Cytomegalovirus Evades TRAIL-Mediated Innate Lymphoid Cell 1 Defenses

Cytomegalovirus (a betaherpesvirus) is a master at manipulating immune responses to promote its lifelong persistence, a result of millions of years of coevolution with its host. Using a one-of-a-kind MCMV mutant unable to restrict expression of the TNF-related apoptosis-inducing ligand death receptors (TRAIL-DR), we show that TRAIL-DR signaling significantly restricts both early and persistent viral replication. Our results also reveal that these defenses are employed by TRAIL-expressing innate lymphoid type I cells (ILC1) but not conventional NK cells. Overall, our results are significant because they show the key importance of viral counterstrategies specifically neutralizing TRAIL effector functions mediated by a specific, tissue-resident subset of group I ILCs.

ABSTRACT

Cytomegalovirus (CMV) establishes a lifelong infection facilitated, in part, by circumventing immune defenses mediated by tumor necrosis factor (TNF)-family cytokines. An example of this is the mouse CMV (MCMV) m166 protein, which restricts expression of the TNF-related apoptosis-inducing ligand (TRAIL) death receptors, promoting early-phase replication. We show here that replication of an MCMV mutant lacking m166 is also severely attenuated during viral persistence in the salivary glands (SG). Depleting group I innate lymphoid cells (ILCs) or infecting Trail^−/− mice completely restored persistent replication of this mutant. Group I ILCs are comprised of two subsets, conventional natural killer cells (cNK) and tissue-resident cells often referred to as innate lymphoid type I cells (ILC1). Using recently identified phenotypic markers to discriminate between these two cell types, their relative expression of TRAIL and gamma interferon (IFN-γ) was assessed during both early and persistent infection. ILC1 were found to be the major TRAIL expressers during both of these infection phases, with cNK expressing very little, indicating that it is ILC1 that curtail replication via TRAIL in the absence of m166-imposed countermeasures. Notably, despite high TRAIL expression by SG-resident ILC1, IFN-γ production by both ILC1 and cNK was minimal at this site of viral persistence. Together these results highlight TRAIL as a key ILC1-utilized effector molecule that can operate in defense against persistent infection at times when other innate control mechanisms may be muted and highlight the importance for the evolution of virus-employed countermeasures.

IMPORTANCE Cytomegalovirus (a betaherpesvirus) is a master at manipulating immune responses to promote its lifelong persistence, a result of millions of years of coevolution with its host. Using a one-of-a-kind MCMV mutant unable to restrict expression of the TNF-related apoptosis-inducing ligand death receptors (TRAIL-DR), we show that TRAIL-DR signaling significantly restricts both early and persistent viral replication. Our results also reveal that these defenses are employed by TRAIL-expressing innate lymphoid type I cells (ILC1) but not conventional NK cells. Overall, our results are significant because they show the key importance of viral counterstrategies specifically neutralizing TRAIL effector functions mediated by a specific, tissue-resident subset of group I ILCs.

INTRODUCTION

Cytomegalovirus (CMV; a betaherpesvirus) infects the majority of the world’s population, with the rate of infection varying from ∼40 to 90% and with infection being based on age, race, geography, and socioeconomic conditions (1). Primary infection is almost always asymptomatic if the subject is healthy but can be life threatening if immunity is compromised or immature (e.g., if the infected individual has had a transplant or congenital infections), and vaccine development is a top priority (2, 3). Like all the Herpesviridae, CMV escapes complete immune clearance, leading to a lifelong persistence/latency. Distinct from its viral cousins, however, CMV establishes a sleepless latency (4) in multiple tissues and cell types, commensurate with significant levels of viral gene expression (5). Consequently, even in healthy individuals, CMV has an enormous impact on shaping the immune system over time (6, 7).

CMV dedicates >50% of its ∼230-kb genome to subverting immune defenses (8), facilitating latency and contributing to its broad impact on immunity. Notably, much of this genomic currency is targeted at the tumor necrosis factor (TNF) superfamily of receptors and ligands (9). The TNF-related apoptosis-inducing ligand (TRAIL) signaling system has been extensively studied for its ability to selectively kill transformed cells, and multiple clinical trials have evaluated the anticancer efficacy of TRAIL death receptor (TRAIL-DR) agonists (10, 11). However, TRAIL is also emerging as an important regulator of antiviral defenses (12). Supporting this, we have reported previously that the UL141 and m166 proteins in human CMV (HCMV) and mouse CMV (MCMV) restrict cell surface expression of the TRAIL-DRs (13, 14). Abolishing m166 expression by inserting two stop codons within the m166 genomic open reading frame (MCMV-m166^stop) revealed that this viral immunomodulatory protein promotes acute infection in multiple organs, most dramatically, in the liver, where replication of this mutant was undetectable (14). Depletion with anti-asialo-GM1 or infection of Trail-dr^−/− mice completely restored the acute replication of MCMV-m166^stop, indicating that m166 restriction of TRAIL-DR expression is critical to block the innate control of early infection by group I innate lymphoid cells (ILCs) (14).

