Lacrimal Gland NK Cells Are Developmentally and Functionally Similar to Conventional NK Cells

Innate lymphoid cells (ILCs) consist of diverse cell types that combat infectious microorganisms and cancer cells and help to maintain tissue homeostasis (1). The different subsets of ILCs are broadly classified as ILC1s, ILC2s, or ILC3s based on their developmental pathways and the cytokines that they produce at maturity (2). Conventional NK (cNK) cells are the prototypical ILC1s (3) that function mainly to induce apoptosis of virally infected cells and tumor cells (4). cNK cells develop from the common lymphoid progenitor in the bone marrow (5, 6) and mature before entering the circulation and traveling to lymphoid and nonlymphoid tissues (7).

In recent years, unique populations of NK cells have been identified in many different tissues. Some are cNK cells that take on altered phenotypes due to signals within the tissue environment (8), whereas others appear to be completely distinct ILC1 lineages. For instance, thymic NK cells have a unique phenotype and developmental pathway compared with cNK cells (9). Recently, populations of tissue-resident NK (trNK) cells have been identified in the liver, skin (10), kidney (11), uterus (12), and salivary gland (13, 14). These trNK cells represent distinct populations of ILC1s, which have unique phenotypes and developmental requirements compared with cNK cells. cNK cells require the transcription factor NFIL3 for development (15–17) and express Eomes. trNK cells in most tissues are Eomes⁻ and develop at least partially independently of NFIL3 (18–20). This is in contrast to evidence that NFIL3 is required for the development of all ILCs (21–24).

The populations of cNK cells, trNK cells, and other ILCs in lymphoid and mucosal tissues have been well characterized (8, 19, 25, 26). Mucosal tissues are varied in structure and function, and their exposure to the external environment results in colonization by a wide variety of commensal and pathogenic microorganisms (27). NK cells and other ILCs are important for maintaining the composition and integrity of mucosal tissues, particularly in response to microbial colonization (28). However, the presence of ILCs in exocrine glands, such as the lacrimal gland (LG), has been relatively understudied. Exocrine glands secrete factors that help to maintain the integrity of mucosal and epithelial surfaces. The LG is essential for eye health, because it is responsible for producing the mucin and aqueous layers of the tear coating. The tear coating is necessary for normal eye function and protection from pathogens, because it supplies the eye with antimicrobial enzymes and protective Igs. Excessive inflammation can damage the LG, which can result in reduced tear production and damage to the ocular surface (29, 30).

The LG is known to contain populations of T cells and B cells (31). In this article, we report that the LG also has a population of NK cells. LG NK cells express Eomes and T-bet and are mostly absent in NFIL3^−/− mice. This suggests that they develop from the cNK cell lineage. In support of this, we found that LG NK cells do not express RORγt during development, which indicates that they are not ex-ILC3s. Although we could not detect viral replication in this organ, LG NK cells mount an effector response during systemic murine CMV (MCMV) infection. However, this response is low in magnitude compared with splenic and liver NK cells. This weak response was found to be tissue specific, because LG NK cells produce similar levels of IFN-γ as splenic NK cells after acclimating to the spleen and liver following adoptive transfer into lymphopenic mice.

C57BL/6 and B6.SJL mice were purchased from Taconic Biosciences (Germantown, NY). A breeding pair of Rag2^−/−IL-2Rγ^−/− mice was purchased from Taconic Biosciences, and these mice were subsequently bred in-house. Rorc.cre and R26R-EYFP mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Rorc.cre mice were bred with R26R-EYFP mice in-house to produce RORγt fate mapping mice. NFIL3^−/− mice were a generous gift from Dr. H. J. M. Brady (15). NFIL3^+/+, NFIL3^+/−, and NFIL3^−/− mice were subsequently bred in-house. All mice were maintained in pathogen-free facilities at Brown University. Mice of both sexes were included, and no differences were observed.

Mice were infected i.p. with 5 × 10⁴ PFU MCMV (strain: RVG102), as previously described (32). In experiments with NK cell depletion, mice were initially injected i.p. with 100 μg of anti-NK1.1 (clone: PK136) 24 h prior to MCMV infection and again every 7 d until takedown. NK cell IFN-γ was measured directly ex vivo without culture following MCMV infection.

