Non‐apoptotic TRAIL function modulates NK cell activity during viral infection

Abstract

The role of death receptor signaling for pathogen control and infection‐associated pathogenesis is multifaceted and controversial. Here, we show that during viral infection, tumor necrosis factor‐related apoptosis‐inducing ligand (TRAIL) modulates NK cell activity independently of its pro‐apoptotic function. In mice infected with lymphocytic choriomeningitis virus (LCMV), Trail deficiency led to improved specific CD8⁺ T‐cell responses, resulting in faster pathogen clearance and reduced liver pathology. Depletion experiments indicated that this effect was mediated by NK cells. Mechanistically, TRAIL expressed by immune cells positively and dose‐dependently modulates IL‐15 signaling‐induced granzyme B production in NK cells, leading to enhanced NK cell‐mediated T cell killing. TRAIL also regulates the signaling downstream of IL‐15 receptor in human NK cells. In addition, TRAIL restricts NK1.1‐triggered IFNγ production by NK cells. Our study reveals a hitherto unappreciated immunoregulatory role of TRAIL signaling on NK cells for the granzyme B‐dependent elimination of antiviral T cells.

TRAIL modulates NK cell function by repressing NK1.1‐triggered IFNγ production. TRAIL also promotes IL‐15 signaling‐dependent granzyme B expression in NK cells, thereby supporting NK cell‐mediated T cell killing during virus infection.

Introduction

Natural killer (NK) cells and CD8⁺ cytotoxic T cells (CTLs) represent a central line of defense against viral infections through their ability to induce apoptotic death in infected cells. Target cell apoptosis involves exocytosis of cytolytic granules containing perforin and granzymes by cytotoxic lymphocytes 1. In addition, NK cells and CTLs may trigger apoptosis through the engagement of death receptor ligands such as Fas ligand (FasL), tumor necrosis factor (TNF), or TNF‐related apoptosis‐inducing ligand (TRAIL) on cognate receptors expressed on target cells 2, 3, 4, 5. Death receptor‐mediated apoptosis exerts distinct effects on viral replication and infection‐associated pathogenesis 5, 6, 7. In particular, the role of TRAIL during infection is controversial. TRAIL is selectively upregulated on influenza‐specific CD8⁺ T cells to eliminate virus‐infected alveolar epithelial cells 2. TRAIL expression on NK cells also appears to limit the in vivo replication of encephalomyocarditis virus 8. However, TRAIL leads to severe inflammation and tissue damage in Listeria‐infected wild‐type (WT) mice, which is associated with impaired pathogen clearance and reduced survival compared with Tnfsf10/Trail ^−/− counterparts 9. Moreover, NK cells induce TRAIL signaling to eliminate hepatitis B virus (HBV)‐specific CD8⁺ T cells 10 or activated CD4⁺ T cells during chronic murine cytomegalovirus (MCMV) infection 11, thereby negatively regulating antiviral immunity. A comparable immunomodulatory effect of TRAIL has also been reported in rheumatoid arthritis, where TRAIL signaling limits pathology and inflammation, possibly independently of its pro‐apoptotic function 12, 13. Indeed, such non‐apoptotic (non‐canonical) TRAIL activity has been mainly reported in cancer, with TRAIL signaling exerting a pro‐tumorigenic effect in resistant tumor cells 14. Non‐canonical TRAIL signaling in cancer cells elicits receptor‐induced kinase activation triggering survival, proliferation, migration, and metastasis 15, 16. Although the ability of TRAIL to induce such non‐canonical signaling in malignant cells is well established, it is currently unclear whether TRAIL may exert a non‐apoptotic function in non‐transformed cells, in particular in the context of viral infection.

Here, we investigated the contribution of TRAIL to the immune response induced by lymphocytic choriomeningitis virus (LCMV). We found an increased specific CD8⁺ T‐cell response and a faster virus control in infected Trail ^−/− versus WT mice. This phenotype was ascribed to a reduced ability of NK cells to limit LCMV‐specific CD8⁺ T cells in Trail ^−/− mice. Further mechanistic studies revealed that Trail blockade mitigated the IL‐15 signaling‐induced granzyme B production in NK cells in a cell‐extrinsic and dose‐dependent manner—thereby accounting for the reduced T‐cell killing. In addition, TRAIL signaling in NK cells repressed IFNγ production induced upon NK1.1 receptor activation. Taken together, these results unveil a previously unappreciated regulatory role of TRAIL for NK cell function during infection, which is independent of TRAIL pro‐apoptotic activity.

Results

LCMV‐infected Trail ^−/− mice show increased CD8⁺ T‐cell response and improved virus clearance

To study the impact of TRAIL on virus immunity, we analyzed virus‐specific CD8⁺ T‐cell responses in WT and Trail ^−/− mice after infection with LCMV strain WE. The frequencies (Fig 1A) and total numbers (Fig EV1A) of IFNγ‐positive and IFNγ/TNF double‐positive CD8⁺ T cells specific for the LCMV glycoprotein (GP) epitope GP_33–41 were increased in spleens and livers of Trail ^−/− mice compared with control animals. We also observed higher frequencies of cytokine‐producing CD8⁺ T cells specific for the H‐2D^b‐restricted nucleoprotein (NP) epitope NP_396–404 in Trail ^−/− mice (Fig EV1B). No differences in CD4⁺ T cells specific for the GP epitope GP_61–80 were detected (Fig EV1C), indicating that TRAIL exerts a CD8⁺ T cell‐restricted effect on the LCMV‐specific T‐cell response. Trail ^−/− mice also showed reduced CD8⁺ T cell‐mediated liver immunopathology, as measured by lower levels of serum alanine transaminase (ALT) (Fig 1B).

Figure 1. Altered response to LCMV in Trail ^−/− mice.

• A
  Frequencies of cytokine‐producing GP_33–41‐specific CD8⁺ T cells were measured on days 8, 10, and 12 after LCMV infection; n = 3 mice per group per day. Data indicate mean ± SEM and show one representative of at least 2–4 experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05; **P < 0.01; ***P < 0.001 between IFNγ⁺ cells. ^## P < 0.01; ^### P < 0.001 between IFNγ⁺ TNF⁺ cells.
• B
  Liver immunopathology was assessed by measuring serum ALT at the indicated time points post‐LCMV infection. Data shown are mean ± SEM of n = 6 mice per group per day, pooled from two independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05; **P < 0.01.
• C
  Virus titers were determined in spleen and liver; n = 3–11 mice per group per time point. Dotted horizontal lines indicate the limit of detection (LOD). Data were pooled from 1 to 4 independent experiments. Statistical analyses were performed using Mann–Whitney test. *P < 0.05; **P < 0.01.
• D, E
  Experimental setup of P14 cell transfer experiments following LCMV infection of WT and Trail ^−/− mice (D). P14 cell expansion was analyzed in spleen 4 days postinfection; n = 3 mice per group (E). Data indicate mean ± SEM and show one representative of at least 2–4 experiments. Statistical analyses were performed using two‐tailed t‐test. **P < 0.01.
• F
  Frequencies of IFNγ⁺ TNF⁺ P14 cells were measured in the indicated organs 7 days postinfection with LCMV; n = 3 mice per group. Data indicate mean ± SEM and show one representative of at least 2–4 experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. **P < 0.01.

Figure EV1. Trail deficiency leads to an altered immune response in LCMV‐infected mice.

• A–C
  Total numbers of cytokine‐producing GP_33–41‐specific CD8⁺ T cells were counted in the spleen at the indicated time points after LCMV infection (A). Frequencies of cytokine‐producing NP_396–404‐specific CD8⁺ T cells (B) or GP_61–80‐specific CD4⁺ T cells (C) were measured 8 days postinfection. Data shown are mean ± SEM of n = 3 mice per group and are from at least 2–3 experiments. Unpaired two‐tailed t‐test was used. *P < 0.05; **P < 0.01 between IFNγ⁺ cells. ^# P < 0.05; ^## P < 0.01 between IFNγ⁺ TNF⁺ cells.
• D
  Cytokine concentrations were measured in the serum 36 h after LCMV infection using a cytokine multiplex assay. Data indicate mean ± SEM of n = 4 mice per group. nd, non‐detectable. One experiment was performed. Statistical analyses were performed using unpaired two‐tailed t‐test.
• E
  Mice were infected with LCMV, and DR5 was measured on splenic monocytes (defined as CD11b⁺CD11c⁻Ly6C⁺Ly6G⁻ cells) 24 h after infection. Naïve monocytes were used as a control, and data show one representative of n = 3 infected mice per group. One experiment was performed.
• F
  Mice were infected with LCMV, and frequencies of IFNγ⁺ GP_61–80‐specific CD4⁺ T cells were measured in the spleen 8 days postinfection. When indicated (aNK1.1), NK cells were depleted. Data shown are mean ± SEM of n = 3 mice per group and are representative of at least two independent experiments. One‐way ANOVA with Tukey post‐test was used. ns, non‐significant; **P < 0.01.