Recently, work by several groups has revealed that group I ILCs are composed not only of conventional natural killer cells (cNK) but also a subset of tissue-resident cells often referred to as innate lymphoid type I cells (ILC1) (15, 16). Both ILC1 and cNK express the transcription factor T-bet, display the NK1.1 and NKp46 cell surface receptors, and can produce gamma interferon (IFN-γ) when activated by various stimuli (15, 16). Consequently, prior to the very recent identification of unique markers segregating ILC1 and cNK, studies of group I ILC antiviral defenses have not assessed the relative importance of these two cellular subsets. This is especially relevant in the case of TRAIL-mediated immunity, as ILC1 have been shown to highly express this TNF-family cytokine under certain conditions (16–19). However, despite several groups showing that TRAIL can be produced by ILC1, virtually nothing is known regarding its importance for their role as immune effectors.

Here we have examined TRAIL expression in total group I ILCs during both the acute and persistent phases of MCMV infection, seeking to determine the relative importance of ILC1 and cNK in TRAIL-dependent innate defense. Our data show that in multiple tissues, ILC1 are by far the major cell type to express TRAIL, while cNK produce almost none. In the absence of m166 blockade, these TRAIL-positive (TRAIL⁺) ILC1 can potently restrict both early and persistent MCMV replication levels. Together, these results highlight ILC1 as key regulators of TRAIL-mediated antiviral immunity.

RESULTS

MCMV m166 inhibits TRAIL-mediated ILC1 defenses during early infection in the liver.

Our past work using a viral mutant lacking expression of the m166 protein (m166^stop) showed that early MCMV replication is severely compromised, as m166 blocks cell surface expression of the TRAIL-DR and subverts immune control by cells sensitive to depletion with anti-asialo-GM1 antibody (14). MCMV-m166^stop replication is normal in the liver of Trail-dr^−/− mice (14), proving the singular importance for restriction of this TNF receptor protein by m166. Together these results suggest that liver-localized, TRAIL⁺ asialo-GM1-positive (asialo-GM1⁺) cells can trigger TRAIL-DR signaling in MCMV-m166^stop-infected cells, killing them and curtailing viral replication.

Natural killer (NK) cells have long been thought to be the primary innate cell subset mobilized to control early MCMV replication (17). However, mature NK cells express very little TRAIL (18), suggesting that they may not be the key regulators of MCMV-m166^stop replication. In the past, liver NK cells have been partitioned into immature and mature subsets based on their expression of CD11b and CD27 (19), and we observed that immature NKp46-positive (NKp46⁺) cells from naive mice (e.g., CD11b⁻ cells) express the highest levels of TRAIL (Fig. 1A and B). More recently, it has become clear that NKp46⁺ cells are comprised of at least two distinct innate cell types, conventional NK cells (cNK) and innate lymphoid type I cells (ILC1), which together comprise the group I ILCs (16). In the liver, ILC1 and cNK can be readily distinguished by the differential expression of the transcription factor Eomes and integrin α2 (CD49b) (20, 21) (Fig. 1C). We found that more than 90% of liver ILC1 (CD19⁻ T cell receptor β-negative [TCRβ⁻] F4/80⁻ CD3⁻ NKp46⁺ Eomes⁻ CD49b⁻) were CD11b⁻, with only ∼50% of cNK (CD19⁻ TCRβ⁻ F4/80⁻ CD3⁻ NKp46⁺ Eomes⁺ CD49b⁺) displaying this phenotype, indicating that the majority of TRAIL-expressing group I ILCs in naive mice were ILC1 and not immature cNK (Fig. 1D).

FIG 1.

TRAIL expression by group I ILCs in the liver. The panels present the results of analysis of total liver mononuclear cells isolated at the indicated times. (A) Total group I ILCs (F4/80⁻ TCRβ⁻ CD19⁻ CD3⁻ NKp46⁺) from naive WT mice were analyzed for expression of CD27 and CD11b. (B) The 4 subsets for which the results are shown in panel A were analyzed for TRAIL expression. (C) Gating strategy for the detection of liver group I ILCs among living CD19⁻ TCRβ⁻ F4/80⁻ cells (left) and of ILC1 and cNK among liver group I ILCs (right). (D) ILC1 (F4/80⁻ TCRβ⁻ CD19⁻ CD3⁻ NKp46⁺ Eomes⁻ CD49b⁻) and cNK (F4/80⁻ TCRβ⁻ CD19⁻ CD3⁻ NKp46⁺ Eomes⁻ CD49b⁻) were analyzed for the expression of CD27 and CD11b. (E, F) Mock- or MCMV-infected mice were analyzed 36 h or 4 days later for the percentage and absolute numbers of ILC1 (E) and cNK (F) among total group I ILCs. Results are representative of those from three independent experiments with at least five mice per group. Graph lines depict the mean, and statistical significance was determined by the two-tailed unpaired Mann-Whitney U test in this and all figures unless otherwise noted. **, P < 0.01; ***, P < 0.001.

