NFIL3 expression distinguishes trNK and cNK-like cells in the mouse submandibular glands

Abstract

The submandibular salivary gland (SMG), a major site of persistent infection for many viruses, contains a large NK cell population. Using NFIL3-deficient mice, PLZF reporter/fate mapping mice, and mixed bone marrow chimeras, we identified two distinct populations of NK cells in the SMG. Although phenotypically unique, the main population relies on NFIL3 but not PLZF for development, and therefore is developmentally similar to the conventional (c)NK cell subset. In contrast, we found that approximately one quarter of the SMG NK cells develop independently of NFIL3. Interestingly, NFIL3-independent SMG tissue-resident (tr)NK cells are developmentally distinct from liver trNK/ILC1 cells. We also demonstrated that the SMG NK cell hyporesponsive phenotype during MCMV infection is tissue specific, and not cell-intrinsic. In contrast, NFIL3-independent SMG trNK cells are intrinsically hyporesponsive. Altogether, our data show that the SMG tissue environment shapes a unique repertoire of NK-like cells with distinct phenotypes.

Introduction

Conventional (c)NK cells are derived from the common lymphoid progenitor (CLP) in the bone marrow (1). From there, they develop into committed NK cell precursors (NKPs), which further develop into immature (i)NK cells upon acquisition of NK1.1 expression. iNK cells progress into mature (m)NK cells with a CD122⁺NK1.1⁺NKp46⁺DX5⁺ phenotype. In addition to cNK cells, several distinct populations of tissue-resident (tr)NK cells have been identified, with unique developmental pathways and phenotypic attributes (2–7). The liver contains a population of cNK cells, as well as a population of trNK cells (phenotypically similar to ILC1) that maintain a CD49a⁺DX5⁻TRAIL⁺ phenotype and develop from a liver-specific precursor pool (3, 8). The skin also harbors a trNK/ILC1 subset, and there is evidence to indicate that skin and liver trNK cells arise from the same developmental lineage (3). Uterine (u)NK cells are another unique population with a distinct phenotype from both cNK and liver/skin trNK cells. uNK cells do not produce an effector or cytotoxic response during encounters with the invading trophoblast cells of the placenta, despite possessing the full complement of activating receptors and cytotoxic machinery (9–11). Thymic NK cells represent another population that develops from unique precursors, and are Ly49^lowCD11b^low and CD127⁺CD69^high, in contrast to cNK cells (4, 5). A unique population of trNK cells has also recently been discovered in the kidneys (6). The current understanding is that cNK cells, together with liver and skin trNK (ILC1), uterine NK cells, thymic NK cells, and kidney trNK cells, account for multiple distinct NK cell lineages (3, 7).

NFIL3 (also called E4BP4) is a basic leucine-zipper transcription factor that has been linked to a number of immune processes, and is crucial for the early development of cNK cells (12–14). The different trNK cell subsets, on the other hand, have unique developmental requirements. While NFIL3 deficiency results in ablation of cNK cells in the periphery, its activity is mostly dispensable for the development of trNK cells in the liver (15), uterus, and skin (3), despite contrasting evidence that NFIL3 is necessary for the development of all ILC lineages (16–19). T-bet and Eomes are also necessary, albeit at different levels, for the development of mature cNK cells (20), and there is evidence that these two transcription factors are regulated by NFIL3 (14). However, liver and skin trNK cells develop independently of Eomes, and uterine trNK cells do not require T-bet (3, 21).

The requirements of these transcription factors for the development of NK cells in the SMG have not been clearly defined. Here, using NFIL3-deficient mice, PLZF reporter/fate mapping mice, and mixed bone marrow chimeras, we show that the murine SMG contains two distinct populations of NK cells: a main cNK-like cell subset that relies on NFIL3 for development, and a smaller trNK cell subset that is NFIL3-independent. Our findings also demonstrate that SMG trNK cells represent another distinct ILC lineage with a unique developmental pathway. Importantly, using the MCMV model of infection, we also show that the hyporesponsive phenotype of NFIL3-dependent SMG NK cells is due to tissue environmental factors, while NFIL3-independent SMG NK cells are intrinsically poor effector cells.