Mice were sacrificed with isoflurane, and cardiac puncture was performed prior to organ removal. Spleens were processed with a GentleMACS Dissociator (Miltenyi Biotec), filtered through nylon mesh, and layered onto a Lympholyte-M gradient (CEDARLANE Laboratories). Lymphocytes were harvested from the gradient interface and washed once in PBS supplemented with 1% FBS (1% PBS–serum). Alternatively, spleens were processed with ammonium chloride to lyse RBCs and enrich for lymphocytes. Livers were perfused with 1% PBS–serum before removal, processed in 1% PBS–serum with the GentleMACS Dissociator, and filtered through nylon mesh. Samples were washed three times with 1% PBS–serum, suspended in 40% Percoll, and layered on 70% Percoll. Lymphocytes were harvested from the gradient interface and washed once with 1% PBS–serum. Extraorbital LGs were processed in Collagenase IV or Liberase-DL (both from Sigma-Aldrich) with the GentleMACS Dissociator, incubated at 37°C for 10 min, filtered through nylon mesh, and washed once with 1% PBS–serum before being layered on a Lympholyte-M gradient. In some experiments, 6–12 LGs from three to six animals were pooled. Lymphocytes were harvested from the gradient interface and washed once in 1% PBS–serum.

Lymphocyte samples were incubated in 1% PBS–serum with the blocking mAb 2.4G2 and stained with specific mAbs for 20 min at 4°C. For intracellular cytokine staining, cells were stained with extracellular mAbs, fixed with Cytofix/Cytoperm (BD Biosciences) for 20 min, and stained with intracellular mAbs in 1× Perm/Wash (BD Biosciences) for 20 min. For intranuclear transcription factor staining, cells were stained with intracellular Abs using Foxp3 transcription factor staining reagents (BD Biosciences). Events were collected on a FACSAria III (BD Biosciences), and the data were analyzed using FlowJo software. Alexa Fluor 488–AsGM1, FITC-CD27, PE-TRAIL, PE–IFN-γ, PE–T-bet, PerCP–Cy5.5–NK1.1, PerCP–Cy5.5–TCRβ, PE–Cy5–DX5, PE–Cy7–NKp46, allophycocyanin-CD3, allophycocyanin-CD19, allophycocyanin-Ly49H, allophycocyanin-KLRG1, allophycocyanin–eFluor 780–CD45, allophycocyanin–eFluor 780–CD45.2, eFluor 450–CD3, eFluor 450–CD11b, eFluor 450–CD45.1, and eFluor 450–Eomes were purchased from eBioscience (Thermo Fisher Scientific). PE-CD49a, allophycocyanin-CD49a, BV421-CD127, BV510-TCRβ, BV570-CD45, BV605-CD3, and BV785-NK1.1 were purchased from BioLegend. FITC-DX5 was purchased from Miltenyi Biotec.

NK cells were sorted under sterile conditions from the spleens of C57BL/6 (CD45.2⁺) mice and the LGs of B6.SJL (CD45.1⁺) congenic mice. Donor NK cells were injected 1:1 into Rag2^−/−IL-2Rγ^−/− recipient mice. Recipient mice were allowed to reconstitute for 7 d before being infected i.p. with 5 × 10⁴ PFU MCMV. At 38 h postinfection, the recipient mice were sacrificed.

Standard plaque assays were carried out, as previously described (33).

Lymphocytes from naive C57BL/6 spleen, liver, and LG were incubated for 4 h in RPMI 1640 complete media, RPMI 1640 with IL-12 (10 ng/ml) and IL-18 (10 ng/ml), or RPMI 1640 with PMA (20 ng/ml) and ionomycin (1 μg/ml). GolgiStop (BD Biosciences) was added at the beginning of the incubation. Cells were washed twice with 1% PBS–serum before Ab staining.

All statistical analyses were performed with Prism Version 7.0 (GraphPad). Unpaired two-tailed Student t tests were used to compare cell populations from different mice. Paired two-tailed Student t tests were used for experiments involving adoptive transfer (****p < 0.0001, ***p = 0.0001–0.001, **p = 0.001–0.01, *p = 0.01–0.05).

