Constitutive but Not Inducible Attenuation of Transforming Growth Factor β Signaling Increases Natural Killer Cell Responses without Directly Affecting Dendritic Cells Early after Persistent Viral Infection

ABSTRACT

Rapid innate responses to viral encounters are crucial to shaping the outcome of infection, from viral clearance to persistence. Transforming growth factor β (TGF-β) is a potent immune suppressor that is upregulated early upon viral infection and maintained during chronic infections in both mice and humans. However, the role of TGF-β signaling in regulating individual cell types in vivo is still unclear. Using infections with two different persistent viruses, murine cytomegalovirus (MCMV) and lymphocytic choriomeningitis virus (LCMV; Cl13), in their natural rodent host, we observed that TGF-β signaling on dendritic cells (DCs) did not dampen DC maturation or cytokine production in the early stages of chronic infection with either virus in vivo. In contrast, TGF-β signaling prior to (but not during) chronic viral infection directly restricted the natural killer (NK) cell number and effector function. This restriction likely compromised both the early control of and host survival upon MCMV infection but not the long-term control of LCMV infection. These data highlight the context and timing of TGF-β signaling on different innate cells that contribute to the early host response, which ultimately influences the outcome of chronic viral infection in vivo.

IMPORTANCE In vivo host responses to pathogens are complex processes involving the cooperation of many different immune cells migrating to specific tissues over time, but these events cannot be replicated in vitro. Viruses causing chronic infections are able to subvert this immune response and represent a human health burden. Here we used two well-characterized viruses that are able to persist in their natural mouse host to dissect the role of the suppressive molecule TGF-β in dampening host responses to infection in vivo. This report presents information that allows an increased understanding of long-studied TGF-β signaling by examining its direct effect on different immune cells that are activated very early after in vivo viral infection and may aid with the development of new antiviral therapeutic strategies.

INTRODUCTION

The coordinated effort of both the innate and the adaptive immune systems is necessary to control invasive pathogens. Rapid viral dissemination has been associated with weak host antiviral responses and viral persistence in both mice and humans (1, 2). Thus, the magnitude of initial innate responses can be critical to limit the establishment of potentially chronic infections. Human chronic infections with hepatitis C virus (HCV), hepatitis B virus (HBV), human immunodeficiency virus (HIV), and Mycobacterium tuberculosis as well as mouse infections with lymphocytic choriomeningitis virus (LCMV) and the malaria parasite are associated with the increased production and/or signaling of the immunosuppressive molecule transforming growth factor β (TGF-β) in immune cells (3–7). However, the precise role of TGF-β-mediated suppression on innate and adaptive immune components during infection has not been fully deciphered.

Among innate cells, dendritic cells (DCs) and natural killer (NK) cells play pivotal roles in antiviral defense. DCs are a heterogeneous population that can be broadly subdivided into conventional DC (cDC) and plasmacytoid DC (pDC) subsets. cDCs are required to prime antigen-specific T cells, while pDCs are specialized to produce large amounts of type I interferons (IFN-Is), a group of potent antiviral mediators (8–10). On the other hand, NK cells sense high levels of activating receptors and/or decreased levels of inhibitory receptors in virus-infected cells to induce cell death through the release of ready-made cytotoxic granules containing granzymes and perforin (11). DC-derived cytokines drive the maturation of NK cells, including upregulating the transcription factors T-bet and Blimp1 to enhance cytotoxic effector functions during inflammation (12). NK cells can also kill DCs and T cells, to limit antiviral immune responses or pathology, establishing a fluid cross talk with other immune cell populations (13–15).

TGF-β is an important negative regulator of the immune system, as best illustrated by the rapid death of TGF-β-deficient mice due to multiorgan inflammatory disease at 3 to 4 weeks of age (16). T cells are the main mediators of this pathology (17), but TGF-β has been shown to also suppress cells of the innate immune system. For example, autocrine TGF-β is required for the development of tissue-resident DC subtypes, including Langerhans cells and tolerogenic CD103⁺ DCs in the gut (18–20). TGF-β also inhibits IFN-I production by pDCs ex vivo (21) and the expression of costimulatory molecules major histocompatibility complex class II (MHC-II), CD40, and CD80/CD86 in bone marrow (BM)-derived DCs in vitro (22). Notably, DCs from mice infected with a persistent strain of LCMV, strain Cl13, have reduced maturation (23), and pDCs produce substantially reduced amounts of IFN-I compared with the amounts produced by pDCs from noninfected hosts (24), similar to the findings for pDCs from HIV-infected individuals (25). These studies raised the possibility that TGF-β signaling may limit DC maturation and/or cytokine production during in vivo viral infection, attenuating early viral containment and promoting viral persistence.

TGF-β also suppresses NK cell function ex vivo during HBV, HIV, and HCV infection (26–28). Treatment of mice with recombinant TGF-β in vivo can suppress early NK cell activation in response to acute infection with LCMV parental strain Armstong53b (29). Furthermore, Laouar and colleagues used mice expressing a dominant negative TGF-β receptor (dnTGF-βR) transgene in both CD11c⁺ DCs and NK cells (here referred to as CD11c-dnTGF-βRII mice) to show that TGF-β suppresses NK cell maturation in neonatal mice in vivo, resulting in enhanced NK cell numbers both before and after infection with murine cytomegalovirus (MCMV) (30, 31). However, whether TGF-β directly limits innate responses to chronic viral infections in adult mice after infection is unknown.

We therefore set out to determine the role that TGF-β signaling may play on DCs and NK cells during the early response to in vivo chronic viral infections in two animal models. First, we used CD11c-dnTGF-βRII mice to attenuate TGF-β signaling only on DCs and NK cells. Second, we crossed mice with a temporally controlled tamoxifen-responsive ERcre transgene (32, 33) with mice harboring a conditional TGF-βRII flox allele (34) (here referred to as ERcre-TGF-βRII mice). We then evaluated the early innate responses in these genetically modified adult mice following infection with the actively replicating persistent RNA virus LCMV Cl13 and/or the latent DNA virus MCMV. We observed that while the TGF-β signaling blockade in innate cells indirectly affected DC numbers, cytokine production, and maturation, direct TGF-β signaling had minimal or no effects on early DC responses during either MCMV or LCMV infections. In contrast, CD11c-dnTGF-βRII adult mice had increased numbers and differentiation of NK cells before and during both MCMV and LCMV infections. This enhanced NK cell response was accompanied by increased viral control and improved survival upon MCMV (but not LCMV) infection. In contrast, deletion of the TGF-βRII conditional allele (ERcre-TGF-βRII) in adult mice (just prior to infection) did not directly affect either DC or NK cell responses to MCMV. Taken together, these data indicate that while cell-intrinsic TGF-β signaling suppresses the NK cell response before infection, its direct signaling after in vivo infection with either persistent DNA or persistent RNA viruses does not play a major role in regulating DC or NK cell responses.

