All mouse experiments were performed following the NIH Guide for the Care and Use of Laboratory Animals recommendations. The animal protocol was approved by the University of Michigan Committee on Use and Care of Animals. The human study was approved by the University of Michigan Institutional Review Board.

Two-day old wild type C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) were exposed to air or 75% oxygen for up to 14 days using a polypropylene chamber coupled to an oxygen controller and sensor (BioSpherix, Lacona, NY) (51). Dams were exchanged between air and hyperoxia daily. In selected experiments, mice were in vivo treated with an anti-Flt3L antibody (R&D, Minneapolis, MN) or a control IgG (BioXCell, West Lebanon, NH) daily intranasally 0.5mg/kg (in 10ul per mouse) during the exposure to normoxia or hyperoxia. This anti-Flt3L antibody is reported to neutralize the Flt3L-Flt3 signaling in vitro and in vivo (38, 52, 53). In selected experiments, upon completion of hyperoxic exposure, mice were treated with 30 µl of RV1B (3 x 10⁸ PFU/ml), or sham infection (HeLa cell lysate) administered intranasally. Lungs were analyzed immediately after hyperoxia or 5 days after RV infection.

Lungs were perfused with PBS containing EDTA (0.5 mM) to remove circulating cells in the vasculature, then minced and digested with Liberase TM (100 µg/mL; Roche, Indianapolis, IN), together with collagenase XI (250 µg/mL), hyaluronidase 1a (1 mg/mL), and DNase I (200 µg/mL; Sigma, St. Louis, MO) for 1 hour at 37°C (54). Cells were filtered and treated with RBC lysis buffer (BD Biosciences, Franklin Lakes NJ) and kept on ice in media containing 10% serum. Dead cells were stained with Pac-Orange Live/Dead fixable dead staining dye (Invitrogen) or DAPI when flow sorting was applied. Lung cells were then stained with fluorescent-labeled antibodies against various leukocyte surface markers (CD45, F4/80, CD11c, CD103, CD11b, CD3ε, CD4, CD8, CD69, and IFN-γ). Appropriate isotype-matched controls and Fluorescence Minus One (FMO) controls were used in all experiments. Antibodies were purchased from Biolegend (San Diego, CA). Cells were fixed and analyzed on a Fortessa (Becton-Dickinson, San Jose, CA) or flow sorted by FACSAria II (BD Biosciences) flow cytometer. Flow sorting purity was >95%. Results were analyzed using FlowJo software (Tree Star, Ashland, OR).

Flow sorted CD103+DCs and CD11b+DCs were used for a gene array (Affymetrix Gene Chip (Affymetrix, CA)) by University of Michigan Sequence Core. RNA was extracted from the sorted cells using the RNeasy Plus Micro Kit (Qiagen, Germantown, MD). The microarray analysis was based on Mouse Gene ST 2.1 strips using the Affy wt-pico kit. The expression values for each gene were calculated using a robust multi-array average (RMA). Weighted linear models were fitted to the data and computed the contrasts of interest. Probesets with variance less than 0.04 were filtered out. Probesets with a fold change of 1.5 or greater and an adjusted p-value of 0.05 or less were selected. P-values were adjusted for multiple comparisons using false discovery rate Benjamini and Hochberg. The oligo and limma packages of Bioconductor were impremented in the R statistical environment.

Mouse whole-lung RNA was prepared using TRIzol (Invitrogen, Carlsbad, CA). Flow sorted cells RNA was extracted using the RNeasy Plus Micro Kit (Qiagen, Germantown, MD). Gene mRNA expression (Flt3, CSF1R, Flt3L, GM-CSF, CD207, SlamF9, Myd88, CXCL16, CXCL9, and IL-12p40) was quantified using SYBR green technology. Primer sequences are available upon request. The level of gene expression was normalized to mRNA of β-actin or GAPDH using the 2^-ΔCT algorithm.

For graphic representation, the mRNAs for selected target genes relative to β-actin were plotted multiplied by 10^-3 as indicated in the figure legends.

Whole mouse lung homogenates in PBS with Protease inhibitor cocktail (Roche, Mannheim, Germany) were centrifuged and supernatants analyzed for proinflammatory cytokines. Flt3L and GM-CSF were measured by ELISA (all from R&D Systems, Minneapolis, MN).

