Low-dose hyperoxia primes airways for fibrosis in mice after influenza A infection

Abstract

It is well known that supplemental oxygen used to treat preterm infants in respiratory distress is associated with permanently disrupting lung development and the host response to influenza A virus (IAV). However, many infants who go home with normally functioning lungs are also at risk for hyperreactivity after a respiratory viral infection. We recently reported a new, low-dose hyperoxia mouse model (40% for 8 days; 40×8) that causes a transient change in lung function that resolves, rendering 40×8 adult animals functionally indistinguishable from room air controls. Here we report that when infected with IAV, 40×8 mice display an early transient activation of TGFβ signaling and later airway hyperreactivity associated with peribronchial inflammation (profibrotic macrophages) and fibrosis compared with infected room air controls, suggesting neonatal oxygen induced hidden molecular changes that prime the lung for hyperreactive airways disease. Although searching for potential activators of TGFβ signaling, we discovered that thrombospondin-1 (TSP-1) is elevated in naïve 40×8 mice compared with controls and localized to lung megakaryocytes and platelets before and during IAV infection. Elevated TSP-1 was also identified in human autopsy samples of former preterm infants with bronchopulmonary dysplasia. These findings reveal how low doses of oxygen that do not durably change lung function may prime it for hyperreactive airways disease by changing expression of genes, such as TSP-1, thus helping to explain why former preterm infants who have normal lung function are susceptible to airway obstruction and increased morbidity after viral infection.

INTRODUCTION

It is well accepted that early oxygen (O₂) exposure in preterm infants can disrupt lung development and function in extremely low gestational age newborns (ELGANs, <29 wk’ gestation), often resulting in bronchopulmonary dysplasia (BPD). Cumulative O₂ exposure strongly predicts BPD diagnoses and severity, but ∼40% of ELGANs escape the BPD “label” because they are weaned off O₂ or respiratory support by 36 wk’ corrected age despite having significant supplemental O₂ exposure (1, 2). Former ELGANs with and without BPD experience increased morbidity that can be linked to cumulative O₂ exposure with increased health care utilization, symptomatic airway disease, and asthma medication use (3–5). Former ELGANs are especially vulnerable to respiratory viral infections with increased early childhood hospitalizations for respiratory illnesses (6–8) through poorly understood mechanisms. Infection-related lung injury results in airway remodeling and longer-term airway hyperreactivity (AHR) (9–12), and increased prescriptions for asthma-related medications (13, 14). Notably, the airway dysfunction in former ELGANs is not bronchodilator responsive, distinguishing it from asthma (15). Thus, there is an urgent need to uncover novel mechanisms responsible for airway dysfunction and wheezing in former ELGANs.

Neonatal hyperoxia is one of the most commonly used exposures in animal models to perturb lung development and model BPD (16). Mice are born in the saccular stage of lung development when airways continue developing and alveolar structure is in its primitive stages, analogous to ELGANs (17). The dose and duration of hyperoxia matter when modeling neonatal O₂ exposure in mice. For example, multiple studies, including several from our own laboratory, show that severe hyperoxia (≥ 60% O₂ for ≥ 4 days) creates a BPD-like phenotype (16, 18) with alveolar simplification, airway remodeling, and viral susceptibility (19–27), even after a long period of room air recovery. These previous models, however, are limited because they often use O₂ doses higher than those seen in real-world NICU settings and cause such profound alveolar simplification that it is difficult to discern physiological changes in the airway. This led our laboratory to develop a translational model of low-dose chronic hyperoxia (40% O₂ for 8 days; 40×8) that causes transiently increased airway resistance and decreased lung compliance with AHR and airway smooth muscle hypertrophy at 4 wk (28), consistent with other studies (29). Interestingly, abnormal lung function, AHR, and smooth muscle hypertrophy all resolve at 8 wk (28), when mice are morphologically and functionally “normal.” Taken together, our model of lower O₂ exposures in mice shows changes in airway function without overt signs of alveolar simplification. We suggest this may more accurately replicate modern preterm infants with increased respiratory morbidity after leaving the NICU without a diagnosis of BPD.

Neonatal hyperoxia also has functional implications when adult mice are challenged with influenza A viral (IAV) infection. Previous studies by our laboratory have shown how infecting adult mice exposed to high doses (100% for 4 days) of hyperoxia as neonates with HKx31 H3N2 IAV causes persistent inflammation, parenchymal fibrosis, and mortality (19–21, 23, 24, 27). Although high doses of hyperoxia cause alveolar simplification in adult mice, it does not affect epithelial infection or clearance of the virus. Instead, increased severity of IAV was mediated in part by hyperoxia reducing the number of adult alveolar epithelial type 2 cells available for regenerating alveolar epithelial type 1 cells. Because 40×8 hyperoxia does not deplete AT2 cells, promote alveolar simplification, or cause long-term changes in lung function, we hypothesized that it would not affect how adult 40×8 mice would respond when infected with the same virus. Here, we present new and compelling evidence with 40×8 mice that seemingly harmless amounts of hyperoxia at birth can prime the lung for hyperreactive airways disease following IAV infection and identify a potential molecular change that can be used as predictive biomarker and/or mediator of lung disease.