Materials and Methods

Mice

C57BL/6, B6.SJL, AhR^−/−, and Rag2^−/−IL-2Rγ^−/− mice were purchased from Taconic Biosciences (Germantown, NY). T-bet^−/−, R26R-EFYP, and PLZF^GFPcre mice were purchased from The Jackson Laboratory (Bar Harbor, ME). R26R-EFYP mice were bred with PLZF^GFPcre mice to produce PLZF^GFPcre+/−ROSA26-floxstop-YFP mice. NFIL3^−/− mice were a generous gift from Dr. Hugh JM Brady (13), and were bred in-house. All mice were maintained in pathogen free facilities at Brown University. Both sexes were included and no differences were observed.

Infection and treatment protocols

MCMV infections were carried out as previously described (2).

Isolation of murine lymphocytes

Mice were sacrificed with isoflurane, and cardiac puncture was performed prior to organ removal. Spleens were processed on the spleen01.01 program on a GentleMACS dissociator (Miltenyi Biotec), filtered through nylon mesh, and layered on Lympholyte-M (Cedarlane Laboratories Ltd., Canada). Lymphocytes were harvested from the gradient interface, and washed once in PBS supplemented with 1% FBS (1% PBS-serum). Livers were perfused with 1% PBS-serum before removal, processed in 1% PBS-serum on the E.01 program on the GentleMACS, and filtered through nylon mesh. The samples were washed 3 times with 1% PBS-serum, resuspended in 40% Percoll and layered on 70% Percoll. Lymphocytes were harvested from the gradient interface and washed once with 1% PBS-serum. SMGs were processed manually to remove lymph nodes, then processed in Collagenase IV (Sigma-Aldrich) on the heart01.01 program on the GentleMACS, incubated at room temperature or 37°C for 10 minutes, filtered through nylon mesh, and washed once with 1% PBS-serum before being layered on a Lympholyte-M gradient. Lymphocytes were harvested from the gradient interface and washed once in 1% PBS-serum. We report that Ly49 markers and TRAIL are sensitive to Collagenase IV, leading to false negatives in some studies. SMGs can be processed without Collagenase in order to ascertain expression of these markers, but the number of lymphocytes recovered is very low. To circumvent this issue, we screened a variety of enzymes and identified Liberase-DL (Sigma-Aldrich), which does not affect these markers. Whenever the expression of these markers was assessed, Collagenase IV was replaced with Liberase-DL.

Flow cytometric analysis, antibodies, and reagents

Lymphocyte samples were incubated in 1% PBS-serum with the blocking monoclonal antibody (mAb) 2.4G2 and stained with specific mAbs for 20 minutes at 4°C. For intracellular cytokine staining, cells were first stained with extracellular mAbs, then fixed with Cytofix/Cytoperm (BD Bioscience) for 20 minutes, and then stained with intracellular mAbs in 1X PermWash (BD Biosciences) for 20 minutes. For intranuclear transcription factor staining, cells were stained with intracellular antibodies using the FoxP3 transcription factor staining reagents (BD Bioscience). Events were collected on a FACSAria (BD), and the data were analyzed using FlowJo (Tree Star Inc.). FITC-DX5, PE-Ly49H, PE-IFN-γ, PE-E4BP4, PE-TCRβ, PE-CD27, PE-NK1.1, PECy5-DX5, PECy7-NKp46, PECy7-T-bet, PECy7-Sca-1, PerCPCy5.5-CD127, PerCPCy5.5-NK1.1, PerCPeFluor710-NKG2A/C/E, APC-CD3, APC-Ly49H, APC-IFN-γ, APC-TNF-α, APC-TRAIL, APC-CD45.1, APC-eFluor780-CD45.2, APC-eFluor780-CD117, eF450-CD11b, eFlour450-IFN-γ, eFlour450-CD3, eFlour450-Eomes, were purchased from eBioscience (San Diego, CA). PE-CD49a, APC-CD49a, Pacific Blue-Lineage, BV421-CD127, BV605-CD3, BV605-NK1.1, BV785-CD3, and BV785-NK1.1 were purchased from Biolegend (San Diego, CA). FITC-Ly49C/I was purchased from BD Pharmingen. FITC-DX5, PE-NK1.1, anti-CD5 magnetic beads, and anti-CD19 magnetic beads were purchased from Miltenyi Biotec. To detect E4BP4 by intracellular staining, activation of NK cells is required.

