Dendritic Cells and NK Cells Stimulate Bystander T Cell Activation in Response to TLR Agonists through Secretion of IFN-αβ and IFN-γ

Mice and injections

C57BL/6 (B6) mice were obtained from Charles River-U.K. or the specific pathogen-free unit at the Institute for Animal Health. IFN-γ-deficient (IFN-γ^−/−) mice, 129SvEv(129) mice, 129 background mice that were IFN-αβ receptor deficient (IFN-αβR^−/−) (16) (originally purchased from B&K Universal), and DO11.10 TCR transgenic mice (17) (originally obtained from Dr. F. Powrie, Sir William Dunn School of Pathology, University of Oxford, Oxford, U.K.) were obtained from the specific pathogen-free unit at the Institute for Animal Health. C3H/HeN and C3H/HeJ mice were obtained from Harlan U.K. All injections were given i.v. in PBS. All animal experimentation was conducted with the approval of the Home Office and the ethical review committee of the Institute for Animal Health.

Reagents

LPS (from Escherichia coli 055:B5) and poly I:C were purchased from Sigma-Aldrich. Peptidoglycan, zymosan, and R-848 were purchased from InvivoGen (San Diego, CA). The sequences for CpG-containing oligonucleotides were supplied by Dr. H. Hochrein (Technische Universitat Munchen, Munich, Germany) and were synthesized by MWG Biotech. Oligo2216 (5′-GGG GGA CGA TCG TCGs GGG GG) and control oligo (5′-GGG GGA GCA TGC TGC GGG GG) were used (phosphothiolated bases in bold, CpG motif underlined) (18). Murine IFN-β protein and rabbit polyclonal Abs against murine IFN-α and IFN-β were purchased from PBL Biomedical Laboratories. mAbs against IFN-α (clone F19) and IFN-β (clone 7F-D3) were purchased from HyCult Biotechnology and Yamasa, respectively. IFN-γ protein and mAbs against IFN-γ (clone 37895.11) and IL-18 (clone 93-10C) were obtained from R&D Systems, and mAbs against IL-12p40/p70 (clone C17.8) and TNF-α (clone MP6-XT3) were purchased from BD Biosciences. rIL-18 was purchased from BioSource International. Recombinant murine IFN-α4 was a gift from A. Le Bon (Edward Jenner Institute for Vaccine Research, Compton, U.K.) (19).

Isolation of DCs and NK cells

Spleen fragments were digested for 20 min at room temperature with 0.14 mg/ml DNase I (Sigma-Aldrich) and either 1 mg/ml collagenase III (Lorne Laboratories) or 0.4 mg/ml Liberase CI (Roche), then treated for 5 min with 7.9 mM EDTA to disrupt T cell-DC complexes. Low density cells were enriched by centrifugation on Nycoprep 1.077A (AXIS-SHIELD PoC) medium. DCs were collected by magnetic cell sorting after labeling with MACS anti-CD11c microbeads (Miltenyi Biotec) or anti-CD11c-FITC (clone HL3; BD Pharmingen) plus anti-FITC microbeads (Miltenyi Biotec). Typically, spleens from three to six mice were pooled to generate CD11c⁺ cells for one experiment. For depletion of CD11c⁺ cells, splenocytes were labeled with anti-CD11c-FITC, and CD11c⁻ cells were sorted with a MoFlo flow cytometer (DakoCytomation). For depletion of DX5⁺ cells, anti-DX5 MACS microbeads were added to the low density cells, and cells were passed over MACS LD depletion columns. CD11c⁺ cells were purified from the resultant NK cell-depleted population (non-DX5⁺ cells, >99%) as described above. Anti-DX5 microbeads with MACS-positive selection columns were used to purify NK cells.

