TLR4 Signaling in Stromal Cells Is Critical for the Initiation of Allergic Th2 Responses to Inhaled Antigen Free

Allergic diseases, including rhinitis, atopic dermatitis, asthma, and food allergies, are driven by Th2 effector cells. Th2 responses are characterized by the production of the cytokines IL-4, IL-5, and IL-13, which contribute to mucus production, eosinophilia, and elevated levels of Ag-specific IgE production (1). Dendritic cells (DCs), which densely line the airways, are critically involved in the pathogenesis of allergic diseases and are known to be potent inducers of CD4 T cell differentiation, expansion, and polarization (2, 3). In the case of allergic sensitization, however, the mechanism by which immature DCs undergo maturation and become Th1- or Th2-inducing APCs is unclear.

Microbial activation of TLRs, which recognize conserved microbial molecular patterns, is known to induce DC maturation, resulting in increased surface MHC expression, upregulation of costimulatory molecules, secretion of proinflammatory cytokines, and the ability to activate naive T cells (4–7). In addition to inducing DC maturation, LPS, a cell wall component of Gram-negative bacteria and TLR4 agonist, induces expression of Notch ligands, such as Delta (e.g., Delta-4) and Jagged (e.g., Jagged-1), which have the capacity to differentially induce either Th1 or Th2 effector cells, respectively (8). Jagged was shown to deliver an instructive signal for Th2 differentiation that is independent of IL-4/STAT6 (8). Furthermore, an increase in the ratio of Delta-4 to Jagged-1 correlates with the enhanced ability of DCs to induce Th1 polarization (8, 9).

We and others have previously demonstrated that generation of both Th1- and Th2-mediated inflammatory responses to inhaled protein Ag is dependent on TLR4 signaling (10, 11), and that during Ag sensitization, the amount of LPS administered with Ag is the critical determinant of CD4 T cell differentiation. Specifically, a high concentration of LPS (LPS^hi) induces a Th1 response and a low concentration (LPS^lo) induces Th2. In these studies, elevated levels of the pro-Th1 cytokine IL-12 were present in the sera of mice exposed to LPS^hi but not in those exposed to LPS^lo, and both Th1 and Th2 responses involved the increased migration of Ag-bearing DCs to the mediastinal lymph nodes (mLNs).

These findings prompted us to investigate how signaling through the same TLR might induce two biologically distinct responses, depending only on the concentration of LPS present at the time of sensitization. One possibility was that DCs do not sense LPS^lo directly. Because airway stromal cells (ASCs) express TLR4 (12–15) and are among the first cells to contact Ag in the airways, we hypothesized that ASCs play a critical role in LPS-dependent sensitization to inhaled Ag.

ASCs produce proinflammatory mediators that can drive inflammatory immune responses (16, 17), and in response to endotoxin, bacteria, and other TLR agonists, both human and murine airway epithelial cell lines have been shown to produce a broad array of proinflammatory factors (18–23). In addition to their involvement in the initiation of innate immune responses, the airway epithelium is also capable of driving the exacerbation of established allergic airway diseases by the production of IL-4, IL-13, mucus, thymus activation-regulated cytokine, thymic stromal lymphopoietin (TSLP), and other Th2 chemoattractant cytokines and lipid mediators (24, 25).

We addressed the role of TLR4 signaling in the SC and hematopoietic cell (HPC) compartments using bone marrow (BM) chimeric mice with TLR4 restricted to either the SC compartment (SC⁺HPC⁻) or the HPC compartment (SC⁻HPC⁺). We demonstrate that although HPC TLR4 is necessary and sufficient for Th1 sensitization to OVA-LPS^hi, TLR4 in both compartments is required for Th2 sensitization to OVA-LPS^lo. Surprisingly, although SC⁺HPC⁻ mice were unable to generate a Th1 response to OVA-LPS^hi, they instead mounted a robust Th2 response, indicating that in the presence of higher concentrations of LPS, SC TLR4 is sufficient for Th2 sensitization. The data further show that stromal TLR4 signaling leads to the maturation of Th2-inducing DCs that fail to produce proinflammatory cytokines or to upregulate the Th1-inducing Notch ligand Delta-4. Following intranasal (i.n.) administration of LPS into the airways, ASCs upregulate mRNA expression of TSLP, suggesting that the SC-dependent instruction of allergic Th2 responses is driven by ASC-mediated induction of Th2-inducing DCs.

BALB/cJ, C.C3H tlr4 lpsd (TLR4d), and C.Cg-Tg (DO11.10)10Dlo/J (D011.10) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred in our facility. BALB/cAnNCr and C57BL/6Ly-Pep3b (Pep3b) were purchased from the National Cancer Institute. C57BL/6J were purchased from The Jackson Laboratory and bred in our facility. TLR4-deficient (TLR4^−/−) mice, provided by Dr. S. Akira (Research Institute for Microbial Diseases, Osaka, Japan), were backcrossed nine generations onto the C57BL/6J background. In all BM chimera experiments, 6- to 12-wk-old female mice were used with four to six mice per group. Male mice, 6–14 wk of age, were used as BM donors in experiments involving BM chimeric mice. Male mice were used as BM donors so that chimerism could be assessed by detecting the presence of the male Y chromosome in female BM recipients using fluorescence in situ hybridization (FISH). All animal experiments were performed in accordance with the guidelines of Yale University’s Institutional Animal Care and Use Committee.

Mice were anesthetized with isoflurane and sensitized i.n. once daily with 100 μg OVA (grade V; Sigma-Aldrich, St. Louis, MO) with LPS^hi (10–15 μg) or LPS^lo (0.01–0.1 μg) in 50 μl PBS on days 0, 1, and 2. Mice were challenged i.n. on days 14, 15, 18, and 19 with 25 μg OVA in 50 μl PBS and sacrificed on day 21.