Generation of PLZF^GFPcre+/−ROSA26-floxstop-YFP bone marrow chimeras

B6.SJL recipient mice (CD45.1⁺) were lethally irradiated with 1050 rad and placed on antibiotic treatment for two weeks. One day post-irradiation, donor bone marrow cells were harvested under sterile conditions from PLZF^GFPcre+/−ROSA26-floxstop-YFP mice (CD45.2⁺/CD45.1⁺), pooled, and stained for Lineage, Sca-1, and cKit. YFP⁻Lin⁻Sca-1⁺cKit⁺ (LSK) cells were sorted and intravenously injected into recipients at approximately 10,000 cells/mouse. The recipients were allowed to reconstitute for at least 8 weeks.

Generation of mixed bone marrow chimeras

Recipient mice were lethally irradiated with 1050 rad and placed on antibiotic treatment for two weeks. One day post-irradiation, recipient mice were injected with a 1:1 mixture of either sorted LSK cells or DX5 and CD5-depleted cells from B6.SJL and NFIL3^−/− bone marrow. The recipients were allowed to reconstitute for 8 weeks.

Adoptive transfer of NK cells

NK cells were sorted under sterile conditions from the spleen of B6 (CD45.2⁺) mice, the SMG of B6.SJL (CD45.1⁺) congenic mice, or the SMG of NFIL3^−/− (CD45.2⁺) mice. Donor NK cells were injected into recipient Rag2^−/−IL-2Rγ^−/− mice. Recipient mice were allowed to reconstitute for 7 days before being infected intraperitoneally with 5×10⁴ pfu MCMV. 38 hours post-infection, recipient mice were sacrificed for experiments.

Statistical Analysis

All statistical analyses were performed with Prism Version 7.0 (GraphPad Software). Unpaired two-tailed Student’s t-tests were used to compare cell populations from different mice. Paired two-tailed Student’s t-tests were used for experiments involving adoptive transfer or chimeric mice. ****p < 0.0001, ***p = 0.0001–0.001, **p = 0.001–0.01, *p = 0.01–0.05.

Results

NFIL3 deficiency significantly reduces the frequency and number of SMG NK cells

While cNK cells depend on NFIL3 for development, trNK cells in the liver, skin, kidneys, and uterus develop mostly independently of NFIL3 (3, 10). It was also recently reported that salivary gland NK cells develop entirely independently of NFIL3 activity (22, 23). In contrast to these findings, we found a significant reduction in the frequency (Fig. 1A and B) and number (Fig. 1C) of SMG NK cells in NFIL3^−/− mice. Intracellular staining with an anti-NFIL3 antibody also shows that a large subset of SMG NK cells express NFIL3 (Fig. 2A). Salivary gland structural development continues for several weeks after birth and is completed when the mice are 10 to 12 weeks old (24, 25). One possible explanation for the discrepancy with previous findings is that NFIL3-dependent NK cells initially seed the SMG and are later replenished by NFIL3-independent trNK cells. However, we found that NFIL3^−/− mice have a significant decrease in SMG NK cell frequency compared to NFIL3^+/+ mice with similar phenotype, regardless of their age (Fig. 2B & 2C).

Figure 1. SMG CD3⁻NK1.1⁺ cells are significantly reduced in NFIL3^−/− mice.

(A) Representative staining of spleen, liver, and SMG NK cells in NFIL3^+/+ and NFIL3^−/− mice. (B) Frequency of SMG NK cells in NFIL3^+/+ (n=23), NFIL3^+/− (n=23) and NFIL3^−/− (n=22) mice. (C) Absolute SMG NK cell number in NFIL3^+/+ (n=14), NFIL3^+/− (n=14), and NFIL3^−/− (n=13) mice. Black circles/bars represent NFIL3^+/+ mice; grey circles/bars represent NFIL3^+/− mice; white circles/bars represent NFIL3^−/− mice. Data are pooled from nine experiments and depicted as mean ± SEM. ****p < 0.0001, ***p = 0.0001–0.001, **p = 0.001–0.01.

Figure 2. SMG NK cells express E4BP4 protein and their number is reduced in aged mice.