Next, we examined the impact of MCMV infection on liver group I ILCs. No changes in the proportions or numbers of ILC1 and cNK were observed at 36 h of infection compared to those in uninfected mice. In stark contrast, the proportion of ILC1 decreased dramatically at day 4 of infection, albeit their absolute numbers did not significantly change, due to an ∼9-fold increase in liver cNK numbers (Fig. 1E and F). TRAIL expression by cNK and ILC1 was also assessed, and we found that the proportion of TRAIL⁺ ILC1 increased from an average of 58% in naive mice to 87% after 36 h of infection, with a commensurate increase in the levels of ILC1 TRAIL expression (Fig. 2A). Interestingly, MCMV-m166^stop induced an increase in the proportions and levels of TRAIL expression by ILC1 similar to that induced by WT virus at 36 h (Fig. 2B). In naive mice, only ∼3% of the liver cNK expressed any detectable TRAIL, and only a slight increase in the proportion was seen in WT or m166^stop-infected mice at 36 h or 4 days postinfection (Fig. 2A and C). In addition, >90% of TRAIL⁺ group I ILCs in both naive and infected mice were Eomes⁻ CD49b⁻, indicating that the vast majority of TRAIL-producing group I ILCs in the liver were ILC1 (Fig. 2D and E).

FIG 2.

ILC1 are the major TRAIL-expressing cells in MCMV-infected livers. The panels present the results of analysis of total liver mononuclear cells isolated at the indicated times from naive or MCMV-infected mice. (A) TRAIL expression by total group I ILCs and ILC1 and cNK subsets at 36 h of MCMV infection. (B) TRAIL expression by cNK and ILC1 subsets at 36 h of infection with either the MCMV WT or m166^stop. MFI, median fluorescence intensity. (C) TRAIL expression by total group I ILCs, ILC1, and cNK at day 4 of infection. (D) ILC1/cNK distribution for TRAIL⁺ and TRAIL⁻ group I ILCs according to their expression of Eomes and CD49b. (E) Percentage of ILC1 among TRAIL⁺ group I ILCs from naive mice and mice infected for 8 days. **, P < 0.01.

To further examine whether TRAIL-expressing group I ILCs were composed largely of ILC1 during MCMV infection, expression of 6 additional markers whose differential expression segregates ILC1 and cNK was assessed (Fig. 3A). At 36 h after infection, TRAIL⁺ group I ILCs expressed higher levels of the tissue residency markers CD69 and CD49a (20, 21) than cNK did, adding further assurance that these cells were ILC1. In addition, the level of expression of CD61, which is selectively expressed by ILC1 from various tissues (22), was also much higher in TRAIL⁺ group I ILCs than cNK. In turn, the cNK-expressed markers CD62L, KLRG1, and CD11b (21) were absent in TRAIL⁺ group I ILCs, similar to the phenotype of ILC1 (Fig. 3A). Both uninfected and MCMV-m166^stop-infected mice showed phenotypes for TRAIL⁺ group I ILCs similar to the phenotype seen in WT virus-infected mice (Fig. 3B and C). Taken together, these results strongly support the conclusion that virtually all liver-resident, TRAIL-expressing group I ILCs that participate in defense against MCMV infection are ILC1 and not cNK.

FIG 3.

TRAIL-expressing group I ILCs display characteristic phenotypic markers of ILC1. (A) Expression of CD49a, CD69, CD61, CD62L, CD11b, and KLRG1 was assessed in total TRAIL⁺ group I ILCs, ILC1, and cNK at 36 h after infection. (B, C) Expression of CD49a, CD69, CD61, CD62L, CD11b, and KLRG1. (B) Expression of each marker by ILC1, cNK, and TRAIL⁺ group I ILC subsets from naive mice. (C) Expression of each marker by cNK and TRAIL⁺ group I ILCs at 36 h after infection with the MCMV WT or m166^stop. (D) ILC1 and cNK from naive mice and mice infected for 36 h were analyzed directly ex vivo for their expression of IFN-γ after 5 h of incubation with brefeldin A and monensin with or without PMA and ionomycin. No Stim, no stimulation; FSCA, forward scatter area. *, P < 0.1; **, P < 0.01.

ILC1 and cNK function distinctly during MCMV infection.

One unifying characteristic of group I ILCs is thought to be their ability to produce IFN-γ upon appropriate stimulation (21, 23), and this was assessed in liver cNK and ILC1 at 36 h after MCMV infection (Fig. 3D). While no significant IFN-γ production by cNK or ILC1 could be detected in the liver of naive mice, a detectable percentage of IFN-γ-positive (IFN-γ⁺) cells was observed for both subsets in MCMV-infected livers when analyzed straight ex vivo without restimulation (Fig. 3D), and this was also true at 4 days following infection (data not shown). For comparison, ∼10 to 15% of splenic cNK produced IFN-γ at this time (data not shown). In contrast, ex vivo restimulation with phorbol myristate acetate (PMA) and ionomycin (Iono) revealed that ∼6% of cNK and ∼30% ILC1 from naive mice were capable of producing IFN-γ, with these proportions increasing to ∼38% and 67%, respectively, 36 h after MCMV infection (Fig. 3D).

MCMV m166 inhibits TRAIL-mediated group I ILC defenses in the salivary gland during viral persistence.