The LG is an exocrine gland that is similar in structure and function to the submandibular salivary gland (SMG) (29), which contains well-characterized populations of ILC1s (13, 14, 32, 34). We isolated lymphocytes from the extraorbital LG of naive C57BL/6 mice and found a population of CD3⁻NK1.1⁺NKp46⁺ cells (Fig. 1A). The maturity of circulating NK cells is a four-stage developmental process, distinguished by expression of CD11b and CD27. NK cell maturity progresses from CD11b^lowCD27^low, to CD11b^lowCD27^high, CD11b^highCD27^high, and, finally, CD11b^highCD27^low (35). In comparison with spleen and liver NK cells, LG NK cells are relatively immature, having a very low frequency of CD11b^highCD27^low cells (Fig. 1B) and low expression of KLRG1 (Fig. 1C). KLRG1 is generally expressed on fully mature CD11b^highCD27^low NK cells (36), which further supports the classification of LG NK cells as relatively immature.

In several organs, such as the liver, skin, uterus (10, 12, 37), and kidney (11), cNK cells can be distinguished from trNK cells based on surface expression of DX5 and CD49a. cNK cells are identified as DX5⁺CD49a⁻, whereas trNK cells are DX5⁻CD49a⁺. Liver trNK cells also express high levels of TRAIL at baseline (10). We found that LG NK cells are mainly DX5⁺ but some are also CD49a⁺. However, they cannot be easily defined as DX5⁺CD49a⁻ and DX5⁻CD49a⁺ subsets, like the NK cells of the liver (Fig. 2A). Much like splenic NK cells and liver cNK cells, LG NK cells are mainly Eomes⁺T-bet⁺ (Fig. 2B) and TRAIL⁻ (Fig. 2C). Liver and kidney trNK cells were also recently shown to be largely negative for the surface receptor asialo-GM1 (AsGM1), which was once used as a marker to identify all NK cells (11). We observed that the majority of LG NK cells are AsGM1⁺ (Supplemental Fig. 1A) and CD127⁻ (Supplemental Fig. 1B). However, differential expression of Eomes, DX5, CD49a, and other surface markers is not sufficient to distinguish cNK cells from trNK cells in all organs. For instance, SMG NK cells are mostly DX5⁺CD49a⁺, as well as Eomes⁺T-bet⁺ (13, 14). The majority of SMG NK cells are cNK cells; however, there is also a trNK cell population. These populations are distinguished based on differential requirements for the transcription factor NFIL3 during development (14). cNK cells are generally dependent on NFIL3 for development, whereas liver, skin, uterus, kidney, and SMG trNK cells develop somewhat independently of this transcription factor (10–12, 20, 38). In the LG, we observed a significant decrease in NK cell frequency in NFIL3^−/− mice compared with wild-type littermate controls (Fig. 3). Together, these data support the conclusion that the vast majority of LG NK cells belong to the conventional lineage.

Recent research has shown that some NKp46⁺ ILC3s downregulate RORγt expression, increase T-bet expression, and gain the ability to produce IFN-γ, effectively taking on an ILC1 phenotype (39, 40). This phenotypic plasticity has made it difficult to classify these ex-ILC3s as either ILC1s or ILC3s (28, 41). We investigated whether any of the LG NK cells were ex-ILC3s based on past RORγt expression by generating RORγt fate mapping mice. Mice that expressed cre recombinase under the control of the Rorc gene were crossed with those carrying the ROSA26–floxstop–YFP allele. In agreement with previous findings (39), we found that the resulting F1 mice had YFP expression in cells that had expressed RORγt during development (Supplemental Fig. 1C).

Using the RORγt fate mapping mice, we found that LG NK cells, as well as spleen NK and liver cNK cells, were mainly YFP⁻ (Fig. 4). This rules out the presence of ILC3s or ex-ILC3s within the LG CD3⁻NK1.1⁺NKp46⁺ population. Interestingly, nearly 20% of liver trNK cells were positive for YFP (Fig. 4B). It is not known whether these cells are ex-ILC3s or whether some liver trNK cells express RORγt as part of an unknown developmental pathway.

