Mouse NK cells express Tcrd transcripts
In Tcrd-H2BEGFP knock-in mice, expression of H2B-EGFP enabled delineation of early events leading to γδ T cell development in the thymus 29. Importantly, germ-line transcription of the Tcrd-H2BEGFP locus in developing thymocytes leads to intermediate levels of reporter fluorescence, whereas high H2B-EGFP levels were confined to successfully rearranged and selected γδ T cells in the thymus and in the periphery.
Unexpectedly, EGFP expression was observed in NK cells in peripheral organs (Fig. 1A). No EGFP expression was observed in granulocytes, monocytes and B cells (data not shown) 29. In contrast to γδ T cells, which express uniform high levels of EGFP (Fig. 1A), NK cells, defined by the surface phenotype NK1.1+CD3–CD5–, from all organs examined were found to express three discrete levels of EGFP (Fig. 1). Two populations of NK cells with low or intermediate levels of EGFP expression (hereafter referred to as EGFPlo and EGFPint, respectively) each contributed approximately 40–60% of the NK cell pool in healthy Tcrd-H2BEGFP mice. An additional population, representing 0.5–6% of NK cells displayed high expression of EGFP (EGFPhi cells). The relative frequencies of EGFPlo and EGFPint populations were similar in all organs examined with the exception of the liver where EGFPlo cells are overrepresented. The EGFPhi phenotype was most frequent in the lymph nodes, the liver and the bone marrow of Tcrd-H2BEGFP mice. In this way, the large majority of NK cells expressed EGFP, indicating that the previously reported Tcrd expression by human NK cells 32, 33 also extends to mouse NK cells.
Figure 1. NK cells express Tcrd transcripts. Expression of EGFP by cell populations of Tcrd-H2BEGFP mice is shown. (A) EGFP expression on splenocytes gated for TCRγδ+ cells and NK1.1+CD3–CD5– NK cells in Tcrd-H2BEGFP mice. Fluorescence profiles (black lines) are overlaid on those obtained from wild-type C57BL/6 controls (dashed lines). (B) Histograms show expression of EGFP on gated NK1.1+CD3–CD5– NK cells from isolates of lymph nodes (inguinal, axillar and mesenteric), liver, lung and bone marrow. The percentages of cells falling into distinct levels of EGFP expression (EGFPlo, EGFPint and EGFPhi) are given. Using identical stainings in spleen [see (A)], the percentages of NK-gated cells with EGFPlo, EGFPint and EGFPhi phenotypes were 38%, 62% and 1.3%, respectively. The results shown are representative of at least three independent experiments.
Download figure to PowerPoint
Germ-line Tcrd-EGFP expression initiates during NK maturation
Development of NK cells from bone marrow Lin– Sca1hic-Kithi hematopoietic stem cells is thought to occur through intermediate populations of lymphoid progenitors, including common lymphoid progenitors and committed NK progenitors, which lack lineage markers but express the IL-2R/IL-15R β chain, CD122 23, 34. We wished to determine whether the different levels of Tcrd expression relate to different stages of NK cell development. Rag-deficient mice were used to facilitate identification of cell populations in the bone marrow, and to restrict our study to that of germ-line Tcrd transcription. Developing NK cells in the bone marrow of Rag-deficient Tcrd-H2BEGFP mice were divided into NK precursors (CD122+NK1.1–DX5–), immature NK (CD122+NK1.1+DX5–) and mature (CD122+NK1.1+DX5+) NK cell populations according to current models 23, 35–37.
Acquisition of EGFP expression at both low and intermediate levels occurred concomitant with DX5 expression, as EGFP was present in mature NK cells but undetectable in immature NK cells (Fig. 2). Immature, DX5– NK cells are present at a high frequency in the liver 24, providing a possible explanation for the overrepresentation of EGFPlo/– cells amongst liver NK cells from Tcrd-H2BEGFP mice (Fig. 1B). These data indicate that germ-line expression of Tcrd by NK cells only occurs following both commitment to the NK cell lineage and further maturation to a DX5+ stage. Mature NK cells include both EGFPlo and EGFPint subpopulations of NK cells (Fig. 2).
Figure 2. Tcrd expression initiates during maturation of immature DX5– NK cells to mature DX5+ cells. Bone marrow cells from Rag1–/– × Tcrd-H2BEGFP mice were stained with antibodies to CD122, DX5 and NK1.1. Cell populations were defined as NK precursors (NKp) (CD122+NK1.1–DX5–), immature NK cells (CD122+NK1.1+DX5–) and mature NK cells (CD122+NK1.1+DX5+). The histograms show expression of EGFP in the indicated NK subsets. Results are representative of three independent experiments.