METHODS

Animal Exposures and Infection

All protocols were approved by the Institutional Animal Care and Use Committee of University of Rochester (Rochester, NY) and were consistent with The Association for Assessment and Accreditation of Laboratory Animal Care International policies (Frederick, MD). Litters of C57Bl/6J (Jackson Laboratory, Bar Harbor, ME) were placed into room air (RA) or 40% oxygen from postnatal day (PND) 0–8 as previously described (28). Nursing dams were rotated every 24–48 h. After exposure, pups of both sexes were allowed to mature until PND 56 under room air conditions where a subset of naïve mice were harvested for pulmonary function, protein, or qRT-PCR analysis. Equal numbers of male and female mice were lightly anesthetized with ketamine/xylazine mixture and given 10⁵ plaque-forming units (PFUs) influenza A (HKx31/H3N2) virus, which was grown and titered in Madin-Darby Canine Kidney (MCDK) cells as previously described (30). Mice were weighed every other day for 2 wk after infection, then weekly thereafter.

Bronchoalveolar Lavage

Bronchoalveolar lavage (BAL) was performed in a subset of animals at postinfection day (PID) 3, 7, 10, and 14 with three separate 1 mL aliquots of ice-cold phosphate-buffered saline (PBS, Thermo Fisher Scientific, Hampton, NH), as previously described (19). The first supernatant was collected for protein analysis and frozen at −80°C for further analysis.

Cell Differentiation

BAL fluid (BALF) from all three aliquots were combined, then separated by centrifugation with removal of erythrocytes in ammonium chloride lysing solution (0.15 M NH₄Cl, 10 mM NaHCO₃, 1 mM EDTA). Total cell count was measured with a TC20 Automated Cell counter (Bio-Rad, Hercules, CA). BALF was then transferred onto slides with a cytological centrifuge (Shandon Cytospin 2, Runcorn, UK) and stained with a Hema 3 Stain Set (Thermo Fisher Scientific, Hampton, NH). Images of stained cells were taken with a Nikon E800 microscope (Nikon Instruments Inc., Melville, NY) using a SPOT RT3 Camera and SPOT Imaging Software (v. 5.2, Diagnostic Instruments, Inc., Sterling Heights, MI). At least 200 cells were counted per slide with ImageJ (NIH, Bethesda, MD). Macrophages/monocytes, neutrophils, and lymphocytes were individually enumerated by two separate investigators. Cell counts were normalized to total BAL fluid recovered.

Protein Analysis

BALF was analyzed using a DuoSet ELISA kit for Mouse CCL2/JE/MCP-1 (R&D Systems, Minneapolis, MN) and TGF-β1 (R&D Systems, Minneapolis, MN) according to the manufacturer’s instructions using a SpectraMax M5 Microplate Reader (Molecular Devices, San Jose, CA) and Softmax Pro 6.4 (Molecular Devices). The detection range for this assay was 3.9–250 pg/mL. In addition, latent TGF-β1 was activated by incubating samples with 1 N HCl and neutralizing with 1.2 N NaOH/0.5 M HEPES, as per kit instructions, and another ELISA was performed to measure immunoreactive TGF-β1.

Pulmonary Function Testing

Naïve (8–10 wk old) and IAV-infected (14 and 56 days postinfection) mice were anesthetized with a ketamine/xylazine mixture [100 mg/kg (Par Pharmaceutical, Chestnut Ridge, NY) and 20 mg/kg (Acorn, Inc., Lake Forest, IL), respectively], immobilized with pancuronium bromide (10 mg/kg, Sigma-Aldrich, St. Louis, MO), and ventilated (SCIREQ Inc., Montreal, Canada) with a tidal volume of 10 mL/kg, 150 breaths/min, PEEP of 3 cm H₂O, and FIO₂ of 21% as previously described (28). Respiratory system resistance (R_rs), Newtonian airway resistance (R_N), respiratory system compliance (C_rs), elastance (H), tissue damping (G), hysteresivity (η, eta) were measured in triplicate at both time points.

Human Tissues

Donor lungs samples were provided through the federal United Network of Organ Sharing via National Disease Research Interchange (NDRI) and International Institute for Advancement of Medicine (IIAM) and entered into the NHLBI LungMAP Biorepository for Investigations of Diseases of the Lung (BRINDL) at the University of Rochester Medical Center overseen by the IRB as RSRB00047606, as previously described (31, 32). Donors with a history of prematurity and bronchopulmonary dysplasia (BPD) were classified as either having “healed BPD” or “established chronic” BPD by a pediatric pathologist who is part of the LungMAP program. Term controls had no history of prematurity and their lung pathology was deemed normal for comparison. No acute viral infections were reported in the past medical history of any infant. The donor characteristics and demographics are available in Table 1. Lung tissue sections were uniformly obtained from the right lower lobe of all infants. Sections (5 µm) of formalin inflated, paraffin embedded RLL parenchymal lung tissue blocks were deparaffinized, rehydrated, and stained for anti-thrombospondin 1 (ab85762, Abcam, Cambridge, UK) and DAPI Fluoromount-G (SouthernBiotech, Birmingham, AL).

Table 1.

