺Tbx21^−/− mice (Fig. 1B). We used mice with a germline Tbx21 deletion because Tbx21^flox/floxNKp46-Cre⁺ mice underwent substantial development or expansion of hILC1s that escaped Tbx21 deletion (Supplemental Fig. 1C, 1D). Compound deficiency of T-bet and Eomes resulted in almost complete elimination of the type 1 ILC compartment in all tissues examined (Fig. 1B). Together with previous findings (1, 6, 7), the present results suggest that both T-bet and Eomes are required to develop the spectrum of lymphoid and nonlymphoid type 1 ILCs. Eomes, furthermore, seemingly plays its essential role in type 1 ILC maturation after the onset of NKp46 expression.

Because loss of Eomes at the NKp46-expressing stage illustrates the necessity of Eomes for cNK development (Fig. 1A), we wished to test if Eomes expression at the onset of hILC1 lineage maturation would be sufficient to redirect hILC1s into a cNK identity. T-bet–dependent hILC1s lack endogenous Eomes expression (1, 2, 4, 6, 7). To enforce Eomes expression in developing hILC1s, we generated Tg mice in which Tbx21 locus control elements drive expression of Eomes codons. Because T-bet is not expressed at substantial levels until type 1 ILCs acquire the lineage Ags NK1.1 and NKp46 (1, 15), we could determine the effect of Tg Eomes expression following divergent lineage specification from NK1.1⁻NKp46⁻ progenitors (6, 8).

We first verified that Tg Eomes is expressed when and where T-bet is expressed by examining the pattern of Eomes expression in three cell populations that ordinarily express T-bet but not Eomes: neonatal type 1 ILCs, NK1.1⁺ T cells, and developing Th1 cells (1, 16). In each case, Tg cells expressed Eomes in proportion to T-bet (Fig. 2A). Analyses of type 1 ILCs revealed that Tg Eomes also perturbed the normal spatial and temporal pattern of hILC1 development. Tg Eomes caused a shift in frequency and absolute numbers toward CD49b (DX5) expression among NK1.1⁺NKp46⁺ type 1 ILCs in both lymphoid and nonlymphoid tissues of adult mice (Fig. 2B, Supplemental Fig. 2A, 2B). Examination of T-bet and Eomes levels revealed repressed T-bet expression in Tg type 1 ILC subsets (Fig. 2C, Supplemental Fig. 2C), suggesting that repression of T-bet by Tg Eomes may at least partly explain the phenotypic shift. In addition to the shift in adult ILC subsets, Tg neonates exhibited precocious development of DX5⁺ type 1 ILCs, compared with the minor population evident in wild-type (WT) neonates (Fig. 2D). Finally, Eomes⁻ hILC1s have been suggested to express higher levels of TNF-α than Eomes⁺ cNKs (1, 6, 7). Consistent with the observed shift in integrin expression, we found that Tg Eomes attenuated TNF-α expression by hepatic type 1 ILCs (Fig. 2E). Eomes expression, thus, appears sufficient to redirect hILC1s into a more cNK-like fate.

Emerging evidence suggests that progenitor cells of all helper ILCs are distinct from progenitor cells of cNKs and that helper ILC progenitors are marked by abundant expression of the transcription factor Id2 (6). We found that the descendant lineages of the distinct precursors, Eomes⁻ hILC1s versus Eomes⁺ cNKs, continue to express high versus intermediate levels of Id2, respectively (Fig. 3A, Supplemental Fig. 2D). Eomes⁻ hILC1s expressed higher levels of Id2 than Eomes⁺ cNKs regardless of their maturation (Fig. 3B). Examination of Tg NK1.1⁺NKp46⁺ type 1 ILCs revealed a subpopulation of Id2^hiEomes⁺ cells, suggesting the possibility that some cells arose from helper ILC precursors and were subsequently diverted into a cNK-like fate (Fig. 3C, Supplemental Fig. 2E). In support of the notion that the Id2^hi subpopulation may have arisen from helper ILC precursors, we detected other characteristics of hILC1s (1, 6, 17), including expression of CD127, integrin α_v (Fig. 3D, Supplemental Fig. 2F), NKG2A/C/E receptors, and a restricted Ly49 receptor repertoire (Fig. 3E, Supplemental Fig. 2G, 2H). The Id2^hi subset of Tg type 1 ILCs, nonetheless, also acquired numerous cNK features. Tg Eomes promoted upregulation of CD11b, CD43, KLRG1, and CD122 expression among Id2^hi cells in the liver and to a lesser degree in the spleen (Fig. 3E, Supplemental Fig. 2G, 2H). Despite increased maturity, Id2^hi Tg type 1 ILCs did not transition to the most terminal CD27⁻ stages (18) of cNK maturation (Fig. 3E, Supplemental Fig. 2G, 2H). Together, these findings suggest that Eomes is sufficient to convert some but not all attributes of hILC1s to those of cNKs.

To determine whether the hILC1-like characteristics in a subset of Tg type 1 ILCs were T-bet dependent, we examined Tg mice with genetic deletion of both endogenous Tbx21 alleles. In the absence of an Id2 reporter allele, we used Ly49⁻NK1.1⁺NKp46⁺ as surrogate markers for hILC1s. Tg Eomes was sufficient to partially rescue the development of cells with hILC1 features (expression of CD127 and integrin α_v) in Tbx21^−/− mice (Fig. 3F), indicating that the residual hILC1-like characteristics of the Tg line are not likely to be T-bet dependent. Therefore, the structural homology between the two transcription factors may enable Eomes to partially substitute for T-bet in hILC1 development. Similarly, T-bet may partially substitute for Eomes during cNK development insofar as Eomes^flox/floxNKp46-Cre⁺ mice contain fewer cNK cells when intercrossed to T-bet–deficient mice (Fig. 1A, 1B).

Our findings suggest that type 1 ILCs with distinct lineage origins have the potential for substantial developmental plasticity. Our data are consistent with a model in which separate type 1 ILC lineages are specified from distinct precursors, and the absence or presence of Eomes subsequently directs or switches, albeit incompletely, their determination as hILC1s or cNKs. In an analogous manner, CD4⁺ T cells can acquire cytotoxic T cell attributes and lose helper attributes when ThPOK expression decreases and Runx3 expression increases, yet they can still continue to express the CD4 marker (19, 20). Understanding the nature of the interrelatedness and plasticity between type 1 ILC subsets will help harness their therapeutic potential in infection, inflammation, and cancer.

We thank Dr. Jeff Zhu for advice in designing the transgene and Dr. Victor Lin of the Transgenic Mouse Shared Resource for microinjection and production of the Tg line.

The authors have no financial conflicts of interest.