Download figure to PowerPoint
NK-gated EGFPhi cells are NK-like γδ T cells with low-level TCR expression
The small proportion of NK1.1+CD3–CD5– (referred to as “NK-gated”) cells that expressed high levels of EGFP (Fig. 1A, B) were selectively absent in Rag-deficient Tcrd-H2BEGFP mice (Fig. 3A). This high level of EGFP expression, comparable with that found in γδ T cells, suggested that these NK-gated cells have high levels of Tcrd transcription. In the thymus, developing γδ T cells only attain these high levels of EGFP expression following V(D)J recombination and surface expression of their TCR and progression through a developmental checkpoint 29. The presence of EGFPhi NK-gated cells not only required V(D)J recombination, but also molecules involved in T cell signaling, such as Lat and CD3ϵ (Fig. 3A).
Figure 3. High EGFP expression in NK-gated cells marks γδ T cells with low surface TCR and CD3. (A) Graphs show the expression of EGFP by spleen and lymph node NK1.1+CD3–CD5– of Tcrd-H2BEGFP mice sufficient and deficient for Rag1, CD3ϵ and Lat. Results from the indicated strains (black line) are overlaid on wild-type C57BL/6 (grey shaded) NK1.1+CD3–CD5– cells. The mean percentages ± SD of EGFPhi NK-like cells for three mice are indicated over the gates. (B) EGFPhi cells have a low cell surface level of TCRγδ. Tcrd-H2BEGFP splenocytes were stained with antibodies against CD3, CD5, NK1.1 and TCRγδ or isotype control. The electronic gating of NK1.1+CD3–CD5– NK cells and EGFPlo, EGFPint and EGFPhi subpopulations is shown. In comparison, whole splenocytes are gated on the population expressing high levels of EGFP. Histograms show the staining of each population with anti-TCRγδ antibodies (heavy black line) overlaid on isotype control staining (grey shaded). (C) Heterogeneous levels of CD3/CD5 and NK1.1 are observed on γδ T cells defined by high levels of EGFP expression in Tcrd-H2BEGFP mice. The contour plot shows the expression of CD3/CD5 and NK1.1 on electronically gated EGFPhigh whole splenocytes from Tcrd-H2BEGFP mice. Results are representative of at least three independent experiments.
Download figure to PowerPoint
These observations raised the question of whether these NK-gated cells belong to the NK or T cell lineages. To answer this, surface staining with anti-TCRγδ was performed on splenic NK-gated cells (Fig. 3B). Staining with TCRγδ mAb was higher on the population of EGFPhi NK-gated cells compared with isotype control antibody, whereas other NK cell populations were unstained. However, the geometric mean fluorescence intensity was only around 15% that of TCRγδ+ EGFPhi γδ T cells (Fig. 3B).
These data indicate that a population of γδ T cells expresses NK1.1 and low levels of surface TCR. Comparison of CD3/CD5 versus NK1.1 expression on total Tcrd-H2BEGFP EGFPhigh splenocytes (Fig. 3C) showed a reduced level of CD3/CD5 expression on a population of EGFPhigh T cells that also express high levels of NK1.1. Demonstrating this inverse correlation, the geometric mean fluorescence intensity of CD3/CD5 staining on NK1.1+ EGFPhigh cells is ∼50% that of NK1.1– EGFPhigh cells, and the 2% of EGFPhigh cells with the most NK1.1 expression have only ∼20% of this level of CD3/CD5 staining (Fig. 3C and data not shown). The poor discrimination of these cells using CD3/CD5, the low level of TCRγδ they express and the general absence of CD4 and CD8 on most γδ T cells mean that EGFPhi NK-gated cells are indiscernible from classical NK cells using surface staining. We thus defined this subset of NK-gated, TCRγδlow cells as NK-like γδ T cells.
NK-like γδ T cells are thymus-dependent CD127+ intracellular CD3+ cells
In an attempt to characterize NK-like γδ T cells further, we performed intracellular staining for CD3ϵ. Consistent with an assignment as T cells, NK-like γδ T cells (gated as EGFPhiNK1.1+CD3–CD5–) could be selectively identified in wild-type and Tcrd-H2BEGFP mice using intracellular CD3 staining (Fig. 4A), comprising around 0.5–2% of NK1.1+CD3–CD5– gated cells. In Tcrd-H2BEGFP mice, the majority of NK-like T cells in the NK gate were EGFPhi, whilst an additional, smaller population lacking EGFP expression was observed.