MATERIALS AND METHODS

Mice and viral stocks.

C57BL/6 (CD45.2), B6.SJL-Ptprc^a Pep3^b/BoyJ (CD45.1), and ERt2-cre mice were purchased from The Jackson Laboratory (Bar Harbor, ME). CD11c-dnTGF-βRII mice on a C57BL/6 background were a kind gift of R. Flavell (Yale University) and were described previously (30). Mice were bred to C57BL/6 or congenic CD45.1 mice and transgene-negative age- and sex-matched littermates, used as wild-type (WT) controls, and maintained as transgene hemizygotes. TGF-βRII^flox/flox mice (35) were generously provided by M. Li (Sloan Kettering) and crossed to hemizygous ERcre mice; both TGF-βRII^flox/flox and ERcre mice had C57BL/6 backgrounds. Chimeras mixed 1:1 were generated by transferring 2 × 10⁶ bone marrow cells of each indicated genotype into irradiated CD45.1 mice, and the cells were allowed to reconstitute for 8 weeks. Tamoxifen (1 mg/day; Sigma) emulsified in sunflower seed oil (Sigma) was injected intraperitoneally (i.p.) for 5 days, and mice were used for experiments 7 days after the last treatment. Mice were maintained in a closed breeding facility in compliance with the requirements of the National Institutes of Health and the Institutional Animal Care and Use Guidelines of the University of California, San Diego. Six- to 12-week-old mice were infected intravenously (i.v.) with 2 × 10⁶ PFU of LCMV Cl13 or 2 × 10⁴ or 2 × 10⁵ PFU of MCMV (Smith strain) i.p. All viruses were grown, identified, and quantified as previously described (36, 37).

Cell purification.

Spleens were treated with collagenase D (1 mg/ml; Roche, Indianapolis, IN) for 20 min at 37°C and were depleted of T and B cells using Thy1.2 and CD19 antibodies and magnetic bead enrichment. Splenic NK cells and pDCs were purified by fluorescence-activated cell sorting (FACS) using a BD Aria apparatus (BD Biosciences, San Jose, CA) for pDCs (CD19⁻ Thy1.2⁻ NK1.1⁻ CD11c⁺ CD11b⁻ B220⁺ PDCA⁺) and NK cells (NK1.1⁺ CD11c⁺ Thy1.2⁻). For TGF-βRII deletion, T cells (Thy1.2⁺), B cells (B220⁺), NK cells (NK1.1⁺ CD11b⁺), monocytes (Thy1.2⁻ B220⁻ NK1.1⁻ CD11b⁺, Ly6C⁺ Gr-1⁻), and neutrophils (Thy1.2⁻ B220⁻ NK1.1⁻ CD11b⁺ Ly6c^lo Gr-1⁺) were sorted from pooled blood by use of a congenic marker (CD45.1 CD45.2). Purity for all cell types was >95%.

Flow cytometry.

The surface and intracellular cytokine staining and gating strategy for DCs was performed as previously described (38). pDCs were gated as CD19⁻ Thy1.2⁻ NK1.1⁻ CD11c⁺ CD11b^lo B220⁺ PDCA1⁺, cDCs were gated as CD19⁻ Thy1.2⁻ NK1.1⁻ CD11c⁺ CD11b^hi/mid B220⁻ CD8α^+/−, and NK cells were gated as CD3⁻ NK1.1⁺ DX5⁺ Ly49H^+/−. In chimeric animals, donor cells were separated by use of the congenic markers CD45.1 and CD45.2, where indicated. Except for antibodies to LIVE/DEAD, which were from Aqua-Invitrogen, and those to TGF-βRII phycoerythrin (PE), which were from R&D Systems, the antibodies to the following were purchased from Ebioscience or BD Pharmingen (San Diego, CA) to stain splenocytes: CD16/CD32 Fc block, CD19 peridinin chlorophyll protein (PerCP) Cy5.5, Thy1.2 PerCP Cy5.5, NK1.1 PE or PE-Texas Red (TR), DX5 PE or Pacific Blue (PB), CD3 PB, granzyme B PE, interferon gamma (IFN-γ) allophycocyanin (APC) or Alexa Fluor 700, Ly49H APC, CD11c APC or PE, PDCA1 fluorescein isothiocyanate (FITC), CD11b PE Cy7 or PB, CD40 PE, PDL-1 PE, B220 Alexa Fluor 780, CD8a Alexa Fluor 700, interleukin-12 (IL-12) Alexa Fluor 647, tumor necrosis factor alpha (TNF-α) eFluor450, CD86 PE Cy7, IAb FITC or PB, KLRG FITC, CD69 PerCP Cy5.5, CD45.1 BV605, and CD45.2 BV650. LCMV-specific tetramers D^b:GP_31–41 (where GP_31–41 represents glycoprotein residues 31 to 41) and I-A^b:GP_67–77 were provided by the NIH Tetramer Core Facility. Samples were acquired on a BD LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar, Inc., Ashland, OR).

Cytokine measurements.

IFN-I bioactivity was measured using L-929 cells transfected with an interferon-sensitive luciferase standardized with recombinant mouse IFN-β (Research Diagnostics, Concord, MA) (39). To measure NK cell and DC intracellular cytokines by flow cytometry, splenocytes were incubated in the presence of 1 μg/ml brefeldin A (BFA; Sigma, St. Louis, MO) without exogenous stimulation for 5 h at 37°C before staining. Secretion of IFN-γ, TNF-α, and IL-2 in LCMV-specific T cells was performed ex vivo using GP_31–41 or GP_67–77 peptide stimulation and BFA.

Quantitative real-time reverse transcription-PCR analysis.

