Single cell transcriptomics identifies distinct profiles in pediatric acute respiratory distress syndrome

Flerlage, Tim; Crawford, Jeremy Chase; Allen, E. Kaitlynn; Severns, Danielle; Tan, Shaoyuan; Surman, Sherri; Ridout, Granger; Novak, Tanya; Randolph, Adrienne; West, Alina N.; Thomas, Paul G.

doi:10.1038/s41467-023-39593-0

Download PDF

Article
Open access
Published: 30 June 2023

Single cell transcriptomics identifies distinct profiles in pediatric acute respiratory distress syndrome

Nature Communications volume 14, Article number: 3870 (2023) Cite this article

3179 Accesses
1 Citations
12 Altmetric
Metrics details

Subjects

Abstract

Acute respiratory distress syndrome (ARDS), termed pediatric ARDS (pARDS) in children, is a severe form of acute respiratory failure (ARF). Pathologic immune responses are implicated in pARDS pathogenesis. Here, we present a description of microbial sequencing and single cell gene expression in tracheal aspirates (TAs) obtained longitudinally from infants with ARF. We show reduced interferon stimulated gene (ISG) expression, altered mononuclear phagocyte (MNP) transcriptional programs, and progressive airway neutrophilia associated with unique transcriptional profiles in patients with moderate to severe pARDS compared to those with no or mild pARDS. We additionally show that an innate immune cell product, Folate Receptor 3 (FOLR3), is enriched in moderate or severe pARDS. Our findings demonstrate distinct inflammatory responses in pARDS that are dependent upon etiology and severity and specifically implicate reduced ISG expression, altered macrophage repair-associated transcriptional programs, and accumulation of aged neutrophils in the pathogenesis of moderate to severe pARDS caused by RSV.

Tracheal aspirate RNA sequencing identifies distinct immunological features of COVID-19 ARDS

Article Open access 26 August 2021

Locally organised and activated Fth1hi neutrophils aggravate inflammation of acute lung injury in an IL-10-dependent manner

Article Open access 13 December 2022

Large scale cytokine profiling uncovers elevated IL12-p70 and IL-17A in severe pediatric acute respiratory distress syndrome

Article Open access 08 July 2021

Introduction

Acute respiratory distress syndrome (ARDS; in children termed pARDS for pediatric acute respiratory distress syndrome) is defined by the acute onset of hypoxemic respiratory failure associated with pulmonary infiltrates in the absence of heart failure. Both lung intrinsic (i.e., lower respiratory tract infection (LRTI)) and lung extrinsic (i.e., sepsis) etiologies are recognized causes of ARDS and pARDS; in children, LRTI is the most common cause of pARDS¹.

A dysregulated immune response is thought to contribute to the development of ARDS, which proceeds through three phases^2,3. The initial stage of ARDS, termed the exudative phase, is mediated by activated innate immune cells, namely resident alveolar macrophages as well as recruited monocytes⁴ and neutrophils^3,5,6,7,8. In this stage, inflammatory mediators (cytokines, including TNF, IL-1β, IL-8, and IL-6, acute phase reactants, high mobility group-1 protein, leukotrienes, defensins, and complement) coupled with oxidative stress mediated by free radical production, local activation of the coagulation cascade and endothelial injury, protease production, and surfactant dysfunction have been implicated in preclinical and human studies in mediating lung damage and the ARDS phenotype^3,8. An important step in the resolution of inflammation is clearance of neutrophils by macrophages via efferocytosis, which has previously been shown to be deficient in human patients with ARDS⁹. Following the exudative stage, the proliferative stage involves reconstitution of barrier integrity via provisional matrix elaboration. The final stage is a fibrotic phase, which occurs in a subset of individuals with ARDS and is characterized by exuberant fibrosis that leads to prolonged impairment in gas exchange². To date, the majority of studies characterizing the immunopathogenesis of ARDS have enrolled adult individuals. Thus, less is currently known about pARDS pathogenesis.

Globally, respiratory syncytial virus (RSV) is a significant cause of respiratory tract infections (RTIs) across the age spectrum as well as mortality caused by RTIs in children less than 5 years of age¹⁰. In a large multicenter study of lower respiratory tract infection (LRTI; pneumonia) leading to hospitalization in the United States, RSV and human rhinovirus (HRV) were the most commonly identified viral etiologies, which together accounted for more than 50% of episodes requiring hospitalization¹¹. Nasal bulk transcriptomic studies have previously identified reduced interferon responses^12,13 and lower upper airway viral load¹³ in infants with higher disease severity RSV bronchiolitis, compared to those with lower severe disease. Airway neutrophilia is commonly seen in RSV infection in infants, and a dysregulated neutrophil response is thought to play a role in disease severity¹⁴. Thus, pathologic immune responses, most notably by neutrophils, contribute to RSV disease severity; however, specific mechanistic insight into neutrophil-mediated disease pathogenesis is lacking in general and particularly in the context of pARDS¹⁵.

In this work, we use single cell transcriptomics and ViroCap sequencing on tracheal aspirate (TA) samples to demonstrate etiology-dependent and severity-dependent alterations in airway cell transcriptional responses involving interferon stimulated gene (ISG) expression, macrophage repair programs, and neutrophil ageing in infants with acute respiratory failure (ARF) requiring endotracheal intubation and invasive mechanical ventilation (IMV).

Results

Description of patients

We prospectively enrolled 24 (14 male and 10 female) patients zero to six years of age (Supplementary Data 1). Because the TA from patient C6 (six years old, the only enrolled patient greater than 2 years of age) had insufficient cell quantity for the primary study objective, these data are not included in final analyses (n = 23 patients). Patients were enrolled into one of three cohorts, including patients with a known or suspected LRTI who met diagnostic criteria for moderate or severe pARDS¹⁶ (P-cohort), patients with known or suspected LRTI who did not meet full diagnostic criteria for pARDS or who developed mild pARDS at any point during follow-up (L-cohort), and patients with ARF without known or suspected LRTI at the time of enrollment (C-cohort; Fig. 1a). Based on the anticipated duration of IMV for each of these cohorts, the study design included collection of up to three TAs from P-cohort patients, up to two TAs from L-cohort patients, and one TA from C-cohort patients. There were no significant differences in timing of sample collection between cohorts at each sampling time point (Fig. 1b).

**Fig. 1: Overview of Study Design and Patient Population.**

The 23 patients were zero to two years of age (mean age 5.8 months (mos), standard deviation (SD) 6.2 mos). Twelve patients were enrolled into the P-cohort, 5 into the L-cohort, and 6 into the C-cohort. C7 presented apnea as the primary indication for IMV and was found to have human rhinovirus (HRV) by subsequent testing but was included in the C-cohort to maintain study group allocation based on the data available at enrollment.

We compared illness severity across cohorts using oxygen saturation index (OSI)^17,18,19 and duration of IMV (Fig. 1c). By OSI, P-cohort patients had significantly higher severity than L- and C-cohort patients. Though there was a difference in mean duration of IMV between P- and L-cohort patients (10.5 vs 7.6 days respectively), this difference was not statistically significant. Notably, L-cohort patients (mean age 1.2 mos, SD 0.44 mos) were significantly younger than P-cohort (mean age 5.5 months, SD 4.5 months) and C-cohort patients (mean age 10.2 months, SD 9.3 months).

ViroCap analysis corroborates and supplements clinical microbiological testing

The clinical infectious etiologic diagnosis of ARF for enrolled patients included identification of a viral etiology alone (n = 9; Supplementary Data 1 and Fig. S1b), bacterial pathogen alone (n = 1), and co-detection (n = 7; both viral and bacterial respiratory pathogens identified). As there is no standardized approach to microbiological testing for patients with ARF, we used ViroCap, a viral capture sequencing technique²⁰ separately on both RNA and DNA extracted from TA supernatant to supplement and further refine clinical diagnoses for patient classification for single cell gene expression (scGEX) analyses based on inciting etiology. In the RNA capture, we identified viral reads in all patients with LRTI as the etiology of respiratory failure (P- and L-cohorts) (Fig. 2a). We were unsuccessful in recovering RNA from any C-cohort TAs, except for HRV-positive C7. The findings from ViroCap sequencing of RNA aligned with clinical testing results, identified additional co-infecting viruses, and suggested an alternative primary etiology of ARF in certain patients.