Mice were sacrificed, and cells in the bronchoalveolar lavage (BAL) were obtained as described previously (26). Briefly, lungs were washed three times with 1 ml ice-cold PBS, between washes cells were centrifuged 500 × g, 5 min, 4°C, and pellets were resuspended and counted. Cytospin preparations were stained with Diff-Quik (Baxter Merz & Dade, Dudingen, Switzerland), and differential cell counts were evaluated by counting at least 200 cells for determination of relative percentage of each cell type in the BAL.

Lungs were perfused with cold PBS through the right ventricle until pulmonary vasculature was clean (∼10 ml PBS). Paraffin-embedded coronal lung sections were prepared as previously described and stained with H&E or periodic acid-Schiff (PAS) (26). H&E images are ×100, and PAS images are ×200 total magnification.

mLN cells were isolated, and single-cell suspensions were prepared and stimulated in vitro with 200 μg/ml OVA and syngeneic T cell-depleted mytomycin C (Sigma-Aldrich)-treated splenocytes (APCs). After 48 h, concentrations in the supernatant of IFN-γ, IL-4, IL-5, and IL-13 were measured using commercially available ELISA kits (R&D Systems), and IL-17 was measured using Bioplex cytokine bead array system (Bio-Rad, Hercules, CA) on the Luminex 100 IS System (Millipore, Bedford, MA) plate reader. Lower detection limits were 25.0 pg/ml (IL-4), 125.0 pg/ml (IL-5 and IL-13), 3.0 pg/ml (IL-17), and 1.9 ng/ml (IFN-γ).

Six- to 12-wk-old female BALB/cJ (wild-type [WT]) and TLR4d mice were lethally irradiated on an X-rad 320 X-ray irradiator with a total of 1200 cGy total body irradiation delivered in two 600 cGy doses spaced 3 h apart. Eighteen to 24 h later, mice were reconstituted with 5–8 × 10⁶ WT or TLR4d BM cells from male donors and maintained on sulfamethoxazole-trimethoprim (5 ml per 250 ml drinking water) for 2 wk, receiving autoclaved food and water. Mice were used in experiments 7–8 wk after BM transfer to allow full reconstitution of BM cells to occur. In some experiments (Supplemental Fig. 3), C57BL/6J and TLR4^−/− mice were used to confirm data obtained in chimeric mice on the BALB/cJ background. These mice received one dose of 1200 cGy total body irradiation and were otherwise treated identically to BALB/cJ mice.

In BM chimeras on a BALB/cJ background, the degree of chimerism of splenic B cells, T cells, CD11c⁺ cells, and lung CD11c⁺ cells was determined by FISH on the Y chromosome of male donor cells in female recipients, 7–8 wk after irradiation.

Splenocytes from mechanically separated spleens were isolated on a density gradient using lymphocyte separation medium (MP Biomedicals, Irvine, CA). T or B cells or CD11c⁺ cells were selected by preincubating splenocytes with anti-FcR (2.4G2) to block nonspecific Ab labeling, followed by labeling with biotinylated anti-TCR (H57), anti-B220, or anti-CD11c. Labeled cells were positively selected using streptavidin-conjugated microbeads (Miltenyi Biotec, Auburn, CA). For lung CD11c⁺ cells, cells from collagenase type IV (Sigma-Aldrich) digested lungs were purified by positive selection using CD11c microbeads (Miltenyi Biotec) according to the manufacturer’s instructions. Isolated cells were dotted onto microscope slides, fixed in 70% ethanol for 10 min at room temperature (RT), and then stored in 100% ethanol at 4°C until use. FISH analysis was performed as described previously (27). Lung mast cell chimerism was determined using BM from BALB/c.129-Il4tm1lky/J (4get) mice, which have mast cells that are constitutively GFP⁺. Digested lung cells were analyzed on a flow cytometer to determine whether lung mast cells were donor or host derived.

For chimeric mice on the C57BL/6J background, male mice that are congenic for CD45.1, C57BL/6Ly-Pep3b (Pep3b) mice, were used as BM donors. Pep3b donor cells were transferred into C57BL/6J mice, which express CD45.2, allowing distinction between host and donor derived cells on all CD45-expressing HPCs. Analysis of chimerism of splenic T cells, B cells, DCs, lung CD11c⁺ cells, and mast cells was performed by flow cytometry using mAbs against CD45.1, CD45.2, CD3e, B220, CD11c, FcεRIα, and cKit (BD Pharmingen, San Diego, CA).

CD4 T cells were isolated from spleens of naive D011.10 mice by negative selection using mAb to I-Ad (212.A1), CD8 (TIB 210), B220 (TIB 164), and FcR (2.4G2), followed by anti–Ig-coated magnetic beads (Polysciences, Warrington, PA). Highly pure (>99%) populations of naive CD4 T cells (CD62L^hi,CD44^lo) cells were obtained by labeling isolated CD4 T cells with FITC-labeled anti-CD44 (Pgp-1) and PE-labeled anti-CD62L (Mel14) Abs and sorted on a MoFlo cell sorter.

For negative selection of DCs, mLN cells were preincubated with anti-FcR (2.4G2) and 2 mg/ml mouse Ig and then labeled with anti-TCRb (H57) and anti-CD62L to predeplete the majority of non-DC mLN cells. DCs were enriched to >85% purity from this population by negative selection using Dynabeads Mouse DC Enrichment kit (Invitrogen, Carlsbad, CA). For positive selection, DCs were selected by magnetic cell sorting using CD11c microbeads (Miltenyi Biotec), labeled with anti-CD11c (clone HL3; BD Biosciences, San Jose, CA), and sorted by flow cytometry to >95% purity.