Total RNA was extracted from purified cells using an RNeasy microkit (Qiagen, Valencia, CA), digested with DNase I, and reverse transcribed into cDNA (SuperScript III; Invitrogen, Carlsbad, CA). cDNA quantification was performed using a SYBR green PCR and real-time PCR detection system (Applied Biosystems, Carlsbad, CA), and the quantity was normalized to that of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) RNA. The primers used were gapdh forward (F) (5′-CCAGTATGACTCCACTCACG-3′) and gapdh reverse (R) (5′-GACTCCACGACATACTCAGC-3′), ifn-α4/6 (ifnα) F (5′-TATGTCCTCACAGCCAGCAG-3′) and ifn-α4/6 (ifnα) R (5′-TTCTGCAATGACCTCCATCA-3′), ifn-β F (5′-CTGGCTTCCATCATGAACAA-3′) and ifn-β R (5′-AGAGGGCTGTGGTGGAGAA-3′), tbx21 F (5′-AGCAAGGACGGCGAATGTT-3′) and tbx21 R (5′-GTGGACATATAAGCGGTTCCC-3′), prdm1 F (5′-ACATAGTGAACGACCACCCCTG-3′) and prdm1 R (5′-CTTACCACGCCAATAACCTCTTTG-3′), and tgfbr2 F (5′-ATGTGGAAATGGAAGCCCAGA-3′) and tgfbr2 R (5′-TGCAGGACTTCTGGTTGTCG-3′).

Statistical analysis.

Statistical differences were determined by Student's t test (or paired t test for mixed chimeras) using InStat (version 3.0) software (GraphPad, La Jolla, CA).

RESULTS

Limited DC response to MCMV infection in CD11c-dnTGF-βRII mice.

TGF-β signaling is rapidly activated in many tissues, especially the spleen and liver, within 6 h following Toll-like receptor (TLR) ligation in vivo (40). To examine the role of innate TGF-β signaling during pathogen challenge, we first infected transgenic CD11c-dnTGF-βRII mice, where CD11c⁺ cells have attenuated TGF-β signaling, or littermate controls (WT mice) with MCMV. Early TLR recognition of MCMV by pDCs and subsequent cytokine (i.e., IFN-I and IL-12) production are critical for pathogen control (8); thus, we initially characterized splenic DC populations by flow cytometry at the peak response, which was at 36 h postinfection (p.i.). As reported previously (41), MCMV infection of WT mice resulted in decreased numbers of DCs (Fig. 1A); this decrease, however, was significantly attenuated in infected CD11c-dnTGF-βRII mice, which showed increased numbers of CD11b⁺ cDCs, CD8α⁺ cDCs, and pDCs compared to infected WT mice (Fig. 1A). We next analyzed the ability of DCs to produce cytokines upon MCMV infection. Because pDCs are a major source of IFN-I at 36 h p.i. (42), we analyzed purified splenic pDCs from both WT and CD11c-dnTGF-βRII mice and measured ifn-α4/6 (ifnα) and ifn-β mRNA transcripts by quantitative PCR (qPCR). While pDCs from MCMV-infected WT mice showed high levels of ifn-α and ifn-β compared to the levels on pDCs from naive mice, pDCs from CD11c-dnTGF-βRII mice had significantly reduced levels of IFN-I transcripts (Fig. 1B). In contrast to WT mice, which exhibited high levels of IFN-I activity in serum, CD11c-dnTGF-βRII mice had substantially reduced levels of systemic IFN-I (Fig. 1C).

FIG 1.

Limited DC response to MCMV infection in CD11c-dnTGF-βRII adult mice. CD11c-dnTGF-βRII mice (DN) or WT littermate control mice (WT) were infected i.p. with 2 × 10⁴ PFU MCMV and sacrificed at 36 h p.i. Naive WT mice were processed as controls. (A) Representative gating and percentages of CD11b⁺ cDCs, CD8α⁺ cDCs, and pDCs (top) and total number of DC subpopulations in splenocytes (bottom) analyzed by flow cytometry. (B) RNA from pooled groups of mice sorted for pDCs and analyzed by qPCR for the numbers of ifn-α and ifn-β transcripts relative to the number of GAPDH mRNA transcripts. (C) IFN-I bioactivity from serum obtained at the indicated time points p.i. (D) IL-12 and TNF-α production by spleen DCs after 5 h of culture in BFA. Representative FACS plots gated on total CD11c⁺ DCs (CD19⁻ Thy1.2⁻ NK1.1⁻) as described in the legend to panel A (left) and the total number of DCs plotted for individual mice (right) are shown. (E) MFI of CD86 and PDL-1 expression on the indicated subpopulations of DCs. Data are representative of those from 3 experiments with 3 to 5 mice per group. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

Direct viral recognition of MCMV also induces the production of IL-12 and TNF-α by all DC subsets. Both the proportion and total number of DCs producing IL-12 and/or TNF-α were substantially reduced in CD11c-dnTGF-βRII mice compared to those in WT mice challenged with MCMV (Fig. 1D). All DC subsets from CD11c-dnTGF-βRII mice also displayed reduced levels of the prototypic activation molecules CD86 and programmed cell death ligand 1 (PDL-1) compared to DCs from littermate control mice (Fig. 1E). Importantly, both DC number and activation were similar in CD11c-dnTGF-βRII and WT mice before infection (not shown). Overall these data indicate that there is an attenuated DC response to MCMV infection in the absence of innate TGF-β signaling.

Increased NK cell response during MCMV infection in CD11c-dnTGF-βRII mice.