**Fig. 2: ViroCap sequencing in tracheal aspirate supernatant confirms and supplements clinical microbiology testing.**

Influenza A virus (IAV), which is included on the institutional clinical multiplex PCR panel but was detected only in clinical testing obtained from a single patient (P11), was identified in TAs from 5/17 (29%) patients with negative clinical testing for IAV by ViroCap. We further analyzed these data by extracting reads assigned to IAV and mapping these to a reference H3N2 IAV genome using Burrows-Wheeler Aligner (BWA)²¹ and VarScan²² (v2.3.9), which confirmed identification of IAV reads mapping across distinct segments of the influenza genome with patient-specific genetic variation. Although patient P1 was diagnosed with a bacterial LRTI and was negative for HRV via clinical testing, ViroCap identified HRV sequences within the TAs across all three sampling time points. We additionally identified RNA capture reads mapping to primate bocaparvovirus 1 in Samples 1 and 2 from patients P7, P8, P9, P10, P12, L3, and L4, which may indicate concurrent or recent infection with human bocavirus 1 or 3 (these are not included on the institutional respiratory virus detection assay). Finally, despite a clinical diagnosis of adenovirus for patient P4, RNA ViroCap lacked adenovirus reads and instead showed enrichment for reads mapping to anelloviruses (torque teno viruses); only a small number of reads from the DNA capture from the second TA supernatant for this patient matched adenovirus, suggesting a potential disconnect between the positive clinical diagnosis and inferred disease etiology.

In addition to viral respiratory pathogens, we identified previously described constituents of the airway virome, including anelloviruses (AVs); torque teno virus (TTV), torque teno mini virus (TTMV), torque teno midi virus (TTMDV), human herpesviruses (HHVs; HHV1 (herpes simplex virus), 4 (Epstein-Barr virus; EBV), 5 (cytomegalovirus; CMV), 6B (HHV6B), and 7 (HHV7)), human papillomaviruses (HPV), and bacteriophages²³.

We further analyzed the proportion of bacterial reads mapping to Haemophilus influenzae (H. flu), Moraxella catarrhalis (M. cat), and Streptococcus pneumoniae (S. pneumo) in capture sequencing performed on DNA extracted from TA supernatant (Fig. 2b). These bacteria were selected because they were the most common clinically identified bacteria in our cohort and are known to colonize the respiratory tract in infants. Of these bacteria, M. cat was the most frequently identified (at least 1 read detected in 35 of 49 TAs (71.4%), including in 3 C-cohort patients). Reads aligning to the H. flu genome were detected in 4 patients from whom clinical airway bacterial cultures were not obtained (P006, P010, P011, and L003; Fig. 2b, black asterisk indicates patients for whom no culture was obtained), suggesting that these patients may have had both RSV and H. flu present in high proportional abundance at the onset of ARF. In response to antibiotic administration, the abundance of H. flu reads decreased in TA samples; however, similar decreases were not witnessed in the frequency of M. cat reads (Fig. 2b, black “a” indicates administration of antibiotics).

Based on clinical diagnoses and results of ViroCap sequencing, which identified RSV as the inciting etiology of ARF in the majority of P-cohort (7/12, 58.3%) and all L-cohort patients, we further divided P-cohort patients into RSVneg (n = 5) and RSVposP (n = 7) groups for comparison with L-cohort patients (RSVposL) and C-cohort patients (Con) (Figs. 2c and S1b) in our scGEX analyses. We noted differences in OSI and ventilator support requirements in Samples 1 and 2 between groups and there was not a statistically significant difference between the ages of patients in the RSVposP and RSVposL groups (Figs. 2d and S1a).

Single cell transcriptomics identifies differences in TA cellular composition and transcriptional state that depend on etiology, illness severity, and time after intubation

We present transcriptional data from 84,495 cells obtained from 51 TA samples (35 from P-cohort, 10 from L-cohort, and 6 from C-cohort patients). The initial unsupervised clustering analysis identified 15 transcriptionally distinct cell clusters (Fig. S2a), which were further grouped by gene expression into 11 distinct cell populations (Fig. 3a). In this process, we separated putative monocyte populations (FCGR3B low, CD14 pos, VCAN hi) from neutrophils (FCGR3B hi, CD14 low, VCAN low) and macrophages (CD68, MARCO, and APOC1 pos). Within macrophage cell clusters, we further divided cells into putative functional phenotypes, including phagocytic macrophages (Ma_phago), which demonstrated increased expression of genes associated with phagocytosis and respiratory burst (SOD2 and CD44), interferon stimulated gene (ISG) expressing macrophages (Ma_ISG), which upregulated ISGs and cytokines (IFIT1, CXCL10), and repair-associated macrophages (Ma_repair), which differentially upregulated genes associated with repair (AREG, IL10) and antigen presentation/adaptive immune cell interaction (CD74, major histocompatibility type 2 (MHC II) genes, and GPR138; Fig. 3a,b).

**Fig. 3: Single cell transcriptomics identifies illness severity and etiology-dependent differences in gene expression.**

Based on clinical and ViroCap data (Fig. 2a), we included the genomes for RSV, enteroviruses B and D, HRV A, B, and C, torque teno virus (TTV), torque teno mini virus (TTMV), IAV, and adenoviruses B and C in a custom reference used to align our single-cell RNA sequencing data. Of the viruses included in the custom reference, we identified reads aligning to RSV and, to a lesser extent, IAV genomes (Fig. S2b). A subset of cells within the epithelial compartment had particularly high expression of transcripts mapping to the RSV genome (Fig. 3b, bottom left) and were analyzed separately as Virus hi cells. This population demonstrated elevated expression of mitochondrial transcripts (Fig. 3b, bottom right) relative to other cell populations, suggesting that they were undergoing lysis or cell death.

After annotating cell populations by lineage and function in the aggregate data set (Fig. 3a), we next began to approach the primary hypotheses of the study by comparing differences in proportional abundance and transcriptional profile of these cell populations by illness severity, inciting etiology of ARF, and time after onset of ARF. Though there was a higher percentage of Sample 1 cells obtained from RSVposL group TAs, proportions of cells from each subgroup were similar at Sample 2 and 3 timepoints (Fig. 3c). The most notable differences between patient groups were in innate immune cell populations in Samples 1 and 2. While RSVneg and RSVposP TAs were significantly enriched in neutrophils, RSVposL TAs were found to have an increase in monocytes and macrophages (Figs. 3d and S2c).

Using CellChat²⁴, an algorithm used to measure ligand-receptor interactions between cells, we found that, in general, cellular interactions were fewer and weaker between cell populations in RSVposP TAs compared to RSVposL and RSVneg TAs. Specifically, Ma_repair and Ma_ISG macrophage populations are major sources of increased outgoing signaling strength in RSVposL TAs compared to RSVposP TAs (Fig. S3d, top) and monocytes, along with Ma_repair populations, are major sources of outgoing signaling strength in RSVneg TAs compared to RSVposP TAs (Fig. S3d, bottom).

Together, these data demonstrate that the development of airway neutrophilia, reduced interferon responsiveness, and reduced ligand-receptor interactions between MNPs and other TA cellular populations communication are key discriminatory features of moderate to severe RSV-associated pARDS compared to milder RSV-associated ARF phenotypes.

Epithelial responses to LRTI-associated ARF depend on primary etiology and illness severity

Respiratory viruses, including RSV, IAV, and HRV, display tropism for epithelial cells lining the airway, from the nasopharynx to the alveoli. Thus, epithelial cell responses to these viral pathogens are of critical importance. We therefore subclustered non-pulmonary neuroendocrine cell (PNEC) populations expressing EPCAM (clusters 11 and 14) from the aggregate data. Among these 3332 epithelial cells, 11 transcriptional states were detected using unsupervised clustering analysis (Fig. 4a). We identified transcriptional states representative of intermediates along the epithelial repair trajectory²⁵, including basal cells (KRT5), secretory cells (MUC5AC, SCGB1A1), deuterosomal cells (HES6), FOXN4 positive cells²⁶, and ciliated epithelial cells (FOXJ1, PIFO). Additionally, there was a population of cells likely originating from tracheal hillocks²⁷ in light of differential upregulation of KRT13 and CXCL17 (Fig. 4b). We also identified distinct clusters of transcriptional states (Fig. 4b), which were composed partially of cells expressing innate immune transcripts (PTPRC, FCGR3B, VCAN). We speculate that these clusters represent immune cells interacting with epithelial cells at the time of cell capture; however, we cannot exclude that these represent technical doublets.