Isolation of T cells

Single-cell suspensions from pooled lymph node (LN) cells were treated with mAbs reactive with CD4 (GK1.5), MHC class II (M5/114), CD45R/B220 (RA3-6B2), and CD11b (M1/70), after which the non-CD8⁺ T cells were removed by immunomagnetic bead depletion using Dynabeads (Dynal Biotech). Resultant CD8⁺ T cells were typically 93–96% pure. In some cases, anti-CD4 was not included, to allow purification of total T cells. Cells were prepared in RPMI 1640 with Glutamax and HEPES (Invitrogen Life Technologies) with 2% FCS. T cells were cultured in complete medium (RPMI 1640 with Glutamax and HEPES (Invitrogen Life Technologies), 10% FCS, 50 μM 2-ME, 5% NCTC medium, 100 U/ml penicillin, 100 μg/ml streptomycin, and 250 μg/ml gentamicin) or with the indicated concentration of cell-conditioned medium overnight. When assessing the ability of neutralizing Abs to inhibit the T cell stimulatory activity of CD11c⁺ cell supernatants, Abs (5 μg/ml) were added to the supernatant 15–20 min before adding the supernatant to the T cells. Except when LPS was added in vitro, all media contained polymyxin B sulfate (100 U/ml; Sigma-Aldrich).

In vitro culture of CD11c⁺ cells and NK cells

For production of CD11c⁺ cell supernatant, cells were cultured for 24 h in complete medium (10⁶ cells/ml). CD11c⁺ cells and NK cells were cocultured for 7–8 h at a 1:1 ratio in 24- or 48-well, flat-bottom plates or in the lower well of Transwell plates (Costar U.K.) at 10⁶ cells/ml. To determine the ability of soluble factors to stimulate NK cell IFN-γ production, CD11c⁺ cells were placed in the upper wells, and NK cells were placed in the lower wells of Transwell plates, separated by a 0.4-μm pore size polycarbonate membrane. When assessing the ability of neutralizing Abs to inhibit NK cell IFN-γ production, Abs (5–10 μg/ml) were added to wells containing NK cells before adding CD11c⁺ cells.

Flow cytometry

Conjugated mAbs against CD4, CD44, CD54 (ICAM), CD69, CD86, CD95, Ly6A/E (Sca-1), CD11c, and CD49b (DX5) were purchased from BD Biosciences. PE-conjugated KJ1-26, which reacts specifically with the clonotypic TCR expressed on CD4⁺ T cells in DO11.10 mice (20), was purchased from Caltag Laboratories. Cy5-conjugated anti-CD8α was prepared in-house from hybridoma YTS.169. Labeled cells were analyzed on a FACSCalibur flow cytometer with the use of CellQuest software (BD Biosciences).

Cytokine detection

IFN-γ concentrations in supernatants were measured using mouse IFN-γ DuoSet or Quantikine M ELISA kits (R&D Systems). To detect intracellular IFN-γ, GolgiStop (BD Biosciences) was added to wells for the last 3–4 h of culture. After staining for cell surface markers, cells were fixed in 2% paraformaldehyde, permeabilized with 0.1% saponin, and then stained with anti-IFN-γ-allophycocyanin (BD Biosciences). To measure total IFN activity, an IFN bioassay was used, in which samples were assayed for their ability to protect L929 cells against encephalomyelocarditis virus infection (21).

LPS induces phenotypic activation of naive and memory phenotype T cells

Previous work has shown that injection of LPS into mice stimulates selective proliferation of memory phenotype (CD44^high) CD8⁺ T cells, but expression of CD69 on a high proportion of all T cells (11). The latter observation suggests that many T cells receive a partial activation signal in response to LPS. To investigate whether T cells also show other signs of activation in LPS-treated mice, LPS was injected i.v. into normal B6 mice, and the expression of a number of cell surface molecules associated with T cell activation was assessed (Fig. 1⇓A). In addition to CD69, CD54, CD86, CD95, and Ly6A/E were consistently up-regulated on CD4⁺ and CD8⁺ T cells present in both LNs and spleen. Phenotypic activation was apparent among CD44^high and CD44^low cells, indicating that both naive and memory T cells received activation signals (Fig. 1⇓B). Moreover, LPS induced phenotypic activation of TCR transgenic CD4⁺ T cells when injected into unprimed DO11.10 mice (note that CD4⁺ T cells in BALB/c background mice are essentially Ly6A/E⁻ in a resting state; Fig. 1⇓C). Because transgenic T cells in these mice (specific for OVA peptide 323–339 in association with I-A^d) (17) are uniformly naive in the absence of immunization with OVA (data not shown), this result provides clear evidence that truly naive T cells receive activation signals in response to LPS in vivo.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