NK cells are critical for the elimination of MCMV-infected cells. Both IFN-I and IL-12 are important activators of NK cells, with IFN-I driving proliferation and cytotoxicity (e.g., granzyme B upregulation) and IL-12 and IL-18 promoting IFN-γ production (43–45). In C57BL/6 mice, MCMV resistance is dictated by direct recognition of infected cells by Ly49H⁺ NK cells (1, 46, 47); however, early activation is cytokine dependent and nonspecific (48, 49). Given the changes in DC-derived IFN-I and IL-12 that we observed in CD11c-dnTGF-βRII MCMV-infected mice (Fig. 1B and C) and given that CD11c-dnTGF-βRII mice were previously reported to exhibit significantly higher numbers of NK cells before and after MCMV infection (30, 31), we next examined NK cell responses in CD11c-dnTGF-βRII mice infected with MCMV. We observed that, indeed, CD11c-dnTGF-βRII mice had substantially higher numbers of NK cells than the WT controls at 36 h p.i. (Fig. 2A). A large proportion of NK cells from WT mice produced IFN-γ and granzyme B after MCMV infection. In contrast, a smaller fraction of NK cells from CD11c-dnTGF-βRII mice than NK cells from WT mice secreted IFN-γ, and the mean fluorescence intensity (MFI) of IFN-γ staining was also reduced in CD11c-dnTGF-βRII mice. Despite the reduced proportion of IFN-γ-producing NK cells, the total number of IFN-γ-producing NK cells was elevated in CD11c-dnTGF-βRII mice compared to the number in WT mice (Fig. 2B). Both a greater proportion and a greater total number of NK cells from CD11c-dnTGF-βRII mice than NK cells from WT mice produced granzyme B (Fig. 2C). Consistent with the lower serum IFN-I levels, NK cells from CD11c-dnTGF-βRII mice had lower levels of the early activation and IFN-sensitive marker CD69 (50) than NK cells from WT mice (Fig. 2D). Lastly, to examine the differentiation state of NK cells, we purified NK cells at 36 h p.i. and analyzed the transcript levels of T-bet (tbx21) and Blimp1 (prdm1). NK cells from CD11c-dnTGF-βRII mice had increased expression of these transcription factors compared to NK cells from WT mice (Fig. 2E). Overall these data indicate that both NK cell number and differentiation toward terminal effector cells are more pronounced during MCMV infection in the absence of TGF-β signaling. These results are consistent with those from a previous characterization of CD11c-dnTGF-βRII mice, where increased differentiation to mature NK cells expressing CD11b and T-bet was present even before infection (31).

FIG 2.

Increased NK cell response to MCMV infection in CD11c-dnTGF-βRII adult mice. CD11c-dnTGF-βRII mice (DN) or WT littermate control mice (WT) were infected i.p. with 2 × 10⁴ PFU MCMV, and spleen NK cells were analyzed at 36 h p.i. Naive WT mice were processed as controls. (A) Gating of NK cells (left) and total percentage and number of NK cells (DX5⁺ NK1.1⁺ CD3⁻) (right). (B and C) IFN-γ (B) and granzyme B (GrzB) (C) production was measured after 5 h of culture ex vivo in the presence of BFA. Representative FACS plots of IFN-γ or granzyme B production in gated NK cells (left) and plots of the corresponding percentages and numbers of individual mice (right) are shown. (D) A representative histogram (left) depicting the results for WT mice (dark fill), CD11c-dnTGF-βRII mice (black line), and naive mice (light fill) (left) and the average MFI of CD69 on NK cells (right). (E) tbx21 (T-bet) and prdm1 (Blimp1) mRNA expression compared to that of GADPH (×100) from FACS-purified splenic NK cells from infected mice. Data are representative of those from 3 experiments with 3 to 5 mice per group. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

Cell-intrinsic alterations of NK cells, but not DCs, in WT–CD11c-dnTGF-βRII mixed chimeras.

To determine the intrinsic effects of TGF-β signaling on both DC and NK cell responses versus the indirect effects due to changes in the surrounding environment, we generated mixed chimeras by cotransferring bone marrow from WT (CD45.2) and CD11c-dnTGF-βRII (CD45.1) mice into irradiated C57BL/6 WT (CD45.2) mice. At 8 weeks postreconstitution, chimeric mice were infected with MCMV and splenic DCs and NK cells were analyzed at 36 h p.i. The overall representation of WT versus CD11c-dnTGF-βRII mouse CD11b⁺ cDCs and pDCs was unchanged compared to that of total chimeric spleen mononuclear cells (SMCs). However, in this competitive setting, CD11c-dnTGF-βRII mouse donor cells developed into CD8α⁺ DCs with a reduced frequency compared to that for donor cells from their WT counterparts (Fig. 3A).

FIG 3.

Direct TGF-β signaling does not modulate DC responses during MCMV infection. Mixed chimeras of C57BL/6 mouse (WT; CD45.2) and CD11c-dnTGF-βRII mouse (DN; CD45.1) BM cells were transferred into irradiated C57BL/6 mice. Eight weeks later, the mice were infected i.p. with 2 × 10⁴ PFU MCMV, and spleens were analyzed at 36 h p.i. (A) The percent chimerism of DC populations from the CD11c-dnTGF-βRII compartment was compared to that of all splenocytes (SMCs), as determined by flow cytometry. (B) Donor RNA from sorted pDCs analyzed for the amounts ifn-α4 and ifn-β transcripts by qPCR and graphed relative to the levels of GADPH mRNA expression. (C) IL-12 production by total CD11c⁺ (CD19⁻ Thy1.2⁻ NK1.1⁻) cells after 5 h of ex vivo culture in the presence of BFA. Representative FACS plots of total CD11c⁺ DCs (left) and percentages plotted for individual mice (right) are shown. (D and E) CD86 and PDL-1 expression on naïve (D) and MCMV-infected (E) total CD11c⁺ cells shown as representative histograms (left) with WT (gray fill) or CD11c-dnTGF-βRII (black line) compartments and the MFI for individual mice (right) are shown. Data are representative of those from 3 experiments with 4 mice per group.

To measure inflammatory responses by DCs, both WT and CD11c-dnTGF-βRII pDCs and cDCs were analyzed. In contrast to pDCs from nonchimeric CD11c-dnTGF-βRII mice (Fig. 1B), WT and CD11c-dnTGF-βRII derived pDCs had similar levels of ifn-α and ifn-β transcripts following MCMV infection (Fig. 3B). Similarly, both WT and CD11c-dnTGF-βRII derived DCs produced similar amounts of IL-12 (Fig. 3C). Furthermore, the levels of expression of the costimulatory molecules CD86 and PDL-1 were similarly low before infection and the molecules were equally upregulated after infection in both WT and CD11c-dnTGF-βRII derived DCs (Fig. 3D and E). These results suggest that TGF-β signaling on DCs does not directly influence acute cytokine production or costimulatory molecule expression in vivo upon MCMV challenge and suggest that the reductions observed in nonchimeric CD11c-dnTGF-βRII mice (Fig. 1) were a result of changes in the environment surrounding the DCs, presumably increased viral control by NK cells.