**Fig. 4: Reduced ISG and cytokine signaling in epithelial cells from patients with moderate to severe pARDS.**

Given the enrichment for RSV LRTI in P- and L-cohorts (Figs. 2a and S1b), we focused further analyses of epithelial cell populations on RSV-host interactions and on comparisons between patient groups. We subsetted cells with any RSV expression (RSV-infected) from those with no RSV infection (bystander). The majority of RSV-infected cells (59.8%) were from RSVposL patients, who had a significantly higher proportion of RSV-infected cells amongst all epithelial cells (Fig. 4c). Interestingly, a small percentage of RSV-infected epithelial cells were from RSV-neg patients (5.9%) and C-cohort patients (1.9%). As persistent RSV infection has been identified previously²⁸, these findings may indicate recent RSV infections.

We next analyzed normalized RSV expression levels within epithelial cells from each comparator group, which demonstrated significantly increased RSV expression levels in RSVposL cells (Fig. 4c, middle top). However, there was no significant difference in normalized RSV expression when comparing subsetted RSV-infected epithelial cells from RSVposL and RSVposP groups (Fig. 4c, middle bottom). We then analyzed percent mitochondrial and RSV gene expression within the subsetted RSV-infected epithelial cells and found that while there was no significant statistical difference between percent of the cellular transcriptome occupied by RSV between RSVposP and RSVposL groups (Fig. 4c, right), there was significantly less mitochondrial gene expression in RSVposL groups compared to all other patient groups (Fig. 4c, right). Consistent with these findings, our ViroCap data demonstrated a non-significant trend towards higher RSV counts in RSVposL TA supernatants compared to RSVposP TA supernatants (Fig. 4d).

We identified significantly increased ISG module²⁹ usage by subsetted epithelial cells from RSVposL patients compared to the other three groups (Fig. 4e). Using pathway enrichment analysis (MSigDB Hallmark 2020 Database) to compare RSVposL and RSVposP epithelial cells, we found that inflammatory, interferon, and epithelial mesenchymal transition pathways were enriched in RSVposL relative to RSVposP epithelial cells (Fig. 4f). Using DGE testing, generated a list of the top 10 transcripts differentially upregulated (by average log₂ fold change) by P and L cohort epithelial cells for module scoring (Supplementary Data 2). We then used these module scores to assess their ability to differentiate P versus L cohort patient origin amongst epithelial cells (Fig. 4g) using receiver operating characteristic (ROC) curves. The area under the ROC curve score (AUROC) score for the P cohort epithelial module was 0.7 and the AUROC score for the L cohort epithelial module was 0.78.

MNP transcriptional response and developmental trajectory is etiology- and severity dependent

Mononuclear phagocytes (MNPs), including macrophages and monocytes, have been implicated in the continuum of host response to pARDS, from initiation to recovery². To study MNP responses in greater detail, we subclustered MNPs from the aggregate data set (Fig. 2a), which included 47,266 cells grouped into 14 clusters by unsupervised clustering analysis (Fig. 5a). We used the transcripts FCN1 (inflammatory monocyte-like phenotype), SPP1 (repair type phenotype), and FABP4 (resident macrophage phenotype)^30,31, along with CXCL10 (interferon stimulated macrophages), FCER1A (dendritic cells), and IL6 (a FCN1 low, SPP1 low cytokine producing macrophage population) to annotate subclustered MNPs. We identified regional upregulation of these transcripts within MNP gene expression space, which was largely represented by FCN high inflammatory populations (Figs. 5a, b, and S3a). Though isolated from a more proximal airway compartment, these populations are reminiscent of those described in bronchoalveolar lavage fluid (BALF) from patients with COVID-19³⁰.

**Fig. 5: Identification of distinct etiology- and illness severity-specific MNP transcriptional programs.**

Given similar MNP gene expression profiles, we used gene lists generated by Liao et al.³⁰ along with those described by Reyes et al.^32,33 to perform module scoring (Fig. S3b, c). We found that the gene module groups described by Liao et al. discretely localized within MNP gene expression space. Group 1 gene module usage was highest in FCN1 high clusters and was significantly increased in RSVneg MNPs compared to other groups, group 1_2 module (enriched in ISGs) usage localized within areas of high CXCL10 and ISG module²⁹ usage (Fig. S3b) and was significantly increased in RSVposL MNPs, and group 2 module usage was enriched in SPP1 high clusters and was also significantly increased in RSVposL MNPs. Together, these data suggest that airway MNP responses are relatively conserved amongst patients with virus induced ARF.

Consistent with previous analyses of proportional abundance of cell phenotypes within patient groups (Fig. 3d), a significantly higher proportion of Samples 1 and 2 MNPs were obtained from RSVposL patients (Fig. 5c). We analyzed the proportional abundance of MNPs within each cluster by patient group and Sample time point (Fig. S3d) and noted a higher abundance of RSVposL and RSVposP MNPs in cluster 7 (corresponding to CXCL10/ISG module high MNPs), RSVneg MNPs in clusters 2, 3, 10 (localized in FCN1 high SPP1 neg gene expression space), RSVposL MNPs in clusters 0, 4, and 11 (most notably from Sample 1 to Sample 2).

Using differential gene expression (DGE) we identified defining characteristics of MNP transcriptional phenotypes enriched amongst patient groups (Supplementary Data 2). Cluster 11 macrophages were transcriptionally distinct from the previously described SPP1-high reparative macrophage populations³¹ and expressed IL10 (Fig. 5b) and the scavenger receptor OLR1 (Fig. 5d), which suggests a particular role for this population in immune regulation and efferocytosis³⁴. Cluster 1 MNPs (Neutrophil-like) demonstrated higher expression of transcripts more commonly associated with neutrophils (CD177, LCN2, PI3, CD24, among others; Fig. 5b). Clusters 2, 3, and 10 were most consistent with an inflammatory monocytic phenotype (expression of S100 transcripts, VCAN, and RETN), which aligns with usage of gene modules from previously published data sets (Fig. S3b). Using the top 20 (by average log₂ fold change) differentially upregulated transcripts, we confirmed that these MNP transcriptional states were enriched in patient groups (i.e., cluster 2/3/10 gene module usage was enriched in RSVneg MNPs) (Figs. 5d and S3e and Supplementary Data 2) and observed spatial localization of MNP derived gene set usage within subclustered MNPs (Fig. 5e).

Finally, using pathway enrichment analysis (MSigDB Hallmark 2020 Database; Fig. 5f), we identified enrichment of inflammatory response and interferon signaling pathways amongst RSVposL MNPs compared to RSVposP MNPs, which is consistent with our interpretation of comparisons of module usage between groups. Taken together, these data identify altered macrophage developmental trajectories that depend on the illness severity and underlying etiology of ARF and specifically demonstrate deficient MNP interferon responses as well development of IL10 expressing repair-associated macrophages in patients with moderate to severe pARDS compared to those with no or mild RSV-induced pARDS.

Specific aged neutrophil responses in moderate to severe RSV-associated pARDS

As airway neutrophilia has been implicated in the pathogenesis of pARDS and severe RSV infection, we analyzed transcriptional responses in neutrophil populations from the aggregate dataset (Fig. 3a) on the basis of FCGR3B expression, which included 23,184 cells. We observed a gradient of CXCR2 and CXCR4 (Fig. 6a), consistent with differences in neutrophil maturity, from mature to aged. Using DGE analyses, we further defined these neutrophil subpopulations and identified increased expression of transcripts encoding S100 proteins, adhesion proteins (VNN2, SELL, GCA), and FOLR3 in mature neutrophils and increased expression of cytokines (CCL3L1, CCL4L2), ISGs (i.e., IFI30), and transcription factors (i.e., HES4) in aging CXCR4-pos neutrophils. The population that we termed intermediate, had higher average expression of IL1 transcripts (IL1RN and IL1B) as well as transcripts that mediate cell survival (TNFAIP3 and IER3; Fig. 6b).

**Fig. 6: Accumulation of aged neutrophils with a unique transcriptional program in moderate to severe pARDS caused by RSV.**

We used previously published²⁹ neutrophil gene sets to perform module scoring to define putative function and responses within subsetted neutrophils (Fig. S4a). Usage of each of these modules was significantly higher in Sample 1 RSVposP neutrophils compared to other patient groups (Fig. S4a, top), but was significantly lower than RSVposL neutrophils at the Sample 2 time point. Module usage for each gene set was localized in gene expression space to mature and intermediate neutrophils, and less so to aged neutrophils. These data provided the initial suggestion of longitudinal differences in neutrophil transcriptional responses in RSVposP patients.