Characteristics of neonatal human donor lungs

ID	Postnatal Age, yr	GA at Birth, wk	Sex	Race	Pathology
090	1.16	Term	F	Unknown	Term born control
011	1.74	Term	F	White	Term born control
032	3.53	Term	M	White	Term born control
056	5.71	Term	M	White	Term born control
139	3.99	Term	F	White	Term born control
053	3.16	24	M	White	Healed BPD
039	3.27	23	F	Unknown	Healed BPD
015	3.89	25	F	White	Healed BPD
055	5.02	32	M	White	Healed BPD
184	1.27	24	M	White	Established, chronic BPD
227	1.58	23	F	White	Established, chronic BPD
115	1.62	26	F	White	Established, chronic BPD
170	1.81	25	F	Black/AA	Established, chronic BPD

Human tissue donors were classified as either term born control, healed BPD, or established chronic BPD by LungMAP pathologists based on lung histology. BPD, bronchopulmonary dysplasia.

Western Blot

Protein was extracted from tissue with protein lysis buffer. Samples were run on an 8% SDS-Page resolving gel with 5% stacking and transferred to a PVDF membrane (IPVH00010, Millipore, Darmstadt, Germany). Membranes were then incubated with a 5% nonfat milk blot in TBS-T, stained with anti-thrombospondin 1 antibody (TSP-1; 1:1,000, Abcam, Cambridge, UK) with IgG(H + L) Mouse/Human ads-HRP (SouthernBiotech, Birmingham, AL). Anti-actin was used as an internal loading control (ACTB, 2066, 1:10,000, Sigma-Aldrich, St. Louis, MO). Blots were analyzed using ImageJ, with each isoform measured separately.

Immunohistochemistry

At PIDs 14 and 56, right lobes of lungs were snap-frozen for qRT-PCR, and left lobes perfused with 10% neutral-buffered formalin (NBF, Thermo Fisher Scientific, Hampton, NH) at 25 mm/Hg, embedded in paraffin wax, and cut to 4 µm thick. Additional samples were taken after saline flush with 1× PBS, before perfusion with NBF. Lung slices were stained with Hematoxylin and Rubens Eosin-Phloxine (H&E; Biocare Medical, Concord, CA) and Gomori’s Trichrome (Richard-Allan Scientific, San Diego, CA) for collagen. At PND56, naïve RA and 40×8 mouse lungs were harvested, perfused with 4% neutral buffered formalin, embedded in OCT compound, frozen, and cut to 8–10 µm. Slides were then fixed in 100% acetone and stained for GP1BB.

Fluorescent immunohistochemistry was performed with primary antibodies S100A4 (1:1,000, PA5-82322, Thermo Fisher Scientific, Waltham, MA), anti-influenza A virus nucleoprotein (NP; NR-43899, BEI Resources, Manassas, VA), anti-thrombospondin 1 (THBS1; ab85762, Abcam, Cambridge, UK), MRC1 (CD206; R&D Systems, Minneapolis, MN), or GP1Bβ (1:300, emfret Analytics, Würzburg, Germany) with appropriate secondary antibody, and DAPI Fluoromount-G counterstain. Negative control slides were processed using the same protocol, but without primary antibody. Stained images were taken with a Nikon E800 microscope (Nikon Instruments Inc., Melville, NY) using a SPOT RT3 Camera and SPOT Imaging Software (v. 5.2, Diagnostic Instruments, Inc., Sterling Heights, MI). Photographs were analyzed with ImageJ.

Activated Macrophage, Fibroblast, and Collagen Quantification

S100A4, MCR1, and Sirius red staining were quantified using ImageJ for fibroblasts, M2 macrophages, and collagen, respectively. Fluorescent images of each airway were taken under red (S100A4, Sirius red), green (MCR1), and blue (DAPI) channels. Airway perimeter was measured in ImageJ, using an upper cut-off of 310 µm so only small airways were analyzed (22). Once the airway lumen was selected, it was radially enlarged by 40 µm (100px) on the Sirius red images to capture collagen and 80 µm (200px) on the S100A4 and MCR1 images to capture fluorescent cells in the peribronchial region. A threshold of fluorescence for each stain was set and used for all images taken. Area of Sirius red stain was measured and normalized to airway perimeter.

qRT-PCR

RNA was extracted from tissue with TRIzol Reagent (Invitrogen, Carlsbad, CA) as previously described (33). Complimentary DNA was run on a C1000 ThermoCycler (Bio-Rad) using a Maxima First Strand cDNA Synthesis kit (Thermo Fisher Scientific). Quantitative real-time PCR was performed using iQ SYBR Green Supermix (Bio-Rad) with CFX96 Real-Time System (Bio-Rad). Genes of interest were run on plates with Gapdh as housekeeping gene, and analyzed using the ΔΔC_T method (34). Three to four samples per treatment were run in duplicate on each plate. Primer sequences can be found in Table 2.

Table 2.