As CD8⁺ T cells are crucial for LCMV clearance 17, we next analyzed virus titers in spleen and liver at different time points. No differences in virus titer were observed until 8 days of infection, suggesting that TRAIL does not affect early control of LCMV‐WE. However, Trail ^−/− mice showed accelerated virus clearance compared with WT animals, with complete virus elimination in spleen and liver 12 days after LCMV infection (Fig 1C).

In the following, we addressed whether TRAIL expression on CD8⁺ T cells may directly contribute to the LCMV‐specific CD8⁺ T‐cell response. To this end, we transferred congenic Trail ^+/+ T‐cell receptor (TCR) transgenic CD8⁺ T cells specific for the LCMV glycoprotein GP_33–41 (P14 cells) into WT and Trail ^−/− mice previously infected with LCMV (Fig 1D). Under these conditions, P14 cells primed in Trail ^−/− recipients expanded at higher frequencies (Fig 1E) and produced more inflammatory cytokines (Fig 1F).

Taken together, these data reveal that TRAIL limits the expansion of LCMV‐specific CD8⁺ T cells in a cell‐extrinsic manner, thereby modulating virus clearance and liver immunopathology.

The increased specific CD8⁺ T‐cell response in LCMV‐infected Trail ^−/− mice depends on NK cells

To further investigate the role of Trail for the LCMV‐specific immune response, we assessed the kinetics of Trail expression in infected mice. There was a substantial increase in Trail transcripts in spleen and liver in the first days of infection, which then progressively declined to naïve levels after 8 days (Fig 2A), thus suggesting a contribution of TRAIL early during LCMV infection. We next measured inflammatory cytokines released systemically to identify immune populations that were possibly altered in recently infected Trail ^−/− versus WT animals. Among the cytokines analyzed, we found in Trail ^−/− mice threefold higher serum levels of IFNγ (Figs 2B and EV1D), a cytokine that is rapidly secreted by innate lymphocytes, in particular NK cells, following viral infection 18. Indeed, early LCMV infection triggered an upregulation of TRAIL and TRAIL receptor (TRAIL‐R or DR5, which is encoded by Tnfrsf10b) on NK cells (Fig 2C), thus indicating a potential effect of TRAIL/TRAIL‐R signaling on NK cell activity during LCMV infection. Of note, other immune cell populations including CD3⁺, CD4⁺, or CD8⁺ T cells; dendritic cells (DCs, defined as CD11c⁺MHCII⁺Ly6C⁻Ly6G⁻, CD11c⁺MHCII⁺Ly6C⁻Ly6G⁻CD8⁺, or CD11c⁺MHCII⁺Ly6C⁻Ly6G⁻CD8⁻ cells); or neutrophils (defined as CD11b⁺CD11c⁻Ly6G⁺) did not express DR5 24 h after infection, while monocytes (defined as CD11b⁺CD11c⁻Ly6C⁺Ly6G⁻ cells) upregulated DR5 (Fig EV1E).

Figure 2. NK cells contribute to the altered CD8⁺ T‐cell response in infected Trail ^−/− mice.

1. WT mice were infected with LCMV, and Tnfsf10/Trail transcript levels were measured in spleen and liver at the indicated time points. Data are represented as fold induction after normalization to levels in naïve tissue and are mean ± SEM of n = 3–9 mice per group per time point, pooled from two independent experiments.
2. Mice were infected with LCMV, and cytokine concentrations (left panel) or type I IFN activity (right panel) was measured in the serum after 24 h using a cytokine multiplex assay or a bioassay, respectively. Data indicate mean ± SEM of n = 4 mice per group. nd, non‐detectable. One experiment was performed. Statistical analyses were performed using unpaired two‐tailed t‐test. **P < 0.01.
3. TRAIL and TRAIL‐R (DR5) surface protein expression were measured by flow cytometry on splenic NK cells. Representative histograms (left‐hand side) and corresponding fold increase of mean fluorescence intensity (right‐hand side) are depicted. Values shown were normalized to naive controls. Data indicate mean ± SEM of n = 4 for groups of infected mice. One experiment was performed. Statistical analyses were performed using one‐sample t‐test. ***P < 0.001; ****P < 0.0001.
4. Frequencies of IFNγ⁺ GP_33–41‐specific CD8⁺ T cells were measured in the indicated organs 8 days postinfection. When indicated, NK cells were depleted (aNK1.1). Data indicate mean ± SEM of n = 3 for spleen and n = 6 for liver. One representative of three independent experiments is shown. Statistical analyses were performed using one‐way ANOVA with Tukey post‐test. Only the indicated groups were compared for statistical analysis. ns, non‐significant; *P < 0.05; ***P < 0.001; ****P < 0.0001.
5. Serum ALT was measured in the indicated groups of mice 8 days postinfection. Data shown are mean ± SEM of n = 3 mice per group. One representative of three independent experiments is shown. Statistical analyses were performed using one‐way ANOVA with Tukey post‐test. Only the indicated groups were compared for statistical analysis. ns, non‐significant; ***P < 0.001.

Previous reports have shown that besides their well‐described antiviral and anti‐tumor functions, NK cells can also regulate T‐cell responses and thereby influence the outcome of viral infections 11, 19, 20, 21. To address a possible role of these cells in our model, we next depleted NK1.1⁺ cells in WT and Trail ^−/− mice before infection with LCMV‐WE. In line with previous studies 19, 20, 21, antibody‐mediated NK cell depletion increased the LCMV‐specific CD8⁺ T‐cell response in WT mice. However, NK cell depletion also abrogated the limiting effect of Trail on T‐cell priming (Fig 2D), and it comparably prevented liver immunopathology in WT and Trail ^−/− mice (Fig 2E). The virus‐specific CD4⁺ T‐cell response was also enhanced in NK cell‐depleted animals, yet to the same extent in the two mouse strains (Fig EV1F).

These results indicate that, during LCMV‐WE infection, Trail contributes to the NK cell‐mediated regulation of the specific CD8⁺ T‐cell response.

Trail controls cytokine production in NK cells during LCMV‐WE infection

We next applied flow cytometry to determine whether NK cells were the source of higher serum IFNγ in LCMV‐infected Trail mice. The frequencies and numbers of IFNγ‐positive NK cells were increased in the spleens and livers of Trail ^−/− versus WT mice, most prominently 24 h after infection (Fig 3A and B). In addition, Trail ^−/− NK cells expressed higher levels of IFNγ (Fig 3C). These data on altered IFNγ in the NK cells of infected Trail ^−/− mice were further corroborated by a gene expression analysis, revealing that several pathways related to inflammation and cytokine production or signaling are differently affected in Trail ^−/− versus WT NK cells upon LCMV infection (Figs EV2A–C and EV3). This was associated with a threefold rise in serum IFNγ in infected Trail ^−/− animals, which could be abrogated by NK cell depletion (Fig 3D). Yet, this increase in serum IFNγ was not due to alterations in absolute numbers of NK cells in spleens and livers of Trail ^−/− mice compared with WT counterparts (Fig EV4A).

Figure 3. Enhanced IFNγ production in Trail ^−/− NK cells does not directly modulate the specific CD8⁺ T‐cell response.