LPS induces phenotypic activation of naive and memory phenotype T cells in vivo. A, B6 mice were injected i.v. with LPS (50 μg) or PBS and were killed 24 h later. Histograms show the expression of the indicated markers on gated CD8⁺ or CD4⁺ T cells from pooled LNs of LPS-injected (thick line) or PBS-injected (thin line) mice. B, B6 mice were injected i.v. with LPS (10 μg) or PBS and were killed 24 h later. Dot plots show CD44 and Ly6A/E staining of gated CD8⁺ or CD4⁺ T cells from pooled LN cells of PBS-injected (left) or LPS-injected (right) mice. C, DO11.10 mice were injected i.v. with LPS (10 μg) or PBS and were killed 24 h later. The histogram shows Ly6A/E expression on gated CD4⁺ KJ1–26⁺ T cells from PBS-injected (▪) or LPS-injected (□) mice. D, B6 mice were injected i.v. with LPS (50 μg) or PBS and were killed 4 h, 24 h, or 7 d later. Data show the average Ly6A/E expression (as mean fluorescence intensity (MFI) ± SE) for three mice per group. E, B6 mice were injected i.v. with PBS or the indicated dose of LPS and were killed 24 h later. Data show the average Ly6A/E expression (as MFI ± SE) for three mice per group.

The increase in expression of activation molecules after injection of LPS was rapid, being detectable by 4 h, and was maximal by 24 h (Fig. 1⇑D). However, these changes were also transient, as expression of all markers had returned to baseline by 7 days. A similar phenotypic response was observed after injection of LPS isolated from many different bacterial strains (including E. coli, Salmonella enteritidis, Pseudomonas aeruginosa, and Neisseria meningitidis) and in mice of several different backgrounds (B6, BALB/c, 129/SvEv, and C3H/HeN; data not shown). T cell expression of activation markers occurred in response to low doses of LPS in vivo, with up-regulation apparent after injection of 10 ng of LPS, and increased in a dose-dependent manner (Fig. 1⇑E). Overall, these experiments indicated that LPS induced widespread phenotypic activation of T cells in vivo.

Cytokines secreted by CD11c⁺ cells mediate LPS-induced T cell phenotypic activation

To investigate how LPS stimulates phenotypic activation of T cells in vivo, we first determined whether its effects could be replicated in vitro. In these and subsequent experiments, Ly6A/E was used as a marker of T cell activation, but similar changes were observed for the other markers described in Fig. 1⇑ (data not shown). As shown in Fig. 2⇓A, phenotypic activation of T cells did occur when total splenocytes were cultured overnight in the presence of LPS; concentrations of LPS as low as 0.1 ng/ml were sufficient to induce this effect (data not shown). However, purified T cells cultured in the presence of LPS did not up-regulate activation markers, indicating that LPS acted indirectly to stimulate T cells (Fig. 2⇓A). Moreover, the indirect effects of LPS on T cells were cytokine mediated, because the supernatant of spleen cells obtained from LPS-injected mice (after overnight culture) induced phenotypic activation when added to CD8⁺ T cells purified from untreated mice (Fig. 2⇓B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

LPS induces T cell phenotypic activation via cytokines produced by CD11c⁺ cells. A, LPS (100 ng) or an equivalent volume of PBS was added to total splenocytes (left) or purified T cells (right), and the cells were cultured overnight. Histograms show Ly6A/E expression on CD8⁺ T cells from cultures with added PBS (thin line) or LPS (thick line). B, Splenocytes from B6 mice injected i.v. 1 h earlier with PBS or LPS (10 μg) were cultured for 24 h. Culture supernatants from these cells were added to purified CD8⁺ T cells, and Ly6A/E expression was measured 18–24 h later. C, Culture supernatants from CD11c⁺ or CD11c⁻ splenocytes, obtained from mice injected i.v. 1 h earlier with PBS or LPS as indicated, were added to purified CD8⁺ T cells at the indicated dilutions, and Ly6A/E expression was measured 18–24 h later. D, CD11c⁺ cells were cultured for 24 h in the presence of added PBS or LPS (2 μg). Supernatants from these cells were added to purified CD8⁺ T cells (1/2 final dilution), and Ly6A/E expression was measured 18–24 h later. Data represent the average Ly6A/E expression (as mean fluorescence intensity (MFI) ± SE) for three mice per group.