qRT-PCR primer sequences

Gene	Primer
Tsp-1	F: 5′-TCC CCT ATT CTG GAG GGT TC-3′ R: 5′-TCC CTG GAA ATA GGC ACA AG-3′
Adamts1	F: 5′-CTC GTA GCT GAC CAG TCC AT-3′ R: 5′-ACT TCT GGT CCC TTC TGC TC-3′
Adamts4	F: 5′-CAG ACG AAG CAC TCA CCT T-3′ R: 5′-CCA GCC TGA GGA ACA TTG A-3′
Adamts9	F: 5′-GCC TGT GCT ACC TTA CCT AAA C-3′ R: 5′-CCA CAA GTC ACG GAA CAA GAG-3′
Adamts15	F: 5′-TGA TCT GTC TCC GAC CCT CA-3′ R: 5′-GAC TCA CCA TGC CCA CT-3′
Tgfβ-R1	F: 5′-AAA ACA GGG GCA GTT ACT ACA AC-3′ R: 5′-TGG CAG ATA TAG ACC ATC AGC A-3′
Tgfβ-R2	F: 5′-AAC ATG GAA GAG TGC AAC GAT-3′ R: 5′-CGT CAC TTG GAT AAT GAC CAA CA-3′
Tgfβ-R3	F: 5′-GGT GTG AAC TGT CAC CGA TCA-3′ R: 5′-GTT TAG GAT GTG AAC CTC CCT TG-3′
Smad1	F: 5′-ACC CCT ACC ACT ATA AGC GAG-3′ R: 5′-TGC TGG AAA GAG TCT GGG AAC-3′
Smad2	F: 5′-ATG TCG TCC ATC TTG CCA TTC-3′ R: 5′-AAC CGT CCT GTT TTC TTT AGC TT-3′
Smad3	F: 5′-CAC AGC CAC CAT GAA TTA CGG-3′ R: 5′-TGG CGT CTC TAC TCT CTG ATA GT-3′
Smad4	F: 5′-CAT TCC AGC GTG CCA TTT C-3′ R: 5′-TTC AAA GTA AGC AAT GGA GCA C-3′
Gapdh	F: 5′-TGT CCG TCG TGG ATC TGA C-3′ R: 5′-CCT GCT TCA CCA CCT TCT TG-3′

F: forward primer; R: reverse primer.

Statistical Analysis

Statistical analyses were performed in GraphPad Prism (GraphPad Software v. 8, San Diego, CA). Pulmonary function data were subjected to D'Agostino and Pearson and Shapiro–Wilk tests for normality, Brown–Forsythe test for variance, and ordinary one-way ANOVA with Tukey’s multiple comparisons test for significance. In instances of failed normality or variance, Kruskal–Wallis nonparametic and Dunn’s multiple comparisons tests were performed for significance. Weight over time was tested for normality and variance, as aforementioned, and multiple t tests performed with Holm–Sidak correction. Cell differentiation data and ELISA data were subjected to Shapiro–Wilk test for normality. Holm–Sidak correction for multiple comparisons was used to further test cell differentiation and Kruskal–Wallis nonparametic and Dunn’s multiple comparisons tests were performed for significance on ELISA data. Human Western blot was subjected to ANOVA quantifying total TSP-1 as a fold change against β-actin. qRT-PCR data were analyzed using the ΔΔCT method and graphed as fold-change normalized to RA = 1, or to RA-naïve animals if comparing qRT-PCR across the infection time course. P values of ≤ 0.05 were considered significant for all analyses performed, and values graphed as means ± SE.

RESULTS

Adult Mice Exposed to Low-Dose Hyperoxia Are Functionally Indistinguishable from Room Air Controls

Lung function and pathology were evaluated in adult (PND56) mice who had been exposed to room air (RA) or 40% oxygen for the first 8 days of life (40×8) (Fig. 1A). We confirmed that RA and 40×8 animals have similar alveolar and airway structure by histology (Fig. 1B), consistent with our laboratory’s previously published study (28). Furthermore, pulmonary function, measurements were similar between RA and 40×8 adult mice at this time point (Fig. 1C).

Figure 1.

Recovered 40×8 hyperoxia mice are phenotypically and functionally “normal”. A: model timeline of hyperoxia exposure and recovery in naïve mice. B: hematoxylin-eosin (H&E) sections of RA and hyperoxia (40% for 8 days; 40×8) exposed mice resemble RA controls at PND 56. C: pulmonary function measurements are similar between RA and O₂-treated mice for respiratory system resistance (R_rs), Newtonian airway resistance (R_N), respiratory system compliance (C_rs), tissue damping (G), elastance (H), and hysteresivity (η, eta). Data represent means ± SE; n ≥ 8 samples/group. Scale bar = 100 µm. RA, room air; PND, postnatal day.

Low-Dose Oxygen Exposed Mice Have Worse Airway Disease after Influenza Infection

We administered IAV to hyperoxia and RA exposed mice as a challenge, to determine if 40×8 mice respond similarly to room air mice after normalization of pulmonary function. Adult RA and 40×8 mice were nasally inoculated with HKx31 IAV (hereafter RA-×31 or O₂-×31) or sham (hereafter RA-PBS or O₂-PBS) (Fig. 2A). Weight loss was not different between the RA-×31 and O₂-×31 groups during the first two postinfection weeks through 8 wk postinfection, whereas sham animals appropriately did not lose weight (Fig. 2B) or experience any inflammation or functional changes (data not shown). Infection was confirmed by positive nucleoprotein (NP) staining in the airway club cells, indicating active IAV infection in both RA-×31 and O₂-×31 animals (Fig. 2C, arrows). At postinfection day (PID) 14, persistent airway inflammation (Fig. 2D, arrows) and fibrosis (Fig. 2E, arrows) were observed in O₂-treated animals compared with RA controls. To confirm increased fibrosis in O₂-×31 animals, Sirius red collagen staining was performed, imaged (Fig. 2F), and quantified (Fig. 2G) in both groups with a focus on the peribronchial spaces, demonstrating increased Sirius red staining in O₂-×31 animals at PID 14. The observed inflammatory and fibrotic changes resolved by PID56 (Fig. 2, H–K). There were no differences in weight loss between male and female mice (data not shown).

Figure 2.