• A
  Frequencies of IFNγ‐positive NK cells were measured 24 and 48 h post‐LCMV infection in spleen and 24 h postinfection in liver. Data indicate mean ± SEM of n = 3 mice per group and show one representative of at least two independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. **P < 0.01.
• B–D
  Total numbers of IFNγ‐positive NK cells (B), IFNγ expression levels in NK cells (C) and serum IFNγ levels (D) were measured 24 h postinfection. Where indicated, mice were pretreated 3 and 1 day before infection with an anti‐NK1.1 antibody (aNK1.1) or 1 day before infection with an anti‐asialo GM1 antibody (aGM1). MFI, mean fluorescence intensity. In (D), baseline IFNγ serum levels in naïve mice (n = 4) are indicated by a dotted line. Data indicate mean ± SEM of n = 3 (B, C) and n = 6 (D) mice per group and show one representative of at least three independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.
• E, F
  Experimental setup of aNK1.1 treatment before or after LCMV infection (E). Frequencies of cytokine‐producing GP_33–41‐specific CD8⁺ T cells were measured 8 days postinfection in the indicated groups of control or aNK1.1‐treated mice (F). Data indicate mean ± SEM of n = 6 mice per group and show one representative of at least two independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. Only the indicated groups were compared for statistical analysis. ns, non‐significant; ****P < 0.0001.

Figure EV2. TRAIL signaling controls the expression of multiple genes in NK cells during early LCMV infection.

• A
  Venn diagram showing the difference and overlap of genes that are differentially expressed in NK cells of WT versus Trail ^−/− mice during LCMV infection (for an adjusted P‐value < 0.01 and absolute log2 fold change ≥ 2).
• B, C
  Gene Ontology graph showing relationships between pathways that are associated with differentially expressed genes in NK cells of WT (B) or Trail ^−/− (C) mice during LCMV infection. Colored dots indicate the most semantically specific pathways, and gray dots represent connection nodes. Pathways related to inflammation are highlighted by a red ring.

Source data are available online for this figure

Figure EV3. TRAIL signaling controls the expression of multiple genes in NK cells during early LCMV infection.

Genes listed in the pathways highlighted by a red ring in Fig EV2 were clustered hierarchically according to their expression pattern. Each row in the dendrogram depicts the gene log2 fold changes during infection of WT or Trail ^−/− mice. In each sample group, dark red indicates higher and dark blue lower transcript expression.Source data are available online for this figure

Figure EV4. Numbers and surface markers of NK and dendritic cells are unaltered in LCMV‐infected Trail ^−/− mice.

• A
  Mice were infected with LCMV and analyzed 24 h postinfection. Total numbers of NK cells were measured in the spleen and liver. Data shown are mean ± SEM of n = 3 mice per group and are representative of at least three independent experiments. Unpaired two‐tailed t‐test was used.
• B–E
  Mice were infected with LCMV and analyzed 24 h postinfection. Frequencies (B) and total numbers of CD11c⁺ dendritic cells (C) were analyzed in the spleen. Frequencies of CD11c⁺ CD86⁺ cells and CD86 expression on CD11c⁺ cells (D), and frequencies of CD11c⁺ I‐A^b MHCII⁺ cells and I‐A^b MHC II expression on CD11c⁺ cells (E) were quantified in the spleen. Data shown are mean ± SEM of n = 3 mice per group and are representative of at least two independent experiments. MFI, mean fluorescence intensity. Unpaired two‐tailed t‐test was used.
• F–J
  Mice were infected with LCMV, and NK cells were analyzed 24 h postinfection. Frequencies of splenic NK cells expressing the indicated markers and mean fluorescence intensity (MFI) levels of these markers were assessed in spleen (F) and liver (G). Frequencies of Ly49H‐positive NK cells were quantified by flow cytometry in the spleens of naïve mice (H). Eomes (I) or T‐bet (J) expression was analyzed by measuring frequencies of positive NK cells or MFI levels 24 h postinfection. (F, G, I, J) Data shown are mean ± SEM of n = 3 mice per group from three independent experiments. (H) Data show one representative of two independent experiments, with 2 representative mice out of 3 tested per strain. Unpaired two‐tailed t‐test was used. *P < 0.05; **P < 0.01.
• K
  NK cells were isolated from naïve mice, and Gzmb transcript levels were quantified. Data are represented as fold induction relative to Gapdh. Data shown are mean ± SEM of n = 3 mice per group from a unique experiment. Unpaired two‐tailed t‐test was used.
• L
  NK cells from naïve mice were analyzed for CD11b and CD27 expression. DN: double‐negative, CD11b^lowCD27^low NK cells; CD11b^low: CD11b^lowCD27^hi NK cells; DP: double‐positive, CD11b^hiCD27^hi NK cells; CD27^low: CD11b^hiCD27^low NK cells. Data shown are mean ± SEM of n = 3 mice per group from three independent experiments. Unpaired two‐tailed t‐test was used.

NK cell‐secreted IFNγ induces the maturation of DCs, leading to enhanced CD8⁺ T‐cell priming 22, 23. However, we did not find differences in cell frequencies, total numbers or expression of activation markers for DCs in WT and Trail ^−/− animals 24 h after infection (Fig EV4B–E). Therefore, we concluded that DCs likely did not account for the altered response to LCMV‐WE in Trail ^−/− mice.

Alternatively, IFNγ has also been reported to directly promote T‐cell responses 24, 25. Therefore, to investigate whether NK cell‐produced IFNγ contributed to T‐cell activation in our model, we depleted NK cells immediately before or early after LCMV infection and analyzed specific CD8⁺ T cells (Fig 3E). This approach allowed discriminating possible effects of NK cells versus NK cell‐derived early IFNγ during T‐cell priming, respectively. Whereas both NK cell depletion regimens improved the virus‐specific T‐cell response 8 days after LCMV infection, they also both abolished the differences between WT and Trail ^−/− animals (Fig 3F).

Taken together, these data suggest that the augmented CD8⁺ T‐cell priming in Trail ^−/− mice is independent of their initial elevated level of systemic IFNγ.

Trail ^−/− NK cell shows impaired cytotoxicity associated with reduced granzyme B expression

Besides cytokine production early during infection, NK cells may also regulate specific T‐cell responses through their cytolytic activity 19, 20, 21, 26, 27. To assess the cytotoxicity of WT versus Trail ^−/− NK cells after LCMV infection, we next performed an in vitro killing assay using TRAIL‐resistant YAC‐1 cells 28 as NK cell targets. Trail ^−/− splenic NK cells isolated from LCMV‐infected animals showed reduced cytotoxicity compared with WT NK cells (Fig 4A). Using Ifnar1 ^−/− cells as in vivo targets, which are particularly susceptible to perforin/granzyme‐triggered NK cell‐mediated lysis 29, 30, we also found that the NK cell‐mediated elimination of antigen‐specific T cells was reduced in LCMV‐infected Trail ^−/− versus WT mice (Fig 4B and C). The underlying mechanisms appeared to be independent of TRAIL pro‐apoptotic function since virus‐specific CD8⁺ T cells do not express DR5 in LCMV‐infected mice (Fig 4D). The altered cytotoxicity of Trail ^−/− NK cells was also not due to impaired degranulation, as assessed by surface CD107a expression (Fig 4E), but rather linked to reduced levels of granzyme B protein (Fig 4F).

Figure 4. Impaired cytotoxicity of Trail ^−/− NK cells is linked to reduced GZMB production.

• A
  WT and Trail ^−/− mice were infected with LCMV, and NK cells were analyzed 5 days postinfection. Splenocytes were incubated with TRAIL‐resistant YAC‐1 cells at the indicated effector/target ratios. Frequencies of annexin V‐ and DAPI‐double‐positive cells are indicated. Prf ^−/− indicates splenocytes from infected perforin‐deficient mice. Data indicate mean ± SEM of n = 3 mice per group and show one representative of three independent experiments. Statistical analyses were performed using two‐way ANOVA with Tukey post‐test. *P < 0.05; ****P < 0.0001.
• B–D
  Experimental setup of in vivo NK cytotoxicity assay using Ifnar1 ^−/− P14 cells as target cells (B). Frequencies of Ifnar1 ^−/− P14 cells were measured in the spleen 6 h after adoptive transfer. Data indicate mean ± SEM of n = 3 mice per group (C). DR5 expression was assessed on adoptively transfer P14 cells 6 days after LCMV infection (D). Naïve P14 cells were used as a control (n = 3 mice). (C, D) One experiment was performed. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05.
• E
  WT and Trail ^−/− mice were infected with LCMV, and NK cells were analyzed 5 days postinfection. CD107a expression on splenic and hepatic NK cells was measured 5 h after incubation with YAC‐1 cells seeded at an effector/target ratio of 1:1. Data indicate mean ± SEM of n = 3 mice per group and show one representative of three independent experiments. MFI, mean fluorescence intensity. Statistical analyses were performed using unpaired two‐tailed t‐test.
• F
  WT and Trail ^−/− mice were infected with LCMV, and NK cells were analyzed 5 days postinfection. Splenic and hepatic NK cells were stained for granzyme B (GZMB). Data indicate mean ± SEM of n = 3 mice per group and show one representative of at least two independent experiments. MFI, mean fluorescence intensity. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05.

