Inhibition of NK cell activity by IL-17 allows vaccinia virus to induce severe skin lesions in a mouse model of eczema vaccinatum

We initially focused on establishing experimental conditions in which infection with VACV induces differential clinical outcomes between mice with and without eczematous skin lesions. Skin lesions were induced on the backs of dermatitis-prone NC/Nga mice (10) by epicutaneous treatment of shaved skin with a mite extract and staphylococcal enterotoxin B (SEB), as described previously (11). This treatment induced elevated serum IgE levels and eczematous skin lesions (Fig. 1 A) (11). Skin lesions with maculopapular rash started to appear on the infected site on day 2–3 after infection in eczematous mice and developed into severe skin erosion. The size of the primary lesion peaked at days 7–8 (Fig. 1, B and C), and the lesion began to subside by day 11. Unlike eczematous mice, most normal mice failed to develop skin lesions after VACV infection, and even when developed, their skin lesions were much milder (Fig. 1, B and C). Virus titers in the lesional skin of eczematous mice were 300–10,000 times higher than those of normal mice over an observation period of 14 d (Fig. 1 D). In erosive skin lesions of eczematous mice, epithelial layers were separated from the rest of the skin and more leukocytes infiltrated the diseased dermis (Fig. 2, A and B). Pock-like satellite lesions distant from inoculation sites were rarely seen (only 3 cases out of 230 eczematous mice and 0 out of 187 normal mice). Although weight loss was observed in a small number of both eczematous and normal mice, there was no correlation with skin conditions (unpublished data). Unlike the intradermal infection at eczematous skin lesions, intranasal infection or intradermal infection at distant normal skin sites failed to induce clinical conditions (e.g., weight loss, survival, and size of skin lesions) distinctly different between eczematous and normal mice (unpublished data). Unlike the Western Reserve strain used throughout this study, intradermal infection with the same dose of ACAM2000, the licensed vaccine cloned from Dryvax, caused much milder skin lesions compared with Western Reserve–induced skin lesions (unpublished data).

Because of the importance of NK cells in rapid antiviral defense (12, 13), we quantified their numbers and activities. NK cells were more abundant in primary skin lesions in eczematous than normal mice (Fig. 2, C and D). Importantly, NK cell cytotoxic activity in the spleen was lower in eczematous mice on days 2 and 3 after infection (Fig. 3 A). We measured expression of molecules involved in NK killing activity by flow cytometry and found that the proportions of splenic NK cells expressing granzyme B, perforin, and IFN-γ were significantly lower in eczematous mice (Fig. 3 B).

IgM and IgG responses against VACV were similar between the eczematous and normal cohorts (Fig. S1 A). Consistent with this, IL-4 mRNA levels in lymph nodes were not reduced in eczematous mice for the initial 7 d after infection (Fig. S1 B). Killing activity of CD8⁺ T cells and their expression of granzyme B, perforin, and IFN-γ in day 7 spleens did not show differences between the two cohorts (unpublished data). These results suggest that adaptive immunity does not play a major role in causing differential skin outcomes of VACV infection between the eczematous and normal mice, although these arms of immunity are critical in the control of virus infection in vaccinia-infected mice (8).

The role of NK cells in this eczema vaccinatum model was assessed by depletion studies. First, dermatitis was induced in NC/Nga mice. 1 d before infection and on d 3 after infection, mice were intravenously injected with anti–asialo GM1 (αAGM1) or control rabbit serum (NRS). Treatment with αAGM1 serum drastically reduced the numbers of NK1.1⁺ cells in the spleens (73–89% reduced as evaluated by flow cytometry) and suppressed NK cell activity in spleens in day 3–infected normal mice (normal/αAGM1 group) compared with NRS-treated normal mice (normal/NRS group; Fig. S2). In contrast, αAGM1 treatment did not significantly reduce the already low NK cell activity in day 3–infected eczematous mice. Substantially higher virus titers were observed in lesional skins (Fig. 3 C) and lungs (not depicted) of αAGM1-treated normal mice than those of NRS-treated normal mice. Importantly, 14 out of 16 mice in the normal/αAGM1 group exhibited erythematous papules at the inoculation site by day 6, whereas only 1 out of 13 mice in the normal/NRS group developed such a lesion. Some normal/αAGM1 and eczematous/αAGM1 mice developed satellite lesions as well (Fig. 3 E). Eczematous mice developed larger erosive skin lesions at the site of virus inoculation than noneczematous mice (Fig. 3 D). These primary lesions in NK-depleted eczematous mice were significantly larger than lesions in control eczematous mice (Fig. 3 D). As αAGM1 treatment might affect other cell types besides NK cells (14), we performed a second experiment in which we depleted NK cells by administering anti-NK1.1 mAb. Results were similar to those with αAGM1 (Fig. S3).