NK cells are crucial for the early control of many viral infections, including CMV (42–44). The CMV family members all have strict species tropism, thus MCMV is often used as a model system to study the pathogenesis and immune response to human CMV. To investigate the effector response of LG NK cells, C57BL/6 mice were infected with MCMV, and NK cell IFN-γ production was assessed at 38 h, day 7, and day 14 postinfection. Previous studies have shown that the effector response of spleen NK cells peaks during the second day of MCMV infection (32). Spleen NK cells, liver cNK cells, and liver trNK cells produced a robust IFN-γ response at 38 h postinfection (Fig. 5). LG NK cells also produced IFN-γ at 38 h postinfection, but at a significantly lower magnitude (Fig. 5). This effector response only occurs during early MCMV infection, because LG NK cell IFN-γ production decreases by day 7 and returns to baseline by day 14 postinfection (Supplemental Fig. 1D). Interestingly, we also found that the frequency of LG NK cells increases dramatically at 38 h postinfection, before decreasing at day 7, as T cells infiltrate the organ. This is in contrast to the spleen, where the NK cell frequency decreases at 38 h postinfection (Supplemental Fig. 1E). However, the effector response of LG NK cells is less robust than that of NK cell populations in the spleen and liver.

We previously reported that SMG NK cells are hyporesponsive to MCMV infection (32). However, we also showed that this phenotype is tissue specific and reversible (14). Thus, we investigated whether the weak effector response of LG NK cells to MCMV is also tissue specific. CD3⁻NK1.1⁺NKp46⁺ lymphocytes were sorted from the spleen of C57BL/6 mice (CD45.2⁺) and the LG of B6.SJL mice (CD45.1⁺) and injected into recipient Rag2^−/−IL-2Rγ^−/− mice, which lack B cells, T cells, and ILCs. After 7 d, the recipient mice were infected with MCMV. Thirty-eight hours postinfection, IFN-γ production was assessed in NK cells recovered from the recipient spleen and liver.

Although donor spleen and LG NK cells were found in the recipient spleen and liver, neither of the donor populations traveled to the LG (Fig. 6A), which indicates that circulating NK cells are not recruited to the LG during MCMV infection. We also found that donor spleen and LG NK cells produced comparable levels of IFN-γ in the recipient spleen and liver (Fig. 6B, 6C). This result indicates that IFN-γ production by LG NK cells during systemic MCMV infection is limited by tissue-specific factors present in the LG but not the spleen or liver. This is further supported by the low level of IFN-γ produced by naive LG NK cells in vitro during IL-12 + IL-18 stimulation (Supplemental Fig. 2A). Naive LG NK cells stimulated with PMA + ionomycin produced similar levels of IFN-γ as did spleen and liver NK cells (Supplemental Fig. 2B).

To determine whether MCMV replicates within the LG, C57BL/6 mice were infected with MCMV, and standard plaque assays were performed on homogenates of the spleen, SMG, and LG at 38 h and days 7, 14, and 21 postinfection. In agreement with previous reports, MCMV was detected in the spleen at 38 h postinfection (45) (Supplemental Fig. 2C) and in the SMG starting at day 7 and continuing to day 21 (46, 47) (Supplemental Fig. 2D). However, our analysis did not reveal viral plaques in the LG at any of the time points (data not shown). Because NK cells are a crucial component of the early immune response to MCMV, we also depleted C57BL/6 mice of NK cells. This treatment resulted in higher levels of viral replication in the spleen (Supplemental Fig. 2C) and SMG (Supplemental Fig. 2D); however, we still did not detect virus in the LG (data not shown).

For decades after their discovery, cNK cells were the only known innate lymphocytes. However, within the last few years, several subsets of ILCs have been identified and characterized in lymphoid tissues, mucosal tissues, and elsewhere (8, 19, 48). The various ILC subsets are broadly classified as ILC1s, ILC2s, and ILC3s. ILC1s constitutively express the transcription factor T-bet, as well as produce type 1 cytokines, such as IFN-γ and TNF-α (49, 50). cNK cells are included in the ILC1 group, along with other subsets of helper-like ILC1s. cNK cells are cytotoxic effector cells. Helper-like ILC1s produce type 1 cytokines but are generally considered noncytotoxic. cNK cells also diverge early from helper-like ILC1s in development (51–53).