Importantly, Trail ^−/− NK cells in LCMV‐infected mice showed WT levels of several activating or inhibiting markers including CD69, thus suggesting that these cells had been properly activated. Only the activating receptor Ly49H was reduced by 20 and 31.7% on splenic and hepatic NK cells of infected Trail ^−/− mice, respectively (Fig EV4F and G). Yet, Ly49H levels were also diminished on NK cells from naïve Trail ^−/− animals (Fig EV4H), implying that LCMV infection did not contribute to this reduced expression.

Furthermore, T‐bet (Tbx21) and Eomesodermin (Eomes), two transcriptional regulators of NK cell development, maturation and function 31, were comparably expressed in NK cells from WT and Trail ^−/− infected mice (Fig EV4I and J). Thus, these results suggest that these transcription factors likely do not play a role for the decreased cytolytic activity of Trail ^−/− NK cells.

Therefore, our findings indicate that the reduced cytotoxicity of NK cells likely underlies the increased specific CD8⁺ T‐cell response in LCMV‐infected Trail ^−/− mice.

Impaired IL‐15 signaling contributes to reduced granzyme B production in Trail ^−/− NK cells

Resting murine NK cells express high levels of granzyme B transcripts. Upon activation, granzyme B protein expression is strongly enhanced in NK cells, yet with minimal changes in granzyme B mRNA abundance 32. To investigate a potential effect of TRAIL on granzyme B transcription, we measured Gzmb expression in naïve NK cells from spleen and bone marrow. There were comparable levels of Gzmb transcripts in naïve Trail ^−/− versus WT NK cells, indicating that Trail deficiency does not affect constitutive Gzmb expression (Fig EV4K). In agreement with these data, frequencies of CD11b^highCD27^low NK cells, which upregulate cytotoxicity‐related transcripts 33, were unchanged in naïve Trail ^−/− mice (Fig EV4L).

Granzyme B protein production in NK cells is induced by engagement of IL‐15/IL‐15 receptor (IL‐15R) signaling 32. To address a possible defect in this signaling pathway, we first measured Il15 and IL‐15Rβ (CD122) expression during LCMV infection. We found comparable Il15 transcript levels in spleen and liver tissues of WT and Trail ^−/− mice 24 h after LCMV infection (Fig 5A). In addition, there was no difference in surface expression of IL‐15β receptor on Trail ^−/− versus WT NK cells, implying intact ability for these cells to bind trans‐presented IL‐15 (Fig 5B and C).

Figure 5. IL‐15 signaling‐dependent GZMB pathway is defective in Trail ^−/− NK cells.

• A
  Mice were infected with LCMV, and Il15 transcript levels were measured in the indicated organs 24 h postinfection. Data are represented as fold induction after normalization to levels in corresponding naïve tissue. Data indicate mean ± SEM of n = 3 mice per group and show one representative of two independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test.
• B–E
  Mice were infected with LCMV, and flow cytometry was applied on splenic NK cells from infected animals to assess frequencies of IL‐15Rβ‐positive cells (B), IL‐15Rβ expression levels (C), AKT phosphorylation (D), or S6 phosphorylation (E). For AKT and S6 phosphorylation, representative histograms and cumulative results are depicted. MFI, mean fluorescence intensity; Isotype, isotype‐matched control antibody. Data indicate mean ± SEM of n = 3 (B–D) or n = 4 (E) mice per group and show one representative of at least three independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05.
• F, G
  Splenic NK cells from naïve donors were stimulated in vitro for 1 h with IL‐15, and phosphorylation of AKT (F) or S6 (G) was measured. Data indicate mean ± SEM of n = 3 (F) or n = 4 (G) mice per group and show one representative of at least three independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test. *P < 0.05; **P < 0.01.
• H
  GZMB expression was measured in splenic NK cells after stimulation with IL‐15 for 20 h. Data indicate mean ± SEM of n = 4 mice per group and show one representative of at least three independent experiments. Statistical analyses were performed using one‐way ANOVA with Tukey post‐test. ***P < 0.001; ****P < 0.0001.
• I
  WT splenocytes were cultured with IL‐15 ± TRAIL‐R2‐Fc chimeric protein, and GZMB expression was measured in NK cells. Values shown were normalized to unstimulated controls. Data show n = 3 mice per group and show one representative of at least three independent experiments. Statistical analyses were performed using paired two‐tailed t‐test. *P < 0.05.

IL‐15/IL‐15R signaling is conveyed through the PI3K‐AKT‐mTOR pathway to induce granzyme B expression in NK cells 34. Flow cytometry analysis of splenocytes isolated 24 h after LCMV infection revealed decreased phosphorylation of AKT and S6—two markers of PI3K‐AKT‐mTOR pathway activation—in NK cells from Trail ^−/− compared with WT mice (Fig 5D and E). These findings were further substantiated by a transcriptomic analysis, indicating that several pathways related to PI3K/AKT signaling and IL‐2 family signaling (to which IL‐15 belongs) are differently affected in WT versus Trail ^−/− NK cells upon LCMV infection (Tables 1 and 2).

Table 1.

REACTOME pathways that are associated with differentially expressed genes in NK cells of WT mice during LCMV infection

Reactome ID	Definition	Enrichment	Hypergeometric P‐value
194306	Neurophilin interactions with VEGF and VEGFR	3/4	0.00042
4641265	Repression of WNT target genes	4/9	0.00055
2029485	Role of phospholipids in phagocytosis	10/66	0.00114
1236975	Antigen processing‐Cross presentation	11/90	0.00383
2029481	FCGR activation	8/57	0.0056
983170	Antigen Presentation: Folding, assembly and peptide loading of class I MHC	6/39	0.01009
392154	Nitric oxide stimulates guanylate cyclase	4/17	0.00758
8985947	Interleukin‐9 signaling	3/10	0.01023
449836	Other interleukin signaling	4/19	0.01143
451927	* Interleukin‐2 family signaling	6/43	0.01604
418346	Platelet homeostasis	8/72	0.02167
389357	* CD28‐dependent PI3K/Akt signaling	4/22	0.01926
139853	Elevation of cytosolic Ca2+ levels	3/13	0.02189
70895	Branched‐chain amino acid catabolism	4/23	0.02245
2173782	Binding and Uptake of Ligands by Scavenger Receptors	8/79	0.03536

REACTOME pathways that are associated with differentially expressed genes in NK cells of WT mice during LCMV infection (with a hypergeometric P‐value ≤ 0.05 and adjusted P‐value ≤ 0.1). Enrichment indicates the number of differently expressed genes among all genes listed in a specific pathway. Pathways related to PI3K/AKT signaling and IL‐2 family signaling are indicated by an asterisk. This table relates to Fig 5.

Table 2.

REACTOME pathways that are associated with differentially expressed genes in NK cells of Trail ^−/− mice during LCMV infection

Reactome ID	Definition	Enrichment	Hypergeometric P‐value
1059683	Interleukin‐6 signaling	5/11	0.00011
391903	Eicosanoid ligand‐binding receptors	5/14	0.00043
447115	Interleukin‐12 family signaling	5/20	0.00259
6799990	Metal sequestration by antimicrobial proteins	2/3	0.00727
879518	Transport of organic anions	4/15	0.00549
202433	Generation of second messenger molecules	5/26	0.00855
425407	SLC‐mediated transmembrane transport	22/250	0.00782
1442490	Collagen degradation	9/71	0.00885
8957275	Post‐translational protein phosphorylation	13/123	0.00885
382551	Transport of small molecules	50/705	0.00932
389948	PD‐1 signaling	4/19	0.01329
1433557	Signaling by SCF‐KIT	6/40	0.01395
199418	* Negative regulation of the PI3K/AKT network	11/104	0.01519
425397	Transport of vitamins, nucleosides, and related molecules	6/41	0.01566
5357801	Programmed Cell Death	11/105	0.01622
425393	Transport of inorganic cations/anions and amino acids/oligopeptides	11/109	0.02088
1475029	Reversible hydration of carbon dioxide	3/12	0.01964
157118	Signaling by NOTCH	6/45	0.02398
352230	Amino acid transport across the plasma membrane	5/33	0.02313
6811558	* PI5P, PP2A, and IER3 regulate PI3K/AKT signaling	10/97	0.02353

REACTOME pathways that are associated with differentially expressed genes in NK cells of Trail ^−/− mice during LCMV infection (with a hypergeometric P‐value ≤ 0.05 and adjusted P‐value ≤ 0.1). Enrichment indicates the number of differently expressed genes among all genes listed in a specific pathway. Pathways related to PI3K/AKT signaling are indicated by an asterisk. This table relates to Fig 5.