In contrast to the findings for DCs and consistent with the findings of recent work (31), NK cells in the CD11c-dnTGF-βRII compartment were overrepresented in proportion to the representation of NK cells in the blood from the WT compartment both before and after infection, whereas the proportions of T cells from each of the WT or CD11c-dnTGF-βRII compartments were normal (Fig. 4A). Before infection, neither WT nor CD11c-dnTGF-βRII derived NK cells spontaneously produced IFN-γ after 5 h of in vitro culture; however, CD11c-dnTGF-βRII derived NK cells did show enhanced granzyme B expression (Fig. 4B). Similar to the findings for NK cells from nonchimeric CD11c-dnTGF-βRII mice (Fig. 2), NK cells derived from CD11c-dnTGF-βRII donor mice produced slightly less IFN-γ and maintained higher levels of granzyme B on a per cell basis than NK cells from WT mice after MCMV challenge (Fig. 4C). CD11c-dnTGF-βRII derived NK cells also had slightly increased levels of CD69 expression prior to infection (Fig. 4D) that changed to slightly reduced levels of CD69 expression compared to the level of CD69 expression by NK cells from WT mice after MCMV infection (Fig. 4E). Lastly, we analyzed the levels of tbx21 and prdm1 to test whether intrinsic TGF-β signaling altered effector differentiation of NK cells. Indeed, CD11c-dnTGF-βRII derived NK cells had higher levels of these transcripts than WT derived NK cells after MCMV infection (Fig. 4F). Together, the aforementioned results suggest that TGF-β signaling had a cell-intrinsic effect on NK cells, enhancing their numbers and effector functions before MCMV infection, as previously described (31), while indirectly affecting DC numbers and function.

FIG 4.

Cell-intrinsic NK cell alteration in WT–CD11c-dnTGF-βRII mixed chimeras. Mixed chimeras of C57BL/6 mouse (WT; CD45.2) and CD11c-dnTGF-βRII mouse (DN; CD45.1) BM cells transferred into irradiated C57BL/6 mice. Eight weeks later, mice were infected i.p. with 2 × 10⁴ PFU MCMV, and splenic NK cells were analyzed at 36 h p.i. (A) The percent chimerism of each compartment in NK cells (DX5⁺ NK1.1⁺ CD3⁻) in the blood before and after infection was determined by flow cytometry (left) and compared to that in T cells (right). (B and C) IFN-γ and granzyme B production by NK cells after 5 h of ex vivo culture with BFA in naive (B) or MCMV-infected (C) animals. Representative FACS plots of IFN-γ or granzyme B production in gated NK cells (left) and the corresponding percentages plotted for individual mice (right) are shown. (D and E) CD69 expression on NK cells in naïve (D) and MCMV-infected (E) animals depicted as a histogram (left) of the WT compartment (gray fill) and the dnTGF-βRII compartment (black line) and the MFI for individual mice (right). (F) Levels of tbx21 (T-bet) and prdm1 (Blimp1) mRNA expression compared to the level of GAPDH mRNA (×100) from FACS-purified splenic NK cells (CD3⁻ Nk1.1⁺ DX5⁺) during MCMV infection. Data are representative of those from 2 to 3 experiments with 4 mice per group. ns, no significant difference; *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

Increased NK cell (but not DC) responses during LCMV Cl13 infection in CD11c-dnTGF-βRII mice.

To compare and contrast the responses of DCs and NK cells from CD11c-dnTGF-βRII mice upon infection with a persistent virus other than MCMV, we inoculated mice with LCMV Cl13. In contrast to the rapid control of replicating MCMV by NK cells (which precedes MCMV latency), LCMV is not highly susceptible to NK cell-mediated clearance and can persist in a continuously replicating state for 60 to 200 days p.i. in multiple tissues (37, 51). However, NK cells have recently been shown to limit adaptive immune responses early during LCMV infection to promote viral persistence or protect against immune pathology (13, 15).

LCMV induces a potent IFN-I response that is dependent on many innate cells, including pDCs (10, 52, 53), where pDCs ultimately contribute to late LCMV control (54, 55). We observed no difference in the number of pDCs or cDCs derived from the WT or CD11c-dnTGF-βRII compartment at 1 to 3 days p.i. (data not shown). Furthermore, we found that WT mice and CD11c-dnTGF-βRII mice exhibited similarly high levels of IFN-I in the blood 24 h after LCMV infection (Fig. 5A). IFN-I is also a main contributor to cDC costimulatory molecule upregulation by 3 days p.i. (56). Accordingly, the levels of CD86 and CD40 expression in DC subsets from CD11c-dnTGF-βRII mice were unchanged from those in WT mice (Fig. 5B). We next measured NK cell number and activity at 1 day p.i. and found that CD11c-dnTGF-βRII mice had increased NK cell numbers compared to WT mice (Fig. 5C); however, the proportions of NK cells producing IFN-γ and granzyme B were similar (Fig. 5D). Despite similar systemic IFN-I levels in the absence of innate TGF-β signaling, NK cells from CD11c-dnTGF-βRII mice had slightly decreased amounts of CD69 compared to NK cells from WT mice following infection (Fig. 5E). Both CD4 and CD8 T cells are critical to control persistent LCMV Cl13 infection, although they become depleted and functionally exhausted, producing low levels of antiviral cytokines (57, 58). To assess the potential impact of innate TGF-β signaling on antiviral T cell responses, we next examined the magnitude and function of immunodominant LCMV-specific T cells in CD11c-dnTGF-βRII mice at day 9 p.i. We found similar numbers of antiviral CD8 D^b:GP_33–41 T cells in WT and CD11c-dnTGF-βRII mice; however, CD11c-dnTGF-βRII mice exhibited fewer LCMV-specific CD4 T cells, as determined by I-A^b:GP_67–77 tetramer staining (Fig. 5F). The secretion of IFN-γ, TNF-α, and IL-2 upon ex vivo GP_33–41 or GP_61–80 peptide stimulation was similar in CD8 and CD4 T cells from WT and CD11c-dnTGF-βRII mice (Fig. 5G). These results indicate that TGF-β signaling influences NK cells but not DCs, therefore limiting NK cell early responses during chronic LCMV infection, and this may in turn affect T cell responses, as previously described (13, 15).