The growing diversity of ILC1s has called into question the dichotomous categorization of the various ILC1 subsets as either cytotoxic or helper-like. For instance, the trNK cells found in the liver are more closely related to other helper-like ILC1s than cNK cells. However, a recent study demonstrated that unlike mucosal helper-like ILC1s, liver trNK cells have high cytotoxic potential (38). Intraepithelial ILC1s have been identified in mucosal tissues and appear to be distinct from cNK cells, trNK cells, and other helper-like ILC1 subsets (54). These findings indicate that ILC1s cannot be simply classified as cytotoxic cNK cells and noncytotoxic/helper-like ILC1s; rather, ILC1 subsets exist along a continuum of diverse phenotypes.

In this study, we identified and characterized a population of ILC1s in the murine LG. LG NK cells appear to be relatively immature and display an unusual expression pattern of DX5 and CD49a. However, as we reported previously in the SMG (14), CD49a is not a definitive marker of trNK cells in all organs. LG NK cells primarily express T-bet and Eomes and are almost completely absent in NFIL3^−/− mice. These findings indicate that LG NK cells are not NFIL3-independent ILC1s. Fate mapping experiments also showed that LG NK cells do not express the transcription factor RORγt during development, which indicates that they do not develop from an ILC3 lineage. Based on these data, we found it prudent to identify LG NK cells as conventional-like NK cells.

cNK cells are homogenous in terms of development, but they are not phenotypically uniform at maturity. Rather, they can take on unique phenotypes after they exit the bone marrow and acclimate to different tissues. For instance, NK cells of the lung appear to be derived from the conventional lineage but are more mature than splenic NK cells (8). They also express higher levels of inhibitory receptors, lower levels of activating receptors, and lower levels of migratory and adhesion molecules than do splenic NK cells (55). Based on our findings, the LG also contains a population of NK cells. These LG NK cells appear to be conventional in development, with a unique phenotype shaped by the tissue environment.

NK cells are crucial for the early defense against viral infections, particularly herpesviruses, such as CMV. Similar to the NK cells of the murine SMG (32), LG NK cells produce a weak effector response to systemic MCMV infection. However, this response is likely mediated by inflammatory cytokines, because we could not detect MCMV in this organ. When LG NK cells were removed from their native environment and allowed to proliferate in the spleen and liver of Rag2^−/−IL-2Rγ^−/− mice, they produced a similar level of IFN-γ as donor splenic NK cells (Fig. 6). Thus, LG NK cells are fully equipped to produce a robust effector response to MCMV infection. Much like SMG NK cells (14), the effector response of LG NK cells is suppressed by factors in their native environment. It is also possible that the weak effector response of LG NK cells in situ is due to the lack of viral replication in the LG during infection. However, we consider this unlikely, because Ly49H⁺ and Ly49H⁻ NK cells of the spleen and LG produce similar levels of IFN-γ during early MCMV infection (56) (Supplemental Fig. 2E). This indicates that direct contact with m157-expressing cells is not necessary for an NK cell IFN-γ response at this early time point in either organ.

In addition to killing virally infected cells and producing proinflammatory cytokines, NK cells can help to limit inflammation (57). This is especially important in secretory tissues, such as the SMG and LG, where extensive inflammation results in a phenotype resembling human Sjögren’s syndrome, an autoimmune disease characterized by a lack of saliva and tear production (58). The LG is a crucial exocrine gland that can be easily damaged by inflammation, and LG NK cells are capable of producing a potent proinflammatory immune response. It is possible that the effector response of LG NK cells is self-modulated or suppressed by other factors within the tissue to prevent inflammatory damage and the resulting lack of tear production. Further work will be necessary to determine the mechanisms behind this.

We thank Kevin Carlson for cell sorting, Céline Fugère for tail vein injections, Courtney K. Anderson for critical reading of the manuscript, and Dr. Hugh J. M. Brady for providing NFIL3^−/− mice.

The authors have no financial conflicts of interest.