To confirm that these differences were associated with IL‐15/IL‐15R signaling and not due to confounding factors related to LCMV infection, we next assessed the ability of Trail ^−/− NK cells to transduce IL‐15 signals in vitro. We indeed found diminished phosphorylation of AKT (Fig 5F) and S6 (Figs 5G and EV5A) in IL‐15‐stimulated NK cells from Trail ^−/− versus WT mice. This resulted in reduced granzyme B levels in these cells (Fig 5H). Of note, basal levels of phosphorylated AKT and S6 were comparable in naïve Trail ^−/− versus WT NK cells (Fig EV5B). Moreover, pharmacological inhibition of PI3K reduced granzyme B production in IL‐15‐stimulated WT NK cells to levels observed in Trail ^−/− counterparts, thereby suggesting a link between TRAIL signaling and PI3K activation in NK cells (Fig EV5C).

Figure EV5. TRAIL regulates IL‐15 and NK1.1 signaling in NK cells.

• A
  NK cells were MACS‐purified from single‐splenocyte suspensions from naïve donors, stimulated in vitro for 1 h with IL‐15, and phosphorylation of S6 was measured by flow cytometry. One representative of two independent experiments is depicted (n = 1 mouse per group).
• B
  AKT and S6 phosphorylation were measured in naive splenic NK cells. Data shown are mean ± SEM of n = 3 mice per group from at least two independent experiments. Statistical analyses were performed using unpaired two‐tailed t‐test.
• C
  Splenocytes were cultured with IL‐15 ± wortmannin (WRM)/LY294002, and GZMB expression was measured in NK cells. Data shown are mean ± SEM of n = 3 mice per group from at least two independent experiments. Statistical analyses were performed using one‐way ANOVA with Dunn's post‐test. *P < 0.05; **P < 0.01.
• D, E
  WT splenocytes were cultured with IL‐15 ± TRAIL‐R2‐Fc chimeric protein, and AKT phosphorylation (D) or S6 phosphorylation (E) was measured in NK cells. Values shown were normalized to unstimulated control. Data show n = 5 mice per group, pooled from two independent experiments. Statistical analyses were performed using paired two‐tailed t‐test. *P < 0.05; **P < 0.01.
• F–H
  Splenocytes from naïve WT and Trail ^−/− mice were cultured with the indicated cytokines, and frequencies of IFNγ⁺ NK cells (F), or IFNγ‐expression levels in NK cells (G) were measured after 5 h. Alternatively, naïve splenocytes were cultured with IL‐18/IL‐12, and IFNγ was measured in the supernatant at the indicated time points (H). Data shown are mean ± SEM of n = 3–4 mice per group and are representative of two (F, G) or three (H) independent experiments. Unpaired two‐tailed t‐test was used. *P < 0.05.
• I, J
  MACS‐purified DX5⁺ cells were cultured in wells coated with an anti‐NK1.1 antibody, and frequencies of IFNγ⁺ NK cells were measured after 5 h (I) or GZMB expression was measured after 24 h (J). Data shown are mean ± SEM of n = 3–4 mice per group and are representative of three (I) or one (J) independent experiments. Unpaired two‐tailed t‐test was used. **P < 0.01.
• K–O
  Flow cytometry was applied on NK‐92 cells to assess expression of TRAIL (K) and the TRAIL receptors DR4 (L) and DR5 (M). NK‐92 cells were stimulated with IL‐2 ± human TRAIL‐R2‐Fc chimeric protein, and S6 phosphorylation (N) or GZMB expression (O) was measured by flow cytometry (n = 1 per condition). (N, O) Data show one representative of 2 independent experiments.

To establish a direct involvement of TRAIL for downstream IL‐15/IL‐15R signaling and granzyme B expression, we next addressed the effect of simultaneous TRAIL blockade and IL‐15 activation. Such combined treatment led to a reduction in the phosphorylation of AKT (Fig EV5D) and S6 (Fig EV5E), and a consequent diminished granzyme B expression in treated WT NK cells (Fig 5I).

Importantly, TRAIL blockade of activated human NK cells also repressed S6 phosphorylation downstream of IL‐15 receptor (Fig 6). Although this effect was less pronounced than in murine NK cells, these results hint at similar TRAIL signaling‐dependent regulatory mechanisms in NK cells of both species.

Figure 6. TRAIL also regulates the signaling downstream of IL‐15 receptor in primary human NK cells.

Human peripheral blood mononuclear cells were cultured with IL‐2 ± human TRAIL‐R2‐Fc chimeric protein, and S6 phosphorylation was measured in CD56^bright NK cells. Data were pooled from three independent experiments (n = 9 donors per condition). Values shown were normalized to unstimulated controls. Representative histograms are also shown (right‐hand side). Statistical analyses were performed using Wilcoxon matched‐pairs signed rank test. **P < 0.01.

As our transcriptomic analysis had disclosed multiple pathways related to inflammation to be differently affected in Trail ^−/− versus WT NK cells upon LCMV infection, we then also investigated the mechanisms underlying the increased IFNγ levels in NK cells of infected Trail ^−/− mice. IFNγ in NK cells can be induced following stimulation with IL‐12 or IL‐18, which exhibit synergistic effects 35. Alternatively, IFNγ may be produced downstream of activating NK receptors 36, 37. Following incubation of WT and Trail ^−/− splenocytes with IL‐12 or IL‐18, we did not find differences in the frequency of IFNγ‐positive NK cells (Fig EV5F) or in the amount of IFNγ produced (Fig EV5G and H). In contrast, splenocyte activation by NK1.1 crosslinking resulted in higher percentages of IFNγ‐positive Trail ^−/− than WT NK cells (Fig EV5I), yet comparable granzyme B expression (Fig EV5J). These results indicate that Trail deficiency promotes NK1.1 receptor‐induced NK cell activation.

Taken together, these findings reveal that TRAIL promotes IL‐15 signaling‐induced granzyme B production in NK cells. The impaired expression of granzyme B in Trail‐deficient NK cells is associated with a reduced cytotoxicity, which likely accounts for the improved virus‐specific CD8⁺ T‐cell response observed in LCMV‐infected Trail ^−/− mice. In addition, TRAIL restricts NK1.1‐induced IFNγ production by NK cells.

Dose‐dependent effect of TRAIL on granzyme B production by NK cells

We next addressed whether the mode of action of TRAIL on NK cells was cell‐autonomous or cell‐extrinsic. To do so, we co‐cultured WT and Trail ^−/− splenocytes at different ratios and stimulated them with IL‐15 to assess downstream IL‐15/IL‐15R signaling (Fig 7A). We found that S6 phosphorylation and granzyme B expression were directly proportional to the frequency of TRAIL‐expressing cells in the culture, irrespectively of the genotype of the analyzed NK cells. Indeed, increased ratios of co‐cultured Trail ^−/− splenocytes reduced S6 phosphorylation and granzyme B expression in WT NK cells (Fig 7B‐left panel). Conversely, augmentation of the proportion of WT, TRAIL‐expressing cells reversed the impaired downstream IL‐15/IL‐15R signaling of Trail ^−/− NK cells in a dose‐dependent fashion (Fig 7B‐right panel).

Figure 7. Dose‐dependent effect of TRAIL on NK cell function.