FIG 5.

Increased NK cell (but not DC) responses in LCMV-infected CD11c-dnTGF-βRII mice. CD11c-dnTGF-βRII mice or littermate controls (WT) were infected with 2 × 10⁶ PFU of LCMV Cl13. (A) Serum IFN-β levels at 1 day postinfection (1 d.p.i.). (B) CD86 (left) and CD40 (right) expression (MFI) in DC subsets was determined by flow cytometry at 3 days p.i. (C) Total number of splenic NK cells at 1 day p.i. (D) IFN-γ and granzyme B production after 5 h of culture in BFA at 1 day p.i. Representative FACS plots gated on total NK cells (left) and the percentage and total number of NK cells plotted for individual mice (right) are shown. (E) CD69 expression on NK cells at 1 day p.i. (F) Total number of D^b:GP_33–41 and I-A^b:GP_61–77 virus-specific CD8 and CD4 T cell responses at 9 days p.i. (9 d.p.i.) in the spleen. (G) Cytokine production as a percentage of virus-specific cells from the assay whose results are presented in panel F upon cognate GP_33–41 or GP_61–80 peptide stimulation in the presence of BFA. Data are representative of those from 2 to 4 experiments with 3 to 5 mice per group. *, P < 0.05; **, P < 0.005.

Enhanced control of MCMV but not LCMV Cl13 loads in CD11c-dnTGF-βRII mice.

Infant CD11c-dnTGF-βRII mice survived an otherwise lethal MCMV challenge, presumably due to the enhanced NK cell maturity that is present in the first week after birth (31). In the next series of experiments, we tested whether this observation could be extended to adult mice. We observed that CD11c-dnTGF-βRII mice were better able to control acute MCMV replication in the liver than their WT counterparts by 3 days p.i. (Fig. 6A). CD11c-dnTGF-βRII mice were consistently protected from a 50% lethal dose of MCMV that was otherwise lethal for WT C57BL/6 mice (Fig. 6B). In contrast, CD11c-dnTGF-βRII mice showed viral loads in the liver similar to those of WT mice at day 3 after LCMV Cl13 challenge (Fig. 6C). Moreover, the levels of viremia that persisted for 60 days in CD11c-dnTGF-βRII mice were comparable to or slightly increased compared with those in WT mice when mice were challenged with LCMV Cl13 (Fig. 6D). These results suggest that heightened NK cell responses due to innate TGF-β blockade likely have the capacity to limit acute viral replication and enhance survival when adult mice are challenged with MCMV but that this is insufficient to influence the course of chronic LCMV infection.

FIG 6.

Enhanced control of MCMV but not LCMV Cl13 loads in CD11c-dnTGF-βRII mice. CD11c-dnTGF-βRII mice (DN) or WT littermate control mice (WT) were infected with 2 × 10⁴ (A) or 2 × 10⁵ (B) PFU MCMV or with 2 × 10⁶ PFU LCMV Cl13 (C and D). (A and C) Viral titers in liver homogenate measured by plaque assay at day 3 p.i. (B) Percentage of surviving mice at the indicated times. Pooled results from 2 experiments with 12 WT mice and 10 CD11c-dnTGF-βRII mice are shown. (D) Viremia determined by plaque assay at the indicated time points. Dotted line, limit of assay detection (200 PFU). Data are representative of those from 2 to 3 experiments with 3 to 5 mice per group. **, P < 0.005.

Cell-intrinsic TGF-βRII deletion in adult mice does not influence DC responses upon MCMV infection.

To assess the role of TGF-β signaling during infection (independently of its developmental roles [31]) and to rule out putative dominant effects of the dnTGF-βRII transgene recently reported in T cells (59), we used ERcre-TGF-βRII mice in which TGF-βRII is ubiquitously deleted upon tamoxifen treatment (32, 60). To analyze the cell-intrinsic effect of TGF-β signaling in DCs and NK cells in vivo (independently of TGF-β signaling on other cells), we created mixed 1:1 ERcre-TGF-βRII BM–CD45.1 WT BM chimeras. At 8 weeks postreconstitution, the mice were treated with tamoxifen for 5 days and gene deletion was confirmed 5 days later in blood leukocytes (Fig. 7A and B). While TGF-βRII protein expression was readily detected by flow cytometry in WT (but not ERcre-TGF-βRII) T cells and B cells, we could not detect minimal surface TGF-βRII expression in NK cells, monocytes, or neutrophils from mice with either the WT or ERcre-TGF-βRII origin (Fig. 7A). However, we did find tgfbr2 mRNA in WT mouse NK cells and monocytes, and this was dramatically reduced to the limit of detection in ERcre-TGF-βRII donor cells (Fig. 7B). These data indicate that 5 days of tamoxifen treatment effectively deleted TGF-βRII from ERcre-TGF-βRII donor leukocytes. Therefore, we next infected the tamoxifen-treated mixed ERcre-TGF-βRII BM–WT BM chimeras with MCMV and analyzed the splenic DC responses at 36 h p.i. Unlike the constitutive expression of dnTGF-βRII, where CD11c-dnTGF-βRII mice had reduced CD8α⁺ DC differentiation in a competitive setting (Fig. 3A), all ERcre-TGF-βRII derived DC subsets had representation similar to that of WT derived DC subsets in the mixed ERcre-TGF-βRII WT BM chimeras (Fig. 7C). In addition, DCs from both WT and TGF-βRII-deficient donors produced equivalent amounts of IL-12p40 (Fig. 7D) and had comparable levels of upregulation of the I-A^b, CD86, and CD40 molecules after MCMV infection (Fig. 7E). These data further support our conclusions from assays with CD11c-dnTGF-βRII mice indicating that TGF-β signaling does not affect early DC responses to viral infections in vivo.

FIG 7.