Consistent with previous analyses of proportional abundance of cell phenotypes (Fig. 3d), there was an increase in the proportion of total neutrophils from patients with moderate to severe pARDS (most notably from RSVneg patients) between Samples 1 and 2 (Fig. S4b). Within neutrophil phenotypes, we identified a significant increase in the proportional abundance of RSVposP neutrophils consistent with an aged transcriptional phenotype over the course of the sampling time points (Fig. 6c), which was not observed in RSVposL or RSVneg neutrophils over the same time points.

To better characterize differences in aged neutrophil gene expression in RSVposP TAs, we identified transcripts differentially upregulated by aged neutrophils between patient groups (Figs. 6d and S4c and Supplementary Data 2). In this analysis we identified differentially increased expression of protease inhibitors (PI3, SLPI, and CSTB) and HLA-C (type I major histocompatibility complex) and differentially decreased expression of cytokines (CCL3L1 and CCL4L2) and mitochondrial genes (MT-CO1) in aged RSVposP neutrophils relative to other patient groups. Together, the pattern of differentially expressed transcripts suggested an aged neutrophil response specific to RSVposP patients, characterized by the upregulation of the protease inhibitors PI3 and SLPI as well as TNFAIP2 (involved in regulation of inflammatory responses) and downregulation of mitochondrial transcripts (MT-CO1) and cytokines (CCL3L1).

We used the list of top transcripts differentially upregulated by aged neutrophils from RSVposP patient groups for module scoring (Supplementary Data 2). This demonstrated increased module usage within aged neutrophil clusters in gene expression space that was distinct in distribution compared to previously described neutrophil gene sets (Fig. S4d). There was significantly increased module usage by RSVposP neutrophils at all 3 Sample time points (Fig. 6e, top), and significantly increased module usage within aged neutrophils over time (Fig. 6e, bottom). We finally tested the ability of aged neutrophil module scores to differentiate between RSVposP and RSVposL group neutrophils by generating a ROC curve (Fig. 6f). We found an AUROC score of 0.75, which further supports specificity of this gene set for RSVposP neutrophils.

FOLR3 as a marker of disease severity in pARDS

To identify early transcriptional markers of pARDS, irrespective of symptomatology and TA cell type, we grouped patients according to the primary study cohort into which they were enrolled at the Sample 1 time point (Fig. 7a). We first performed differential gene expression testing between P and L cohorts in Sample 1 cells (Fig. 7b). This analysis identified differential upregulation of VCAN, mitochondrial transcripts (MT-CO2), ribosomal genes (RPL34), and FOLR3, and differential downregulation of cytokines (CCL3L1, CCL3, IL6) and ISGs (ISG15 and IFITM3) in P-cohort TAs.

**Fig. 7: Identification of FOLR3 as a potential airway marker of severity in pARDS.**

We then analyzed expression of these transcripts between the three study groups in Sample 1 TAs and confirmed upregulation of FOLR3, S100A12, and VCAN in P-cohort TAs and ISG15, CCL4, and CCL3 in L-cohort TAs. While expression of mitochondrial and ribosomal transcripts was higher in P-cohort relative to L-cohort TAs, expression of these were higher in C-cohort TAs (Fig. 7c). Further analysis demonstrated specificity of FOLR3 expression by neutrophils, S100A12 by neutrophils and monocytes, VCAN by monocytes and “inflammatory” macrophage populations, ISG15 by Ma_ISG cells, and cytokines (CCL3L1, CCL3, CCL4) by neutrophils and mononuclear phagocytes (Fig. 7d).

We then measured concentrations of proteins encoded by these transcripts in TA supernatants obtained from an independent cohort of patients enrolled at another children’s hospital (Fig. 7e and S5). Samples included in this analysis were chosen to match our age and pre-admission health status inclusion criteria. We assigned disease severity to patients in this cohort using the same consensus definitions¹⁶ (Supplementary Data 3). We found that concentrations of FOLR3 were significantly higher (p = 0.04) in first sample TAs obtained from patients with moderate to severe pARDS compared to patients with no or mild pARDS. Additionally, we identified a non-significant trend towards increased concentrations of Resistin (p = 0.054) and S100A12 (p = 0.15) as well as decreased concentrations of Interferon alpha (p = 0.15) in patients with moderate to severe pARDS compared to patients with no or mild pARDS.

Together, these analyses suggest severity-associated responses detectable in TA samples from patients with pARDS that are consistent with a neutrophil predominant response in patients with moderate to severe pARDS and a MNP predominant response in patients with no or mild pARDS.

Discussion

This study details single cell transcriptional profiles in TA samples obtained from infants with ARF requiring IMV. Consistent with the known epidemiology of ARF in this age group¹¹, we identified viral LRTI, and in particular RSV-associated LRTI, as the most common cause of ARF in our cohort. Our analyses suggest etiology-dependent and illness severity-dependent alterations of TA cellular composition and transcription in epithelial cells, MNPs, and neutrophils. We present several findings that provide insight into the pathogenesis of ARF in infants.

We characterized RSV expression using two distinct methods, including by viral capture sequencing and by RSV expression in single cell transcriptomic data. By ViroCap sequencing, there was a trend towards more TA supernatant reads mapping to RSV in patients with no or mild pARDS related to RSV compared to those with moderate to severe pARDS related to RSV. In our single cell transcriptomic data, we consistently identified a significantly higher proportion of RSVposL cells infected with RSV (including epithelial cells, MNPs, and neutrophils) compared to RSVposP cells; however, RSV expression, represented as either a normalized abundance or as a proportion of the cellular transcriptome, was not statistically different between groups, but was generally higher in RSVposP cells. RSV viral load (VL) has been associated previously with disease severity, albeit with conflicting results depending on the patient population and study design^35,36,37. Our study suggests that host response, as opposed solely RSV VL, makes as an important contribution to differences in illness severity.

In keeping with this interpretation, divergent severity-associated epithelial responses to ARF are apparent within our data. Amongst subclustered epithelial cells, we identified significantly increased ISG gene module²⁹ usage by as well as interferon and inflammatory response pathways within RSVposL compared to RSVposP epithelial cell populations. Gene lists generated using DGE testing to identify differences between P and L cohort epithelial cells had AUROC scores of 0.70 to 0.78, respectively, suggesting these gene modules are potentially informative classifiers. These gene lists include transcripts involved in epithelial defense (SCGB3A1) and cellular detoxification (GPX4, GSTP1, PRDX5, and SERF2) for the P cohort gene list and monocyte chemoattractants (CCL3/4), ISGs (IFI6, ISG15, and IFITM3), and TIMP1 (an inhibitor of metalloproteinases) for the L cohort gene list. Together, these data suggest epithelial cells, and potentially interacting innate immune cells, from lower illness severity TAs (RSVposL) are coordinating a type I interferon response to infection, whereas epithelial cells from patients with higher illness severity are less interferon-responsive and are responding to mitochondrial and oxidative stress. These findings agree with previous nasal transcriptome studies of RSV disease severity, in which patients with more severe disease demonstrated lower ISG expression^12,13 and lower viral loads compared to a comparator group demonstrating a less severe disease phenotype¹³.

We identified illness severity and etiology-dependent differences in MNP transcriptional responses to ARF. We were most intrigued by the significantly higher proportional abundance of macrophages expressing repair-associated transcripts (i.e., AREG), the scavenger receptor involved in recognition of apoptotic cells OLR1³⁸, IL10, and the glycoprotein THBS1 in Sample 2 TAs obtained from patients in the RSVposL group compared to patients in the RSVposP and RSVneg groups. Overall, this gene expression profile suggests a regulatory role played by this macrophage population, potentially with a particular focus on phagocytosis of apoptotic cells. IL-10 is a key anti-inflammatory cytokine that reduces excessive immunopathology, potentially at the cost of incomplete pathogen control in certain contexts³⁹. IL-10 has been implicated in human ARDS severity^40,41 and has been shown to regulate neutrophil chemotaxis to the lung in a preclinical model of lower respiratory tract bacterial infection⁴². Macrophages upregulate IL-10 following exposure to proinflammatory mediators (i.e., lipopolysaccharide (LPS)) as well as after contact with apoptotic cells⁴³. IL-10 producing macrophages have increased phagocytic capacity with preferential specificity for early apoptotic cells^38,44. Thrombospondin-1, an extracellular matrix glycoprotein recognized by CD36 participates in interactions between phagocytes and apoptotic cells. Indeed, in a preclinical ARDS model, Thbs1^−/− mice demonstrate impaired resolution of pulmonary inflammation characterized by reduced macrophage IL-10 production, increased inflammatory cytokine expression, and exacerbated neutrophilic pulmonary infiltrates⁴⁵. The oxidized low-density lipoprotein receptor 1 (LOX1; encoded by OLR1) is another scavenger receptor expressed by phagocytes (in addition to CD36) involved in recognition of apoptotic cells³⁸. Intriguingly, OLR1 expression is significantly upregulated in dendritic cells generated in vitro from peripheral blood monocytes in the presence of interferon alpha⁴⁶. As noted above, we consistently identified increased ISG module usage across multiple cell types in RSVposL TA samples. Taken together, we hypothesize that this regulatory macrophage population, which is enriched in TAs from patients with lower illness severity potentially due to increased type I interferon signaling, plays a key role in mediating resolution of inflammation via clearance of early apoptotic cells and reduction of neutrophil chemotaxis to the airways.