(A) Representative intracellular E4BP4 staining in the NK cells of the spleen, liver, and SMG of NFIL3^+/+ and NFIL3^−/− mice. To detect E4BP4 by intracellular staining, activation of NK cells is required and mice had been infected with MCMV for 38 hours. Data are representative of two experiments. (B) Frequency of NK cells in the SMG of NFIL3^+/+ mice and NFIL3^−/− mice at different ages. Data are pooled from two experiments and are depicted as mean ± SEM. ****p < 0.0001, **p = 0.001–0.01. (C) Representative staining of DX5 and CD49a expression on SMG NK cells from NFIL3^+/+ mice and NFIL3^−/− mice at different ages.

To unequivocally determine the origin of these two subsets, we generated PLZF reporter/fate mapping mice by crossing PLZF^GFPcre+/− reporter mice with mice carrying the ROSA26-floxstop-YFP fate mapping allele, as described recently (26). Bendelac and colleagues have reported that most ILCs, including ILC1 but not cNK cells, are YFP positive in these mice (26). It should be noted that in the resulting mice (PLZF^GFPcre+/−ROSA26-floxstop-YFP), ~30% of the cells become YFP positive before hematopoiesis ((26) and data not shown). To circumvent this issue, we generated chimeric mice reconstituted with sorted YFP⁻ LSK bone marrow precursors (26) (See diagram, Supplemental Fig. 1A). In agreement with previous findings (26), iNKT cells express YFP in the chimeric mice (Supplemental Fig. 1B), while B cells and conventional T cells are mostly unlabeled (Fig. 3A). In the liver, approximately 60% of trNK cells were YFP⁺, while only ~ 20% of the cNK cells were labeled (Fig. 3B), which is in agreement with previous studies (26). Importantly, in the SMG of these mice, approximately 90% of the NK cells were YFP negative, indicating that they originated from the cNK lineage (Fig. 3C). Interestingly, the remaining YFP positive NK cells had lower DX5 expression than the YFP negative SMG NK cells, consistent with a possible NFIL3 independency (Fig. 3D). To exclude potential extrinsic factors, which could explain the difference with the study from Cortez and colleagues, we also generated B6.SJL/NFIL3^−/− mixed bone marrow chimeras (See diagram, Supplemental Fig. 1C). We found that approximately 90% of SMG NK cells were derived from B6.SJL donor bone marrow in all organs tested, including the SMG (Fig. 4A). In contrast, non-NK lymphocytes were derived evenly from B6.SJL and NFIL3^−/− donor bone marrow (Fig. 4B). Altogether, these data demonstrate that the SMG harbors at least two populations of NK cells: a preponderant NFIL3-dependent population developmentally similar to the cNK lineage, and an NFIL3-independent tissue-resident population.

Figure 3. PLZF reporter/fate mapping mice demonstrate that SMG CD3⁻NK1.1⁺ cells are mostly conventional NK cells.

(A) YFP expression by indicated cell types in the spleen of PLZF^GFPcre+/−ROSA26-floxstop-YFP bone marrow chimeras. (B) YFP expression by cNK and trNK cells from PLZF^GFPcre+/−ROSA26-floxstop-YFP chimeras. (C) YFP expression by CD3⁻NK1.1⁺ cells in the spleen, liver, and SMG of PLZF^GFPcre+/−ROSA26-floxstop-YFP chimeras. (D) Representative staining of DX5 and CD49a expression on YFP⁺ and YFP⁻ SMG NK cells from PLZF^GFPcre+/−ROSA26-floxstop-YFP chimeras. Data are representative of three experiments. Four chimeric mice were pooled per experiment.

Figure 4. SMG NK cells originate preponderantly from NFIL3⁺ bone marrow.

(A) Percent of spleen, liver, and SMG NK cells derived from donor B6.SJL and NFIL3^−/− bone marrow. (B) Percent of spleen, liver, and SMG non-NK cells derived from donor B6.SJL and NFIL3^−/− bone marrow. Data are pooled from two experiments. Four chimeric mice were analyzed individually per experiment. Data are represented as mean ± SEM. ***p = 0.0001–0.001.