RV1B (from American Type Culture Collection, Manassas, VA) was grown in HeLa cells, concentrated, partially purified, and titrated as described (55). Viral titers were measured by plaque assay (56). RV copy number was quantified by a Taqman quantitative PCR method with primers and FAM labeled probe that specifically recognizes RV (57). The primer and the probe sequences are available upon request.

We examined tracheal aspirates from infants admitted to the C.S. Mott Children’s Hospital Newborn Intensive Care Unit. Entry criteria included gestational age at birth ≤ 32 weeks, mechanical ventilation for respiratory distress, and age ≤ 7 days. Aspirates were collected during routine tracheal suctioning of mechanically ventilated premature infants as described (58). Patient demographics are included in Table 1 . Human Flt3L, IL-12p40, IL12-p70, IFNr and IL-5 were measured using the tracheal aspirate with a customized Multiplex ELISA (MilliporeSigma, Burlington, MA).

Data are represented as mean ± SEM or median and interquartile range (IQR). Statistical significance was determined by unpaired t-test, nonparametric rank-sum test, one-way ANOVA, as appropriate. Differences were pointed by Newman-Keuls’ multiple comparisons test.

Neonatal hyperoxia impairs alveolar and vascular development (18, 59, 60), however little is known about the effects on neonatal lung immune cell development. We have previously reported that in neonatal mice exposure to hyperoxia for 14 days, a model of BPD, increases the number of lung IL-12-producing CD103+ DCs, and these cells are further increased after RV infection (49). We have also shown that CD103+ DCs are required for hyperoxia-induced proinflammatory responses and airway hyperresponsiveness following RV infection (50). These results implicate hyperoxia-activated neonatal lung CD103+ DCs as a mediator of the inflammatory responses induced by hyperoxia. We have not determined yet the timeline and mechanisms regulating the development of lung CD103+ DCs during early-life hyperoxia.

In this study, we investigated the longitudinal effects of neonatal hyperoxia on lung CD103+ DCs and compared with the effects on lung CD11b^hi DCs. We exposed two-day old mice to normoxia or hyperoxia (0.75 FiO2) continuously for up to 14 days as described previously (49, 50). We monitored the abundance of lung CD103+DCs and CD11b^hi DCs after 4, 7 and 14 days of exposure ( Figure 1 ). This period corresponds to rapid alveolar growth in mice (59), and is reminiscent of the exponential increase in alveolar surface area in young human infants (60), thus making this model relevant to understand human BPD pathology. We also examined the long-term effects of neonatal hyperoxia on lung cDC abundance 30 days after hyperoxia was discontinued, on day of life (DOL) 45, corresponding to young adult age. To quantify lung cDCs and critical macrophage populations, we examined lungs by flow cytometry and performed cell sorting of CD103+DCs and CD11b^hi DCs to determine the total number of these cell populations per lung ( Figure 1A ). After gating on Live, CD45+ cells, we differentiated lung DCs from alveolar macrophages and interstitial macrophages based on F4/80 and CD11c expression as described previously (49, 61). The fraction of total DCs (CD11c+F4/80- cells) was increased as early as 7 days after hyperoxia began and further increased after 14 days of exposure ( Figure 1B ). Additionally, we examined the two lung cDC subsets, CD103+ DCs and CD11b^hi DCs. Under normoxia, the total number of lung CD103+ DCs and CD11b^hi DCs increased between exposure day 4 (DOL 6) and exposure day 14 (DOL16) ( Figure 1A , open bar). In the adult lung (DOL45), the total number of CD103+ DCs remained stable, whereas the total number of CD11b^hi DCs further increased. Compared to normoxia, as early as 7 days after hyperoxia the total number of CD103+ DCs and CD11b^hi DCs increased significantly, and this number further increased after 14 days of exposure ( Figure 1A , black bar). Interestingly, the effect of hyperoxia on the number of CD103+DCs, but not CD11b^hi DCs persisted in adult lungs 30 days after hyperoxia was discontinued. The effects of hyperoxia on other lung myeloid cells included an increase in the exudative macrophages (F4/80+CD11c+CD11b+, ExMfs) after 14 days exposure and a decrease in alveolar macrophages (F4/80+CD11c+CD11b-, AMs) after 7- and 14-day exposure ( Figure 1B ; Supplemental Figure 1 ). There were no differences in the fraction of interstitial macrophages (IMs) and neutrophils. These results show that neonatal hyperoxia alters the lung innate immune cell milieu with selective expansion of lung CD103+ DCs and CD11b^hi DCs.

