Human upper airway epithelial cell line BEAS-2B (CRL-9609) (21) and human lung fibroblast cell line MRC-5 (CCL-171) (22) were purchased from the American Type Culture Collection (Manassas, VA, USA). Cells were cultured at 37°C under 5% CO₂/95% air in Eagle’s minimum essential medium (MEM; Gibco, Grand Island, New York, NY, USA) containing 10% fetal calf serum and 100 µg/ml kanamycin (Meiji Seika, Tokyo, Japan). Human peripheral blood monocytes were cultured in RPMI 1640 medium (Sigma-Aldrich, St. Louis, MO, USA) containing 10% fetal calf serum and 100 µg/ml kanamycin.

Three chemical skin sensitizers, oxazolone (4-ethoxymethylene-2-phenyl-2-oxazolin-5-one; OXA, purity ≥90%, E0753), formaldehyde (FA, purity 36.5–38%, F8775), and 2,4-dinitrochlorobenzene (DNCB, purity 97%, 138630), and three respiratory sensitizers, ortho-phthaldialdehyde (OPA, purity ≥97%, P1378), hexamethylene-1,6-diisocyanate (HDI, purity ≥98%, 52650), and trimellitic anhydride (TMA, purity 97%, B4600), were purchased from Sigma-Aldrich (Table S1 in Supplementary Material). A magnetic bead conjugated with monoclonal antibody (mAb) against CD14 was purchased from Miltenyi Biotec (Bergisch Gladbach, Germany). mAbs for human CD14 (HCD14), CD11c (cllne 3.9), and CD11c (EP1347Y) were obtained from BioLegend (San Diego, CA, USA), eBioscience (La Jolla, CA, USA), and Abcam (Cambridge, UK), respectively. Alex Fluor 647 anti-rabbit IgG was purchased from Thermo Fisher Scientific (Waltham, MA, USA). Human recombinant granulocyte macrophage colony-stimulating factor (GM-CSF) was purchased from Miltenyi Biotec. Human recombinant IL-4 was kindly provided from Schering-Plough (currently, Merck, Kenilworth, NJ, USA).

Fresh human peripheral blood was collected from healthy volunteers and mononuclear cells were immediately purified by using Lympholyte-H (Cedarlane, Burlington, ON, Canada) density gradient centrifugation. Monocytes were further purified from peripheral blood mononuclear cells by positive selection using an AutoMACS Pro with a magnetic bead conjugated with mAb against CD14 (Miltenyi Biotec). The purity was analyzed by flow cytometry with PE anti-CD14 using a FACS Canto II (BD Bioscience, San Jose, CA, USA) followed by analysis with FlowJo Software (Tree Star, Ashland, OR, USA), and was routinely more than 99%.

This study was approved by the institutional review board of Tokyo Medical University (no. 3323). Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

Human CD14⁺ monocytes (1 × 10⁶ cells/ml) were stimulated in a 24-well plate by GM-CSF (50 ng/ml) and IL-4 (10 ng/ml) for 6 days. The purity of resultant immature DCs was analyzed after staining with anti-CD11c and was routinely more than 99%.

The 3D coculture was prepared by using Alvetex scaffold 12-well inserts or 24-well plates, which were purchased from ReproCell (Glasgow, UK) (23). The scaffold was initially washed sequentially with ethanol, PBS, and medium according to the manufacture’s instructions. A 75-µl aliquot of cell suspension of BEAS-2B or MRC-5 cells (1.5 × 10⁶ cells) was gently seeded directly onto the center of the scaffold in the 12-well insert and left for 4 h to allow cell attachment. Then, medium was gently added on the scaffold and incubated for 3 days. A 50-µl aliquot of cell suspension of immature DCs (0.7–1.0 × 10⁶ cells) was gently seeded directly onto the center of the scaffold in the 24-well plate and left for 6 h. Then, an equal mixture of MEM and RPMI 1640 containing GM-CSF (50 ng/ml) and IL-4 (10 ng/ml) was gently added on the scaffold and incubated for 24 h. After incubation, the individual scaffolds were gently detached from the 12-well insert or 24-well plate and piled up in the order of MRC-5 cells (bottom), immature DCs (middle), and BEAS-2B cells (top), and attached to the bottom of a new sterile 12-well insert (Figure 1). Subsequently, the 12-well insert was placed in a 12-well plate, and MEM was gently added and incubated for another 4 h. Then, the medium was gently removed, and a 5-µl aliquot of a chemical sensitizer, which was initially dissolved in DMSO and then diluted with MEM, was gently added at six places on the top scaffold and left for 30 min followed by addition of MEM medium (2 ml). After stimulation for 9 h, the piled scaffold was disassembled, and RNA was extracted from the respective scaffolds for real-time RT-PCR analysis. After stimulation for 24 h, each scaffold was immersed in OCT embedding compound, snap-frozen in liquid nitrogen, and stored at −80°C until use.