O₂-treated mice given influenza virus have worse pulmonary morbidity. A: experimental timeline – at PND 56, mice previously exposed to RA or O₂ at PND 0–8, were intranasally infected with 10⁵ PFU of HK×31 (H3N2) IAV or PBS (sham). B: weight loss curve was similar among O₂/RA sham and O₂/RA-infected groups. C: fluorescent NP (red), counterstained with SCGB1A1 (green) and DAPI (blue), at PID 3 indicates viral infection in small airways of both treatment groups (arrows). Hematoxylin-eosin (H&E; D) and trichrome stains (E) at PID14 indicate increased inflammation and fibrosis around the small airways of O₂-×31 animals (arrows). Sirius red staining (arrows) (F) and quantification (G) at PID 14 show increased collagen deposition in the O₂-×31 treatment group. H&E (H) and trichrome (I), Sirius red staining (J), and quantification (K) at PID 56 showed a recovered phenotype with resolved inflammation and fibrosis for both groups. Data represent means ± SE; n ≥ 4 samples/group, *P ≤ 0.05. Scale bars = 100 µm. PID, postinfection day; PND, postnatal day; RA, room air.

Comprehensive assessments of pulmonary function were performed at both PID 14 and PID 56 to determine if the observed histologic findings had a physiologic correlate. At PID 14, O₂-×31 animals had higher total respiratory system resistance (R_rs, Fig. 3A), Newtonian airway resistance (R_N, Fig. 3B), and elastance (H, Fig. 3E) with decreased respiratory system compliance (C_rs, Fig. 3C). Tissue damping (G, Fig. 3D) and hysteresivity (η, Fig. 3F) were unchanged. The magnitude of the changes in R_N (which contributes to R_rs) signifies that the majority of resistance change was due to changes in airway resistance. Interestingly, at PID56, the resistance changes normalized (Fig. 3, A and B), whereas the compliance and elastance (Fig. 3, C and E) remained persistently abnormal in the 40×8 animals. The RA animals returned to their previous levels. This indicates that the lung is persistently stiffer after IAV infection resolves, and it has reached a new, lower physiologic baseline.

Figure 3.

O₂-treated mice given influenza virus have pulmonary function deficits that do not recover. Pulmonary function testing at PID 14 and 56. R_rs (A) and R_N (B) were higher in the O₂-×31 group at PID 14 indicating more hyperactive airways. C: C_rs was decreased in O₂-×31 animals at both time points. D: G was unchanged at both time points indicating no difference in alveolar energy dissipation. E: H was increased at both time points indicating increased tissue stiffness F: η remained unchanged at both time points indicating homogenous lung disease. Data represent means ± SE; n ≥ 10 samples/group. *P ≤ 0.05.C_rs, respiratory system compliance; η (eta), hysteresivity; G, tissue damping; H, tissue elastance; PID, postinfection day; R_N, Newtonian resistance; R_rs, respiratory system resistance.

Molecular Differences Persist in Low-Dose Hyperoxia-Exposed Mice after Room Air Recovery

Given the increased morbidity after IAV infection in 40×8 mice, we sought to determine whether uninfected adult RA and recovered 40×8 lungs had “silent” gene expression differences that could induce a worse phenotype after infection. To do this, we examined a previously published Affymetrix array in adult mice who received high-dose (100×4) oxygen at birth for candidate genes (35). Out of 45,109 probes present on the array, 54 transcripts were differentially expressed between the RA and the 100×4 mice using a false discovery rate of 10%. Neonatal hyperoxia reduced expression of 43 genes, most of which reflected a loss of pulmonary cardiomyocytes (35). We analyzed the upregulated transcripts for genes regulating inflammation using qRT-PCR and found that the extracellular matrix glycoprotein thrombospondin 1 (TSP-1) and several members of the A disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS) family of proteinases in the high dose 100×4 model. TSP-1 and ADAMATS share common functions in their ability to activate latent TGFβ through conformational change of the latent binding protein that opens the binding site of TGFβ to its receptor (36). We confirmed that TSP-1 expression is increased in 40×8 adult (PND56) mice but not at the end of oxygen exposure on PND8 (Fig. 4D). We did not identify increased ADAMTS proteinases at high enough levels to justify further study (Supplemental Fig. S1; all Supplemental material is available at https://doi.org/10.6084/m9.figshare.14051579), thus choosing to focus on the source and function(s) of TSP-1.

Figure 4.

Thrombospondin-1 prevalence in mouse and human tissues. Lung samples were obtained from C57Bl/6J mice that were exposed to oxygen at PND 0–8. A: PND 56 mouse lung sections stained with antibodies to THBS1 (red), GP1BB (green), and DAPI (blue). Dashed boxes indicate zoomed images (dashed outset boxes) that show colocalization of THBS1⁺ and GP1BB⁺ (platelet/MK) cells. TSP-1 Western blot (B) more TSP-1+ cells (C) are present in lungs per high-powered field (HPF). D: Tsp-1 mRNA is increased over twofold by PND56, but no difference at PND8. E: analysis of protein extracted does not show differences among RA and O2 mice. F: representative images from control (left) and BPD (right) human lung sections stained with antibody to THBS1 (red), GP1BB (green), and DAPI (blue). Western blot (G) and analysis of protein extracted (H) show increasing TSP-1 with more severe BPD (left to right on the blot). I: BPD samples have more TSP-1+/DAPI+ cells per HPF compared with control. Data represent means ± SE; n ≥ 7 (mouse) and n ≥ 4 (human) samples/group. *P ≤ 0.05. Scale bar = 100 µm. aw, airway; BPD, bronchopulmonary dysplasia; HPF, high-powered field; MK, megakaryocyte; PND, postnatal day; RA, room air; TSP-1, thrombospondin-1.