In considering which cells within the spleen were responsible for production of the T cell stimulatory cytokines, APCs (DCs and macrophages) seemed likely candidates given their key role in the innate recognition of infectious agents and, in particular, their ability to respond to LPS. We used a DC marker, CD11c, to investigate the potential role of these cells in the stimulation of T cells by LPS. Spleen cells were obtained from mice 1 h after LPS injection and depleted of CD11c⁺ cells by cell sorting. The remaining CD11c⁻ cells were then cultured overnight, and the supernatant from these cells were assayed for its ability to induce T cell phenotypic activation. As shown in Fig. 2⇑C, supernatant from CD11c⁻ cells did not stimulate up-regulation of Ly6A/E when added to purified CD8⁺ T cells. In contrast, CD11c⁺ cells isolated from the spleens of LPS-injected mice secreted cytokines that induced T cell phenotypic activation (Fig. 2⇑C). Therefore, CD11c⁺ cells were necessary and sufficient for production of the T cell stimulatory cytokines in response to LPS. This appeared to be due to direct effects of LPS, because CD11c⁺ cells purified from control mice secreted T cell stimulatory cytokines when treated with LPS in vitro (Fig. 2⇑D).

Agonists of several different TLRs trigger CD11c⁺ cells to secrete T cell stimulatory cytokines

Previous work has shown that CpG ODNs also induce phenotypic activation of T cells in vivo (14). We investigated whether CpG ODNs as well as agonists of other TLRs acted similarly to LPS in stimulating the production of T cell stimulatory cytokines by CD11c⁺ cells. Agonists of TLR2 (zymosan and peptidoglycan) (22, 23), TLR3 (poly I:C) (5), TLR4 (LPS) (3, 4), TLR7 (R848) (24), and TLR9 (CpG ODNs) (15) were injected into mice, and CD11c⁺ splenocytes were isolated 1 h later. After culturing these cells overnight, the supernatants were tested for their ability to induce up-regulation of Ly6A/E when added to purified CD8⁺ T cells. As shown in Fig. 3⇓A, only zymosan and peptidoglycan failed to induce production of T cell stimulatory cytokines by CD11c⁺ cells; an identical pattern of response was observed when in vivo T cell activation was assessed 24 h after injection of the various TLR agonists (Figs. 3⇓B and 4⇓). Therefore, triggering of TLR3, -4, -7, and -9, but not TLR2, leads to the production of cytokines by CD11c⁺ cells that induce widespread phenotypic activation of T cells.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Signaling through TLR3, -4, -7, and -9, but not TLR2, induces phenotypic activation of T cells. Mice were injected i.v. with peptidoglycan (20 μg), zymosan (20 μg), poly I:C (100 μg), LPS (10 μg), R-848 (25 nmol), CpG (10 nmol), control CpG (10 nmol), or PBS. A, After 1 h, CD11c⁺ cells were purified and placed in culture overnight. CD11c⁺ cell-conditioned medium was diluted 1/2 and added to purified CD8⁺ T cells. Ly6A/E expression was determined after culturing the T cells for 18–24 h, and the results are shown as a ratio of mean fluorescence intensity (MFI) for TLR ligand-induced, supernatant-exposed cells to that of control supernatant-exposed cells. Results represent the mean ± SE for two to four independent experiments. B, Twenty-four hours after injection, the expression of Ly6A/E on LN cells was assessed. Data represent the average Ly6A/E expression on CD8⁺ T cells (as MFI ± SE) for three mice per group.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

IFN-αβ and IFN-γ contribute to TLR agonist-induced phenotypic activation of T cells in vivo. Mice were injected i.v. with poly I:C (100 μg), LPS (10 μg), R-848 (25 nmol), or CpG ODNs (10 nmol) or were left uninjected. Twenty-four hours later, Ly6A/E expression on LN T cells was assessed. A, Comparison of response in WT (129) vs IFN-αβR^−/− mice. B, Comparison of response in WT (B6) vs IFN-γ^−/− mice. Data represent the average Ly6A/E expression on CD8⁺ (left) or CD4⁺ (right) T cells (as mean fluorescence intensity (MFI) ± SE) for three mice per group.