Airway neutrophilia has previously been demonstrated in the context of critical RSV and is thought to play a role in disease severity^14,47. Consistent with this, we observed a progressive increase in proportional representation of neutrophils across sampling times within pARDS patients. Aged neutrophils were enriched in RSVposP patient TAs and demonstrate a unique transcriptional profile, enriched with protease inhibitors (i.e., SLPI, PI3, and CSTB), ORM1 (an acute phase reactant of uncertain immunoregulatory function), and TNFAIP2 (an inhibitor of TNF signaling). Aged neutrophils, which upregulate CXCR4, have been shown previously to demonstrate inflammatory responses via Toll-like Receptor 4 (TLR4) and Integrin signaling compared to non-aged neutrophil phenotypes⁴⁸. Using an RSVposP aged neutrophil gene list, we identified a AUROC of 0.75, which suggests that this module may perform well as a classifier of disease severity amongst infants with RSV infection. Together, these data support the conclusion that moderate to severe RSV-associated pARDS is associated with the accumulation of aged neutrophils displaying a unique transcriptional phenotype, which may contribute to the development of clinically severe disease.

We identified increased expression of FOLR3 in patients with pARDS compared to patients with no or mild pARDS and provide initial potential validation of tracheal aspirate FOLR3 protein concentration as a marker of pARDS severity using samples obtained from an independent study at another hospital. FOLR3 encodes the secreted Folate Receptor γ, one of three human folate receptors. It is known to be a component of neutrophil secondary granules⁴⁹ and is the predominant folate receptor detectable in plasma⁵⁰. As it has no murine orthologue, less is known about FOLR3 function in the context of an immune response; however, it may contribute to antimicrobial defense via binding and depletion of extracellular folate⁵⁰. Along with validation in a larger cohort, future study of whether FOLR3 concentrations are a marker of increased TA neutrophil numbers or an indicator of increased secondary granule production and degranulation may be of interest for future study.

Due to the dynamic nature of clinical treatment and illness presentation of ARF, we note unavoidable limitations to our study. TA samples, which contain mucous secretions, RNAses, and extracellular nucleic acid, inherently present technical difficulties with respect to single cell RNA sequencing. Thus, obtaining single cell suspensions free of mucous and extracellular nucleic acid is potentially problematic. To mitigate these potential technical issues, we processed all samples immediately and utilized a Percoll gradient to purify cells from mucus. Despite these efforts, we acknowledge that these unavoidable technical limitations may have contributed to less-than-ideal QC metrics for several samples included in the final dataset (Supplementary Data 4). A potential explanation for observed differences in gene and RSV expression between groups hinges upon differences in time from symptom onset to enrollment, a known limitation for studies of naturally acquired infections in humans. However, the relevant findings are generally consistent with our current understanding of ARDS-associated immunopathology. Finally, there were differences in demographic variables, including age, sex, and race, between patient cohorts, which may contribute to differences in immune responses between individuals. Thus, our findings, which were developed using analyses of samples obtained from patients admitted to a single center, will require future validation in a larger cohort in which these demographic variables can be better balanced between comparator groups.

Despite these limitations, this study details unique single cell transcriptomic characterization of responses to ARF occurring in the infant airway. A particular strength of this study is the prospective design, wherein patients contributed samples longitudinally over the course of the first week of hospitalization. We identified illness severity, etiology, and sample time point dependent changes in cell composition and transcriptional response, even in patients presenting similarly clinically. Our most salient findings are reduced interferon stimulation within epithelial and MNP cell populations, reduced proportional abundance of IL10 expressing regulatory macrophages, and increased airway neutrophilia associated with a unique transcriptional profile in patients with moderate to severe pARDS compared to patients with no to mild pARDS. We go on to identify increased FOLR3 concentrations in TAs collected in an independent study at another center. Together, these data implicate type I interferons, macrophage-mediated clearance of early apoptotic cells, and neutrophil ageing as important contributors to disease severity in infants with RSV-associated ARF requiring endotracheal intubation and IMV.

Methods

Patient recruitment and ethics approval

Patients were enrolled in this prospective observational cohort study conducted at LeBonheur Children’s Hospital in Memphis, Tennessee from March 2018 to February 2020. This study was approved by the University of Tennessee Institutional Review Board (IRB) (IRB Number 17-05744-XP STJUDE) and the St. Jude Children’s Research Hospital IRB (IRB Number Pro00008351). Signed informed consent for participation was obtained from the patient’s parent or legal guardian prior to inclusion in the study.

Inclusion and exclusion criteria

Children aged 0–12 years with acute respiratory failure requiring endotracheal intubation and invasive mechanical ventilation (IMV) were enrolled in one of three groups: patients with a known or suspected lower respiratory tract infection (LRTI) who met consensus criteria for the pediatric acute respiratory distress syndrome (pARDS)¹⁶ (P cohort), patients with a known or suspected LRTI who do not meet consensus criteria for pARDS¹⁶ (L cohort), and control patients with respiratory failure without known or suspected LRTI or findings consistent with a diagnosis pARDS¹⁶ (C cohort). Consensus criteria were applied based on review of the clinical history, results of clinically obtained laboratory tests, and results of chest imaging by the principal investigator (TF) in consultation with the pediatric intensive care unit (PICU) team managing the patient’s care. Patients were excluded from enrollment if they had previously been diagnosed with chronic lung disease of prematurity, had not been discharged from the hospital following delivery, uncorrected congenital heart disease, pulmonary hemorrhage, and primary or secondary immunodeficiency. Patients were additionally excluded if they were prescribed inhaled or systemic corticosteroids prior to hospitalization in the previous four weeks, had a tracheostomy, required home supplemental oxygen, or required home non-invasive mechanical ventilatory support. Finally, patients were excluded if they required extracorporeal membrane oxygenation (ECMO) support prior to being approached for enrollment.

Study design

We designed the study to collect TA samples from enrolled patients at one of three timepoints. The timing of sample collection depended on the cohort into which they were enrolled based on the a priori assumption that P-cohort patients would have the longest duration of IMV (Fig. 1a). Three TA samples were collected from patients enrolled in the P-cohort. The first was between 6 and 24 h after endotracheal intubation (Sample 1), the second was 48–72 h after Sample 1 (Sample 2), and the third was 48–72 h after Sample 2 (Sample 3). Two TA samples were collected from patients in the L-cohort at the same time as Samples 1 and 2 in the P-cohort. One TA sample was collected from patients in the C-cohort at the same time as Sample 1 in the P and L-cohort. At the same time as sample collection, patient data, including demographics, ventilator settings, results of microbiologic testing, and interventions, including antibiotics, corticosteroids, and inhaled nitric oxide, was abstracted from the electronic medical record (EMR) and stored using Microsoft Excel. Patients were followed for 60 days after enrollment to determine duration of IMV, PICU length of stay (LOS), and PICU mortality. Patient sex was not considered in the study design; however, these data were collected as listed in the EMR for the patient after obtaining informed consent for participation on the study.

Sample collection

After enrollment, tracheal aspirates (TAs) were collected by the patient’s respiratory therapist or registered nurse as per routine practice and timing for routine endotracheal tube care. This was intended to obtain samples that would be comparable to those able to be obtained in the course of standard patient care. 2–4 milliliters (mL) of sterile saline were instilled into the patient’s endotracheal tube (ETT). The TA was then collected using in-line endotracheal tube suctioning into a sterile specimen suction trap at the time when the patient’s ETT would have otherwise needed to be cleared for routine care. After collection, the specimen was placed on ice and immediately transported to be processed fresh for downstream assays (Fig. 1b).