After host defenses resolve the initial, systemic replication phase of MCMV within ∼7 days, the virus disseminates to the salivary glands (SG), where it replicates persistently for several more weeks. To determine the extent to which m166 might also impact TRAIL-mediated control of MCMV persistence by group I ILCs, the SG replication levels of WT and m166^stop were compared (Fig. 4A). A dramatic reduction in replication of m166^stop was observed at all times between days 4 and 15 (e.g., ∼70-fold at day 8 and day 10). We postulated that this might result from reduced early replication, which could potentially restrict seeding of the SG, and tested this directly by varying the dose of initial infection. A low-dose infection (10³ PFU) yielded undetectable replication in the liver and spleen at day 4, while a higher-dose infection (10⁵ PFU) yielded readily detectable levels by plaque assay (Fig. 4B). In contrast, both doses resulted in identical SG replication at day 12 (Fig. 4C), indicating that persistent replication levels are largely independent of early-phase levels and that reduced m166^stop replication in the SG is likely due to direct immune control at this mucosal site and not decreased dissemination to it.

FIG 4.

M166 restricts TRAIL-dependent control of MCMV by group I ILCs during viral persistence in the salivary gland. MCMV replication levels in various organs are shown as the number of PFU determined in NIH 3T3 cells. (A) Mice were infected with the MCMV WT or m166^stop, and SG replication levels were determined 4 to 15 days later. (B, C) Mice were infected with 10³ PFU (black dots) or 10⁵ PFU (gray dots) of WT MCMV, and replication was determined at day 4 in the liver and spleen (B) or day 12 in the SG (C). (D) Mice were infected with either the WT (black) or m166^stop (red), and replication in the SG was assessed 8 days later. WT mice were depleted (or not) of CD4 T cells prior to infection. (E) Proportion of group I ILCs in the SG 9 days after mice were injected with either PBS or anti-asialo-GM1 antibody. (F, G) Mice were infected with either the WT (black) or m166^stop (red), and replication in the SG was assessed 8 days later. (F) WT mice were depleted (or not) of group I ILCs prior to infection. (G) WT and Trail^−/− mice were infected with either WT or m166^stop MCMV. Results are representative of those from three independent experiments with at least four mice per group. The dashed line denotes the limit of assay detection. *, P < 0.1; **, P < 0.01; ****, P < 0.0001.

CD4 T cells are critical for controlling MCMV persistent replication in the SG (24). To determine to what extent the reduced persistence of m166^stop was dependent on CD4 T cells, we depleted CD4 T cells using the anti-CD4 monoclonal antibody GK1.5 (which produces >95% depletion in SG), and MCMV replication was then measured at day 8 (Fig. 4D). This depletion enhanced the replication of both the WT and m166^stop, highlighting the known requirement for CD4 T cell control in the SG, but indicated that these cells are not selectively responsible for the enhanced control of m166^stop. Very little is known regarding how group I ILCs contribute to MCMV SG defenses, particularly with regard to the relative roles of ILC1 and cNK. To probe this question, mice were injected with anti-asialo-GM1 prior to infection with either the WT or m166^stop. Approximately 80% of group I ILCs were depleted in the SG (Fig. 4E), and this completely restored the replication of m166^stop to WT levels (Fig. 4F). In addition, m166^stop replication was also fully normalized in Trail^−/− mice (Fig. 4G). Together these data show that group I ILCs can mediate TRAIL-dependent antiviral defenses to restrict viral persistence in the salivary gland and that MCMV m166 can subvert this mechanism of host immune control.

MCMV induces specific, high-level TRAIL expression by ILC1 in the salivary gland.

Results of assays in which group I ILCs were depleted indicated that either cNK or ILC1 in the SG were responsible for enhanced control of the m166^stop mutant. Therefore, TRAIL expression in total NKp46⁺ SG-resident cells was assessed at day 8 after infection. Notably, unlike the liver, TRAIL expression was virtually undetectable in total group I ILCs isolated from the SG of naive mice but was significantly upregulated following both MCMV WT and m166^stop infection (Fig. 5A), albeit to a lower extent with m166^stop. Additionally, MCMV induced a significant increase in both the absolute number and the proportion of TRAIL⁺ group I ILCs at day 8 (Fig. 5A). We then determined whether these TRAIL⁺ SG cells were ILC1 or cNK. Recent studies in C57BL/6 (B6) mice have shown that, unlike in the liver, these two subsets of group I ILCs cannot be differentiated by their expression of Eomes and CD49b (25, 26). To determine if this was also true in BALB/c mice, TRAIL expression by total NKp46⁺ SG-resident cells was measured. Unlike in the liver, TRAIL⁺ group I ILCs distributed equally between Eomes- and CD49b-expressing subsets (Fig. 5B). Additionally, expression of the tissue residency markers CD69 and CD49a was equivalent in Eomes⁺ CD49b⁺ and Eomes⁻ CD49b⁻ SG group I ILCs, again distinct from the findings for the liver (Fig. 5C) (27, 28). These data indicate that, like in B6 mice, Eomes and CD49b expression does not discriminate between ILC1 and cNK in BALB/c mouse SGs.

FIG 5.