TSP-1 is a calcium binding ECM glycoprotein first discovered in platelet granules (37) and later localized to many other tissues including the lung (38). TSP-1 is synthesized by endothelial cells, fibroblasts, smooth muscle cells, monocytes, and macrophages (39). To evaluate the prevalence and distribution of TSP-1, tissue sections of naïve 40×8 mice were stained with TSP-1 antibody and counterstained with DAPI. In 40×8 mice, TSP-1 was increased in the alveolar spaces (Fig. 4A) but less noticeable around the airways. Colocalization with GP1Bβ (CD42c), a platelet/megakaryocyte (MK) antibody showed complete overlap of staining. Notably, these lungs were not perfused before harvest, so the double-positive cells are likely a mixture of smaller platelets and larger MKs (Fig. 4A, outset dashed rectangles). More TSP-1+ cells were present in 40×8 mice compared with RA per high powered field (Fig. 4C). TSP-1 mRNA was increased at PND56 but unchanged at PND8 after cessation of oxygen exposure where it was colocalized with platelet-derived growth receptor alpha (PDGRFα) and fibroblasts (Fig. S2). Increased expression of TSP-1 was not detected by Western blot, perhaps because it is a semiquantitative assay (Fig. 4, B and E).

TSP-1 Expression Is Increased in Humans with BPD

In human autopsy samples (n = 5 term healthy controls, n = 4 recovered BPD, n = 4 established chronic BPD), increased TSP-1+ cells were similarly detected in the alveolar spaces of BPD infants where staining was much sparser in control infants (Fig. 4, F and I). We confirmed this by immunoblot of protein extracts showing that although there was some variability between subjects, there was a significant difference in TSP-1 protein abundance with worsening BPD pathology (Fig. 4, G–H). The characteristics and demographics of each donor included in the TSP-1 immunoblot are shown in Table 1.

Platelet Activation, TSP-1, and MCP1 Expression Change across IAV Infection

After localizing increased TSP-1 to MKs and platelets in naïve lungs, we hypothesized that platelets would be activated after IAV infection and localize to the lung because they are a known acute phase reactant. TSP-1 mRNA was highest at PID3 at a 7–10 fold increased levels compared with naïve animals, where PID3 increases O₂-×31 animals compared with RA-infected controls. TSP-1 mRNA levels subsequently decreased as infection resolved (Fig. 5B). Immunohistochemical staining for TSP-1 demonstrated the same temporal relationship as mRNA with TSP-1+ cells distributed throughout the airway and alveolar space at PID3 and later resolving (Fig. 5A). Separate from MKs/platelets, TSP-1 expression was noted in rare airway cells at PID7 (Fig. 5A, arrows), but appeared similar in RA and 40×8 infected animals at the time point where SCGB1A1 expression is near its nadir. The spike in TSP-1 mRNA and staining correlated with the peak TGFβ activation levels in the lung, suggesting that platelets were likely the source of active TGFβ during infection (Fig. 5C). Further, more active TGFβ was noted at PID7 in O₂-×31 animals compared with PID7 RA-×31 mice. Finally, CCL2 levels also followed a similar time course peaking at day 3, with increased PID7 CCL2 in O₂-×31 animals (Fig. 5D). There was a slight increase in CCL2 levels at PID14 in 40×8-×31 mice, but this did not reach statistical significance.

Figure 5.

Infection-related time course of TSP-1 and inflammatory mediators. A: immunofluorescent staining for THBS1 (red), SCGB1A1 (green, and DAPI (blue) during IAV infection shows increased THBS1 staining at PID 3 and less in PID 7 in blood vessels, airways (white arrows), and alveolar spaces of infected mice. THBS1⁺ cells (platelets) clear as infection resolves. B: THBS1 gene expression peaks at PID 3, decreasing as infection resolves with clearance of platelets; O₂-×31 animals had higher THBS1 at PID 3 compared with RA controls. C: BALF active TGFB1 levels peak at PID 3, decreasing as infection resolves. O₂-×31 animals have higher TGFB1 levels at PID 7 compared with RA-×31. D: BALF monocyte chemoattractant protein 1 (CCL2/MCP1) levels also peak at PID 3, with higher PID 7 CCL2 levels in O2-×31 animals. Data represent mean ± SE; n = ≥ 3 samples/group, *P ≤ 0.05. Scale bar = 100 µm. PID, postinfection day; RA, room air; TSP-1, thrombospondin-1.

Given that TSP-1 activates TGFβ and TGFβ promotes fibrosis, we hypothesized that key TGFβ signaling pathway mediators were different in naïve RA and 40×8 animals that would prime the lung toward fibrosis. To our surprise, differences in expression of TGFβ receptors (Fig. S3, A–C), any of the canonical SMAD genes (Fig. S3, D–F), or common TGFβ reporters (Fig. S3, G– I) were not detected by qRT-PCR. Furthermore, there was a strong trend for decreased Tgfb-R2 and Tgfb-R3 in naïve mice before infection (Fig. S3, B and C); however, the peribronchial inflammation (Fig. 2) and pulmonary function (Fig. 3) differences did do not temporally associate with the TSP-1/TGFβ spike (PID14 vs. PID3, respectively). This suggests TSP-1/TGF-b signaling may not be the direct mediator of airway inflammation following IAV infection.