Sample processing

The TA was first transferred from the sterile specimen suction trap to a 15 mL conical tube. After pipetting the sample into the 15 mL conical tube, 1 mL of phosphate buffered saline (PBS) was used to rinse the sterile specimen suction trap and then added to the 15 mL conical tube. Samples were then centrifuged at 500 × g for 5 min at 4 degrees Celsius (°C). The supernatant was collected into a 15 mL tube and mixed thoroughly before dividing into 500 µL aliquots and stored at −80 °C for later nucleic acid extraction for microbiologic sequencing (see below). We initially found that mucous in the samples complicated achieving a clean single cell suspension for single cell RNA-seq protocols. To remove mucous and other debris, we used a Percoll gradient consisting of four different concentrations. First, 100% Percoll was added to a 15 mL conical tube. A 60% Percoll gradient was overlaid on top of the 100% Percoll. After the initial centrifugation step and removal of supernatant, the pellet was resuspended in 40% Percoll and overlaid on top of the 60% layer. Finally, a 30% Percoll layer was overlaid on top of the 40% Percoll layer. The Percoll gradient (60/40/30) containing the resuspended pellet in the 40% layer was centrifuged at 2000 × g for 20 min at room temperature without a brake. After centrifugation, cells were collected from the 60/40 and the 40/30 interface and transferred to a new 15 mL conical tube. Percoll was diluted with PBS to a 15 mL total volume and centrifuged at 500 × g for 5 min at 4 °C. The supernatant was removed and the pellet was resuspended in 1–2 mL of PBS and 0.04% ultrapure bovine serum albumin (BSA). This suspension was then filtered through a 30 µm cell strainer to remove any residual debris. At this point, cells were counted using a hemocytometer with filtered 0.4% trypan blue serving as a viability count. After confirming successful creation of gel bead in emulsion (GEM) in the initial step of single cell gene expression library preparation (described below), any cells remaining were frozen down at 1.5–5 million cells/mL in BAMBANKER Freezing Medium (Fujifilm Wako Chemicals Corporation, Richmond, VA).

Single-cell gene expression library preparation

Libraries for single cell gene expression analyses were generated individually on the fresh TA single cell suspensions generated as above using the 3’ Gene Expression Kit (version 2, PN 120235 (Gel Bead Kit) and PN 120264 (Library Kit), 10X Genomics, Pleasanton, CA) according to the manufacturer’s instructions. Because cell yield varied by sample, we targeted final single cell suspensions to a concentration of 1000–1200 cells/µL and then loaded up to 17,000 cells onto a 10X Genomics chip for each sample. Due to a technical failure during loading of the freshly collected sample, Sample 1 from patient P1 was rerun using cryopreserved cells. This was the only sample processed and analyzed using cryopreserved cells. Libraries were sequenced on the Illumina NovaSeq platform (Illumina, San Diego, CA).

Single-cell analysis

Sequencing data were processed using CellRanger (v6.0.0, 10X Genomics) with a custom reference that included the human genome (GRCh38-2020-A; 10X Genomics) and the genomes of Enterovirus B (NC_038307.1), Enterovirus D (NC_038308.1), Respiratory syncytial virus (NC_038235.1), Rhinovirus A (NC_038311.1), Rhinovirus B (NC_038312.1), Rhinovirus C (NC_038878.1), Torque teno mini virus (NC_020498.1), Torque teno virus (NC_015783.1), Adenovirus B (NC_011203.1), Adenovirus C (NC_001405.1), and each segment of Influenza A (A/Michigan/45/2015(H1N1)). Quality control metrics, including median number of genes per cell and fraction of reads within cells, were recorded and are reported in Supplementary Data 4. Due to the heterogeneous nature of samples, the independent libraries were subsequently aggregated using CellRanger, without normalization, into a single matrix. Downstream analyses were performed using the Seurat (v4.0.3) framework⁵¹.

To exclude potential doublets, we removed cells that exhibited UMI counts or feature counts greater than the 99.5% probability quantile for each sample. Cells with greater than 20% of gene expression owing to mitochondrial genes were excluded as putatively dead or dying cells. We inferred cell cycle using the Seurat CellCycleScoring algorithm and as described elsewhere⁵². Each library was normalized using the SCTransform algorithm, with percent of mitochondrial expression and cell cycle effects regressed out. The fastMNN algorithm⁵³ was then utilized to integrate datasets from distinct libraries, effectively minimizing patient- and sample-specific differences in order to identify similar transcriptional subsets; this approach was repeated independently across all major cell subsets, including epithelial cells, MNPs, and MNPs with neutrophils. In each case, the first 15 fastMNN dimensions were used for UMAP dimensionality reduction and for nearest-neighbor graph construction for identifying transcriptional clusters in Seurat.

Differential gene expression (DGE) testing was performed between patient groups using the FindAllMarkers function in the Seurat R package with default settings, which identifies differentially expressed genes between two groups of cells using a Wilcoxon Rank Sum test. Significant differences between cell groups were determined using adjusted p-values, which are adjusted using a Bonferroni correction using all genes in the dataset⁵¹. Gene lists were generated from previously published manuscripts^29,30,32,33 and were used to generate module scores using Seurat’s AddModuleScore. The only change that was made to any gene list was the addition of IFI6 to the ISG module described by Wilk et al. We performed this analysis with the original gene set and with the addition of IFI6 and reached the same conclusion with regard to differences in module usage between etiology and illness severity patient groups. Pathway enrichment analysis was performed with Seurat’s DEenrichRplot command, which returns bar plots with top enriched terms from the selected EnrichR database. For these analyses, the enrich.database was set to “MSigDB_Hallmark_2020”, max.genes was set to 2000, and num.pathway was set to 10. ROC curves and AUROC scores were generated from module scores using the R package pROC (v1.18.0). We implemented CellChat (v1.1.3)²⁴ by first extracting CellChat input files from our Seurat object and subsequently creating a CellChat object. We set the ligand-receptor interaction database to CellChatDB.human for these analyses with no changes made to the database. This is a database of ligand-receptor interactions taken from previously published literature or from the Kyoto Encyclopedia of Genes and Genomes (KEGG) and includes 1939 validated molecular interactions. We performed all analysis using the standard workflow. To identify differences in ligand-receptor interactions between groups, we created a CellChat object for each patient group and merged them using the mergeCellChat command. Subsequent comparisons between groups followed the standard workflow.

ViroCap enrichment and sequencing

RNA and DNA were extracted from cryopreserved TA supernatants using the quick-DNA/RNA Miniprep Plus Kit (Catalog Number D7003; Zymo Research, Irvine, CA) according to the manufacturer’s protocol in two batches. Briefly, between 200 µL and 450 µL of TA supernatant (depending on initial yield after sample processing; 44 of 51 (86%) TAs started with 450 µL of TA supernatant) was combined with 300 µL to 550 µL (to a uniform starting total volume of 750 µL) of RNA/DNA Lysis Buffer and centrifuged using the Spin-Away Filter in a Collection Tube. 450 µL of nuclease free water combined with 300 µL of RNA/DNA Lysis Buffer was used as a negative control for each batch of extractions. The Spin-Away Filter was placed in a new collection tube and stored at 4 °C while performing RNA purification. For this, 200 proof ethanol (ETOH) (Sigma Aldrich, St. Louis, MO) was added to the flow-through from the Spin-AwayFilter centrifugation step. 750 µL of the mixture of ETOH and flow-through was added to the Zymo-Spin IIICG Column in a collection tube and centrifuged. The flow-through was discarded and the remaining 750 µL was added to the same Zymo-Spin IIICG Column and centrifuged again. DNA and RNA were then prepared and washed according to the manufacturer’s protocol and eluted into a final volume of 30 µL of water. DNA and RNA were quantified using Qubit High Sensitivity dsDNA and RNA Assay Kits (Catalog Number Q32851) on a Qubit 4 Fluorometer (ThermoFisher, Waltham, MA). Between two and four nucleic acid extractions were performed for five of the TA samples (P003_0, P003_2, P005_1, P005_2, P006_2) due to low initial RNA yield.