Perfused lungs were finely minced in Dispase and swirled 45 min at RT. HBSS/5% FCS/40 μg/ml DNase type I (Sigma-Aldrich) was added and incubated at RT at 10 min. Single-cell suspensions were prepared and RBCs were lysed. Cell pellets were resuspended in PBS plus 2 mM EDTA/0.5% FCS and labeled with predetermined optimal concentrations of biotinylated 2.4G2, anti-CD11b, and anti-CD45 mAb (clone 30-F11) 25 min on ice, washed with PBS plus 2 mM EDTA (no FCS), and incubated in MACS streptavidin-conjugated magnetic beads per the manufacturer’s instructions. Cells were magnetically separated. The enriched SCs (flow-through cells) contained >98% CD45.2⁻ cells, determined by staining the enriched cells with anti-CD45.2 (clone 104; BD Pharmingen) and performing flow cytometry (Supplemental Fig. 4). We confirmed that CD45.2 Ab is capable of binding cells labeled with biotinylated Abs to 2.4G2, CD11b, and CD45 and streptavidin-conjugated magnetic beads by staining an aliquot from each group just prior to placement on the magnetic column for cell separation (predepletion). Although the enriched fraction (postdepletion) containing SCs was negative for CD45.2, the predepletion fraction contained cells that stained positive for CD45.2, indicating that epitope masking by the biotinylated Abs or streptavidin-conjugated beads was not occurring (Supplemental Fig. 4).

Enriched mLN DCs from sensitized mice (72 h after first of three daily i.n. sensitizations) were cocultured with naive DO11.10 T cells at 1:20 (DC:T) in complete Bruffs media (C’ Bruffs) (Click’s medium plus 50 μM 2-ME plus 20 μg/ml gentamicin; Life Technologies, Carlsbad, CA) containing 10% FCS, 1% penicillin/streptavidin, and 2 mM l-glutamine. After 24 h, cells were pulsed with 25 U/ml IL-2. On day 4, cells were washed and rested 48 h in C’ Bruffs. Rested cells were restimulated at a ratio of 1:1 with APCs with or without 5 μg/ml OVA peptide (OVA_323–339). Culture supernatants were collected after 48 h, and cytokine levels were analyzed by Bioplex cytokine bead array system (Bio-Rad) on the Luminex 100 IS System (Millipore) plate reader.

RNA was extracted using TRIzol Reagent (Invitrogen) and gene expression determined by quantitative PCR (qPCR) on Stratagene Thermocycler (Stratagene, La Jolla, CA). Expression was normalized to β-actin or HPRT. Primers: β-actin, 5′-GAAGTCCCTCACCCTCCCAA-3′ and 5′-GGCAT-GGACGCGACCA-3′; HPRT, 5′-CCAGCAAGCTTGCAACCTTAACCA-3′ and 5′-GTAATGATCAGTCAACGGGGGAC-3′; TSLP, 5′-CGACAG-CATGGTTCTTCTCA-3′ and 5′-ATTTGCTCGAACTTAGCC-3′; Dll4, 5′-AGGTGCCACTTCGGTTACACAG-3′ and 5′-CAATCACACACTCG-TTCCTCTCTTC-3′; and Jag1, 5′-AGAAGTCAGAGTTCAGAGGCGT-CC-3′ and 5′-AGTAGAAGGCTGTCACC-AAGCAAC-3′. Forward and reverse primers for Delta-4 (Dll4) and Jagged-1 (Jag1) were described previously (8).

Analyses were performed using 100 μg FITC-OVA (Sigma-Aldrich), 50 μg DQ-OVA (Molecular Probes, Eugene, OR), or 50 μg Eα peptide 52–68 (ASFEAQGALANIAVDKA). Eα peptide presentation on I-Ab was detected using the biotin-conjugated anti–I-Ab:Eα complex-specific Ab, YAe (eBioscience, San Diego, CA), as described previously (28). Labeled OVA or PBS was administered on days 0, 1, and 2 to naive BALB/c mice with LPS^hi or LPS^lo in 50 μl PBS. Seventy-two hours after the first i.n. sensitization, mice were sacrificed, and mLN cells were isolated. Cells were preincubated on ice with anti-FCR (2.4G2), labeled with MAb specific for CD11c (HL3), CD40 (HM50-3), I-Ad (AMS-32.1) (BD Pharmingen), CD86 (GL-1; eBioscience), and biotin-conjugated YAe mAb (eBioscience) (for Eα-peptide experiments) to detect Eα peptide presentation on I-Ab and then analyzed by flow cytometry.

Data are presented as means ± SEM, indicated by error bars. Statistical differences between groups in in vivo experiments were calculated using the Mann-Whitney U test in GraphPad Prism, version 4.0. Statistical differences between groups in in vitro experiments were calculated using the Student t test. All values of p ≤ 0.05 were considered statistically significant.

We have previously published that C.C3H tlr4 lpsd (TLR4d) mice, in which functional TLR4 signaling is absent because of a spontaneous point mutation in the signaling domain of TLR4, fail to be sensitized to inhaled OVA and do not mount an immune response to OVA challenge (10). Furthermore, OVA-LPS^lo sensitization induces a Th2 response, whereas OVA-LPS^hi drives Th1. Fig. 1A shows a representative BAL cell differential of these previously published results.

We compared the ability of DCs to migrate to the mLN in OVA-LPS^hi– and OVA-LPS^lo–sensitized mice. Following 3 d of once daily exposure to inhaled OVA and LPS, or PBS alone, the percentage of CD11c⁺ DCs in isolated mLNs was determined by flow cytometric analysis of CD11c-labeled cells, and the total number of DCs in the mLN was determined. The total number of DCs in the mLN was similar between OVA-LPS^hi and OVA-LPS^lo groups, and both were significantly increased compared with saline control mice (Fig. 1B). Similar results have also been obtained at the earlier time points of 24 and 48 h (unpublished data).