• A, B
  Experimental setup of co‐culture assays (A). WT and Trail ^−/− splenocytes were co‐cultured and at the indicated ratios and stimulated with IL‐15 to assess NK cell pS6 and GZMB expression (B). Data were normalized to mean fluorescence intensity (MFI) levels of cultures containing only WT (left panel) or Trail ^−/− (right panel) cells. Data shown are mean ± SEM and are representative of three independent experiments (n = 6 for GZMB and n = 3 for S6). Statistical analyses were performed using one‐way ANOVA. **P < 0.01; ****P < 0.0001.
• C, D
  Experimental setup of adoptive transfers (C). WT (Ly5.1) donor splenocytes (10⁷) were transferred into WT (Ly5.2) or Trail ^−/− (Ly5.2) recipient mice that were then infected with LCMV. Expression of pS6 was assessed after 24 h on donor and host NK cells (D). Data shown are mean ± SEM and are representative of two independent experiments (n = 4). Statistical analyses were performed using one‐way ANOVA with Tukey post‐test. ns, non‐significant; **P < 0.01; ****P < 0.0001.
• E
  NK‐92 cells were cultured at various concentrations of plate‐bound or soluble recombinant TRAIL followed by stimulation with IL‐2. Phosphorylation of S6 was measured by flow cytometry. Data are representative of two independent experiments (n = 1 per condition).

To evaluate the physiological relevance of these findings, we transferred congenic WT (Ly5.1) splenocytes into recipient WT (Ly5.2) and Trail ^−/− (Ly5.2) mice, which were then infected with LCMV (Fig 7C). Similar to the in vitro co‐culture studies, donor WT NK cells showed decreased S6 phosphorylation when activated in Trail ^−/− compared with WT recipient mice (Fig 7D).

As these results implied a cell‐extrinsic contribution of TRAIL to the control of NK cell function, we next examined the role of soluble versus membrane‐bound TRAIL for this regulatory mechanism. For this, we used human NK‐92 cells that show constitutive expression of TRAIL and its receptors DR4 and DR5, and which are thus inherently endowed with the capacity to engage TRAIL signaling (Fig EV5K–M). Of note, NK‐92 cells were found to behave similarly to primary murine and human NK cells, since TRAIL blockade also repressed the signaling downstream of IL‐15 receptor in this cell line (Fig EV5N and O). Remarkably, addition of plate‐coated—but not soluble—recombinant TRAIL promoted, in a dose‐dependent manner, IL‐15 signaling in NK‐92 cells (Fig 7E). Therefore, cell‐extrinsic membrane‐bound TRAIL activates TRAIL signaling in NK cells during inflammation to dose‐dependently regulate granzyme B production.

Collectively, our findings uncover a non‐canonical role of TRAIL signaling for the modulation of essential NK cell functions, which restrains T‐cell response and thereby determines the course of LCMV infection (Appendix Fig S1).

Discussion

In this study, we identified a novel, non‐apoptotic role of TRAIL as a modulator of cytokine production and cytotoxic granule content in NK cells, during viral infection. Our data indicate that Trail‐deficient NK cells have reduced granzyme B expression that is associated with impaired NK cell‐mediated restriction of T‐cell response. This resulted in better virus control and reduced liver pathology in LCMV WE‐infected Trail ^−/− animals.

A previous study reported no alteration in the primary CD8⁺ T‐cell response of Trail ^−/− mice infected with the Armstrong (ARM) strain of LCMV 38. In agreement with these findings, we found that, compared to WT control recipients, Trail ^−/− mice only showed an increased expansion of donor Trail‐competent transgenic P14 cells when they were infected with LCMV‐WE, but not with LCMV‐ARM (Appendix Fig S2). These Trail‐dependent differences in the T‐cell response induced by LCMV‐WE compared to LCMV‐ARM are likely due to distinct virus‐specific CD8⁺ T‐cell expansion magnitude or kinetics, which are directly related to different kinetics of virus replication between LCMV‐WE and LCMV‐ARM strains 39. Thus, it is conceivable that the Trail‐controlled modulation of NK cell activity may differently regulate the virus‐specific CD8⁺ T‐cell response in dependence of its extent or kinetics.

NK cells directly interact with T cells via the engagement of activating and inhibitory ligands whose integration determines overall NK cell function 40, 41, 42. Although granzyme B production in Trail ^−/− NK cells was markedly altered, there were no substantial differences in NK cell receptor expression in Trail ^−/− versus WT NK cells. Of the NK cell receptors analyzed, only Ly49H was affected in its expression, with decreased levels on Trail ^−/− NK cells. Ly49H is an activating NK receptor that recognizes the m157 glycoprotein of murine cytomegalovirus (MCMV) on infected cells. Engagement of Ly49H triggers activation, robust expansion and differentiation of NK cells into a long‐lived memory subset 43, 44. While the function of this receptor during LCMV infection is unknown, Ly49H‐positive NK cells from LCMV‐infected mice—but also from naïve animals—produce more IFNγ than their Ly49H‐negative counterparts when stimulated ex vivo by pro‐inflammatory cytokines 45. In contrast, our results indicate that whereas Trail ^−/− NK cells exhibit lower Ly49H expression than WT cells during LCMV‐WE infection, they show increased IFNγ production. Together, these findings suggest that TRAIL has only minor effects—if any—on NK cell development and that it likely does not alter the ability of NK cells to recognize activated T cells.

Our data establish that TRAIL positively modulates IL‐15 signaling‐induced granzyme B production in NK cells. A previous study reported no change in the cytotoxicity of NK cells from MCMV‐infected Tnfrsf10b/TRAIL‐R2 ^−/− mice 46. These different findings might be explained by the fact that upon MCMV infection Il15ra‐ or Il15‐deficient NK cells are still capable to secrete cytokines and exert cytotoxicity. Indeed, NK cell effector functions during MCMV infection chiefly rely on the m157 viral glycoprotein‐induced activation of Ly49H receptor on NK cells 47.

IL‐15 not only shapes the function of NK cells. It is also of central relevance for the development, homeostasis, and proliferation of these cells 48, which is induced downstream of IL‐15/IL‐15R signaling by phosphorylation of the transcription factor STAT5 49. In spite of a reduced ability to engage IL‐15/IL‐15R signaling upon activation, NK cells in Trail ^−/− mice were found to develop similarly as in WT animals, with comparable numbers at steady state and upon LCMV infection. As mTOR activation requires higher concentration of IL‐15 than STAT5 phosphorylation 35, we therefore conclude that TRAIL only regulates IL‐15/IL‐15R signaling associated with PI3K‐mTOR pathway.

Our results indicate that DR5 is upregulated on NK cells during LCMV infection, but not on virus‐specific CD8⁺ T cells. Although this strongly suggests that in our model NK cells control T cells independently of TRAIL pro‐apoptotic function, we cannot formally exclude that TRAIL‐induced cell death may be engaged in other cells in vivo, which might subsequently affect NK cell function.

Besides the well‐characterized induction of caspase‐dependent apoptosis, TRAIL/TRAIL‐R signaling may also activate non‐canonical pro‐survival pathways through activation of NF‐κB, ERK1/2, and PI3K/AKT pathways 50, 51. Furthermore, TRAIL treatment induces rapid phosphorylation of Akt and mTOR in TRAIL‐resistant cancer cell lines 52, 53. While this non‐canonical TRAIL‐PI3K‐AKT signaling has been described in transformed cells, its precise underlying molecular mechanisms are still unclear 54. Here, we provide evidence that a similar mechanistic link between TRAIL and the PI3K‐mTOR pathway exists in primary NK cells. Hence, TRAIL appears to enhance IL‐15/IL‐15R signaling by synergistic induction of AKT phosphorylation and subsequent mTOR activation.

Our finding of a negative contribution of TRAIL signaling to IFNγ secretion by NK cells relates to a previous report indicating increased serum IFNγ and IL‐12 in MCMV‐infected TRAIL‐R2 ^−/− mice. However, TRAIL‐R2 ^−/− mice did not show higher frequency of IFNγ⁺ NK cells in this infection model. Rather, enhanced IL‐12 producing by TRAIL‐R2 ^−/− DCs likely induced increased numbers of IFNγ‐secreting NK cell upon MCMV infection 46.