Figure 4. T cells in the NK cell gate are thymus-dependent cells identified by expression of intracellular CD3ϵ and CD127. (A) Intracellular expression of CD3ϵ on populations of NK-gated (NK1.1+CD3–CD5–) cells from spleens of the indicated strains of mice. Following surface staining, intracellular staining of CD3ϵ was performed using the same CD3ϵ mAb conjugated to a different fluorochrome. Contour plots show the expression of EGFP and intracellular levels of CD3 (ic CD3) or isotype control antibody (ic isotype). The percentages of cells within the indicated gates are shown. Results shown are representative of at least three independent experiments. (B) Splenic, lymph node (inguinal and axillar) and thymic cells from wild-type and athymic Foxn1nu/nu mice on a C57BL/6 background were gated on NK cells (NK1.1+CD19–CD3/CD4/CD8–). The fluorescence profiles of staining for intracellular CD3 and CD127 or isotype control are shown. Percentages of cells within the indicated quadrants are given. (C) The figure shows quantification of CD127 expression by intracellular CD3– NK cells in thymus, spleen and lymph nodes of C57BL/6, and nu/nu mice on C57BL/6 (B6-con.Foxn1nu/nu) and nude (Nude.Foxn1nu/nu) backgrounds. C57BL/6 background NK cells were gated as in (B), whereas NKp46+CD19–CD3/CD4/CD8– was used to identify nude mouse NK cells. Two samples of B6-con.Foxn1nu/nu mice are shown individually. Other data are means ± SD for the number of mice indicated.
Download figure to PowerPoint
EGFPhi NK-like T cells correspond to γδ T cells as they were absent from Rag1-deficient Tcrd-H2BEGFP mice, but present in Tcrd-H2BEGFP mice deficient in the enhancer for TCR β gene (Eβ). In contrast, EGFP– NK-like T cells were absent from both Rag1-deficient and Eβ-deficient Tcrd-H2BEGFP mice, assigning them as αβ T cells. It is possible that NK-like αβ T cells comprised CD1d-restricted Vα14i NKT cells, MR1-restricted MAIT cells, or other populations of NK1.1+ T cells that, similarly, have low levels of TCR expression. However, the small size of this subset (<0.5% of NK-gated cells) precluded its detailed analysis in the absence of selective genetic tagging.
The striking resemblance of NK-like γδ T cells to NK cells raised the possibility that these cells have previously been described as NK cells. This situation may be problematic when distinctions between T cell and NK cell populations are being made. In particular, NK-like γδ T cells could contaminate an NK cell population when their developmental origin is under investigation. We therefore wished to examine the developmental origins of NK-like γδ T cells. Athymic C57BL/6 mice congenic for the nude mutation (B6-con.Foxn1nu/nu) were examined to determine a possible thymic origin of NK-like T cells (Fig. 4B). B6-con.Foxn1nu/nu mice lacked discernable intracellular CD3 staining on NK-gated cells from spleen and lymph node, strongly arguing for a thymic origin of these cells. Importantly, the large majority of NK-like T cells in C57BL/6 mice also expressed the IL-7Rα chain, CD127 (Fig. 4B and data not shown), a receptor that has been described as a distinguishing marker of thymic NK cells 38. Without intracellular staining with CD3, these cells are therefore indistinguishable from NK cell populations that express CD127.
Whilst NK-like T cells make only a minor contribution to the population of NK cells in lymph node and spleen, they constitute a major fraction of NK-gated cells in the thymus: Approximately 0.04% of thymic cells fall into an NK gate defined by expression of NK1.1 and lack of CD3, CD4, CD8 and CD19. Of these, around 75% (Fig. 4B) express intracellular CD3ϵ, and in the Tcrd-H2BEFGP model, they express either high levels of EGFP or intracellular TCRβ (data not shown), suggesting that they are T cells. This results in a large over-evaluation of the CD127+ NK compartment in the thymus when intracellular CD3 staining is not used to exclude NK-like T cells. Indeed, once intracellular CD3+ cells have been factored out, 0.012 ± 0.004% of adult thymocytes are NK cells. Given a thymus of 150 million cells, these figures mean that only 17 400 ± 5600 thymic cells are NK cells. Only 20% (equivalent to 4 000 cells) of these express CD127. These statistics are inconsistent with a large thymic output of NK cells.
Identification of NK-like T cells through intracellular staining with CD3 thus permitted us to re-examine CD127+ NK cells after gating to remove NK-like T cells (Fig. 4C). CD127 is selectively expressed by human CD56bright NK cells, leading to its proposition as a marker for an equivalent subset in the mouse 38, 39. Human CD56bright NK cells are not NK-like γδ T cells since they do not have intracellular CD3ϵ antibody staining (data not shown). The highest frequency of CD127 expression on C57BL/6 NK cells was observed in the thymus (12.9 ± 6.9%), a similar representation was observed in the lymph nodes (9.6 ± 3.8%), but much lower frequencies were seen within the splenic NK cell population (3.2 ± 1.5%), consistent with previously published data 38.