To determine whether low and high concentrations of LPS induce similar levels of costimulatory molecule expression on DCs in the mLN, we analyzed CD40 and CD86 expression on the total DC population and on the specific population of DCs that had taken up Ag in mice sensitized for 3 d with the labeled Ag DQ-OVA. DQ-OVA is a self-quenched conjugate of OVA that only fluoresces upon proteolytic degradation, thereby allowing us to characterize the specific population of DCs that had captured and processed labeled Ag, in addition to the total DC population.

We found that, 72 h after the first i.n. sensitization, mLN DCs from mice sensitized with DQ-OVA-LPS^hi or with DQ-OVA-LPS^lo expressed increased levels of CD40 and CD86 compared with DCs from PBS-treated mice. More DCs expressed high levels of CD40 in DQ-OVA-LPS^hi–sensitized mice, and the peak of CD86 expression was brighter, although a similar percentage of DCs in the DQ-OVA-LPS^lo group also expressed high levels of CD86 (Fig. 1C). These findings are consistent with patterns of costimulatory molecule expression observed on lung DCs 48–72 h following an identical sensitization protocol (unpublished data). DCs that had captured Ag (CD11c⁺FL-1⁺ cells; Fig. 1D, Supplemental Fig. 1B) from OVA-LPS^lo– and OVA-LPS^hi–sensitized mice expressed similar levels of CD40 and, in both groups, upregulated CD86; however, as was observed in the total DC population, FL-1⁺ DCs from OVA-LPS^hi mice expressed brighter levels of CD86 (Fig. 1E).

In OVA-LPS^lo–sensitized mice, FL-1⁺ DCs represented 4.2% of all CD11c⁺ cells, compared with 1% in OVA-LPS^hi–sensitized mice (Fig. 1D, Supplemental Fig. 1B). The percentage of FL-1⁺ DCs in OVA-LPS^hi was surprisingly low, at just under twice background fluorescence (PBS control, 0.6%; Fig. 1D, Supplemental Fig. 1B).

Similar studies were carried out using FITC-labeled OVA (to characterize Ag uptake and migration; Supplemental Fig. 1A) or Eα-conjugated OVA (to characterize Ag presentation on surface MHC molecules using YAe mAb (Supplemental Fig. 1C). As was observed with DQ-OVA (Fig. 1D, Supplemental Fig. 1B), whereas similar total numbers of DCs were present in the mLNs of LPS^hi and LPS^lo groups (FITC-OVA: 96.2 × 10³ [LPS^hi] versus 123. 7 × 10³ [LPS^lo]; Eα-OVA: 82.99 × 10³ [LPS^hi] versus 73.8 × 10³ [LPS^lo]; unpublished data), there was an increased percentage of FITC⁺ (Ag-bearing) and YAe⁺ (Ag-presenting) DCs in LPS^lo groups compared with LPS^hi groups (FITC-OVA: 7.55% [LPS^hi] versus 17.2% [LPS^lo]; Eα-OVA: 2.4% [LPS^hi] versus 6.0%[(LPS^lo]) (Supplemental Fig. 1A, 1C).

CD40 and CD86 expression levels on all mLN DCs and on Ag-presenting (YAe⁺) DCs were evaluated (Supplemental Fig. 1D), and the costimulatory molecule expression profiles were similar to those in Ag-containing DCs (DQ-OVA⁺ DC; Fig. 1C, 1E). In conclusion, exposure to both low and high levels of LPS in the airways was sufficient for DCs to capture and process Ag, migrate to the mLN, upregulate costimulatory molecules, and present Ag. Furthermore, these experiments show that OVA-LPS^lo sensitized mice reproducibly had increased percentages of Ag-containing mLN DCs.

To evaluate the ability of DCs exposed to OVA and LPS^lo or LPS^hi to activate and polarize naive, Ag-specific CD4 T cells, we set up an ex vivo assay to coculture OVA-specific naive D011.10 TCR transgenic CD4 T cells (DO11.10 CD4 T cells) with isolated mLN DCs that had captured Ag in vivo (i.e., DCs were not pulsed with any additional Ag following isolation).

Cell suspensions were made from the mLNs of WT mice that had been sensitized i.n. with OVA-LPS^hi or OVA-LPS^lo for 3 d, and DCs were enriched by positive magnetic selection of CD11c⁺ cells, followed by cell sorting to further enrich the DCs (>97%). mLN cells that were depleted of DCs were also tested in the coculture assay to determine whether the DCs in the lymph node (LN) population were required for both Th1 and Th2 differentiation of naive CD4 T cells. The highly enriched DCs, or mLN cells that were depleted of DCs, were cocultured with DO11.10 CD4 T cells at increasing ratios (DC:T cell) from 1:160 to 1:10, and T cell proliferation ([³H]thymidine incorporation) was measured. As expected, T cells cultured alone did not proliferate (Fig. 2A). mLN cells isolated from OVA-LPS^hi– or OVA-LPS^lo–sensitized mice that were depleted of DCs, or from mice sensitized with LPS^hi only (no Ag), did not stimulate DO11.10 CD4 T cell proliferation, indicating that DCs were absolutely required for T cell activation and that this was an Ag-dependent mechanism of CD4 T cell activation (Fig. 2A). Significant levels of T cell proliferation were observed when T cells were cocultured with enriched DCs isolated from either OVA-LPS^hi– or OVA-LPS^lo–sensitized mice (Fig. 2A). However, mLN DCs from OVA-LPS^lo–sensitized mice induced increased proliferation of DO11.10 CD4 T cells compared with OVA-LPS^hi (Fig. 2A).