To complement the NK depletion experiments, we performed adoptive transfer of NK cells to determine whether activated NK cells could rescue NC/Nga mice from eczema vaccinatum. NK cells were obtained by culturing splenocytes in IL-15 for 4 d. The cultured cells, composed of a >95% CD3⁻ NK1.1⁺ population (Fig. 3 F, inset), were intravenously transferred to eczematous or normal mice. Transfer of NK cells either totally suppressed the development of erosive skin lesions or greatly reduced skin lesion sizes (Fig. 3 F). The activated NK cells also delayed the kinetics of lesion development in the subset of mice that eventually developed erosive skin lesions. Therefore, the NK depletion and transfer experiments demonstrate a critical role for NK cells in protecting mice from developing VACV-induced erosive skin lesions and satellite skin lesions in this NC/Nga mouse model of eczema vaccinatum.

NK cell function is under the control of various cytokines, including IL-6 and IL-10, which each inhibit NK cell activity. The proinflammatory cytokine IL-17 is produced by Th17 cells (15). In eczematous mice, real-time PCR analysis of splenocytes showed increased mRNA expression of IL-17A and the cytokines involved in Th17 development (IL-6, TGF-β, IL-21, and IL-23) and effector functions (IL-21 and IL-22) (Fig. 4 A). IL-17A and IL-6 mRNAs were also increased in lesional skins of uninfected eczematous mice, whereas IL-17A, IL-6, and IL-23 mRNAs were increased in draining lymph nodes of eczematous mice (Fig. 4 A). Consistent with these mRNA results, lymph nodes contained an increased number of Th17 cells in eczematous mice (Fig. 4 B). In contrast with Th17-related cytokines, surface expression of NK cell receptors such as NKG2D, NKG2A/C/E, Ly49A/D, and Ly49I/G was comparable in eczematous and normal mice (unpublished data).

Neutralization of IL-17A in eczematous mice with anti–IL-17 antibody caused a delay in the onset of skin lesions after virus infection, and the lesion size was significantly smaller on days 2 and 3 after infection (Fig. 4 C). Although the number and the percentage of NK cells in the spleen and at the lesion site were not changed by anti–IL-17 antibody treatment (Fig. 4 D), the proportions of NK cells expressing granzyme B, perforin, and IFN-γ were increased in IL-17–neutralized mice (Fig. 4 E). Consistent with these changes, viral loads in the spleen and lesional skin were lower in IL-17–neutralized mice (Fig. 4 F). Furthermore, when the NK cells were depleted by αAGM1 antibody, the effect of anti–IL-17 antibody treatment on the incidence and lesion size (Fig. 4 G) and viral titers (Fig. S4) was almost abrogated, indicating that effects of IL-17 neutralization are exerted through the regulation of NK cells. Consistent with these in vivo findings, the expression of killing effector molecules in cultured splenic NK cells was reduced by IL-17A in a dose-dependent manner (Fig. S5), but not by IL-17F (Fig. 5). IL-17A reduced the expression of killing effectors induced by IL-4 (Fig. 5), IL-2, IL-12, IL-15, or IL-18 (Fig. S6). The survival of these NK cells was not affected by IL-17A or IL-17F (unpublished data). These results collectively suggest that IL-17A plays a critical role in lowering NK cell activity in eczematous mice.

IL-15 is required for the proliferation and activation of NK cells (16). Antibody-mediated neutralization of IL-15 caused more severe skin lesions in VACV-infected normal mice compared with the control cohort (Fig. S7). However, IL-15 neutralization in eczematous mice did not induce significant differences in skin lesion development. Although the mRNA level of IL-15 is not significantly different between normal and eczematous mice (unpublished data), the results of IL-15 neutralization further confirm that NK cell activity is critical for early protection from skin lesion development.

Our NC/Nga infection model does not exhibit all of the expected features of human eczema vaccinatum. For instance, NC/Nga mice with eczematous skin lesions exhibited functional but not numerical defects in NK cells, unlike atopic dermatitis patients, who have defects in both number and function (17, 18). Nevertheless, this model exhibits key features of atopic dermatitis observed in humans, including defective NK cell killing activity (17, 18) and high IL-17A expression (5, 6). IL-6 and TGF-β are required for induction of Th17 cells, and IL-23 is required for the establishment of Th17 cells (19, 20). IL-21 is produced by Th17 cells and exerts critical functions in Th17 cell differentiation (21–23). Th17 cells were more abundant and the Th17-related cytokines were increased in eczematous mice, suggesting that Th17 cells may be involved in reducing NK cell killing activity. The NK cell–suppressive function of IL-17A observed in our in vitro and in vivo studies was consistent with an earlier IL-17 study (24), although it is possible that the increased IL-17A and Th17-related cytokines might also contribute to VACV-induced inflammation via the enhanced immunopathology. Our results also support the conclusion that NK cells are important in controlling early local and systemic spreading of VACV in mice (25, 26). Although atopic dermatitis is still only partially understood in humans, there are strong indications that NK cell defects are involved (17, 18, 27). Our data now show that critical failures in NK cell–mediated immunity allow for disastrous early spread of vaccinia after cutaneous infection, and these NK cell defects are related to the immunosuppressive effects of IL-17A.