IFNγ signaling is critical to restrict the spread of various LCMV strains at the onset of infection and therefore promote the virus‐specific CD8⁺ T‐cell response and later pathogen clearance 55, 56, 57, 58. Yet, our data from NK cell depletion experiments indicate that the higher levels of IFNγ that are produced by NK cells in LCMV‐infected Trail ^−/− mice do not account for the increased CD8⁺ T‐cell response in these animals. This conclusion is further supported by the findings from the in vivo NK cell‐mediated killing assay, suggesting that reduced NK cell cytotoxicity, rather than increased IFNγ production, regulates the expansion of virus‐specific CD8⁺ T cells in infected Trail ^−/− mice. Although IFNγ can affect donor T‐cell survival, motility, and migration 59, 60, 61, this effect is unlikely to happen within the 6‐h window of this in vivo cytotoxicity assay, as also suggested by a previous study using as a readout transgenic CD8⁺ T cells and antibody‐mediated IFNγ blockade 59, 60, 61.

We found that crosslinking of the activating NK cell receptor NK1.1 leads to enhanced IFNγ production in Trail ^−/− compared with WT NK cells, implying an inhibitory effect of the TRAIL/TRAIL‐R pathway on the signaling downstream of NK1.1. Activation of NK1.1 engages a CARMA1/BCL10/MALT1 complex that induces JNK and MAPK phosphorylation and promotes canonical NF‐κB pathway. While this pathway was shown to be largely dispensable for NK cell‐mediated cytotoxicity, it specifically controls the production of cytokines including TNF and IFNγ 62. Absence of TRAIL‐R is associated with prolonged NF‐κB pathway activation in stimulated DC and macrophages, possibly via regulation of IκB‐α degradation or stability 46. Therefore, the TRAIL/TRAIL‐R axis might similarly regulate activating NK cell receptor‐dependent cytokine production via modulation of NF‐κB signaling.

NK cell activation by cytokines or engagement of NK cell receptors can promote cytokine secretion or cytotoxicity either conjointly or independently 48, 63. Our results indicate that TRAIL constrains the ability of activated NK cells to secrete cytokine while enhancing their cytotoxic potential.

NK cells may show divergent roles during infection and subsequent pathophysiology, in a microbe‐ or organ‐specific manner 64. NK cell‐derived cytokines can promote infection‐induced immunopathology 65, 66, 67, 68. Yet, IFNγ produced by NK cells during infection can be also beneficial for recovery—while at the same time NK cytotoxic activity is detrimental 69. The dual regulatory role of TRAIL in NK cells may have evolved in such context to fine‐tune NK effector function and balance protective immunity versus immunopathology.

In conclusion, these findings reveal an unprecedented and unexpected contribution of TRAIL to the control of NK cell function, independently of its pro‐apoptotic role. Furthermore, they also show the relevance of this novel regulatory mechanism for the modulation of CD8⁺ T‐cell response and subsequent virus control. Our results warrant further investigation on whether manipulation of TRAIL signaling may be exploited for NK cell‐based immunotherapy.

Materials and Methods

Mice

C57BL/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and subsequently bred in‐house. Tnfsf10/Trail ^−/− 70 (referred to as Trail ^−/− mice), Prf1 ^−/− 71, congenic B6.SJL‐Ptprc^aPepc^b/BoyJ (Ly5.1), and Tg(TcrLCMV)327Sdz;Rag1 ^tm1Mom;B6.SJL‐Ptprc^aPepc^b/BoyJ (P14) 72 mice were housed and bred in specific pathogen‐free facilities. Annette Oxenius (ETH Zurich, Switzerland) kindly provided us with spleens from B6.129S7‐Ifnar1 ^tm1Agt;Tg(TcrLCMV)327Sdz;B6.PL‐^Thy1a/CyJ (Thy1.1⁺ Ifnar1 ^−/− P14) mice 73. Trail ^−/− mice were backcrossed at least ten times onto a C57BL/6 background, and all mice were maintained on a C57BL/6 genetic background. For all experiments, non‐randomized groups of 8‐ to 12‐week‐old females or males were used. Animal experiments were carried out in compliance with the ARRIVE reporting guidelines. All experiments were performed in accordance with Swiss Federal regulations and were approved by the Cantonal Veterinary Office of Bern, Switzerland.

Antibodies and reagents

All fluorescent‐labeled antibodies used in this study are indicated in Appendix Table S1. LCMV‐specific peptides GP_33–41 (KAVYNFATC), GP_61–80 (GLKGPDIYKGVYQFKSVEFD), and NP_396–404 (FQPQNGQFI) were purchased from Eurogentec (Lüttich, Belgium). Cell proliferation dye eFluor 670 was obtained from eBioscience (Santa Clara, CA, USA).

Virus and virus titration

LCMV strain WE was obtained from Stefan Freigang (Institute of Pathology, University of Bern, Switzerland) and was propagated at a low multiplicity of infection on L929 fibroblast cells. Virus titers were measured using a plaque‐forming assay, as previously described 72. Mice were injected intravenously (i.v.) with 10⁵ plaque‐forming units (pfu) LCMV strain WE, a dose range shown to induce liver immunopathology 20. In the Fig 1C, group size is as follows: for each group of mice, virus titers in the spleen: day 1, n = 9; day 2, n = 9; day 3, n = 3; day 5, n = 3; day 7, n = 3; day 8, n = 9; day 10, n = 11; day 12, n = 6; virus titers in the liver: day 1, n = 3; day 2, n = 3; day 3, n = 3; day 5, n = 3; day 8, n = 6; day 10, n = 8; day 12, n = 3.

NK cell depletions

NK cells were depleted either with 200 μg of anti‐NK1.1 (clone PK136, BioXcell, New Hampshire, USA), as performed in a previous study 26, or 20 μl of anti‐asialo GM1 antibody, as recommended by the manufacturer (Wako Pure Chemical Industries, Virginia, USA). Depletions were performed by intraperitoneal (i.p.) injection of the respective depleting antibodies 3 and 1 days before LCMV infection. Saline (PBS) was injected as a negative control. The efficacy of NK cell depletion was verified.

Adoptive cell transfers

CD8⁺ T cells were purified from naïve P14 spleens by immune‐magnetic negative selection (Miltenyi Biotec, Bergisch Gladbach, Germany), and 10⁶ cells were transferred per recipient mouse 48 h after LCMV infection. P14 cells were analyzed for expansion and activation 2 and 5 days after transfer, respectively.

Alternatively, 8 × 10⁶ wild‐type (Ly5.1) splenocytes were transferred per wild‐type or Trail ^−/− (Ly5.2) recipient mouse 16 h before infection and both donor and recipient NK cells were analyzed by flow cytometry after LCMV infection.

Quantification of alanine aminotransferase and cytokine levels in serum

Alanine aminotransferase (ALT) concentration in the serum was measured at the Department of Clinical Chemistry of the Inselspital/Bern University Hospital using a Roche Modular P800 Analyzer (Roche Diagnostics, Rotkreuz, Switzerland).

Serum cytokine levels were measured using a mouse IFNγ ELISA set kit (BD Biosciences, San Jose, CA, USA) or by flow cytometry using a bead‐based multiplex assay (LEGENDplex, BioLegend, San Diego, CA, USA). Type I IFN activity was measured using an L‐929 cell line transfected with an interferon‐sensitive luciferase construct 74.

Isolation of spleen and liver lymphocytes

Spleens were homogenized to single‐splenocyte suspensions using a 70 μm cell strainer (Falcon Technologies BD, NY, USA). For livers, lymphocytes were isolated from single‐cell suspensions by Percoll gradient centrifugation (800 × g, 15 min).

Flow cytometry

For intracellular cytokine staining, lymphocyte preparations were first incubated for 5 h at 37°C with 20 μg/ml Brefeldin A (Sigma‐Aldrich, St. Louis, MO, USA) (for NK cells), or with 20 μg/ml Brefeldin A and 10⁻⁷ M of a particular LCMV‐specific peptide (for T cells). After incubation, cells were stained for surface markers for 20 min. Cells were then washed, fixed, and permeabilized using BD Bioscience Cytofix/Cytoperm solution, followed by intracellular staining with anti‐IFNγ and anti‐TNF antibodies.

For intracellular staining of phosphorylated proteins, cells were fixed using paraformaldehyde 4% w/v in PBS and methanol 90% v/v in water.

Data were acquired on a LSRII flow cytometer (BD Bioscience) and analyzed using a FlowJo software (Tree Star Inc., Ashland, Oregon, USA).