TRAIL-expressing group I ILCs in MCMV-infected salivary glands express the classical ILC1 phenotypic markers. The panels present the results of analysis of total SG mononuclear cells isolated from naive WT or Trail^−/− mice and at 8 days postinfection with MCMV WT or m166^stop. (A) TRAIL expression by total group I ILCs from naive mice or mice infected for 8 days with the MCMV WT or m166^stop. (B) Distribution of TRAIL⁺ and TRAIL⁻ group I ILCs at 8 days after infection with the MCMV WT according to their expression of Eomes and CD49b. (C) Expression of CD49a and CD69 by Eomes⁺ CD49b⁺ and Eomes⁻ CD49b⁻ group I ILCs at 8 days after infection with the MCMV WT. Gray histograms represent the liver cNK of mice infected for 36 h. (D) Expression of CD49a, CD69, CD61, CD62L, CD11b, and KLRG1 by the TRAIL⁺ (red) and TRAIL⁻ (blue) group I ILC subsets of mice infected for 8 days. Gray histograms represent liver cNK from mice infected for 8 days. (E) Expression of CD49a, CD69, CD61, CD62L, CD11b, and KLRG1 by TRAIL⁺ group I ILCs at 8 days after infection with MCMV WT or m166^stop. Gray histograms represent liver cNK of mice infected for 36 h. **, P < 0.01.

It has been reported that, similar to the liver, SG-resident ILC1 express low/no CD62L, CD11b, and KLRG1 and also highly express CD61 (22, 25, 27). Therefore, expression of these molecules by TRAIL⁺ and TRAIL-negative (TRAIL⁻) group I ILCs in the SG was compared to that by liver cNK after infection with the MCMV WT and m166^stop (Fig. 5D and E). After infection with both the WT and m166^stop, TRAIL⁺ group I ILCs highly expressed the canonical ILC1 markers (CD69, CD49a, CD61) but not the cNK markers (CD62L, CD11b, KLRG1) (Fig. 5D and E). Interestingly, TRAIL⁻ group I ILCs displayed an intermediate phenotype between the phenotypes of the TRAIL⁺ subset and liver cNK, suggesting that in the SG they may be composed of both cNK and ILC1 (Fig. 5D). Nevertheless, these results strongly suggest that the vast majority of TRAIL⁺ group I ILCs in the SG are ILC1. As a second approach, we first subdivided total group I ILCs based on their CD61 expression, followed by analysis of the same canonical cNK and ILC1 markers. This analysis revealed that TRAIL⁺ group I ILCs in the SG show a phenotype very similar to that of total CD61⁺ cells, while both differed significantly from CD61⁻ group I ILCs, which more closely resembled liver cNK (Fig. 6A). These data strongly suggest that CD61 expression can be used to identify ILC1 (CD61⁺) and cNK (CD61⁻) in the SG of BALB/c mice. Interestingly, the absolute numbers of both ILC1 and cNK increased significantly in the SG after infection with both WT and m166^stop (Fig. 6B). Of these, 72% of ILC1 and 27% of cNK expressed TRAIL in WT-infected mice, whereas 43% and 12%, respectively, expressed TRAIL in m166^stop-infected mice (Fig. 6C and D). Overall, our results support the conclusions that (i) CD61 expression identifies SG-resident ILC1, (ii) these CD61⁺ ILC1 are the major subset of group I ILCs producing TRAIL during MCMV persistence, and (iii) this innate cell subset can control MCMV persistent replication if the virus is unable to restrict TRAIL-DR expression.

FIG 6.

ILC1 are the major TRAIL-expressing cells in MCMV-infected salivary glands. The panels present the results of analysis of total SG mononuclear cells isolated from naive WT or Trail^−/− mice and at 8 days postinfection with MCMV WT or m166^stop. (A) Group I ILCs from mice infected with the MCMV WT for 8 days were divided according to their expression of CD61. Cell surface expression of CD49a, CD69, CD62L, CD11b, and KLRG1 was assessed on CD61⁺, CD61⁻, and TRAIL⁺ total group I ILCs. (B) Absolute number of ILC1 (CD61⁺) and cNK (CD61⁻) group I ILCs from naive mice (yellow) and at 8 days after infection with either the MCMV WT (purple) or m166^stop (green). (C) TRAIL expression by cNK (CD61⁻ group I ILCs) and ILC1 (CD61⁺ group I ILCs) from naive mice and mice infected for 8 days. (D) The proportion of TRAIL⁺ ILC1 and cNK from naive mice (yellow) and at 8 days after infection with MCMV WT (purple) or m166^stop (green). (E) ILC1 (CD61⁺ group I ILCs, blue) and cNK (CD61⁻ group I ILCs, gray) from naive mice and mice infected for 8 days were analyzed ex vivo for their expression of IFN-γ after 5 h of incubation with brefeldin A and monensin with or without PMA and ionomycin. *, P < 0.1; **, P < 0.01.

Finally, we assessed the relative expression of IFN-γ by SG-resident group I ILCs in naive and MCMV-infected BALB/c mice (Fig. 6E), as they produce low levels of this effector cytokine in B6 mice compared to their splenic counterparts (29). At day 8 of infection, neither cNK nor ILC1 produced detectable IFN-γ (Fig. 6E). Even following ex vivo stimulation with PMA and ionomycin, only a small proportion of both cNK and ILC1 was able to produce IFN-γ (∼5%).