To further define the effects of neonatal hyperoxia on lung cDCs, we examined the transcriptional profile of lung CD103+ DCs and CD11b^hi DCs isolated from normoxia- and hyperoxia-exposed neonatal lungs. First, we ensured that we can precisely sort the two lung cDC populations from normal neonatal lungs. We established that the differentially expressed genes between neonatal lung CD103+ DCs and CD11b^hi DCs closely matched the reported gene signature for lung and other nonlymphoid organ cDC types (30, 62). CD103+ DCs differentially expressed CD207 (Langerin), CD103 (Itgae), Clec9a, Xcr1, Tlr3, Irf8, Flt3, CD86 and other CD103+ DC-specific genes, while CD11b^hi DCs differentially expressed CD11b (Itgam), CD209d (DC-SIGN), Clec4a1, Tlr7, Tlr2, Cx3cr1, Irf4, CD172a and other CD11b DC-specific genes ( Figures 2A–E ). Both DC subsets expressed high levels of CD26 and low levels of CD88, consistent with a preDC- and not monocyte-derived origin ( Figure 2E ) (63). Next, we evaluated the effects of hyperoxia on the two lung cDC types. Compared to normoxic lung CD103+ DCs, hyperoxic lung CD103+ DCs differentially expressed 471 genes ( Figure 2F ). We analyzed the differentially expressed genes in hyperoxic lung CD103+ DCs for over-representation of pathways and gene ontology terms using the ConcensusPathDB database (http://cpdb.molgen.mpg.de/) (64, 65). We identified 60 enriched pathway-based sets of genes as defined by the KEGG and Reactome databases ( Supplemental Table S1 ). Relevant over-represented pathways reflecting functional changes in the CD103+ DCs after hyperoxia included pathways related to generation of second messenger molecules (8 genes, p-value = 1.85e-09, q-value = 9.54e-07), NF-kappa B signaling pathway (12 genes, p-value = 1.65e-06, q-value = 7.06e-05), signal transduction (64 genes, p-value = 0.000425, q-value = 0.00723), MAPK signaling pathway (14 genes, p-value = 0.00321, q-value = 0.0405) and RAS signaling pathway (12 genes, p-value = 0.00321, q-value = 0.0405) ( Table 2 ). We also found 65 enriched gene ontology-based sets of genes ( Supplemental Table S2 ). The over-represented sets included gene ontology terms for biological processes related to immune cell activation and development such as leukocyte activation (52 genes, p-value = 1.29e-12, q-value = 4.83e-11), cell motility (68 genes, p-value = 8.76e-12, q-value = 1.46e-10), anatomical structure development (162 genes, p-value = 9.25e-11, q-value = 1.26e-09), response to stress (115 genes, p-value = 4.39e-10, q-value = 4.39e-09) and cytokine production (37 genes, p-value = 1.04e-08, q-value = 9.72e-08) ( Table 3 ). Among the top 25 highest upregulated genes were CD207 (Langerin, an endocytic receptor involved in cross-presentation to CD8+ T cells), Slamf9 (CD2-F10, involved in T cell and NK cell activation), the purinergic receptors P2ry12 and P2rx5, oxidative-stress responsive 1 (Oxsr1), Cxcl9, Myd88, Stat2 and others ( Figure 2G ). Among the most downregulated genes were genes involved in cell migration and in vasculature development, including Itga6, Ang, Aqp1, Igfbp5, Ldb2, Enpp2, Plk2 and others ( Figure 2H ). Comparison of normoxic lung CD11b^hi DCs and hyperoxic lung CD11b^hi DCs revealed 550 differentially expressed genes ( Figure 2I ). We identified 37 enriched pathway-based sets of genes as defined by the KEGG and Reactome databases ( Supplemental Table S3 ). Over-represented pathways reflecting functional changes in the CD11b^hi DCs after hyperoxia included pathways related to innate immune system (53 genes, p-value = 4.15e-08, q-value = 4.51e-06), cell surface interactions at the vascular wall (12 genes, p-value = 7.45e-05, q-value = 0.00276) and immunoregulatory interactions between lymphoid and non-lymphoid cells (11 genes, p-value = 7.63e-05, q-value = 0.00276) ( Table 4 ). Additionally, we found 55 enriched gene ontology-based sets of genes ( Supplemental Table S4 ). The over-represented sets included gene ontology terms for biological processes related to response to stress (143 genes, p-value = 3.26e-12, q-value = 1.61e-10), leukocyte activation (54 genes, p-value = 5.14e-10, q-value = 1.3e-08) and cytokine production (45 genes, p-value = 5.25e-10, q-value = 1.3e-08) ( Table 5 ). Among the top 25 most upregulated genes in CD11b^hi DCs from hyperoxia-exposed neonatal lungs were predominantly genes involved in regulation of cellular metabolic processes, including Coq3, Scimp, Mapre3, Usp1, Fnip2, Creb5, Epcam, Pla2g4a, Zfp948, Spint1, Nbn, Mvp and Lpar3 ( Figure 2J ). The most downregulated genes in hyperoxic lung CD11b^hi DCs included genes regulating immune responses (Cd177, Ace, Cd6, Cd79a, Gadd45a, Cd79b, Pglyrp1, Treml4) ( Figure 2K ).