NC/Nga mice (10) were used in all animal experiments. Eczematous skin lesions were induced as described previously (11). In brief, mice were shaved on the back and dermatitis was induced by two rounds of a week-long treatment with Dermatophagoides farinae extract (Der f; Greer Laboratories) and SEB (Sigma-Aldrich). During this treatment, their back was occluded with a bandage, which was removed the following week. Mice with a clinical score of ≥8 (eczematous group) were intradermally injected with 10⁶ PFU per 3 µl of VACV (Western Reserve strain) at the center of skin lesions. A cohort (normal group) of age and sex-matched mice with healthy skin was also infected at the same anatomical site. Clinical scores of eczematous skin lesions are based on severity (0, no signs; 1, mild; 2, intermediate; 3, severe) of four signs (redness, bleeding, eruption, and scaling). Scoring was performed in a blind manner. The virus was prepared by repeated (three times) freeze–thaw cycles of infected HeLa cells in DMEM/1% FCS followed by centrifugation. Uninfected HeLa cell extract was used as control. Virus titers were measured by plaque-forming assays on Vero cells. All of the mouse experiments were approved by an Institutional Review Board of the La Jolla Institute for Allergy and Immunology.

CD4⁺ and CD8⁺ T cells, Mac-1⁺ monocytes/macrophages, and Ly49G2 (4D11)⁺ NK cells were detected by immunochemical staining. Mast cells were stained by toluidine blue, and eosinophils and neutrophils were detected by hematoxylin and eosin or Congo red staining.

NK cell activity was measured using YAC-1 cells as target cells, with effector-to-target ratios (12.5:1, 25:1, and 50:1) in spleen cells isolated on day 2 or 3 after infection.

Splenocytes from NC/Nga mice were negatively selected by MACS beads (Miltenyi Biotec) or the EasySep mouse NK cell enrichment kit (StemCell Technologies Inc.). These NK-enriched cells were cultured in RPMI 1640 with 10% FBS containing 500 ng/ml IL-15 for 4 d. The purity of NK cells was checked by flow cytometry after staining with anti-NK1.1 and anti-CD3 antibodies before the use for adoptive transfer. For in vitro NK cell assays, purified splenic NK cells or whole splenocytes were cultured in RPMI 1640 with 10% FBS containing 20 ng/ml IL-4 and/or 50 ng/ml IL-17A or IL-17F with or without 2 µg/ml anti–IL-17RA for 2 d, followed by flow cytometry.

Single-cell suspensions of splenocytes or lymph node cells were surface stained with anti-NK1.1 and CD3. The cells were then fixed, permeabilized, and stained with anti–granzyme A, –granzyme B, -perforin, or –IFN-γ. Data were acquired with a FACSCalibur (BD) and analyzed using FlowJo software (Tree Star, Inc.).

Skin tissues were taken from infection sites or erosive areas by punch biopsy, and axillary lymph nodes and spleens were also isolated. Total RNAs were isolated with TRIzol reagent (Invitrogen) and used as a template to prepare cDNAs. PCR products were analyzed by agarose gel elctrophoresis. Quantitative PCR was performed using a LightCycler 480 (Roche).

Statistical analysis in each independent experiment was performed with an unpaired, two-way analysis of variance using Prism software (GraphPad Software, Inc.), otherwise noted. P < 0.05 was considered statistically significant.

Fig. S1 shows antibody responses. Fig. S2 depicts αAGM1 effects on NK cells. Fig. S3 shows anti-NK1.1 effects. Fig. S4 depicts viral titers in anti–IL-17– and αAGM1-treated mice. Fig. S5 shows in vitro effects of IL-17A on NK cell mediators. Fig. S6 depicts the effects of various cytokines on NK cell mediators. Fig. S7 shows skin lesions in anti–IL-15–treated mice.

The authors thank M. McCausland for excellent technical assistance, and Drs. W.M. Yokoyama and L.L. Lanier for helpful discussion.

This study was supported in part by Atopic Dermatitis and Vaccinia Network National Institutes of Health (NIH)/National Institute of Allergy and Infectious Diseases contract N01 AI40030 to T. Kawakami, and NIH grants AI63107 and AI77953 to S. Crotty. This study is publication no. 1071 from the La Jolla Institute for Allergy and Immunology.

The authors declare no competing financial interests.