Inducible deletion of TGF-βRII in adult mice does not affect DC responses to MCMV. A 1:1 mix of BM from C57BL/6 mice (WT; CD45.1) and ERcre-TGF-βRII mcie (RII^flox; CD45.2) was transferred into irradiated CD45.1 hosts. Eight weeks later, 1 mg/day of tamoxifen (TAM) was injected i.p. for 5 days to induce Cre activity. Five days later, blood leukocytes were analyzed for receptor expression (A to C); mice were then infected with 2 × 10⁴ PFU of MCMV, and splenic DCs were studied by flow cytometry at 36 h p.i. (D and E). (A) TGF-βRII protein expression in the indicated leukocyte populations was determined by flow cytometry (left), and the MFI after tamoxifen treatment is shown (right). (B) tgfbr2 mRNA levels relative to GAPDH mRNA levels in FACS-purified leukocyte populations were quantified by qPCR. Numbers in boxes indicate the fold change in expression. QTY, quantity. (C) Percent chimerism of the indicated DC subsets from each compartment. (D) IL-12 production by DCs after 5 h of culture in BFA. A representative FACS plot gated on total CD11c⁺ DCs (left) and the percentages of IL-12-positive DCs for individual mice (right) are shown. (E) I-A^b, CD86, and CD40 expression on DC subsets. (Top) Representative histograms for CD11b⁺ DCs with WT (gray fill) and ERcre-TGF-βRII (black line) overlays are shown. (Bottom) The MFI for each molecule in the indicated DC subsets is plotted for individual mice. Data are representative of those from 2 experiments with 4 mice per group. *, P < 0.05; ***, P < 0.0005.

Cell-intrinsic TGF-βRII deletion in adult mice does not affect NK cell responses upon MCMV infection.

In the next series of experiments, we measured early NK cell responses to MCMV infection in the mixed ERcre-TGF-βRII WT bone marrow chimeras described above. We first observed similar proportions of WT and ERcre-TGF-βRII donor-derived NK cells in the blood before and after infection, indicating no preferential expansion without TGF-β signaling (Fig. 8A). We further found that the proportion of NK cells from the ERcre-TGF-βRII compartment in spleen and liver after infection was similar to that in the blood preinfection (Fig. 8B). Early activation at 36 h p.i. is cytokine dependent and independent of the Ly49H response (48); however, we further found that the proportion of blood Ly49H⁺ NK cells increased during infection in both WT and ERcre-TGF-βRII derived NK cells and that this increase was only slightly greater in NK cells from ERcre-TGF-βRII mice than in NK cells from WT mice at day 6 p.i. (but not day 1.5 p.i.) relative to their corresponding starting points (Fig. 8C), indicating little MCMV-specific repression of Ly49H⁺ NK cells by TGF-β signaling. We next analyzed spleen NK cell cytokine production in WT ERcre-TGF-βRII bone marrow chimeras at 36 h p.i. We observed that NK cells from the ERcre-TGF-βRII compartment had levels of production of IFN-γ and granzyme B equivalent to those of NK cells from the WT compartment (Fig. 8D). Given that NK cells are recruited to the liver rapidly after MCMV infection (61, 62), we next examined NK cell function in the livers of infected WT ERcre-TGF-βRII bone marrow chimeras at 36 h p.i. Again, we found that NK cells from WT and TGF-βRII-deficient mice exhibited similar levels of IFN-γ and granzyme B expression in the liver (Fig. 8E). Furthermore, we measured the expression of CD69 in splenic NK cells and found that the acute deletion of TGF-βRII did not affect CD69 upregulation after infection (Fig. 8F). Finally, T-bet expression, which was intrinsically upregulated in CD11c-dnTGF-βRII derived NK cells (Fig. 4D), was unaltered in splenic ERcre-TGF-βRII derived NK cells compared to T-bet expression in NK cells from their WT counterparts (Fig. 8G). Overall, these data indicate that (in contrast to the constitutive expression of dnTGF-βRII) the inducible deletion of TGF-βRII in adult mice does not directly affect early NK cell responses upon MCMV infection.

FIG 8.

Inducible deletion of TGF-βRII in adult mice does not affect NK cell responses to MCMV. A 1:1 mix of BM from C57BL/6 mice (WT; CD45.1) and ERcre-TGF-βRII mice (RII^flox; CD45.2) was transferred into irradiated CD45.1 hosts. Eight weeks later, 1 mg/day of tamoxifen (TAM) was injected i.p. for 5 days; mice were then infected with 2 × 10⁴ PFU of MCMV, and NK cells were studied by flow cytometry. (A and B) Chimerism of NK cells in the blood during infection (A) and in tissues at 36 h p.i. (B). (C) Percentage of Ly49H⁺ cells within blood NK cells. Representative FACS plots (top) and average values at the indicated time points (bottom) are shown. (D and E) IFN-γ and granzyme B production in spleen (D) and liver (E) NK cells after 5 h of culture with BFA. Representative FACS plots gated on total NK cells (left) and percentages plotted for individual mice (right) are shown. (F and G) CD69 (F) and T-bet (G) expression on NK cells depicted as a histogram (left) for the WT compartment (black line) and ERcre-TGF-βRII compartment (gray line) upregulated over naive NK cells as a reference (gray fill) and the MFI for individual mice (right). Data are representative of those from 2 experiments with 4 mice per group.

DISCUSSION

The early presence of host resistance factors can influence the balance between host and microbe to favor rapid pathogen eradication. For example, a strong innate response via Ly49H⁺ NK cells controls MCMV in C57BL/6 mice, whereas susceptible BALB/c mice, which lack a protective MCMV-specific Ly49H⁺ haplotype, demonstrate a higher mortality rate or establish more latent virus, depending on the inoculation dose (1, 63). Similarly rapid and robust adaptive CD8 T cell responses correlate with improved outcomes of LCMV infection in mice and simian immunodeficiency virus (SIV) infection in macaques (2). TGF-β is a major immunosuppressive cytokine that limits innate (DC and NK cell) and adaptive immunity in vitro and in vivo in different systems, with high levels of TGF-β correlating with enhanced susceptibility to virus and viral persistence for HCV in humans and SIV in rhesus macaques (64, 65) as well as LCMV and the malaria parasite in mice (6, 7).

In this study, we show that while constitutive genetic attenuation of TGF-β signaling in innate cells (via expression of the dnTGF-βRII transgene in CD11c⁺ cells) did not directly affect key DC functions, it resulted in increased NK cell responses before and after infection and therefore promoted the early control of MCMV but not LCMV chronic infections in vivo. In contrast, genetic deletion of TGF-βRII signaling in adult mice (just before infection) did not change early innate responses to MCMV infection. Our findings therefore support two conclusions. First, despite previous evidence that TGF-β suppresses DCs in vitro (21, 22) and in inflammatory disease models (18, 66–68), cell-intrinsic TGF-β signaling does not modulate key DC responses early after infection with potentially persistent viruses in vivo. Second, that cell-intrinsic TGF-β signaling on NK cells before (but not after) viral infection limits their early responses in vivo.