Next, we sought to confirm the microarray results. First, we focused on examining the expression of the two receptors that regulate cDCs development, Flt3 and CSF1R in normal conditions (33). Flt3 was among the differentially expressed genes between neonatal lung CD103+ DCs and CD11b^hi DCs ( Figures 2B, E ). Using qPCR, we confirmed that Flt3 mRNA expression in flow sorted neonatal lung CD103+ DCs was significantly higher compared to flow sorted CD11b^hi DCs. ( Figure 3A ). As previously shown in adult CD11b^hi DCs (33), neonatal lung CD11b^hi DCs expressed higher levels of Csf1r ( Figure 3A ). Flow cytometry analysis of lung cDCs at DOL 9 confirmed that most CD103+ DCs expressed Flt3 protein compared to less than 50% of the CD11b^hi DCs ( Figure 3B ). The average Flt3 protein expression per cell presented as mean fluorescence intensity (MFI) was also higher in CD103+ DCs compared to CD11b^hi DCs ( Figure 3B ).

Additionally, the microarray showed that hyperoxia induced the expression of proinflammatory genes in CD103+ DCs and CD11b^hi DCs. Using qPCR, we confirmed that hyperoxic CD103+ DCs expressed higher levels of CD207, Slamf9 and Myd88 mRNA compared to normoxic CD103+ DCs. As for the CD11b^hi DCs, hyperoxia induced the mRNA expression of Cxcl9 and Cxcl16 ( Figure 3C ).

Together these results indicate that early-life hyperoxia not only numerically increases lung cDCs, but also induces unique proinflammatory transcriptional signatures in each cDC subset. These results also show higher Flt3 expression in CD103+ DCs, which may confer a developmental advantage upon stimulation with Flt3 ligand.

To investigate the mechanisms by which early-life hyperoxia induces lung cDC expansion and activation, we quantified the expression of the DC growth factors Flt3L and granulocyte-macrophage colony-stimulating factor (GM-CSF) in normoxia- and hyperoxia-exposed neonatal lungs. Flt3L plays a critical role in CD103+ DC development and partially regulates the development of CD11b^hi DCs (33). GM-CSF directs differentiation of DCs in vitro (66), and regulates nonlymphoid tissue DC development (67). Additionally, GM-CSF is a key driver of chronic lung inflammation contributing to asthma development (68). We found that exposure of two-day old mice to hyperoxia for 1 and 7 days significantly increased Flt3L mRNA and protein levels ( Figure 4A ), whereas the expression pattern of GM-CSF was different. GM-CSF mRNA expression was significantly induced after 1 day but not after 7 days of hyperoxia and GM-CSF protein levels were not affected by hyperoxia ( Figure 4B ). These results indicate that early-life hyperoxia selectively induces the Flt3L/Flt3 pathway.