CD11c⁺ cells induce phenotypic activation of T cells through secretion of IFN-αβ and IFN-γ

Previous work has shown that IFN-αβ contributes to the bystander T cell phenotypic activation stimulated by CpG ODNs (14). To determine whether IFN-αβ plays a role in T cell activation induced by LPS-stimulated CD11c⁺ cells, we examined the ability of T cells from IFN-αβ receptor-deficient (IFN-αβR^−/−) mice to respond to CD11c⁺ cell supernatant in vitro. As shown in Fig. 5⇓A, up-regulation of Ly6A/E was greatly reduced on IFN-αβR^−/− compared with wild-type (WT) T cells, indicating that IFN-αβ contributed to the induction of T cell activation. This type I IFN-dependent effect appeared to be due to both IFN-α and IFN-β, because neutralizing Abs against either reduced the degree of Ly6A/E up-regulation by WT T cells (Fig. 5⇓B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

CD11c⁺-derived IFN-αβ and IFN-γ induce phenotypic activation of T cells. A, WT (B6) or IFN-γ^−/− mice were injected i.v. with LPS (10 μg) or PBS, and CD11c⁺ cells were purified 1 h later. After overnight culture, supernatants of CD11c⁺ cells were obtained and added (at 1/2 dilution) to CD8⁺ T cells from WT (129) or IFN-αβR^−/− mice. B, B6 mice were injected i.v. with LPS (10 μg) or PBS, and CD11c⁺ cells were purified 1 h later. After overnight culture, supernatants of CD11c⁺ cells were obtained. Neutralizing Abs against the indicated cytokines were added to the CD11c⁺ supernatants before adding the conditioned medium (at 1/2 final dilution) to CD8⁺ T cells (from B6 mice). IFN-αβγ, All three Abs were added. C, The indicated concentrations of IFN-α, -β, or -γ were added to purified B6 CD8⁺ T cells, and Ly6A/E expression was measured after 24-h culture. All data show Ly6A/E expression as the mean fluorescence intensity (MFI) and are representative of at least two independent experiments.

Because the phenotypic response of IFN-αβR^−/− T cells was not completely abrogated, we investigated the consequences of blocking the activity of other cytokines. Addition of neutralizing Abs against TNF-α, the IL-2/IL-15R β-chain (CD122), IL-18, or IL-12 to the CD11c⁺ cell supernatant had no effect on T cell up-regulation of Ly6A/E (data not shown). Conversely, anti-IFN-γ partially inhibited the T cell response (Fig. 5⇑B). Furthermore, when CD11c⁺ cells were purified from LPS-injected, IFN-γ-deficient (IFN-γ^−/−) mice, the supernatant derived from these cells had a reduced ability to up-regulate Ly6A/E expression on control T cells (Fig. 5⇑A). Notably, no up-regulation of Ly6A/E was observed when a combination of Abs against IFN-α, -β, and -γ was added to the culture (using supernatant from WT CD11c⁺ cells and WT T cells; Fig. 5⇑B), or when supernatant derived from LPS-stimulated IFN-γ^−/− CD11c⁺ cells was added to IFN-αβR^−/− T cells (Fig. 5⇑A). Taken together, these data indicated that IFN-αβ and IFN-γ accounted for all the T cell stimulatory activity present in the CD11c⁺ cell supernatants. In agreement with this, IFN-α, -β, and -γ were all capable of stimulating up-regulation of Ly6A/E by purified T cells (Fig. 5⇑C).

To determine whether IFN-αβ and IFN-γ also contribute to TLR agonist-induced T cell activation in vivo, T cell phenotype was examined after injection of poly I:C, LPS, R848, or CpG ODNs into IFN-αβR^−/− or IFN-γ^−/− mice (Fig. 4⇑). In all cases, up-regulation of Ly6A/E was reduced in both types of knockout mice compared with controls. The extent of the reduction in T cell activation varied with the different agonists; for some, a greater effect was observed in IFN-αβR^−/− mice. For example, for R848, Ly6A/E up-regulation was reduced by 82–95% (for CD4⁺ or CD8⁺ T cells, respectively) in IFN-αβR^−/− mice, but only by ∼50% in IFN-γ^−/− mice, whereas for CpG ODNs, a greater reduction was observed in IFN-γ^−/− than in IFN-αβR^−/− mice (44–51 vs 22–37%). Nevertheless, the data indicated that both IFN-αβ and IFN-γ participate in bystander T cell activation induced by all stimuli tested.