Frozen samples were cut into10-μl sections by using a cryostat (Microm HM550, Thermo Fisher Scientific), fixed in 10% formalin, and stained with hematoxylin and eosin (HE). For DC staining with anti-CD11c (EP1347Y, Abcam), the fixed sections were incubated in a 0.05% Tween 20/citrate buffer (pH 6.0) solution at 60°C overnight to retrieve antigenicity according to the manufacture’s instructions. Then, the sections were stained with anti-CD11c followed by Alexa Fluor 647 anti-rabbit IgG and counterstained with Hoechst 33258 (Dojindo, Kumamoto, Japan). Resultant sections were examined by using a confocal laser microscope (Fluoview FV10i, Olympus, Tokyo, Japan) or a fluorescence microscope (Biozero, Keyence, Osaka, Japan).

Total RNA was prepared from individual scaffolds using an RNeasy Mini Kit (Qiagen, Hilden, Germany), and cDNA was prepared using oligo(dT) primer and SuperScript III RT (Thermo Fisher Scientific). Real-time quantitative PCR was performed using SYBR Premix Ex Taq II and a Thermal Cycler Dice real-time system according to the manufacturer’s instructions (Takara, Otsu, Shiga, Japan). Hypoxanthine phosphoribosyltransferase (HPRT) was used as a housekeeping gene to normalize mRNA. Relative expression of real-time PCR products was determined by using the ΔΔCt method to compare target gene and HPRT mRNA expression. Primers used in this study are listed in Table S2 in Supplementary Material.

Data are presented as mean ± SD in triplicate. Statistical analyses were performed by using two-tailed Student’s t-test for comparisons of two groups and one-way analysis of variance and the Tukey–Kramer multiple comparison test for comparing more than three groups using GraphPad Prism 7 (GraphPad Software Inc., La Jolla, CA, USA). The receiver-operating characteristic curve and the area under the curve were used to assess the cut-off level between two groups. P < 0.05 was considered to indicate a statistically significant difference.

To mimic human upper airway epithelium, we first established a 3D coculture system consisting of airway epithelial cell line BEAS-2B, peripheral blood mononuclear cell-derived immature DCs, and lung fibroblast cell line MRC-5 (Figure 1), using the Alvetex scaffold, which is made of 200-μm-thick polystyrene and is porous with approximately 36–40 µm voids (23). The scaffold in the bottom of the insert well can be detached easily. In addition, the detached scaffolds can be attached sterilely to the bottom of the insert even after piling up, which is the best way to use this scaffold in this system. Histochemical analysis with HE staining revealed that the three sets of cells migrated into the respective scaffold and spread throughout it (Figure 2).

A previous study showed that DCs migrate into the upper epithelial scaffold in response to challenge by a protein allergen (12), but the physiological relevance and molecular mechanism of this phenomenon remain elusive. After stimulation with OXA or OPA for 24 h, immunohistochemical analysis was performed using anti-CD11c mAb together with nuclear staining with Hoechst to evaluate the migration of DCs. In the scaffold layer of DCs, no obvious decrease in the number of DCs was observed after stimulation with either chemical sensitizer (Figure 2A). Consistent with this, almost no DCs were detected in the BEAS-2B and MRC-5 scaffolds (Figures 2B,C). When higher concentrations of sensitizers were used, although the number of DCs in the scaffold decreased, no migration of DCs into other scaffolds was observed (data not shown). Likewise, no migration of DCs into other scaffolds was observed after stimulation with DNCB or TMA, although some cytotoxicity was seen under these conditions (Figure S1 in Supplementary Material). Thus, in our 3D coculture system, DCs did not migrate into other scaffolds containing BEAS-2B or MRC-5 cells.