Double stranded cDNA (dscDNA) was created from total RNA extracted and purified from TA supernatants using the Maxima H Minus Double-Stranded cDNA Synthesis Kit (Catalog Number K2562; ThermoFisher Scientific, Waltham, MA) according to manufacturer’s protocol. Briefly, 13 µL of total RNA was added to 1 µL of random hexamer primer for first strand cDNA synthesis. Following second strand cDNA synthesis, 100U of RNase I was added prior to proceeding with dscDNA cleanup. A 1.8X AMPure (Beckman Coulter, Brea, CA) bead cleanup was used for dscDNA purification and the purified product was eluted into a final volume of 40 µL in Buffer EB (Qiagen, Hilden, Germany). Double-stranded cDNA concentrations were then quantified using Qubit High Sensitivity dsDNA Assay Kits (Catalog Number Q32851) on a Qubit 4 Fluorometer (ThermoFisher, Waltham, MA). Double-stranded cDNA was created and purified from total RNA from each aliquot of TA supernatant for these samples. Following dscDNA purification, dscDNA from each sample was combined and concentrated using a 1.8X AMPure XP (Beckman Coulter, Brea, CA) cleanup into 10 µL Buffer EB (Qiagen, Hilden, Germany).

Double-stranded cDNA and genomic DNA were each used to construct sample-specific libraries with KAPA HyperPrep kits (Catalog Number 07 962 363 001; Roche) with mechanical shearing. Approximately 100 ng of each individually indexed library was pooled by nucleic acid type (cDNA or gDNA), with up to 14 samples in a pool, and pools were enriched for viral content using the SeqCap EZ Developer kit (Roche) designed with ViroCap probes (IRN40000²⁰;). Library generated from purified Seal Influenza A hemagglutinin cDNA was used as a positive control. Pools were blocked with human COT and IDT blockers before a 20-h hybridization. Enriched library pools were sequenced on the Illumina NovaSeq platform (Illumina, San Diego, CA) at 100 × 100 base pairs.

ViroCap analysis

Raw reads were concatenated by sample and trimmed using Trimmomatic⁵⁴ to remove sequencing adapters. Kraken2⁵⁵ (v2.0.7b) was used to taxonomically classify reads using the Standard reference (which includes archaea, bacteria, viral, plasmid, human, and UniVec_Core sequences) downloaded from the Kraken website in September of 2020. Bracken⁵⁶ (v2.5) was subsequently used to estimate relative abundances of taxonomic annotations identified by Kraken2. Additionally, Kraken2 annotations were processed with Pavian (v0.8.4)⁵⁷ to summarize the abundance of classified reads.

Endotracheal Aspirate Protein Quantification

All samples were stored at −80 °C before they were diluted 1:2 for Cytokine analysis using Human Magnetic Luminex Assay (PN LXSAHM-19; R&D Systems) with the manufacturer’s set protocol. Data was collected on a Luminex 200 (Luminex) and analyzed on Milliplex analyst software (Vigene tech). In addition, after 1X thaw, samples were tested with individual Quantitative ELISA Kit per manufacturer’s protocol as follows: Human Thrombospondin 1:20(PN ab193716; Abcam); SLP-I 1:20(PN DP100; R&D Systems); Annexin A1 1:20(PN ab222868; Abcam); IL-8/CXCL8 1:20(PN D8000C; R&D Systems); CCL20/MIP-3 alpha 1:20(PN DM3A00; R&D Systems); CXCL10/IP-10 1:50(PN DIP100; R&D Systems); Human EN-RAGE 1:100(PN D41052-05; R&D Systems); Human S100A8/S100A9 Heterodimer 1:1000(PN D48226-05; R&D Systems). After 2× thaw samples were diluted for another set of individual Quantitative ELISA Kits as follows; Human FOLR3 1:20(PN DY5319; R&D Systems); Human CCL3L1/LD78 Beta 1:20(PN NBP2-7065; Novus); Human CCL4L1/LAG-1 1:20(PN NBP2-75067; Novus). All ELISAs were read on ELISA Reader (Molecular Probes) using dual wavelengths of 450/540 and data was analyzed using SOFTMAX Pro software 7.1.4.

Analysis of endotracheal aspirate protein concentrations

For measurements above the upper limit of detection (ULOD) or below the lower limit of detection (LLOD) in the Luminex multiplex assay we used the value of the ULOD or the LLOD in our analysis. For ELISA-based measurements we averaged the analyte concentrations extrapolated from the standard curve for Samples, which had been run in duplicate. Samples for which there was no fluorescence detected in either well were treated as not detected (value of 0 pg/mL). We noted that even at a 1:20 dilution the concentrations of SLPI exceeded the range of the standard curve. Concentrations were plotted and statistical testing between patient groups was performed using Graph Pad Prism (version 8.4.3) Software.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The raw sequencing data generated in this study have been deposited in the Sequence Read Archive database under accession code PRJNA971535. All other data are available in the article and its Supplementary files or from the corresponding author upon request. Source data are provided with this paper.