IFN-αβ and IFN-γ are produced by distinct cell populations

Although we had used CD11c as a marker to enrich for DCs, the identification of IFN-γ as an active factor in the CD11c⁺ cell supernatant raised the question of whether DCs were actually responsible for secreting these cytokines. To investigate the identity of the IFN-γ-secreting cells more definitively, we combined intracellular cytokine staining with analysis of additional cell surface molecules on the CD11c⁺ cell population. As shown in Fig. 6⇓A, staining with the NK cell marker DX5 revealed a population of DX5⁺ CD11c⁺ cells; the same cells were also positive for another NK cell marker, NK1.1 (in B6 mice), and for CD11b, but were negative for CD3, CD14, CD19, MHC class II, Gr-1, and F4/80 (data not shown). This phenotype is consistent with these cells being NK cells (25). Strikingly, virtually all IFN-γ-producing cells were contained within the DX5⁺ CD11c⁺ cell population (Fig. 6⇓A).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

LPS-induced IFN-αβ and IFN-γ are produced by distinct cell types. A, LPS (10 μg) was injected i.v. into B6 mice, and CD11c⁺ cells were purified 2 h later. CD11c⁺ cells were cultured in vitro, and IFN-γ expression by was visualized by intracellular cytokine staining (8-h total time post-LPS injection). Dot plots on the right show DX5 and CD11c expression on total (upper) and IFN-γ⁺ (lower) cells. Data are representative of four experiments. B, PBS or LPS (10 μg) was injected i.v. into B6 mice. Twenty minutes later, total CD11c⁺, DX5⁺, and CD11c⁺DX5⁻ cell populations were purified and placed in culture for 24 h. The amount of bioactive IFN was determined by bioassay. Data represent mean IFN activity (units per milliliter) ± SE for two independent experiments.

To determine whether the same cells were also producing IFN-αβ, CD11c⁺ DX5⁻ and DX5⁺ cells from LPS-injected mice were cultured for 24 h, and IFN-αβ activity in the supernatants was measured by bioassay. As shown in Fig. 6⇑B, bioactive IFN was produced predominantly by the CD11c⁺DX5⁻ cell population. Importantly, all the bioactive IFN produced by these cells appeared to be IFN-αβ, because IFN-γ was essentially absent in the supernatant of CD11c⁺DX5⁻ cells (Fig. 7⇓A). Therefore, IFN-αβ and IFN-γ were produced by discrete cell populations after LPS injection: IFN-αβ by DCs and IFN-γ by NK cells.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 7.

DCs activated by TLR agonists induce NK cell IFN-γ production. A, CD11c⁺, DX5⁺, and CD11c⁺DX5⁻ cell populations were cultured as described in Fig. 5⇑B, and the amount of IFN-γ in the supernatant was determined by ELISA. B, PBS or LPS (10 μg) was injected i.v. into C3H/HeN mice. Twenty minutes later, CD11c⁺ cells were purified and placed in culture with CFSE-labeled DX5⁺ cells from C3H/HeJ mice. CD11c⁺ and DX5⁺cells were either mixed together in the same well (upper) or separated by a porous membrane in Transwells (lower). Numbers show the percentage of IFN-γ⁺ cells among CFSE⁺ cells. C, CD11c⁺ cells derived from mice injected i.v. with ligands of TLR3, -4, -7, and -9 (as in Fig. 3⇑) were cultured with CFSE⁺ DX5⁺ cells from untreated mice, and the amount of IFN-γ was determined by intracellular cytokine staining; CD11c⁺ and DX5⁺ cells were either mixed together in the same well or separated by a porous membrane in Transwells as indicated. Data show the percentage of IFN-γ⁺ cells among DX5⁺ CFSE⁺ cells (mean ± SE from two or three experiments).