After disassembling the 3D coculture system, we performed mRNA expression analysis on individual scaffolds. We first evaluated the cytotoxicity of chemical sensitizers by analyzing the expression level of HPRT. The skin and respiratory chemical sensitizers, OXA and OPA, respectively, were added on the 3D coculture system. After stimulation for 3, 6, 9, and 12 h, RNA was extracted from individual scaffolds and subjected to real-time RT-PCR analysis. A dose-dependent decrease of mRNA expression of HPRT was observed in the DC scaffold after stimulation for 9 h (Figure S2 in Supplementary Material) and at other times (data not shown). A higher concentration of sensitizers greatly decreased the mRNA expression of HPRT, which was probably caused by degradation of RNA due to cell damage. Therefore, we did not use samples whose HPRT mRNA expression level was less than 1/10 that of untreated sample.

The induction level of molecules important for DC maturation, such as co-stimulatory molecules CD86 and CD80, major histocompatibility antigen class II HLA-DR, and chemokine receptor CCR7, was examined after stimulation for 9 h (Figure 3A), which appeared to be the optimal time (data not shown). Both chemical sensitizers increased the expression of the co-stimulatory molecules, but not HLA-DR and CCR7, and no difference in the induction level was observed between chemical skin and respiratory sensitizers. In contrast, the expression of OX40L (also known as tumor necrosis factor ligand superfamily member 4, TNFSF4) was preferentially increased by stimulation with OPA but not with OXA (Figure 3B). This is noteworthy because OX40L was demonstrated to play a critical role in induction of Th2 differentiation through interaction with OX40 expressed on activated T cells (24, 25). However, thymic stromal lymphopoietin (TSLP), IL-12p35, IL-12p40, intracellular adhesion molecule-1 (ICAM-1), IL-10, IL-1β, and IL-8 did not show any significant and preferential upregulation by treatment with these chemical sensitizers (Figures 3B,C).

Thus, OXA and OPA similarly upregulate CD86 concomitant with DC maturation. However, during the DC maturation, OPA (but not OXA) preferentially augments the expression of OX40L.

Next, the effects of chemical sensitizers on other molecules in upper airway epithelial cell line BEAS-2B and lung fibroblast cell line MRC-5 were examined. After stimulation with OXA or OPA for 9 h, RNA was extracted from the respective scaffolds and subjected to real-time RT-PCR analysis. However, we observed no significant and preferential upregulation of molecules including TSLP, IL-33, IL-10, and IL-4 (Figure 4), which were reported to be critically involved in the induction of Th2 differentiation and exertion of Th2 immune responses (20, 26, 27). Thus, no preferential expression was observed by stimulation with OXA or OPA in BEAS-2B and MRC-5 cells.

To expand the applicability of this 3D coculture system to discriminate chemical skin and respiratory sensitizers, we used another set of typical skin and respiratory sensitizers, FA and HDI, respectively. FA and HDI both upregulate CD86 in DCs, and no difference was observed in the induction level of CD86 in DCs, suggesting similar induction of DC maturation (Figure 5A). Consistent with the results obtained by using OXA and OPA, HDI but not FA preferentially upregulated OX40L in DCs. Moreover, preferential upregulation of ICAM-1 and IL-1β was also observed by stimulation with FA as compared to that with HDI (Figures 5B,C). ICAM-1 was reported to be involved in the promotion of Th1 differentiation in T cells (28). No other preferential expression was observed by stimulation with them except for the preferential upregulation of IL-4 by FA in MRC-5 cells, which is opposite to what is expected (Figure S3 in Supplementary Material). Similar preferential upregulation of OX40L but comparable upregulation of CD86 in DCs was observed when the third set of chemical sensitizer, DNCB and TMA, was used (Figure 6). Collectively, these results suggest that both chemical skin and respiratory sensitizers similarly upregulate CD86 and induce DC maturation, but only respiratory sensitizers preferentially and consistently upregulate OX40L in DCs, indicating that OX40L expression in DCs is a good marker for discrimination between chemical respiratory and skin sensitizers.

Finally, to validate the applicability of this 3D coculture system to predict chemical skin and respiratory sensitizers, the responsiveness of all six chemical sensitizers was examined simultaneously using respective optimum concentrations. The sensitizers similarly upregulated CD86 in DCs and induced DC maturation (Figure 7A), but only the three respiratory sensitizers tended to more strongly upregulate OX40L. Moreover, an average of mean-fold of the induction of CD86 expression by sensitizers relative to the untreated expression was comparable between the sets of skin and respiratory sensitizers (Figure 7B). However, the average of mean-fold of the induction of OX40L expression of the respiratory sensitizers relative to untreated expression was significantly increased compared with that of the skin sensitizers; receiver-operating characteristic curve analysis revealed that the cut-off level to discriminate skin and respiratory sensitizers is 6.81-fold (Figure 7B).