SMG trNK cells are phenotypically and developmentally different from liver trNK cells

We next examined whether SMG trNK cells were developmentally related to liver trNK cells (3). In the liver, cNK cells develop under the coordinated influence of the transcription factors NFIL3, T-bet, and Eomes, while the trNK population develops independently of NFIL3 and Eomes, but still requires T-bet (3). Moreover, there is a clear distinction between Eomes⁺ cNK and Eomes⁻ trNK cells, with NFIL3 deficiency causing a bias toward liver Eomes⁻ trNK cells ((15), Fig. 5A). In contrast, in the SMG of wild-type mice, a large proportion of the NK cells are Eomes⁺T-bet⁺, and their relative frequency in NFIL3^−/− animals remains mostly unchanged (Fig. 5A). Liver NK cells are also clearly divided between DX5⁺CD49a⁻ cNK and DX5⁻CD49a⁺ trNK, while the majority of SMG NK cells are CD49a⁺, regardless of DX5 expression (Fig. 5B). In addition, NFIL3-independent Eomes⁻ liver trNK cells are mostly DX5⁻, whereas >50% of the NK cells in the SMGs of wild-type and NFIL3^−/− mice are DX5⁺ (Figure 5C).

Figure 5. SMG trNK cells have a unique phenotype compared to liver trNK cells.

(A) Representative staining of T-bet and Eomes expression on spleen, liver, and SMG NK cells from wild-type control (C57BL/6 or NFIL3^+/+ littermate controls) and NFIL3^−/− mice. Data are representative of four experiments. 2–3 mice were pooled in each experiment. (B) Representative staining of DX5 and CD49a expression on spleen, liver, and SMG NK cells from wild-type control (C57BL/6 or NFIL3^+/+ littermate controls) and NFIL3^−/− mice. Data are representative of four experiments. 2–3 mice were pooled in each experiment. (C) Representative staining of DX5 and Eomes expression in spleen, liver, and SMG NK cells from wild-type control (C57BL/6 or NFIL3^+/+ littermate controls) and NFIL3^−/− mice. Data are representative of four experiments. 2–3 mice were pooled in each experiment. Spleens from NFIL3^−/− mice were enriched for NK cells using anti-CD5 and anti-CD19 magnetic beads.

Regarding their effector functions, it has been shown that liver Eomes⁻DX5⁻ NFIL3-independent trNK cells express TRAIL (3, 20, 21, 27). Similarly, we found that approximately 30% of SMG NK cells constitutively express TRAIL, independently of NFIL3 expression (Fig. 6A). However in contrast to the liver, TRAIL expression does not mark a specific subset of cells (i.e DX5⁻Eomes⁻), as the SMG TRAIL⁺ NK cells are mostly DX5⁺Eomes⁺. In NFIL3^−/− animals, the DX5⁺Eomes⁺TRAIL⁺ SMG NK cell population is retained (data not shown). Altogether, these data demonstrate that the NFIL3-independent population of NK cells in the SMG is distinct from the liver trNK cells, both in cell surface phenotype and developmentally.

Figure 6. SMG trNK cells are less mature compared to cNK-like cells.

(A) Representative staining of TRAIL expression on NK cells from the spleen, liver, and SMG of wild-type control (C57BL/6 or NFIL3^+/+ littermate controls) and NFIL3^−/− mice. Spleens from NFIL3^−/− mice were enriched for NK cells using anti-CD5 and anti-CD19 magnetic beads. Data are representative of three experiments. 2–3 mice were pooled in each experiment. (B) Representative staining of CD127 and NK1.1 on NKp46⁺ lymphocytes in the SMG and lamina propria of C57BL/6 and AhR^−/− mice. Data are representative of three experiments. Three mice were pooled in each experiment.

The SMG does not contain Group 3 ILCs

Having identified a novel population of NFIL3-independent NK cells in the SMG, we sought to determine if Group 3 ILCs were also present in this organ. Group 3 ILCs develop under the influence of both RORγt and aryl hydrocarbon receptor (AhR) (28). Therefore, we examined whether the absence of AhR affects the development of SMG lymphocytes. We did not find NKp46⁺CD127⁺NK1.1⁻ lymphocytes in the SMG from either B6 or AhR^−/− mice (Fig. 6B). However, we found as previously reported (28) that these cells are reduced in the LPL of AhR^−/− mice (Fig. 6B). These results are consistent with our previous findings showing that no significant number of RORγt⁺ cells were found in this organ using RORγt reporter mice (2).