At variance with a recent report 38, we found that CD127+ NK cells were still present in the lymph nodes and spleen of athymic Foxn1nu/nu mice at frequencies comparable or greater than those in euthymic controls (Fig. 4B, C). Therefore the thymus is not essential for development of CD127+ NK cells. As has been suggested for the CD56bright population of human NK cells, which are also CD127+, selective development of CD127+ NK cells may occur in other organs such as the lymph nodes 40. Alternatively, the non-uniform representation of CD127+ NK cells in different organs may be due to other mechanisms connected to homeostatic regulation by IL-7.
Therefore, CD127 expression is not sufficient to unambiguously identify a single population of NK-gated cells. Whilst early thymic progenitors can differentiate into NK cells 41, evidence for its occurrence in wild-type mice, which relies on the use of “thymic” NK markers not found in athymic mice, including CD127 38 and TCR gene rearrangements 42, must consider the similarity of NK-like γδ T cells to bona fide NK cells.
NK-like γδ T cells share multiple phenotypes and functions with NK cells
Above, we show that gating on NK cells identifies bona fide NK cells and a population of NK-like T cells that are best classified as T cells because of CD3/TCR expression and a requirement for TCR signaling during their thymic development. Their similarity to NK cells raised the question of how they compare in function and phenotype to classical populations of NK and T cells. To address this, Tcrd-H2BEGFP mice were used to facilitate the identification of NK cells, NK-like γδ T cells and γδ T cell populations expressing or lacking NK1.1.
Strikingly, a large proportion of peripheral NK-like γδ T cells (EGFPhi NK-gated cells) expressed NK-activating receptors including NKp46, NKG2D, 2B4 and CD16 (Fig. 5A), with greater than 60% of NK-like γδ T cells bearing each receptor. Few NK1.1– γδ T cells expressed these receptors, but NK1.1+ γδ T cells (with higher levels of CD3/CD5 than NK-like γδ T cells) had an intermediate phenotype. Amongst the molecules tested, NKp46 showed the greatest NK specificity 21. NKp46 expression by NK-like γδ T cells therefore highlights the extent of their NK programming, as it was present on 60% of NK-like γδ T cells (Fig. 5A) and similar frequencies of NK-like αβ T cells (as determined by NKp46 and intracellular CD3 staining of Tcrd-H2BEGFP NK-gated cells, data not shown).
Figure 5. A partial NK programming is associated with NK1.1 expression and low surface CD3 expression in γδ T cells. (A, B) Phenotypic analysis of NK and γδ T cell populations by FACS. To identify populations, whole Tcrd-H2BEGFP splenocytes were divided into NK-gated cells (NK1.1+CD3–CD5–) and CD3/CD5+ γδ T cells (CD3/CD5+, EGFPhigh). NK-gated and γδ T populations were then subdivided according to expression of EGFP and NK1.1 as shown in Fig. 3. The percentage of cells staining with antibodies against the indicated receptors and markers is shown. (A) Top graph shows NK cell markers and main activating receptors. Bottom graph shows expression of MHC class I receptor family members. (B) Other receptors and markers of activation. (C) IFN-γ response of indicated populations to the plate-bound antibodies against NK1.1, Ly49D and NKp46 or isotype controls in a 4-h stimulation assay. Experiments were performed on enriched populations of pooled splenic and lymph node NK and γδ T cells from Tcrd-H2BEGFP mice. Following 4 h of culture, cells were harvested and stained for surface markers and intracellular IFN-γ. Populations of cells were identified by surface markers and EGFP expression as in (A, B) with the exception of γδ T cells (here defined as CD3/CD5+, TCRγδ+). (D) Response of NK and γδ T cell populations to canonical NK stimuli in a 4-h assay. (Top graph) Enriched NK and γδ T cells were incubated with or without YAC-1 target cells as indicated in the presence of anti-CD107a (LAMP1) antibody. Cells were harvested after 4-h incubation and identified as in (A, B). (Bottom graph) IFN-γ production in response to 4-h incubation with IL-12 and IL-18 was detected using intracellular staining. All graphs show mean percentage ± SD for three to eight independent data points.