We sought to determine if Flt3L is required for neonatal lung cDC development under normoxic and hyperoxic conditions. Neonatal mice were exposed to normoxia or hyperoxia from DOL2 to DOL11 and treated daily with anti-Flt3L or isotype control IgG. On DOL14, lung CD103+ DCs and CD11b^hi DCs were quantified by flow cytometry ( Figures 5A, B ). Total lung DCs were identified by sequential gating of live, CD45+, F4/80-, CD11c+ cells. The two types of lung cDCs were identified based on CD103 and CD11b expression. Like the results shown in Figure 1 , in IgG-treated mice hyperoxia increased the fraction of CD103+ DCs and CD11b^hi DCs. Anti-Flt3L treatment significantly reduced the lung CD103+DCs in normoxic and hyperoxic conditions ( Figures 5A, B ). Anti-Flt3L did not affect the fraction of CD11b^hi DCs in normoxic neonatal lungs but attenuated the effect of hyperoxia on these cells. These results are consistent with the notion that Flt3L is critically required for neonatal lung CD103+ DC development and for the effects of hyperoxia on this cell population.

We have previously shown that neonatal hyperoxia induces the expression of the proinflammatory IL-12p40 cytokine in whole lung and this increase is dependent on Batf3, a transcription factor required for CD103+ DC development (49, 50). We examined the effects of Flt3L neutralization on whole lung IL-12p40 expression under normoxic and hyperoxic conditions. We found that anti-Flt3L treatment blocked hyperoxia-induced IL-12p40 mRNA expression ( Figure 5C ). These results indicate that Flt3L-dependent DCs are required for hyperoxia-induced proinflammatory IL-12p40 expression.

In extensive prior studies, we have reported that neonatal hyperoxia induces proinflammatory lung CD4+ T cell activation following RV infection (49), and that CD103+ DCs are required for these effects (50). Above we showed that anti-Flt3L treatment significantly reduced the number of lung CD103+DCs in normoxic and hyperoxic neonatal mice. We sought to evaluate the effect of in vivo blockade of Flt3L on lung T cell activation in hyperoxia-exposed, RV-infected neonatal mice. We exposed neonatal wild-type mice to hyperoxia from DOL 2 to 11 and treated them with anti-Flt3L or isotype control IgG daily. Upon discontinuing the exposure we inoculated the mice with RV (RV1B, 3X10⁷ PFU) or sham, administered intranasally. Using flow cytometry, we examined lungs five days after RV infection and quantified the frequencies of CD4+ T cells and CD8+ T cells expressing CD69, T cell activation marker. Lung T cells were identified by sequential gating for cells that are Live CD45+ CD3ε+ and either CD4+ or CD8+ ( Supplemental Figure 2 ). Anti-Flt3L treatment did not affect the frequencies of total CD4+ T cells and CD8+ T cells ( Figure 6A ). Compared with IgG-treated mice, anti-Flt3L-treated neonatal mice had significantly lower frequency of activated CD69+ CD4+ and CD8+ T cells ( Figure 6B ). There was no difference in RV copy numbers between IgG and anti-Flt3L treatment groups ( Supplemental Figure 3 ). The results indicate that the effect of hyperoxia on lung CD4+ and CD8+ T cell activation following RV infection is mediated by Flt3L-expanded/activated CD103+DCs.

To establish the relevance of our findings to human disease, we measured levels of FLT3L, IL-12p70, IL-12p40 and IFN-γ in tracheal aspirates from human preterm infants mechanically-ventilated for respiratory distress in the first week of life. Patient characteristics are described in Table 1 . We defined BPD based on the need for supplemental oxygen at 36 weeks postmenstrual age. For prematurely born infants, death before 36 weeks postmenstrual age is a competing outcome for BPD and may confound a cause-specific analysis, therefore we combined both outcomes. We found that tracheal aspirates from premature infants who go on to develop BPD have significantly higher FLT3L and IL-12p70 protein levels compared to aspirates from infants who do not develop BPD ( Figure 7A ). IL-12p40 and IFN-γ protein levels also showed a similar trend with borderline statistical significance, whereas levels of IL-5, a type 2 cytokine, were similar ( Figure 7A ). Additionally, FLT3L protein levels positively correlated with levels of IL-12p70, IL-12p40 and IFN-γ ( Figure 7B ). These results suggest that higher levels of FLT3L and proinflammatory cytokines IL-12p70, IL-12p40 and IFN-γ are associated with BPD diagnosis, a predictor of higher chronic respiratory morbidity.