Together these results reveal that SG-resident ILC1 produce low levels of the canonical effector cytokine IFN-γ in response to MCMV but express high levels of TRAIL, suggesting that TRAIL-mediated defenses may play a particularly important role in regulating host immune responses at mucosal sites of viral persistence.

DISCUSSION

ILCs are important first-line defenders against viral infection due to their ability to mount fast and robust responses in an antigen-independent manner (30, 31). NK cells were the first described and prototypical ILCs (32, 33). Since then, additional subsets have been defined and subdivided into 3 groups based largely on distinct transcription factor requirements for development and unique cytokine production (15, 16, 20). NK cells fall within the group I ILCs and were long believed to be a largely homogeneous cell population expressing NK1.1 and NKp46, with CD11b and CD27 expression varying based on their activation/maturation state (19). Very recently, however, work from several groups has revealed that NK cells defined by these parameters are composed of at least 2 developmentally distinct subsets, cNK and ILC1, together comprising the group I ILCs (16, 20, 22). Consequently, the relative importance of ILC1 and cNK in mounting specific antiviral responses is only beginning to be deciphered, especially with regard to their key effector mechanisms. The fact that ILC1 reside permanently within multiple tissues suggests that they may play a particularly important role in frontline defense against infection, perhaps partnering with tissue-resident memory T cells to provide a rapid one-two punch. ILC1 can rapidly produce IFN-γ in response to both viruses and bacteria (22, 23, 34), but the roles that other immune effector molecules may play is understudied, and our data suggest that their expression of TRAIL may be particularly important. Using the most recently described markers to differentiate ILC1 and cNK, we investigated the phenotype and function of these 2 subsets in naive and MCMV-infected mice. Consistent with what has been reported for B6 mice (16), we found that SG-resident ILC1 have major phenotypic differences compared to their liver counterparts in BALB/c mice. Illustrating this, liver-resident ILC1 do not express Eomes or CD49b, while a significant proportion of ILC1 in the SG do. This is consistent with the findings of several studies showing that the ILC1 phenotype can vary significantly based upon the specific tissue. However, these differences seem to be particularly notable for SG-resident ILC1, prompting them to be proposed as a unique ILC1 subset (16).

CMV has evolved a multitude of strategies to avoid immune detection, with restriction of TRAIL-DR expression by m166 being one example (13, 14). TRAIL can be induced in multiple innate and adaptive immune cells, with its death receptors also being broadly expressed (35). The use of MCMV-m166^stop and several recently described phenotypic markers for group I ILCs allowed us to assess the relative role of cNK and ILC1 in TRAIL-mediated antiviral defense. Additionally, as this work was performed in BALB/c mice, we have confirmed that the cNK and ILC1 phenotypes are largely conserved between these and the widely used B6 strain. Notably, BALB/c mice depleted of group I ILCs showed no increase in persistent replication levels of WT MCMV in the SG, similar to the findings seen in the liver and spleen (14), indicating that this virus can completely neutralize all effector mechanisms used by these innate cells, including TRAIL. This raises interesting questions regarding the relative importance of TRAIL and other group I ILC effector pathways dampened by CMV, such as NKG2D activation (17). It would have been desirable to specifically assess the capacity of ILC1 and cNK to control m166^stop in the absence of the other subset, as opposed to depleting all group I ILCs with anti-asialo-GM1. However, to the best of our knowledge, no currently available antibody/chemical reagents can achieve this. A second approach would be to use genetically deficient mice specifically lacking either subset. Potentially suitable mice do exist on the B6 genetic background, as Zfp683^−/− mice are reported to lack liver-resident ILC1, and cNK are absent or reduced in the SG of Nfil3^−/− mice and in the liver of NCR1^CreEomes^−/− mice (22, 26, 36). Unfortunately, none of these knockout strains are currently available on the BALB/c background, and the presence of Ly49H⁺ group I ILCs in B6 mice abrogates the ability to study m166 function in the absence of m157-mediated hyperactivation of these cells (37–39).

Past work has shown that MCMV can induce TRAIL expression by group I ILCs in the liver and SG, but whether this was by cNK or ILC1 was not explored at that time (14, 40). More recently, splenic cNK have been shown to be capable of expressing TRAIL in response to transforming growth factor β signaling both in vitro and in vivo (25, 41). Taking advantage of recently identified markers differentiating ILC1 and cNK, we have found that liver-resident cNK express virtually no TRAIL, irrespective of MCMV infection. In contrast, ∼60% of ILC1 are TRAIL⁺ in naive livers, with this proportion increasing to >80% by 36 h of infection. While ILC1 from multiple tissues can upregulate TRAIL upon stimulation, liver ILC1 uniquely express very high constitutive levels (20), perhaps emanating from basal IFN-γ signaling in this organ (18). Using CD61 to identify SG ILC1, 34% of ILC1 in naive mice were seen to express TRAIL, while only 3% of SG cNK were TRAIL⁺. Notably, if CD61 was used to segregate liver-resident ILC1 and cNK, results similar to those seen when CD49b and Eomes were used as discriminating markers were seen (Fig. 7). Our data reveal that downregulation of TRAIL-DR cell surface expression by m166 in MCMV-infected cells prevents TRAIL-expressing ILC1 from inducing their apoptosis. Interestingly, MCMV-m166^stop is essentially dead in the liver, while in the spleen and SG, replication is reduced ∼10- to 100-fold but remains detectable (14). Perhaps this TRAIL poised state in the liver (∼60% of ILC1 from naive liver express TRAIL) allows ILC1 to respond even more rapidly than other organs with lower constitutive TRAIL expression.