TLR agonist-stimulated DCs induce IFN-γ production by NK cells

The production of IFN-αβ by DCs and IFN-γ by NK cells after LPS injection raised the question of whether these responses were triggered independently or were in some way linked. Consistent with the latter idea, little or no IFN-γ was produced when CD11c⁺ cells from mice injected 1 h previously with LPS were separated into DX5⁺ and DX5⁻ subsets before culture (Fig. 7⇑A). This observation suggested that signals derived from LPS-stimulated DCs were required for triggering NK cell IFN-γ production.

To investigate whether TLR agonist-stimulated DCs were capable of directly inducing NK cells to produce IFN-γ, DCs were purified from LPS-injected mice and mixed with DX5⁺ NK cells from untreated mice; NK cells were labeled with CFSE to permit discrimination between the two cell populations. To eliminate any effects due to possible carryover of LPS from the injected mice, either polymyxin B was added to cultures or NK cells were derived from LPS-hyporesponsive C3H/HeJ mice (26) (in which case these cells were mixed with DCs from LPS-responsive C3H/HeN mice). As shown in Fig. 7⇑B, NK cells mixed with DCs from LPS-injected, but not PBS-injected, mice produced IFN-γ. Similarly, DCs from mice injected with poly I:C, R848, or CpG ODNs stimulated NK IFN-γ expression (Fig. 7⇑C). (Note that no CD11c⁺ cell IFN-γ secretion was detectable after injection of peptidoglycan or zymosan (data not shown).) Direct contact between DCs and NK cells was not essential, because NK cells expressed IFN-γ when separated from LPS-exposed DCs by a porous membrane, although minimal cell contact-independent IFN-γ production was observed in response to other TLR stimuli (Fig. 7⇑, B and C). Interestingly, greater induction of LPS-induced, contact-independent IFN-γ production was observed when cells were derived from C3H as opposed to B6 mice (compare Fig. 7⇑, B and C), indicating that strain differences may exist in this respect. Overall, however, it was evident that IFN-γ production was much higher when DCs and NK cells were mixed together. Hence, TLR agonist-stimulated DCs provide both soluble and cell-associated factors that play a role in inducing resting NK cells to produce IFN-γ.

Role for IFN-αβ in DC-induced IFN-γ production

Because DCs produce type I IFN in response to LPS injection, we investigated the possibility that IFN-αβ contributed to the stimulation of NK cell IFN-γ production. NK cells were obtained from untreated WT (129) or IFN-αβR^−/− mice and cultured with DCs from LPS-injected WT mice. In cultures where DCs and NK cells were separated by porous membranes, IFN-αβR^−/− NK cells failed to produce IFN-γ, indicating an absolute requirement for IFN-αβ in DC cytokine-induced IFN-γ expression (Fig. 8⇓A). IFN-γ production by IFN-αβR^−/− NK cells was also greatly reduced compared with that by WT NK cells when these cells were allowed to come into direct contact with DCs (Fig. 8⇓A), although the response was not completely eliminated. Similarly, addition of neutralizing anti-IFN-αβ to cocultures of WT DCs and WT NK cells partially inhibited IFN-γ expression (Fig. 8⇓B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 8.

IFN-αβ and IL-18 contribute to DC-stimulated NK cell IFN-γ production. LPS (10 μg) was injected i.v. into WT (129) or IFN-αβR^−/− mice (A) or B6 mice (B), and CD11c⁺ cells were purified 20 min later. CD11c⁺ cells were placed in culture with CFSE⁺ DX5⁺ cells from untreated WT or IFN-αβR^−/− mice (A) or B6 mice (B). A, CD11c⁺ and DX5⁺ cells were either mixed together in the same well or separated by a porous membrane in Transwells as indicated. B, Neutralizing Abs against the indicated cytokines were added to cocultures of CD11c⁺ and CFSE⁺ DX5⁺ cells. Data show the percentage of IFN-γ⁺ cells among DX5⁺ CFSE⁺ cells (mean ± SE from two (A) or three (B) experiments).