We next asked whether mLN DCs isolated from OVA-LPS^hi– or LPS^lo–sensitized mice were sufficient to induce Th1 and Th2 differentiation, respectively. Naive (CD62L^hiCD44^lo) DO11.10 CD4 T cells were isolated from the spleens and enriched to >98% purity by cell sorting for coculture with mLN DCs. Following 4 d of coculture and 2 d of rest in fresh media, we restimulated DO11.10 CD4 T cells with mytomycin C-treated, T-depleted splenocytes (APCs) pulsed with OVA peptide and measured the cytokines present in the culture supernatant after 48 h. mLN DCs from naive mice, or from LPS^hi-only sensitized mice (no Ag), did not induce CD4 T cells to produce Th1 or Th2 cytokines (unpublished data and Fig. 2B), which is consistent with the failure of these DCs to induce CD4 T cell proliferation. DCs isolated from OVA-LPS^lo–sensitized mice induced naive CD4 T cells to secrete significantly higher concentrations of the Th2 cytokines IL-4, IL-5, and IL-13. A significantly decreased concentration of IFN-γ was detected, compared with OVA-LPS^hi–sensitized mice. DO11.10 CD4 T cells cocultured with DCs from OVA-LPS^hi–sensitized mice produced high levels of IFN-γ and low levels of Th2 cytokines (Fig. 2B). These data demonstrate that DCs that have captured Ag in the lung and migrated to the mLN are sufficient for instructing naive CD4 T cell differentiation into both Th1 and Th2 effector cells, depending only on the concentration of LPS.

To determine whether high and low concentrations of LPS administered with OVA differentially induced Notch ligands in DCs activated in vivo, we isolated mLN DCs from OVA-LPS^hi– and OVA-LPS^lo–sensitized mice at 0, 24, and 48 h after sensitization and determined the expression of Delta-4 and Jagged-1 over time by qPCR. At 48 h after the first i.n. sensitization, we observed a >10-fold increase in the Delta-4/Jagged-1 ratio in mLN DCs from OVA-LPS^hi–sensitized mice but no increase in the Delta-4/Jagged-1 ratio in DCs from OVA-LPS^lo-sensitized mice (Fig. 2C). In OVA-LPS^lo–sensitized mice, the Delta/Jagged ratio was significantly lower than the ratio in mLN DCs isolated from OVA-LPS^hi–sensitized mice at 48 h. The Delta/Jagged ratio was low because Delta-4 mRNA expression was only weakly upregulated in OVA-LPS^lo mLN DCs (Fig. 2D), while concurrently, Jagged-1 mRNA expression was upregulated following sensitization and maintained at elevated expression at 48 h (Fig. 2E). The upregulation of Jagged-1 in both OVA-LPS^hi and OVA-LPS^lo DCs indicated that exposure to even low levels of LPS is sufficient to lead to the upregulation of Jagged-1. Taken together, these results indicate that the Delta-4/Jagged-1 ratio is an indicator of the Th1- or Th2-inducing capacity of an mLN DC, and this striking difference demonstrates that other factors beside the ability to produce proinflammatory cytokines distinguish these functionally distinct, Th1- and Th2-inducing mLN DCs.

To test whether TLR4 signaling in DCs is required for both Th1 and Th2 sensitization, we restricted TLR4 signaling to either the hematopoietic or stromal compartment using BM chimeric mice on the BALB/cJ background (Fig. 3A). Some experiments were also carried out using mice on the C57BL/6 background (Supplemental Fig. 3).

To restrict TLR4 signaling to the HPC compartment, we lethally irradiated recipient TLR4d mice and reconstituted them with WT BM. In these mice, referred to as SC^-HPC⁺, only HPCs derived from transferred WT BM are TLR4 sufficient (+), whereas the radioresistant SCs derived from the recipient TLR4d strain do not express functional TLR4 (−). Conversely, in SC⁺HPC^- mice, we restricted TLR4 signaling to only the SC compartment by lethally irradiating WT mice and reconstituting the HPC compartment with TLR4d BM. As controls, WT or TLR4d mice were lethally irradiated and reconstituted with their own BM type (SC⁺HPC⁺ and SC⁻HPC⁻, respectively). Seven to 8 wk following BM reconstitution, chimeric mice were sensitized with OVA and LPS^hi or LPS^lo, and 2 wk later, their responses to i.n. OVA challenge were evaluated.

Mice on the BALB/cJ background that were congenic for a cell surface marker expressed on all HPCs (e.g., CD45.2 [Ly5.2]), as required for assessment of chimerism of many different cell lineages, were not available to us. Thus, we opted to generate BM chimeras using female mice as recipients and male mice as BM donors. We then determined chimerism by FISH to detect the Y chromosome of male donor cells in female recipients. An exemplary FISH analysis on lung DCs from male (i), female (ii), and chimeric (iii) lung CD11c⁺ cells is shown in upper panels of Fig. 3B. Using FISH, we determined that full chimerism (>95%) of lung CD11c⁺ cells (Fig. 3B) and splenic B cells, T cells, and DCs (Fig. 3D) had been established.

To determine chimerism of BALB/cJ lung mast cells, which we find to be present at a very low frequency (<0.05%) in the lungs of naive mice, we transferred BM from BALB/c.129-Il4tm1lky/J (4get) mice or from BALB/c mice (as negative control) to BALB/cJ mice. Mast cells from 4get mice are constitutively GFP positive (29) and can be detected by flow cytometric analysis of autofluorescence (AF) low FceRIa⁺,cKit⁺ cells. In two experiments, it was determined that 7 wk following lethal irradiation, mast cells were 88–89% BM donor derived (Fig. 3C), compared with the positive control (4get FceRIa⁺cKit^hi cells). Background GFP⁺ fluorescence was detected in 4% of negative control cells.