Isolation and RNA sequencing of NK cells

Single‐cell suspensions were prepared from spleens isolated from naive (n = 3 per group) or from LCMV‐WE‐infected (n = 4 per group) WT or Trail ^−/− mice, 1 day postinfection. NK cells (defined as NK1.1⁺ and CD3^‐ cells) were sort‐purified by flow cytometry and resuspended in TRI‐reagent (Sigma‐Aldrich). RNA was isolated according to the manufacturer's instructions, and RNA concentration and integrity were assessed using a Bioanalyzer 2100 (Agilent, Santa Clara, CA).

Barcoded stranded mRNA sequencing libraries were prepared from high‐quality total RNA samples (~10 ng/sample) using combination of the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB #E7490, Ipswich, MA, USA) to enrich the samples for polyadenylated RNA transcripts and the Ultra II Directional RNA Library Prep Kit (NEB #E7760). Obtained libraries that passed the quality check step were pooled in equimolar amounts, and 1.8 pM solution of this pool was loaded on the Illumina sequencer NextSeq 500 and sequenced uni‐directionally, generating ~500 million reads, each 85 bases long. Library preparation and sequencing was performed at the EMBL Genomics Core Facilities (GeneCore, Heidelberg, Germany).

Computational analysis of RNA sequencing data

RNA sequencing (RNA‐seq) data processing was performed on the SevenBridges platform 75. Between 39.9 and 47.6 million reads were obtained per sample. Read quality was assessed using FastQC 76, and STAR 77 was applied to align the reads to the reference genome (Ensembl m38, build 93) 78. We then used HTSeq‐count 79 to count the number of reads per gene. Differential expression analysis was performed with DESeq2 80 to identify genes for which the relative frequency of transcripts differed in NK cells upon infection with LCMV‐WE (i.e., transcripts were compared before and after infection and selected for adjusted P‐value < 0.01 and absolute log2 fold change ≥ 2). Samples of WT and Trail‐deficient populations were then compared separately to extract two lists of differentially expressed genes.

Pathway analysis

BioInfoMiner 81 was applied for the functional interpretation of the differently expressed genes, using the Biological Process domain of Gene Ontology (GO). BioInfoMiner is an automated tool embedded on the SevenBridges platform that was created for the translational analysis of genomic data. It takes into account the topological organization of terms, targeting to correct the ontological annotation, while it adopts a non‐parametric statistical correction during the enrichment analysis in order to ensure the promotion of non‐trivial, system‐level terms. For the analysis, two lists of significant pathways were extracted, one for each comparison. To disclose differences between these pathway lists, the parent–child relations of Gene Ontology graph were exploited. The difference between a set of terms N to a set M could be defined as those terms which belong to N, so that neither their descendants nor themselves are included in M. Using this approach, unique terms were revealed for each pathway list. In order to remove potential redundancy, terms whose descendants were also significant were filtered out. Thereby, starting from the functional interpretation of differentially expressed genes, the analysis ended to disclose the most specific, uniquely associated biological processes found to be affected during infection for each genotype.

Alternatively, REACTOME 82 was used instead of Gene Ontology as a reference database, with BioInfoMiner as an analysis tool.

Data from the RNA‐seq analysis with detailed gene and pathway lists are provided in the source data files related to Figs EV2 and EV3.

NK cell cytotoxicity assays

For the ex vivo NK cell cytotoxicity assays, total splenocytes were isolated on day 5 post‐LCMV‐WE infection and incubated for 5 h and at different effector/target ratios with eFluor 670‐stained YAC‐1 cells. The number of effector cells was adjusted based on NK cell frequencies among splenocytes. Live/dead discrimination of YAC‐1 cells was performed using annexin V (BioLegend) and DAPI (BioLegend).

For the in vivo NK cell cytotoxicity assay, 5 × 10⁷ splenocytes from Ifnar1 ^−/− P14 mice were transferred into NK cell‐depleted WT recipient mice that were then infected with LCMV. Four days after infection, splenic CD8⁺ T cells were MACS‐purified and transferred into groups of day 4 LCMV‐infected recipient mice. Frequencies of Ifnar1 ^−/− P14 cells in the spleen were measured 6 h after transfer.

Quantitative PCR analysis

Total RNA was isolated from mouse spleen and liver tissue using TRI‐reagent (Sigma‐Aldrich) according to the manufacturer's instructions. RNA was reverse‐transcribed into cDNA using M‐MLV Reverse Transcriptase (Promega, Fitchburg, WI, USA). FastStart SYBR Green Master (Roche, Basel, Switzerland) and commercial primers (Qiagen, Venlo, Netherlands) were used to detect Trail, Il15, and Gapdh transcript levels. All PCR products were run and analyzed on a StepOnePlus Real‐Time PCR System (Life Technologies, Carlsbad, CA, USA). Expression levels of the tested genes were normalized to Gapdh mRNA, or fold induction was calculated using the 2^−ΔΔCT method 83.

In vitro cytokine stimulation of NK cells and TRAIL signaling inhibition studies

Splenocytes were isolated from naïve mice and cultured with recombinant murine IL‐15 (100 ng/ml) (PeproTech, NJ, USA). For PI3K pathway inhibition, 1 μM wortmannin and 50 μM LY294002 were added 1 h prior to activation with IL‐15. TRAIL signaling was blocked using 3 μg/ml of mouse TRAIL‐R2‐Fc chimeric protein (Enzo Life Sciences, NY, USA) added 1 h prior to activation with IL‐15. Kinase phosphorylation was assessed 1 h and granzyme B expression 20 h after the addition of IL‐15.

Whole blood samples from health donors were obtained from Interregional Blood Transfusion SCR Ltd, Bern, Switzerland, under the signed consent of the donors and in agreement with the local legislation. Human peripheral blood mononuclear cells were isolated by Ficoll gradient centrifugation and were stimulated for 1 h at 37°C with 1,000 UI/ml of recombinant human IL‐2, as previously described 35. TRAIL signaling was blocked using 3 μg/ml of human TRAIL‐R2‐Fc chimeric protein (Enzo Life Sciences).

Co‐culture of wild‐type and Trail ^−/− splenocytes

Splenocytes isolated from naïve wild‐type (Ly5.1) and Trail ^−/− (Ly5.2) mice were co‐cultured at different ratios and stimulated with IL‐15 (100 ng/ml) for 1 h to assess kinase phosphorylation or for 20 h to evaluate granzyme B expression.

In vitro experiments using NK‐92 cells

Human NK‐92 cells were initially obtained from ATCC and kindly provided to us by Eva Szegezdi (NUI, Galway, Ireland). Cell lines were tested negative for mycoplasma. NK92 cells were maintained in medium supplemented with 100 UI/ml of IL‐2. To assess the mechanisms underlying the immunoregulatory effect of TRAIL signaling on NK cells, NK‐92 cells were first washed and cultured for 16 h in medium not supplemented with IL‐2. NK‐92 cells were next incubated at 37°C with different concentrations of plate‐coated or soluble human TRAIL (BioLegend) for 30 min, and IL‐2 (2,000 UI/ml) was then added to these cultures for one or 20–24 more hours before flow cytometry analysis of S6 phosphorylation or GZMB expression, respectively.

Statistical analysis

Sample size for in vivo studies was estimated by power analysis and adjusted for β = 0.2, with the assumption that differences between the groups were 1.5‐ to 2‐fold. Statistical tests were selected based on the variation in each data group and on whether multiple comparisons were performed. Groups with similar variance were compared using parametric tests; groups with significantly different variations were analyzed using non‐parametric tests. For datasets including several time points, tests were selected based on the type of variance present in the majority of the time points. Statistical tests are two‐tailed and indicated in the figure legends. All statistical evaluations were performed using GraphPad Prism v.7.0b for Mac or v.6.03 for Windows (GraphPad Software, La Jolla, CA, USA). Unless specified, only statistically significant differences are indicated in the figures. For all statistical analyses: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Author contributions

LCA conceived and performed experiments and wrote the manuscript. MDB, NK, and CL performed experiments. TK and AC performed computational analysis. RS provided key reagents. NC provided expertise and feedback. PK conceived experiments, wrote the manuscript, and secured funding.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

EMBO Reports (2020) 21: e48789

Data availability

The datasets produced in this study are available in the following database: RNA‐Seq data: ArrayExpress E‐MTAB‐7562 (https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-7562/).