SMG cNK-like cell hyporesponsiveness is dependent on the tissue environment, while SMG trNK cells are intrinsically hyporesponsive

We and others have previously shown that SMG NK cells are hyporesponsive during MCMV infection (2, 22). Having identified two subsets of NK cells in this organ, we sought to revisit these findings and examine if this phenotype was reversible. To determine whether SMG NK cells were capable of producing an effector response in new tissue environments, NK cells were sorted from the SMGs of B6.SJL mice (CD45.1⁺) and the spleens of C57BL/6 mice (CD45.2⁺). The sorted NK cells were mixed in a 1:1 ratio and adoptively transferred into Rag2^−/−IL-2Rγ^−/− mice. The adoptively transferred NK cells were allowed to reconstitute for 7 days before the mice were infected intraperitoneally with MCMV. We first found that the frequency of the two donor populations is unchanged and roughly at a 1:1 ratio (Fig. 7A). Importantly, at 38 hours post-infection, the magnitude of the IFN-γ response from the spleen and SMG-derived NK cells was comparable (Fig. 7B). This result indicates that the hyporesponsive phenotype seen in SMG NK cells in situ is not cell-intrinsic, but is caused by properties of the SMG microenvironment. To assess the effector capacity of SMG NFIL3-independent trNK cells, SMG trNK cells were sorted from NFIL3^−/− mice and adoptively transferred into Rag2^−/−IL-2Rγ^−/− mice. We found that NFIL3^−/− SMG NK cells produced significantly less IFN-γ than B6.SJL SMG NK cells under the same conditions (Fig. 7C). Therefore, in contrast to NFIL3-dependent SMG NK cells, the NFIL3-independent trNK cells appear not to respond optimally to MCMV infection, regardless of the tissue environment.

Figure 7. SMG cNK-like cell hyporesponsiveness but not the SMG trNK cell hyporesponsiveness can be reversed by tissue environment.

(A) NK cell frequency from the two different donors. (B) Frequency of donor IFN-γ⁺ B6 splenic derived NK cells and IFN-γ⁺ B6.SJL SMG derived NK cells in the spleen and liver of recipient Rag2^−/−IL-2Rγ^−/− mice, 38 hours after MCMV infection. (C) Frequency of donor IFN-γ⁺ SMG NK cells from B6.SJL and NFIL3^−/− mice in the spleen and liver of recipient Rag2^−/−IL-2Rγ^−/− mice, 38 hours after MCMV infection. C57BL/6 + B6.SJL data are pooled from seven experiments. NFIL3^−/− data are pooled from five experiments. Two recipient Rag2^−/−IL-2Rγ^−/− mice were pooled in each experiment. Data are represented as mean ± SEM. *p = 0.01–0.05.

Discussion

Although NK cells have been studied and characterized for decades, our understanding of their developmental pathways is still incomplete. cNK cells were discovered decades ago, but the last ten years have seen the emergence of several new classes of trNK cells, each with unique properties and developmental pathways. Moreover, NK cells as a whole comprise one subset of ILCs, a diverse group of immune lymphocytes that may represent the innate analog of T cells (29). Here we show that the SMG in naïve C57BL/6 mice contains at least two distinct populations of NK cells. Although phenotypically unique (most likely due to their tissue location), the main population of SMG NK cells expresses Eomes, T-bet, and requires NFIL3 for development, indicating they are developmentally similar to cNK cells. The remaining NK cells in this organ develop independently of NFIL3, and therefore can be classified as trNK cells, yet they are phenotypically different from the recently described resident NK cells in other organs. Although our results are in agreement with recent reports from Colonna’s group regarding the phenotype and the functions of the SMG NK cells, they differ on the NFIL3 dependence (22, 23). The differences between these two studies can potentially be explained by variations in extrinsic parameters such as microbiota or housing dependent inflammation. In fact, the identification of an MCMV-driven population of peripheral NK cells in NFIL3^−/− mice supports this possibility (30). However, we found only residual seeding of the NFIL3^−/− derived NK cells in the salivary glands in mixed bone marrow chimeras (Fig. 4), ruling out host derived extrinsic factors independent of microbiota. In addition, we detect NFIL3 protein in SMG NK cells (Fig. 2A) from wild-type animals, and approximately 90% of CD3⁻NK1.1⁺NKP46⁺ cells never express PLZF during their development (Fig. 3C). Altogether, these data advocate for the presence of cNK-like cells in this organ. In support of this conclusion, a recent report showed that the transcription factor Runx3 similarly affects splenic cNK and SMG NK cells (31).