Download figure to PowerPoint
Receptors from the Ly49 and NKG2 families of MHC class I receptors showed a more complicated picture (Fig. 5A). Ly49G2 followed the trend of other NK receptors, with frequency of expression following the pattern NK cells>NK-like γδ T cells>NK1.1+ γδ T cells>NK1.1– γδ T cells. In contrast, Ly49C/I expression was more specific to NK cells, and expression of Ly49D was only observed on populations of bona fide NK cells. This fits with previous reports showing that activating Ly49 receptors are not expressed by T cells 43, 44. On the other hand, NKG2A/C/E was found on the vast majority of NK1.1+ γδ T cells (including NK-like γδ T cells), but not NK1.1– γδ T cells.
The cell surface expression of CD43, CD44 and CD122, markers associated with memory-phenotype T cells, also followed the trend of NK receptors, with high frequencies of high-level expression in NK cells and NK1.1+ γδ T cells, and lower frequencies on NK1.1– γδ T cells (Fig. 5B). In contrast, CD69, a marker of acute activation, was not significantly expressed by any population in Tcrd-H2BEGFP mice. CD11b and CD27, markers of maturation status on NK cells 37, 39, showed a different expression pattern on NK cells compared with γδ T cells. NK cells were mostly CD11bhigh, whilst NK-like γδ T cells and other γδ T cells were low or negative for this marker. In contrast, CD27 was expressed by most γδ T cells from each population, but high expression was absent from approximately 50% of NK cells, indicating a highly differentiated phenotype (Fig. 5B).
Together, these data suggest that NK-like γδ T cells, which express NK1.1 and have low levels of TCR, have undergone a differentiation program giving them multiple characteristics of classical NK cells. However, the MHC class I receptor repertoire remains distinct from that of NK cells, and NK-like γδ T cells remain CD11blow and CD27high, in a similar manner to immature NK populations 37, 39, suggesting that their NK programming is limited under steady-state conditions.
We therefore tested whether NK-like γδ T cells were responsive to classical T cell and NK stimuli. Consistent with classification of NK-like γδ T cells as T cells, stimulation of all populations of γδ T cells (NK1.1–, NK1.1+ and NK-like γδ T cells) with plate-bound anti-CD3 and anti-CD28 led to IFN-γ production in a 4-h assay (data not shown). Akin to NK cells, NK-like γδ T cells became IFN-γ+ in response to 4-h stimulation with mAb against NK1.1 (Fig. 5C). In contrast, antibodies recognizing Ly49D and NKp46 did not stimulate NK-like γδ T cells. The former is explained by a lack of Ly49D expression by NK-like γδ T cells, but the absence of NKp46 responsiveness was surprising given that ∼60% of these cells expressed this receptor (Fig. 5A). One possible explanation for this differential response is the lower level of NKp46 expressed by NKp46-positive NK-like γδ T cells than by NK cells (data not shown), suggesting that the level of receptor may be insufficient for activation.
Two major innate immune effector functions of NK cells are natural cytotoxicity and the rapid production of cytokines in response to inflammatory stimuli such as IL-12 and IL-18 8, 45. We therefore tested the cytolytic response against NK-sensitive YAC-1 target cells by NK-like γδ T cells and other γδ T cells using a 4-h CD107a (LAMP1) assay (Fig. 5D). In this assay, the up-regulation of CD107a on the surface of cells acts as a proxy marker for degranulation of cytolytic granules, providing a convenient means of testing specific cytotoxicity at the single-cell level 46. EGFPlo and EGFPint NK cells potently responded to YAC-1 targets, with ∼15% becoming CD107a+ during the assay. However, EGFPhi NK-like γδ T cells and NK1.1+ γδ T cells responded only very weakly and no response was detectable by NK1.1– γδ T cells (Fig. 5D).
The cytolytic activity of each population, as measured in this assay against YAC-1 targets, paralleled expression of granzyme B. Approximately 70% of EGFPlo and EGFPint NK cells expressed granzyme B at discernable levels, whereas approximately 20% of NK-like γδ T cells and 10% of NK1.1+ γδ T cells had very modest staining (data not shown). In contrast, when exposed to cytokines IL-12 and IL-18 for 4 h, NK-like γδ T cells responded in a similar manner to NK cell populations (Fig. 5D). Approximately 60% of NK-like γδ T cells (59.0 ± 10.1%) and each NK population (EGFPlo, 63.7 ± 7.5%; EGFPint, 62.0 ± 11.5%) produced intracellular IFN-γ in response to this stimulation, in comparison with only 6.1 ± 3.6% of NK1.1– and 36.1 ± 3.0% of NK1.1+ γδ T cells. Together, these data show that NK-like γδ T cells have acquired IL-12/IL-18 monokine responsiveness and the ability to rapidly produce the cytokine IFN-γ during the course of their development and maturation.