Oxygen Exposed Mice Have Persistent Peribronchial Inflammation Characterized by Profibrotic Macrophages

We next sought to determine whether low-dose O₂ exposure is associated with increased inflammatory cell recruitment similar to our previous studies using high-dose (100×4) oxygen (23). Bronchoalveolar lavage fluid (BALF) was sampled, spun, and quantified with differential cell counts at four time points during infection: days 3, 7, 10, and 14. We found a slight increase in total cell counts and neutrophils in PID7 RA-×31 animals (Fig. 6, A and B) which may be attributable to leakier lungs in the O2-×31 group inhibiting BALF recovery during peak infection. More interestingly, we observed increased trends for total and macrophage counts among O₂-×31 animals at PID14 (Fig. 6, A and D). Given our finding that CCL2/MCP-1 was increased at PID7, we hypothesized that there may be later macrophage-mediated inflammation in our model. To better characterize the inflammatory cells present around the airways, tissue sections were stained for fibroblast stimulatory protein 1 (S100a4, FSP1), a marker shown to identify myofibroblasts and profibrotic (M2 subtype) subpopulations of macrophages (40, 41), and MRC1 (CD206), another M2 macrophage marker. Increased S100a4 staining was concentrated around the small airways at PID 14 (Fig. 6, E and F) with rare MRC1 overlap, indicating myofibroblast activation at PID14. S100a4 was also increased in alveolar spaces of O2-×31 mice, but staining was much less prevalent than in airway sections (Fig. S4). Taken together, the low-dose hyperoxia model followed by influenza infection has identified airway-specific inflammation and myofibroblast activation leading to airway hyperreactivity.

Figure 6.

Lung Inflammation during Influenza Infection. BALF was collected at PID 3, 7, 10, and 14 from RA-×31 and O₂-×31 animals. Total cells (A), neutrophils (B), lymphocytes (C), and macrophages were enumerated (D). Total cells were increased in the RA-×31 group at PID 7 and trends for increased total cells and macrophages were present at PID 14. E: PID14 lung slices were stained with S100A4 antibodies (red), MRC1 (CD206, green), and DAPI (blue). F: more S100A4⁺ cells were present around O₂-×31 airways group compared with RA-×31 controls. G: similar MRC1+ cells were identified around airways from each group. Data represent means ± SE; n ≥ 3 samples/group. #P ≤ 0.10 (trend), *P ≤ 0.05. Scale bar = 100 µm. BALF, BAL fluid; PID, postinfection day; RA, room air.

DISCUSSION

Airway disease in former ELGANs is characterized by increased airflow obstruction in infancy (42), specifically in the mid-to-later forced expiratory flows (43, 44). These functional deficits predispose ELGANs to wheezing in infancy and childhood irrespective of their BPD status (45–51). Because BPD is often defined by need for supplemental O₂ near term corrected gestational age (2), many infants who were exposed to O₂ escape the BPD diagnosis, but still have significant pulmonary morbidity associated with their extreme prematurity. Our laboratory has performed several clinical studies quantifying cumulative O₂ exposure in ELGANs with and without BPD and have shown that ELGANs with increased O₂ exposure have worse obstructive lung disease (FEV_0.5/FVC ratio), but “high” and “low” O₂ exposed ELGANs have significant airflow obstruction (FEF₇₅) compared with term infants at 1 yr of age (52). Those functional studies were performed in asymptomatic, well ex-preterm infants, but there is an abundance of evidence that when challenged with a respiratory infection, former ELGANs have more significant lower airway symptoms with wheezing (suggesting airflow obstruction) and increased rehospitalization rates (12, 53, 54). These clinical studies emphasize that there is a spectrum of lung disease present in former ELGANs, and thus justifies studying a spectrum of O₂ exposures in the laboratory that more closely model variant neonatal exposures, focusing on airway pathology to determine their impacts on lung development, function, and response to infection.

Using these ELGAN studies, we sought to create a translationally relevant paradigm of low-dose neonatal hyperoxia to show that mice are primed for airway disease when challenged with respiratory viral infection in adulthood. We chose 40% oxygen for 8 days to distinguish it from other models, wherein prolonged high O₂ concentrations cause significant alveolar simplification, making it difficult to discern and isolate differences in airway pathology. Relatively low-dose 40×8 oxygen does not cause significant alveolar simplification, though if left in 40% oxygen for longer periods subtle changes in alveolar structure are detectable (55). We wanted to test if the “repaired” lung after hyperoxia would respond similarly to an RA animal when challenged with influenza A virus and hypothesized based on the lack of alveolar simplification and unaffected type II pneumocyte number that they would respond normally, distinguishing it from our other published studies (19–21, 23, 24, 27). We chose HKx31 strain of influenza because it causes lower airway symptoms, usually without mortality, and mice respond to IAV similarly compared with humans (56–58); unlike some other viruses (e.g., respiratory syncytial virus) more commonly seen in infants that are difficult to model in mice. Furthermore, our infection-related changes are evident at baseline and without bronchoconstrictors such as methacholine, implicating alternate mechanisms of airway pathology in O₂ exposed mice apart from smooth muscle bronchospasm, and further distinguishing it from asthma. Finally, inflammatory changes resolve by PID56, but the lung is stiffer and less compliant. We speculate that the profibrotic/proreparative airway macrophages may facilitate myofibroblast activation around the airway (Fig. 6) and, to a lesser extent, alveolar spaces (Fig. S4) may permanently remodel the lung structure and increase susceptibility to repeated viral/injurious insults.