Although NFIL3-independent SMG trNK cells share similarity with other trNK cells, they have unique phenotypic and effector characteristics. In contrast to the liver and skin trNK cells, but similarly to uterine trNK cells, SMG trNK cells from NFIL3-deficient animals express both DX5 and Eomes. Several transcription factors are known to play critical roles during ILC development. A committed α4β7⁺PLZF⁺ precursor to all helper-like ILCs (excluding cNK and LTi) was identified by Bendelac and colleagues (26, 32), while the Diefenbach group identified the common helper-like innate lymphoid precursor (CHILP) as α4β7⁺ID2^high (26, 32). The development of this common ILC precursor is dependent on NFIL3, which directly regulates Id2 to promote the development of the CHILP from the CLP (19). Under this paradigm, the NFIL3-independent NK cells of the SMG would represent yet another ILC subset, independent from helper-like ILC1s. Beside cNK-like cells and NFIL3-independent trNK cells, other small NK1.1⁺ subsets are found in the SMG. This includes a subset of NK cells strictly dependent on T-bet and similar to liver trNK cells (Supplemental Fig. 1D, see gate DX5⁻CD49a⁺). In addition, a subset of T cells not detected in B6 mice can be observed in the SMG of NFIL3-deficient animals (Supplemental Fig. 1E). Although the characterization of these CD3⁺NK1.1⁺NKP46⁺ T cells is beyond the scope of this manuscript, our preliminary data indicate that these T cells are not semi-invariant iNKT cells (data not shown).

We have also previously shown that SMG NK cells are hyporesponsive to MCMV infection, both in situ and during in vitro cytokine stimulation assays (2). However, we show here that when wild-type SMG NK cells are isolated from their native environment and allowed to reconstitute in peripheral tissues, they regain the ability to produce an effector response to MCMV. cNK cell effector plasticity has been reported in other contexts (33, 34). The ability of the SMG NK cells (Fig. 7) and splenic NK cells (33, 34) to regain effector functions reinforces our finding that the majority of the NK cells (~75%, Fig. 1) in the B6 SMG are developmentally and functionally similar to cNK cells. These findings also indicate that environmental factors in the SMG influence NK cell effector potential. A role for TGF-β, which is 100 times more abundant in the SMG than in the spleen, has recently been demonstrated in the salivary glands (23). The phenotypic change induced by TGF-β appears to be reversible. Indeed, addition of TGF-β to splenic NK cells induces them to differentiate into tissue resident-like NK cells, while blocking TGF-β signaling (23), or SMG NK cell relocation into low TGF-β environments (Figure 7) reinstates their effector functions. Cortez and colleagues propose that TGF-β drives the progressive differentiation of CD49a⁻ NFIL3-dependent SMG ILCs into CD49a⁺ NFIL3-independent mature SMG ILCs. Although our data and their data support a linear differentiation model for SMG NK cells, it is unclear how NFIL3-dependent SMG NK cells become independent in this model. Instead, we propose this model is better explained by the existence of two distinct populations in this organ.

NK cells have also been shown to limit salivary gland inflammation and tissue damage during MCMV infection (35), indicating that they play an immunoregulatory role. Recent studies have begun to address whether NK cell regulation occurs via a viral load decrease, or if it is mediated by a more complex mechanism. In support of the second possibility, a recent report demonstrated that TRAIL⁺ NK cells in the SMG specifically eliminate CD4⁺ T cells, which are critical for the clearance of active MCMV from the salivary glands (27). The authors argue that NK cell-mediated T cell killing would prolong MCMV infection, but also reduce inflammatory damage to the delicate SMG tissues, allowing the virus to be cleared slowly without causing irreversible damage to the host. Our data add to these findings and show that NK cells are rendered hyporesponsive by the salivary gland environment, presumably benefiting the host.