FIG 7.

CD61 expression segregates ILC1 and cNK in the liver and SG. TRAIL expression by liver ILC1 (F4/80⁻ TCRβ⁻ CD19⁻ CD3⁻ NKp46⁺ CD61⁺) and cNK (F4/80⁻ TCRβ⁻ CD19⁻ CD3⁻ NKp46⁺ CD61⁻) at 36 h of MCMV WT infection. **, P < 0.01.

In contrast to MCMV inducing high numbers of TRAIL-expressing ILC1 in both the liver and SG, <5% of both ILC1 and cNK produced IFN-γ (without ex vivo restimulation). This may be due to a general tolerogenic environment of these organs compared to the spleen, where a much higher proportion of cNK are IFN-γ⁺ at similar infection times (42, 43). However, a significant proportion of both ILC1 and cNK isolated from infected livers were capable of producing IFN-γ following PMA-Iono restimulation, with ILC1 producing more than cNK (∼70% and 40%, respectively). In contrast, only ∼5% of both subsets isolated from infected SG were IFN-γ⁺ following this strong ex vivo restimulation. This is consistent with past results showing that SG-resident group I ILCs are hyporesponsive compared to their liver counterparts (29). Notably, at the same day 8 time point, T cells produced high IFN-γ levels following PMA-Iono restimulation (Fig. 8), indicating that not all IFN-γ-dependent immune cell responses are dampened in this mucosal organ.

FIG 8.

IFN-γ hyporesponsiveness in the SG is group I ILC specific. T cells (CD3⁺ TCRβ⁺) isolated from the SG of naive mice and mice infected for 8 days were analyzed ex vivo for their expression of IFN-γ after 5 h of incubation with brefeldin A and monensin with or without PMA and ionomycin.

The existence of an HCMV protein, UL141, exerting the same function as m166 strongly suggests that blocking group I ILC-mediated TRAIL-induced apoptosis is critical for viral replication in various hosts (13). Whether analogous populations of mouse ILC1 and cNK exist in humans is an important question. Phenotypically distinct NK cells are present in human peripheral blood and tissues, but whether these represent ILC1 and cNK remains controversial due to overlapping phenotypes and conflicting reports (44, 45). HCMV-infected humans often show expansions of CD57⁺ NKG2C^high NK cells in blood, and these cells play an important role in controlling viremia in transplant patients (46). These NK cells have been postulated to be counterparts to memory-like, Ly49H⁺ cNK that expand in response to m157-mediated activation (46). Distinct populations of CD56^bright human NK cells exist and are present at higher levels in tissues than in blood, and whether these CD56^bright human NK cells are analogous to ILC1 in mice is being explored (47, 48). Interestingly, a subset of CCR6⁺ CD56^bright NK cells residing in human livers expresses TRAIL when appropriately stimulated, and despite showing phenotypic differences from mouse ILC1, they also highly express the tissue residency marker CD69 (48). Similar to m166, the HCMV UL141 protein inhibits cell surface expression of the human TRAIL-DRs and blocks TRAIL-mediated killing of infected cells (13). These data support the hypothesis that CMV has evolved to block TRAIL-mediated ILC1 defenses in both people and mice, a model that can be directly tested once human group I ILC subsets are better characterized.

MATERIALS AND METHODS

Mice.

All the mice used in the study were generated on a BALB/c background. BALB/c mice were originally purchased from The Jackson Laboratory (Bar Harbor, ME) and subsequently bred in-house. They are defined throughout the article as WT mice. Trail^−/− BALB/c mice were obtained from Tom Griffith (University of Minnesota, Minneapolis, MN, USA). They are defined in the article as Trail^−/− mice. The Institutional Animal Care and Use Committee (IACUC) of the La Jolla Institute for Immunology approved all the animal experiments.

Reagents and antibodies.