By using either CD11c-dnTGF-βRII mice or inducible TGF-βRII deletion in adult mice, our results indicate that TGF-β does not directly affect early DC numbers, cytokine production, or maturation in spleens after viral infection. This is true for both LCMV and MCMV infections, regardless of the different route and doses of inoculation, the RNA versus DNA nature of the viral genomes, and the different pathogen recognition receptors activated upon these infections (10). Our findings are unexpected, given the numerous previous reports supporting a role for TGF-β regulation of DCs in different settings. In vitro TGF-β can suppress pDC IFN-I production (42, 54) and cDC maturation and cytokine production (22). In addition, a number of studies using different models support an in vivo role for TGF-β on DCs. For instance, in long-term chronic inflammatory models using CD11c-dnTGF-βRII mice, such as experimental autoimmune encephalitis (68) and atherosclerosis (69), innate TGF-β signaling suppresses disease. Blockade of TGF-β on innate cells also decreases latent herpes simplex virus 1 (70). Ramalingam et al. recently showed that CD11c-Cre-mediated DC deletion of TGF-βRII causes increased IFN-γ and TNF-α levels in DCs in long-term spontaneous autoimmunity but did not affect steady-state MHC-II costimulatory expression and caused no apparent increase in NK cell numbers (71), highlighting the differences in genetic models as well as infectious versus noninfectious contexts.

pDCs and IFN-I are essential for the control of MCMV (42, 54), and IFN-I induction by pDCs is tightly linked to the presence of virus and TLR9 activation in this infection (72). Previous work by Andrews et al. has shown that a protective haplotype of Ly49H in NK cell responses limits the pDC cytokine response during MCMV infection, clearly indicating feedback between NK cell and DC activation (63). This seems to be mediated both indirectly though viral control, reducing TLR9 stimulation, and directly by DC killing by NK cells. Thus, the limited DC activation and IFN-I production that we saw in MCMV-infected CD11c-dnTGF-βRII mice was most likely due to initial enhanced viral control by NK cells, as assays with mixed chimeras showed that DCs with attenuated TGF-β signaling had the same activation potential and cytokine production as WT DCs when they were in the same environment in vivo. Consistently, in LCMV Cl13 infection, where viral loads are not altered in CD11c-dnTGF-βRII mice, we did not detect differences in IFN-I or DC activation early after infection.

DC-derived IL-12 and IFN-I are potent inducers of IFN-γ and granzyme B in NK cells, respectively (10, 44), alongside MCMV m157-Ly49H recognition in C57BL/6 mice (46). In this study, the constitutive attenuation of intrinsic TGF-β signaling on NK cells resulted in increased granzyme B expression and decreased IFN-γ and CD69 expression. Consistently, Laouar and colleagues have recently shown that TGF-β signaling suppresses the number of NK cells and the maturation toward terminally differentiated NK cells without altering the distribution of Ly49H, Ly49C/I, Ly49D, or NKG2D and that infant CD11c-dnTGF-βRII mice are resistant to MCMV for this reason (31). Terminally differentiated NK cells express the most granzyme B and Blimp1 (12). In line with this, we found that CD11c-dnTGF-βRII NK cells were overrepresented in competition with WT cells and had higher levels of granzyme B even before infection. CD11c-dnTGF-βRII mouse NK cells maintained higher levels of T-bet, Blimp1, and granzyme B expression after MCMV infection, confirming the presence of a more mature phenotype. Furthermore, we also observed enhanced MCMV resistance in adult CD11c-dnTGF-βRII mice likely resulting from an increased NK cell response.

Interestingly, even in the presence of this heightened NK cell frequency and effector function, LCMV Cl13 was able to rapidly establish infection, indicating the context-specific capacity of NK cells to control viral loads. Recent reports demonstrated that NK cells are capable of directly killing CD4 and CD8 T cells during LCMV Cl13 infection to limit fatal immune pathology and promote viral persistence (13, 15). In line with these studies, we did observe reduced virus-specific CD4 T cells in CD11c-dnTGF-βRII mice at day 9 p.i.; however, the enhanced NK cell numbers in CD11c-dnTGF-βRII mice appears to be insufficient to alter CD8 T cell responses or the ultimate outcome of chronic LCMV infection. It should be noted, however, that the dominant negative TGF-βRII transgene has recently been shown to enhance T cell homeostatic proliferation and activation when a CD4 promoter drives expression even in the absence of the normal TGF-βRII (59). Therefore, the possibility that the TGF-βRII transgene driven by the CD11c promoter has off-target effects in CD11c-dnTGF-βRII mice cannot be completely ruled out and represents a potential limitation of this model.

To assess the role of direct TGF-β signaling during infection independently of the developmental effects of TGF-β in NK cells and the putative off-target effects of the dnTGFβRII transgene, we utilized mixed ERcre-TGF-βRII WT bone marrow chimeras in which TGF-βRII was deleted in adult mice just before MCMV infection. In this system, we observed minimal differences in MCMV-specific spleen or liver NK cell accumulation, IFN-γ and granzyme B secretion, and CD69 or T-bet expression between WT cells and cells in which TGF-βRII was deleted. This is in contrast to the findings of a previous study showing that injection of recombinant TGF-β suppresses NK cell numbers during acute LCMV infection (29), possibly due to differences between endogenous and inoculated TGF-β doses and target cells. Furthermore, the absence of an endogenous TGF-β effect on NK cells after MCMV infection may result, at least in part, from the low level of TGF-βR expression that we observed in WT NK cells.

When the findings of this study are taken together, our study supports a role for TGF-β during the development of NK cells, as previously proposed (31), and contrasts this TGF-β function with no apparent direct effect on DCs. Furthermore, neither mature NK cells nor DCs in the periphery of adult mice appear to be susceptible to TGF-β-mediated suppression, as the limitation of cell-intrinsic TGF-β signaling did not affect their subsequent early antiviral responses. However, TGF-β is known for its context-dependent effects, and therefore, signaling on innate cells could have different outcomes during tissue-specific infections, in tumor models, or in long-term chronic inflammatory environments.