The failure to totally abolish NK cell IFN-γ production in DC-NK cell cocultures by blocking IFN-αβR-mediated signaling in vitro implied that IFN-αβ-independent mechanisms also contributed to the response. However, although the approaches described above should block any IFN-αβ stimulation of NK cells, they would not preclude IFN-αβ-mediated effects on DCs themselves, which could occur before placing these cells in culture. To address this possibility, we measured IFN-γ expression when various combinations of IFN-αβR-positive and -negative DCs and NK cells were cocultured (Fig. 8⇑A). A low percentage of NK cells expressed IFN-γ even when both LPS-stimulated DCs and NK cells were derived from IFN-αβR^−/− mice, providing definitive evidence that DCs were capable of stimulating IFN-γ production by IFN-αβ-independent mechanisms under these culture conditions. However, it was also evident that the percentage of IFN-γ-expressing NK cells was appreciably reduced when IFN-αβR^−/− DCs were used. Therefore, IFN-αβ stimulation of both DCs and NK cells contributes to the induction of NK cell IFN-γ expression.

To determine whether other cytokines acted together with IFN-αβ in DC-mediated stimulation of NK cell IFN-γ production, various neutralizing Abs were added to DC-NK cell cocultures (Fig. 8B⇑). Abs against the IL-2/15R β-chain (CD122) had no effect when added alone or in combination with other Abs, indicating no role for IL-2 or IL-15. Conversely, although anti-IL-12 had no effect on IFN-γ production when added on its own, diminished IFN-γ production was observed when anti-IL-12 was added in combination with anti-IFN-αβ; this result suggested that a positive effect of IL-12 on IFN-γ production was only evident when IFN-αβ activity was blocked, consistent with an antagonistic relationship between these cytokines, as previously reported (27).

Addition of anti-IL-18 alone substantially inhibited NK cell production of IFN-γ, causing a reduction similar to that mediated by anti-IFN-αβ. IFN-γ production was further reduced when a combination of Abs against IFN-αβ and IL-18 was added to cultures, showing that these two cytokines could act independently of each other. Lastly, anti-IL-12 caused an additional decrease in IFN-γ expression when combined with anti-IFN-αβ and anti-IL-18, but not with anti-IL-18 alone, again compatible with the view that IL-12 contributes to LPS-stimulated DC induction of IFN-γ expression only when IFN-αβ activity is blocked. Nearly complete inhibition of IFN-γ production was observed when Abs against IFN-αβ, IL-18, and IL-12 were added. Taken together, the results show a strong requirement for cytokines even in the cell contact-dependent induction of NK cell IFN-γ expression by LPS-stimulated DCs, with IFN-αβ and IL-18 playing predominant roles.

Finally, for each of the TLR agonists that were shown to induce bystander T cell activation, we investigated the importance of IFN-αβ in the in vivo stimulation of IFN-γ production. Poly I:C, LPS, R848, or CpG ODNs was injected into IFN-αβR^−/− or control mice, and 1 h later splenic CD11c⁺ cells were purified and placed in culture. Supernatants obtained after overnight culture were then assayed for IFN-γ by ELISA (Fig. 9⇓). Equivalent amounts of IFN-γ were produced after injection of CpG ODNs into either WT or IFN-αβR^−/− mice, indicating that this TLR9 agonist was capable of stimulating IFN-γ expression by an IFN-αβ-independent mechanism. Conversely, IFN-γ induction by poly I:C, LPS, and R848 was markedly reduced in IFN-αβR^−/− mice. Poly I:C stimulated only low level IFN-γ production (∼100 pg/ml) in WT mice on a 129 background and completely failed to induce IFN-γ in IFN-αβR^−/− mice. LPS and R848 each stimulated much higher amounts of IFN-γ in WT mice (1–2 ng/ml), but this was reduced by >90% in IFN-αβR^−/− mice. Therefore, in vivo stimulation of NK cell IFN-γ expression by agonists of TLR3, -4, and -7, but not TLR9, is highly dependent on IFN-αβ.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 9.

Contribution of IFN-αβ to TLR agonist-induced NK cell activation in vivo. WT (129) or IFN-αβR^−/− mice were injected i.v. with poly I:C (100 μg), LPS (10 μg), R-848 (25 nmol), or CpG ODNs (10 nmol). One hour after injections, CD11c⁺ cells were purified and placed in culture overnight. Data show the concentrations of IFN-γ in the supernatants as determined by ELISA (mean ± SE for three mice per group).