To facilitate analysis of chimerism in C57BL/6J mice, whose HPCs express CD45.2, CD45.1 congenic mice were used as BM donors for irradiated C57BL/6J and TLR4^−/− mice. In chimeric mice, CD45.1⁺ cells derived from the donor BM are distinguished from any remaining recipient CD45.2⁺ HPCs by flow cytometric analysis. Individual chimeric mice, four per group, were analyzed 7 wk after lethal irradiation, and the degree of chimerism (%) was determined to be as follows: (97.8 ± 0.6 [lung mast cells], 95.6 ± 0.6 [lung DCs], 81.0 ± 0.9 [splenic T cells], 95.5 ± 0.8 [splenic DCs], and 97.8 ± 0.6 [splenic B cells] [Fig. 3E–G]).

Upon OVA challenge, the OVA-LPS^lo–sensitized positive control group (SC⁺HPC⁺) exhibited typical, Th2-type, eosinophilic BAL cell differentials, whereas SC⁻HPC⁻ chimeras had very low numbers of BAL cells (Fig. 4A). When TLR4 signaling was restricted to either the hematopoietic or the stromal compartment (SC⁻HPC⁺ and SC⁺HPC⁻, respectively), the total number of BAL cells was significantly reduced (p < 0.001) to levels similar to the negative control (SC⁻HPC⁻) (Fig. 4A).

To further establish that the Th2 response was abrogated in SC⁺HPC⁻ and SC⁻HPC⁺ chimeric mice, we evaluated the draining LN response following challenge. The total number of cells in the mLN was only moderately increased compared with the negative control group (SC⁻HPC⁻ chimeric mice) in either group with compartment-restricted TLR4 signaling, confirming that sensitization to OVA in these chimeric mice was diminished compared with the WT control (Fig. 4B). The data shown in Fig. 4B are representative of the average number of cells in each group (mLNs from all mice were pooled before cell suspensions were made and cells counted) in each of three experiments. Upon restimulation with OVA-pulsed APCs in vitro, mLN cells isolated from SC⁺HPC⁺ mice produced Th2 cytokines, as is typical of WT Th2 responses. Consistent with the BAL data indicating that no Th2 response was initiated, LN cells in SC⁺HPC⁻ and SC⁻HPC⁺ mice produced decreased levels of Th2 cytokines compared with the positive control (Fig. 4C). Although some IFN-γ was detected in the SC⁺HPC⁺ and SC⁻HPC⁺ groups, the levels were very low overall (compared with the IFN-γ concentration associated with a Th1 response).

We next asked whether, following Ag challenge, chimeric mice sensitized with OVA-LPS^hi would also exhibit a dual requirement for TLR4 signaling in both the stromal and hematopoietic compartments for the initiation of Th1 responses. As expected, the SC⁻HPC⁻ control group did not respond to Ag challenge, whereas the total number of BAL cells and the BAL cell differential in SC⁺HPC⁺ chimeras were characteristic of a WT Th1 response (Fig. 4D). SC⁻HPC⁺ mice exhibited a Th1-like BAL cell differential comprising neutrophils and very few eosinophils, although the total BAL cell number was reduced by ∼50% compared with SC⁺HPC⁺ mice, suggesting that Th1 induction and/or recruitment of cells to the lung following Ag challenge was impaired in these mice (Fig. 4D). Strikingly, SC⁺HPC⁻ mice mounted a robust BAL response to OVA-LPS^hi (Fig. 4D). The response induced by stromal TLR4 was Th2 driven, as the BAL made up >50% eosinophils. This experiment was also carried out in mice on the C57BL/6 background, using TLR4^−/− mice, with identical results, indicating that this finding was not limited to a single mouse strain (Supplemental Fig. 3).

Analysis of the total numbers of cells in the mLN revealed that high numbers of cells were present in SC⁻HPC⁺, SC⁺HPC⁻, and SC⁺HPC⁺ mice, indicating that LN response to challenge was robust in all groups, except the negative SC⁻HPC⁻ control group (Fig. 4E). Like mLN cells from SC⁺HPC⁺ mice, upon restimulation with OVA-pulsed APCs, LN cells from SC⁻HPC⁺ produced IFN-γ but low levels of Th2 cytokines. IFN-γ production in SC⁺HPC⁻ mice was reduced compared with SC⁺HPC⁺ mice, which is consistent with our observation that the BAL responses in SC⁻HPC⁺ chimeras were reduced by ∼50% in total cell number but were still characteristic of a Th1-driven inflammatory response. The mLN cells from SC⁺HPC⁻ mice, in contrast, produced dramatically elevated levels of the Th2 cytokines IL-4, IL-5, and IL-13 and failed to produce IFN-γ (Fig. 4F).

We also determined the expression levels of IL-17. In WT control mice (SC⁺HPC⁺), IL-17 was detected at the highest level in the OVA-LPS^hi group but was detected at or near background levels (of SC⁻HPC⁻ and PBS control groups) in OVA-LPS^lo groups, indicating that high amounts of LPS are required for IL-17 production in our system (Supplemental Fig. 2). In the OVA-LPS^hi groups, IL-17 was also detected in both the SC⁻HPC⁺ and SC⁺HPC⁻ groups, indicating that TLR4 signaling in either compartment was sufficient for IL-17 production; however, the levels of production were lower in SC⁺HPC⁻ mice, suggesting that HPC TLR4 signaling induces a more robust IL-17 response.

To further confirm our novel finding that stromal TLR4 signaling is sufficient to induce Th2 responses to OVA-LPS^hi, we performed H&E and PAS staining of fixed lung tissue isolated from these chimeric mice. H&E staining revealed that peribronchiolar inflammation was present in the lungs of SC⁺HPC⁺, SC⁻HPC⁺, and SC⁺HPC⁻ mice, whereas mucus production, indicated by positive PAS staining, was present only in SC⁺HPC⁻ mice (Fig. 4G).