Anti-asialo-GM1 antibody, purchased from Wako, was injected 24 h before infection to deplete group I ILCs. GK1.5, purchased from Bio X Cell, was injected at day 3 preinfection and day 2 postinfection to deplete CD4 T cells. Cellular depletion was verified in spleen and organs by flow cytometry. The following antibodies were purchased from BioLegend: anti-CD3-Alexa Fluor 700 (clone 17A2), anti-CD11b-Brilliant Violet 650 (clone M1/70), anti-CD19-Brilliant Violet 570 (clone 6D5), anti-CD49a-phycoerythrin (PE; clone HMα1), anti-CD49b-fluorescein isothiocyanate (clone DX5), anti-CD61-PE-Cy7 (antibody 2C9.G2; clone HMβ3-1), anti-CD62L-allophycocyanin-Cy7 (clone MEL-14), anti-TCRβ-Brilliant Violet 570 (clone H57-597), anti-TRAIL-biotin (clone N2B2), and anti-streptavidin-Brilliant Violet 605. The following antibodies were purchased from BD Horizon: anti-CD69-PE-CF594 (clone H1.2F3), anti-KLRG1-Brilliant Violet 786 (clone 2F1), and anti-F4/80-Brilliant Violet 570 (clone T45-2342). The following antibodies and reagents were purchased from eBiosciences: anti-IFN-γ-ef450 (clone XMG1.2), brefeldin A, and red blood cell (RBC) lysis buffer. The following antibodies and reagents were purchased from Invitrogen: anti-Eomes-ef450 (clone Dan11mag), Yellow LIVE/DEAD, fixation/permeabilization buffer, and permeabilization buffer. The following antibodies and reagents were purchased from BD Biosciences: anti-NKp46-Brilliant Violet 711 (clone 29A1.4) and GolgiStop protein transport inhibitor. PMA and ionomycin were purchased from Sigma-Aldrich. Percoll was purchased from GE Healthcare. The following reagents were purchased from Gibco: Dulbecco's modified Eagle’s medium (DMEM), 200 mM l-glutamine, penicillin (Pen; 10,000 units/ml), streptomycin (Strep; 10,000 μg/ml), 0.01 M HEPES buffer, phosphate-buffered saline (PBS), and RPMI 1640. The fetal bovine serum (FBS) was purchased from Gemini. The newborn calf serum (NCS) was purchased from Omega Scientific. Collagenase D and DNase were purchased from Sigma. The sodium azide (NaN₃) was purchased from Fisher Scientific.

Cell culture.

NIH 3T3 cells were obtained from the ATCC (CRL1658) and were cultured in Dulbecco's modified Eagle’s medium (DMEM) supplemented with 10% NCS, 1% 200 mM l-glutamine, 1% Pen-Strep, and 1% 0.01 M HEPES buffer. All cell cultures were verified to be mycoplasma free. All plaque assays used NIH 3T3 cells.

Generation of CMV stock preparation.

The K181 m166^stop mutant virus was generated as described in previous work (14). K181 WT, referred to in this article as MCMV WT, and m166^stop, referred to in this article as m166^stop, bacterial artificial chromosome-derived viral stocks were generated in NIH 3T3 cells (49), and stock titers were determined in NIH 3T3 cells.

Tissue cell isolation.

Harvested livers were perfused with phosphate-buffered saline (PBS), mashed though a 70-μm-pore-size cell strainer, washed with 1× PBS, and pelleted. Cell slurries were resuspended in 33.75% Percoll solution (in 1× PBS) and centrifuged at room temperature for 12 min (700 × g) to pellet the mononuclear cells. The cells were washed, treated with red blood cell (RBC) lysis buffer for 5 min, washed again, and then counted. Spleens were mashed through a 70-μm-pore-size cell strainer and washed with complete medium prior to RBC lysis, washed again, and counted. Salivary gland (SG) lymphocytes were isolated by mincing submaxillary glands devoid of contaminating sublingual, parotid, or fat tissue, prior to incubating at 37°C with rotation (85 rpm) in complete medium (RPMI 1640 supplemented with 10% fetal bovine serum [FBS], 1% 200 mM l-glutamine, 1% Pen-Strep) with collagenase D (100 μg/ml) and DNase (10 μg/ml) for 20 min. Cells were then mashed through a 70-μm-pore-size strainer, washed, pelleted, and treated with RBC lysis buffer for 5 min, washed again, and counted. To detect cytokine expression, cells isolated from various tissues were incubated in a 96-well plate in complete medium with brefeldin A and GolgiStop protein transport inhibitor (BD) with or without PMA (0.1 μg/ml) and ionomycin (0.5 μg/ml) for 5 h at 37°C in 5% CO₂.

Determination of virus replication level in mouse organs.

Age- and sex-matched mice were infected with 10³ to 10⁶ PFU of the K181 WT or K181 m166^stop virus intraperitoneally (i.p.). The numbers of PFU from the infected spleen, liver, or total SG were quantified at different days following the infection by doing a standard plaque assay in NIH 3T3 cells as described previously (50).

Flow cytometry staining.

Cells isolated from tissues were washed in fluorescence-activated cell sorting (FACS) buffer (1× PBS, 2% FBS, 0.1% NaN₃), incubated at 4°C for 20 min with antibodies, and washed again. If streptavidin addition was needed, a second similar incubation step was performed. Cells were then fixed with fixation/permeabilization buffer at 4°C for 20 min, washed twice with FACS buffer, washed once with permeabilization buffer, and then incubated at 4°C for 30 min with the antibodies diluted in permeabilization buffer to detect intracellular cytokine and transcription factor expression. Finally, cells were collected in FACS buffer prior to analysis by flow cytometry.

Statistical analysis.

Acquisition of flow cytometric data was done with an LSR II flow cytometer (BD Biosciences), with data analysis being performed using FlowJo (version X) software (TreeStar). GraphPad Prism (version 7.0c) software (GraphPad Software) was used to create graphs and perform statistical analyses. Differences between groups were evaluated for statistical significance by the two-tailed unpaired Mann-Whitney U test. Results are expressed as the mean ± standard deviation (SD).