We showed that 40×8 mice have increased airway-specific fibrotic repair resulting in hyperactive airways, peribroncihal fibrosis associated with TGFβ hyperactivation, higher TSP-1, and delayed inflammatory resolution compared with room air controls. Our investigation of upregulated TSP-1 led us to investigate the prevalence and function of lung MKs and platelets in infection-related airway disease. Recent evidence implicates MKs and platelets are important immune mediators that can influence the organ microenvironment. Lung megakaryocytes are gaining appreciation as important stem cells that contribute to the overall circulating platelet population (59), which through secretion of immune effector molecules can affect the organ-specific and systemic environment. Mouse models of myocardial infarction (MI) suggest a coordinated interplay between platelet-secreted TGFβ and β2 microglobulin (β2M) which influence the polarization of macrophages toward the M2 proreparative/profibrotic phenotype (60, 61). There may be similarities between the MI and 40×8 hyperoxia IAV paradigm because both characterize late inflammation associated with macrophage recruitment to tissue injury sites, but further investigation is needed to determine the overlap between these two model systems.

TSP-1 may provide mechanistic insight into IAV-related morbidity in hyperoxia exposed mice and humans with BPD as it interacts with several ECM components including integrins, fibronectin, cell receptors, growth factors (like TGFβ-1), cytokines, and proteases (62, 63). Antiangiogenesis, smooth muscle proliferation, nitric oxide signaling antagonism, and inflammation regulation are known functions of TSP-1 relevant to the lung (38). TSP-1-null animals have bronchial epithelial hyperplasia, proximal mucous metaplasia, vascular smooth muscle hyperplasia, club cell hyperplasia, and uncontrolled inflammation (36). The strikingly similar phenotype of TSP1-null and TGFβ-1-null mice suggest that TSP-1 is the main TGFβ activator in vivo (36). Notably, the TGFβ signaling machinery changes both its expression and localization within the lung across development and dysregulation of the TGFβ pathway and its machinery has been implicated in high-dose hyperoxia (64–66) with receptors relocating to the airway epithelium and upregulation of certain SMAD proteins involved in canonical signaling (64). However, our study does not show priming of the TGFβ pathway by itself in noninfected conditions, suggesting that the TSP-1/TGFβ interplay may be relevant in the setting of MK/platelet activation and release after infection. In contrast, upregulation of TSP-1 was noted in the preterm ventilated lung on autopsy (67), an in utero model of tracheal occlusion (68) (lung stretch), and other profibrotic diseases, but to our knowledge this has not been further explored in BPD. The BPD model is of particular interest because TSP-1 may play a role in capillary rarefication (also seen in mouse hyperoxia models (69)) and contribute to other hyperoxia-induced diseases such as pulmonary hypertension (70). Interestingly, hyperoxia increases TSP-1 at the same magnitude (∼ twofold) in both low- and high-dose oxygen models, indicating the TSP-1 is a marker of both low- and high-neonatal O₂ exposure. Our human studies (Fig. 4) confirm this, with increasing TSP-1 with more severe BPD. Thus, TSP-1 is an intriguing candidate for further study in rodent BPD models and the extremely preterm infant.

Our results suggest low-dose neonatal O₂ causes “silent” changes in whole lung gene expression with long-term functional consequences after a respiratory viral infection, and that the “repaired” lung still reacts abnormally to a profibrotic stimulus such as IAV. This 40×8 hyperoxia model causes an airway-specific phenotype observed at baseline and drives increased respiratory morbidity after infection, thus recapitulating airway diseases observed in former ELGANs. We can now exploit this model to better understand how the MK/platelet environment primes the lung that may help explain infection-related morbidity in vulnerable former ELGANs.

SUPPLEMENTAL DATA

Supplemental Figs. S1–S4: https://doi.org/10.6084/m9.figshare.14051579.

GRANTS

This work was supported by a grant from the Strong Children’s Research Center at the University of Rochester (A.M.D.), the American Lung Association Catalyst Award (A.M.D.) and National Institutes of Health grant R01 HL091968 (M.O.R.). NIH Center Grant P30 ES001247 supported the animal inhalation facility and the tissue-processing core. The human subject studies were supported by NHLBI Molecular Atlas of Lung Development Program Human Tissue Core Grant U01HL122700 and U01HL148861 (G.H.D., T.J.M., G.S.P.). Additional support was obtained through T32 training grants GM068411 and AI118689 (R.W.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.M.D., J.H., M.Y., and M.A.O. conceived and designed research; A.M.D., J.H., R.W., M.Y., and G.S.P. performed experiments; A.M.D., J.H., R.W., and M.Y. analyzed data; A.M.D., J.H., R.W., and M.A.O. interpreted results of experiments; A.M.D. and J.H. prepared figures; A.M.D. and J.H. drafted manuscript; A.M.D., J.H., R.W., M.Y., G.S.P., and M.A.O. edited and revised manuscript; A.M.D., J.H., R.W., M.Y., G.S.P., and M.A.O. approved final version of manuscript.