To determine whether SCs instruct lung DCs to become Th2 inducing, we characterized DC phenotype and function in SC⁺HPC⁺, SC⁻HPC⁻, and SC⁺HPC⁻ mice, examined the production of proinflammatory cytokines in the lungs of sensitized chimeric mice, and determined the expression of Delta-4 and Jagged-1 in mLN DCs isolated from sensitized BM chimeric mice. First, to ascertain whether stromal TLR4 signaling induces DC migration, we determined the number of DCs that migrated to the mLN following i.n. sensitization with OVA-LPS^hi in SC⁺HPC⁺, SC⁻HPC⁻, and SC⁺HPC⁻ mice. Increased numbers of DCs were observed in the mLNs of SC⁺HPC⁺ and SC⁺HPC^- mice compared with SC⁻HPC⁻ mice (Fig. 5A). We have also observed increased numbers of DCs that had captured FITC-labeled Ag in SC⁺HPC⁻ mice (unpublished data), further indicating that stromal TLR4 signaling induces a high number of DCs in the lung to mature and migrate. We observed similar levels of CD86 expression on SC⁺HPC⁺ and SC⁺HPC⁻ mice, confirming that the TLR4d DCs in SC⁺HPC⁻ mice were being indirectly activated by SC TLR4 signaling (Fig. 5B).

To evaluate whether proinflammatory cytokines were produced in the lung in the absence of HPC TLR4 signaling, we measured the level of proinflammatory cytokines present in the BAL fluid (BALF) of OVA-LPS^hi–sensitized SC⁺HPC⁺, SC⁻HPC⁻, and SC⁺HPC⁻ chimeric mice. We detected significant levels of IL-1α, IL-1β, IL-6, TNF-α, MIP-1β, and MCP-1 in the BALF of SC⁺HPC⁺ mice but not in the BALF of SC⁺HPC⁻ or SC⁻HPC⁻ mice, indicating that HPC TLR4 signaling is required for the production of detectable levels of these proinflammatory cytokines (Fig. 5C). This was also a further confirmation that 7–8 wk postirradiation, LPS-responsive HPCs, which are capable of secreting high levels of these cytokines, are not present in the lungs of lethally irradiated WT mice reconstituted with TLR4d BM. IL-25 was not detected in the BALF of any of these groups.

To determine whether the SC TLR4 response to LPS can stimulate DCs to become Th2-inducing, we used an ex vivo coculture assay to assess the ability of freshly isolated DCs that had captured Ag in vivo to polarize naive, Ag-specific CD4 T cells. Highly enriched mLN DCs isolated from OVA-LPS^hi–sensitized BM chimeric mice were cocultured with naive (CD62L^hiCD44^lo) D011.10 CD4 T cells. mLN DCs from sensitized SC⁺HPC⁻ chimeric mice induced the differentiation of polarized CD4 T cells that, upon restimulation, produced decreased levels of IFN-γ and increased levels of IL-4 compared with CD4 T cells that were cocultured with DCs isolated from the sensitized SC⁺HPC⁺ chimeric mice (Fig. 5D).

To further confirm that stromal TLR4 signaling led to the conditioning of Th2-inducing DCs, we determined the ratio of expression of Delta-4 and Jagged-1 in mLN DCs from BM chimeric mice sensitized with OVA-LPS^hi. Similar to Th1-inducing WT DCs, the Delta/Jagged ratio in DCs from sensitized SC⁺HPC⁺ mice was significantly increased 48 h after the first of two i.n. sensitizations (Fig. 5E). Conversely, mLN DCs from sensitized SC⁺HPC⁻ mice failed to upregulate the Delta/Jagged ratio, exhibiting a ratio that was >10-fold lower than that of SC⁺HPC⁺ DCs.

To identify a potential mechanism of Th2 induction by stromal TLR4 signaling, we examined the expression of the known, pro-Th2, proallergic cytokine TSLP in ASCs. WT or TLR4d mice were sensitized i.n. with OVA-LPS^hi, OVA-LPS^lo, or PBS alone. Three hours following sensitization, mice were sacrificed, and ASCs were isolated and enriched to >98% purity by negative depletion of CD45⁺, CD11b⁺, and FcR⁺ HPCs (Supplemental Fig. 4).

RNA was isolated from the purified stromal fraction, and the expression of TSLP mRNA was determined by qPCR and normalized to HPRT expression. WT mice significantly upregulated TSLP gene expression after exposure to OVA-LPS^hi or OVA-LPS^lo compared with the PBS control, with significantly increased expression in the OVA-LPS^hi group compared with OVA-LPS^lo. TLR4d mice did not upregulate TSLP in any of the groups (Fig. 6). Thus, exposure to LPS in the airways, even at low concentrations, is sufficient to stimulate a pro-Th2 response in ASCs that includes the upregulation of TSLP mRNA, strongly suggesting that SCs of the airways can condition Th2-inducing DCs.

The critical factor driving both Th1 and Th2 adaptive immune responses to inhaled Ag is the ability to detect and respond to the presence of pathogen-associated molecular patterns (4, 30). We show that for sensitization to inhaled OVA-LPS^lo, TLR4 signaling in both the HPC and SC compartments is required for Th2 sensitization, whereas, surprisingly, SC TLR4 signaling was sufficient for Th2 sensitization in the presence of high concentrations of LPS. We found that SC TLR4 signaling led to DC Ag uptake, maturation, and migration to the mLN. Furthermore, we found that mLN DCs from SC⁺HPC⁻ mice were capable of inducing naive CD4 T cells to differentiate into IL-4–producing Th2 cells in vitro.

In a recent study using an endotoxemia model to study the role of TLR4 signaling in the HPC and SC compartments, the authors oppositely concluded that stromal TLR4 signaling did not lead to the induction of DC maturation and migration (31). We note that the authors studied splenic DCs. Considerable differences exist between mucosal and nonmucosal DCs, and this may be particularly important for how DCs respond to signals from their microenvironment (32). Thus, our contrasting findings may be explained by the difference in the types of DCs studied. Furthermore, we studied migration of DCs from the lung to the mLN, whereas the authors of this study evaluated migration of resident splenic DCs within the spleen.

Another group published that TLR4 signaling in lung structural cells was necessary and sufficient to induce a Th2 response following intratracheal exposure to house dust mite (HDM)/LPS and suggested that Th1 and/or Th17 sensitization to OVA-LPS was also dependent on SC TLR4 (33). With respect to the latter conclusion, the authors reported that SC TLR4 signaling was necessary for production of IFN-γ and IL-17 by restimulated mLN cells from OVA-LPS^hi–sensitized mice. In contrast, we found that SC TLR4 is neither sufficient nor necessary for Th1 sensitization and show that both IFN-γ and IL-17 are produced by mLN cells from SC⁻HPC⁺ chimeric mice. We note that IFN-γ levels were consistently decreased in the SC⁻HPC⁺ group compared SC⁺HPC⁺ but were still increased compared with the negative control. We found no decrease in IL-17 production in the SC⁻HPC⁺ group. Furthermore, other parameters were consistent with a Th1 and/or Th17 response, such as neutrophilic BALs and similar levels of OVA-specific serum IgG2a (unpublished data) in SC⁺HPC⁺ and SC⁻HPC⁺ mice.

We note that the authors of the HDM study adoptively transferred naive OVA-specific transgenic CD4 T cells to chimeric mice and examined cytokine production by mLN cells only 5 d after Ag sensitization, whereas we evaluated cytokine production by mLN cells isolated from mice 21 d after the first sensitization and 2 d after the last of four airway Ag challenges. This full sensitization and challenge protocol allows full differentiation and expansion of effector T cells and, in our view, is a more accurate determinant of the type of effector response generated by the sensitization conditions.

The authors’ finding that SC TLR4 is necessary and sufficient for Th2 sensitization is in contrast to our data in OVA-LPS^lo–sensitized SC⁺HPC⁻ chimeric mice showing that HPC TLR4 signaling is also required. However, their finding is consistent with our data in OVA-LPS^hi–sensitized mice. The authors stated that HDM contains only very low levels of endotoxin that, alone, are insufficient to induce a Th2 response. They therefore attributed the SC TLR4-dependent Th2 sensitization to HDM. HDM is known to have cysteine protease activity, which has been shown to drive Th2 sensitization in the lung (34). The cysteine protease activity may directly activate SCs, or have other nonspecific effects in the lung that lead to maturation of lung DCs and Th2 sensitization. The authors also proposed that the Der p 2 allergen may enhance the response to endotoxin by acting as an MD2-like chaperone that promotes TLR4 signaling (35). This protease activity and/or TLR4 signal-enhancing property of HDM, coupled with the low levels of endotoxin present in the extract, may have provided a sufficient SC-activating signal to drive the response independently of HPC TLR4 signaling.

We propose a model in which SCs require a threshold level of TLR4 activation to independently drive Th2 sensitization. When this threshold level is not met, endotoxin sensitivity is increased by TLR4-expressing cells of the HPC compartment, thereby permitting Th2 sensitization to low levels of LPS. This model is supported by our finding that although TSLP mRNA expression in ASCs was induced by both low and high levels of LPS, significantly increased TSLP expression was detected in LPS^hi-exposed ASCs. Thus, in SC⁺HPC⁻ mice, we propose that only LPS^hi-induced levels of TSLP were sufficient to drive an HPC TLR4-independent Th2 response. Overexpression of TSLP (i.e., high levels of TSLP) in the airways is known to lead to Th2-mediated inflammation (36).

Furthermore, we propose that despite TSLP upregulation following OVA-LPS^hi sensitization in WT mice, a Th2 response is not induced because the Th2-inducing signal provided by TSLP is overcome by the Th1-favoring HPC response to LPS^hi. Our data show that Delta-4 expression is induced in TLR4-competent DCs following sensitization with OVA-LPS^hi but not OVA-LPS^lo. We note that Delta-4 has been shown to inhibit Th2 cell differentiation by inhibiting IL-4 signaling (37). Thus, in OVA-LPS^hi–sensitized WT mice, upregulation of Delta-4 by DCs may allow the DCs to overcome the pro-Th2 signal provided by TSLP.

Our data further demonstrated that Th2-inducing mLN DCs upregulate Jagged-1 but not Delta-4. Upregulation of Jagged-1 but not Delta-4 was also observed in mLN DCs from OVA-LPS^hi–sensitized SC⁺HPC⁻ mice, strongly suggesting that SCs induce DCs to upregulate Jagged-1, thereby conditioning DCs to drive Th2 sensitization. Further studies should be carried out to directly assess the mechanism of SC-driven Th2 sensitization and regulation of DC function.

In conclusion, we show that TLR4 signaling in both the HPC and SC compartments is critical for Th2 sensitization to OVA and low levels of LPS; however, in the presence of high concentrations of LPS, stromal TLR4 signaling is sufficient for the induction of Th2 responses. Our data further suggest that ASCs condition DCs to drive allergic Th2 responses by the expression of the pro-Th2 cytokine TSLP.

We appreciate Patricia Ranney, Lan Xu, and Paula Preston-Hurlburt for excellent technical assistance. We thank Octavian Henegariu for help with FISH, primers, and qPCR analysis, and Orkun Tan for assistance with generation of BM chimeras. We also thank Anna M. Dittrich, Kathryn Wilkinson, and Sean Connolly for helpful discussions regarding experimental design and the manuscript.

Disclosures The authors have no financial conflicts of interest.