References

Khemani, R. G. et al. Paediatric acute respiratory distress syndrome incidence and epidemiology (PARDIE): an international, observational study. Lancet Respir. Med. 7, 115–128 (2019).
Article PubMed Google Scholar
Thompson, B. T., Chambers, R. C. & Liu, K. D. Acute respiratory distress syndrome. N. Engl. J. Med. 377, 562–572 (2017).
Article CAS PubMed Google Scholar
Bellingan, G. J. The pulmonary physician in critical care * 6: the pathogenesis of ALI/ARDS. Thorax 57, 540–546 (2002).
Article CAS PubMed PubMed Central Google Scholar
Rosseau, S. et al. Phenotypic characterization of alveolar monocyte recruitment in acute respiratory distress syndrome. Am. J. Physiol. Lung Cell. Mol. Physiol. 279, L25–L35 (2000).
Article CAS PubMed Google Scholar
Schütte, H. et al. Bronchoalveolar and systemic cytokine profiles in patients with ARDS, severe pneumonia and cardiogenic pulmonary oedema. Eur. Respir. J. 9, 1858–1867 (1996).
Article PubMed Google Scholar
Chollet-Martin, S. et al. High levels of interleukin-8 in the blood and alveolar spaces of patients with pneumonia and adult respiratory distress syndrome. Infect. Immun. 61, 4553–4559 (1993).
Article CAS PubMed PubMed Central Google Scholar
Aggarwal, N. R., King, L. S. & D’Alessio, F. R. Diverse macrophage populations mediate acute lung inflammation and resolution. Am. J. Physiol. Lung Cell. Mol. Physiol. 306, L709–L725 (2014).
Article CAS PubMed PubMed Central Google Scholar
Matthay, M. A. & Zemans, R. L. The acute respiratory distress syndrome: pathogenesis and treatment. Annu. Rev. Pathol. 6, 147–163 (2011).
Article CAS PubMed PubMed Central Google Scholar
Grégoire, M. et al. Impaired efferocytosis and neutrophil extracellular trap clearance by macrophages in ARDS. Eur. Respir. J. 52, 1702590 (2018).
GBD. 2016 lower respiratory infections collaborators. Estimates of the global, regional, and national morbidity, mortality, and aetiologies of lower respiratory infections in 195 countries, 1990–2016: a systematic analysis for the global burden of disease study 2016. Lancet Infect. Dis. 18, 1191–1210 (2018).
Article Google Scholar
Jain, S. et al. Community-acquired pneumonia requiring hospitalization among U.S. children. N. Engl. J. Med. 372, 835–845 (2015).
Article CAS PubMed PubMed Central Google Scholar
Chu, C.-Y. et al. Airway gene expression correlates of respiratory syncytial virus disease severity and microbiome composition in infants. J. Infect. Dis. 223, 1639–1649 (2021).
Article CAS PubMed Google Scholar
Thwaites, R. S. et al. Reduced nasal viral load and IFN responses in infants with respiratory syncytial virus bronchiolitis and respiratory failure. Am. J. Respir. Crit. Care Med. 198, 1074–1084 (2018).
Article CAS PubMed PubMed Central Google Scholar
Kirsebom, F. C. M., Kausar, F., Nuriev, R., Makris, S. & Johansson, C. Neutrophil recruitment and activation are differentially dependent on MyD88/TRIF and MAVS signaling during RSV infection. Mucosal Immunol. 12, 1244–1255 (2019).
Article CAS PubMed PubMed Central Google Scholar
Sebina, I. & Phipps, S. The contribution of neutrophils to the pathogenesis of RSV bronchiolitis. Viruses 12, 808 (2020).
Pediatric Acute Lung Injury Consensus Conference Group. Pediatric acute respiratory distress syndrome: consensus recommendations from the Pediatric Acute Lung Injury Consensus Conference. Pediatr. Crit. Care Med. 16, 428–439 (2015).
Article Google Scholar
Khemani, R. G. et al. Pulse oximetry vs. PaO2 metrics in mechanically ventilated children: Berlin definition of ARDS and mortality risk. Intensive Care Med. 41, 94–102 (2015).
Article PubMed Google Scholar
Khemani, R. G. et al. Comparison of SpO2 to PaO2 based markers of lung disease severity for children with acute lung injury. Crit. Care Med. 40, 1309–1316 (2012).
Article CAS PubMed Google Scholar
Leteurtre, S., Dupré, M., Dorkenoo, A., Lampin, M.-E. & Leclerc, F. Assessment of the pediatric index of mortality 2 with the Pao2/Fio2 ratio derived from the Spo2/Fio2 ratio: a prospective pilot study in a French pediatric intensive care unit. Pediatr. Crit. Care Med. 12, e184 (2011).
Article PubMed Google Scholar
Wylie, T. N., Wylie, K. M., Herter, B. N. & Storch, G. A. Enhanced virome sequencing using targeted sequence capture. Genome Res. 25, 1910–1920 (2015).
Article CAS PubMed PubMed Central Google Scholar
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
Article CAS PubMed PubMed Central Google Scholar
Koboldt, D. C. et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res. 22, 568–576 (2012).
Article CAS PubMed PubMed Central Google Scholar
Mitchell, A. B., Oliver, B. G. G. & Glanville, A. R. Translational aspects of the human respiratory virome. Am. J. Respir. Crit. Care Med. 194, 1458–1464 (2016).
Article CAS PubMed Google Scholar
Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat. Commun. 12, 1088 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Pardo-Saganta, A. et al. Injury induces direct lineage segregation of functionally distinct airway basal stem/progenitor cell subpopulations. Cell Stem Cell 16, 184–197 (2015).
Article CAS PubMed PubMed Central Google Scholar
Plasschaert, L. W. et al. A single-cell atlas of the airway epithelium reveals the CFTR-rich pulmonary ionocyte. Nature 560, 377–381 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Montoro, D. T. et al. A revised airway epithelial hierarchy includes CFTR-expressing ionocytes. Nature 560, 319–324 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Yeshuroon-Koffler, K., Shemer-Avni, Y., Keren-Naus, A. & Goldbart, A. D. Detection of common respiratory viruses in tonsillar tissue of children with obstructive sleep apnea. Pediatr. Pulmonol. 50, 187–195 (2015).
Article PubMed Google Scholar
Wilk, A. J. et al. Multi-omic profiling reveals widespread dysregulation of innate immunity and hematopoiesis in COVID-19. J. Exp. Med. 218, e20210582 (2021).
Liao, M. et al. Single-cell landscape of bronchoalveolar immune cells in patients with COVID-19. Nat. Med. 26, 842–844 (2020).
Article CAS PubMed Google Scholar
Morse, C. et al. Proliferating SPP1/MERTK-expressing macrophages in idiopathic pulmonary fibrosis. Eur. Respir. J. 54, 1802441 (2019).
Reyes, M. et al. An immune-cell signature of bacterial sepsis. Nat. Med. 26, 333–340 (2020).
Article CAS PubMed PubMed Central Google Scholar
Reyes, M. et al. Plasma from patients with bacterial sepsis or severe COVID-19 induces suppressive myeloid cell production from hematopoietic progenitors in vitro. Sci. Transl. Med. 13, eabe9599 (2021).
Xu, W. et al. IL-10-producing macrophages preferentially clear early apoptotic cells. Blood 107, 4930–4937 (2006).
Article CAS PubMed Google Scholar
El Saleeby, C. M., Bush, A. J., Harrison, L. M., Aitken, J. A. & Devincenzo, J. P. Respiratory syncytial virus load, viral dynamics, and disease severity in previously healthy naturally infected children. J. Infect. Dis. 204, 996–1002 (2011).
Article PubMed PubMed Central Google Scholar
Hasegawa, K. et al. Respiratory syncytial virus genomic load and disease severity among children hospitalized with bronchiolitis: multicenter cohort studies in the United States and Finland. J. Infect. Dis. 211, 1550–1559 (2015).
Article PubMed Google Scholar
Garcia-Mauriño, C. et al. Viral load dynamics and clinical disease severity in infants with respiratory syncytial virus infection. J. Infect. Dis. 219, 1207–1215 (2019).
Article PubMed Google Scholar
Roberts, A. W. et al. Tissue-resident macrophages are locally programmed for silent clearance of apoptotic cells. Immunity 47, 913–927.e6 (2017).
Article CAS PubMed PubMed Central Google Scholar
Ouyang, W. & O’Garra, A. IL-10 family cytokines IL-10 and IL-22: from basic science to clinical translation. Immunity 50, 871–891 (2019).
Article CAS PubMed Google Scholar
Donnelly, S. C. et al. The association between mortality rates and decreased concentrations of interleukin-10 and interleukin-1 receptor antagonist in the lung fluids of patients with the adult respiratory distress syndrome. Ann. Intern. Med. 125, 191–196 (1996).
Article CAS PubMed Google Scholar
Armstrong, L. & Millar, A. B. Relative production of tumour necrosis factor alpha and interleukin 10 in adult respiratory distress syndrome. Thorax 52, 442–446 (1997).
Article CAS PubMed PubMed Central Google Scholar
Sun, L. et al. Effect of IL-10 on neutrophil recruitment and survival after Pseudomonas aeruginosa challenge. Am. J. Respir. Cell Mol. Biol. 41, 76–84 (2009).
Article CAS PubMed Google Scholar
Chung, E. Y. et al. Interleukin-10 expression in macrophages during phagocytosis of apoptotic cells is mediated by homeodomain proteins Pbx1 and Prep-1. Immunity 27, 952–964 (2007).
Article CAS PubMed PubMed Central Google Scholar
Ogden, C. A. et al. Enhanced apoptotic cell clearance capacity and B cell survival factor production by IL-10-activated macrophages: implications for Burkitt’s lymphoma. J. Immunol. 174, 3015–3023 (2005).
Article CAS PubMed Google Scholar
Zhao, Y. et al. Thrombospondin-1 triggers macrophage IL-10 production and promotes resolution of experimental lung injury. Mucosal Immunol. 7, 440–448 (2014).
Article CAS PubMed Google Scholar
Parlato, S. et al. LOX-1 as a natural IFN-alpha-mediated signal for apoptotic cell uptake and antigen presentation in dendritic cells. Blood 115, 1554–1563 (2010).
Article CAS PubMed Google Scholar
Habibi, M. S. et al. Neutrophilic inflammation in the respiratory mucosa predisposes to RSV infection. Science 370, eaba9301 (2020).
Uhl, B. et al. Aged neutrophils contribute to the first line of defense in the acute inflammatory response. Blood 128, 2327–2337 (2016).
Article CAS PubMed PubMed Central Google Scholar
Rørvig, S., Østergaard, O., Heegaard, N. H. H. & Borregaard, N. Proteome profiling of human neutrophil granule subsets, secretory vesicles, and cell membrane: correlation with transcriptome profiling of neutrophil precursors. J. Leukoc. Biol. 94, 711–721 (2013).
Article PubMed Google Scholar
Holm, J. & Hansen, S. I. Characterization of soluble folate receptors (folate binding proteins) in humans. Biological roles and clinical potentials in infection and malignancy. Biochim. Biophys. Acta. Proteins Proteom. 1868, 140466 (2020).
Article CAS PubMed Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e29 (2021).
Article CAS PubMed PubMed Central Google Scholar
Tirosh, I. et al. Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science 352, 189–196 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Haghverdi, L., Lun, A. T. L., Morgan, M. D. & Marioni, J. C. Batch effects in single-cell RNA-sequencing data are corrected by matching mutual nearest neighbors. Nat. Biotechnol. 36, 421–427 (2018).
Article CAS PubMed PubMed Central Google Scholar
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
Article CAS PubMed PubMed Central Google Scholar
Wood, D. E., Lu, J. & Langmead, B. Improved metagenomic analysis with Kraken 2. Genome Biol. 20, 257 (2019).
Article CAS PubMed PubMed Central Google Scholar
Lu, J., Breitwieser, F. P., Thielen, P. & Salzberg, S. L. Bracken: estimating species abundance in metagenomics data. PeerJ Comput. Sci. 3, e104 (2017).
Article Google Scholar
Breitwieser, F. P. & Salzberg, S. L. Pavian: interactive analysis of metagenomics data for microbiome studies and pathogen identification. Bioinformatics 36, 1303–1304 (2020).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

This study was funded by a Children’s Infection Defense Center Grant (St. Jude Children’s Research Hospital) awarded to Paul Thomas and Tim Flerlage, LeBonheur Children’s Hospital Clinical Fellow Research Award to Tim Flerlage, AI121832 and AI144616 to Paul Thomas, ALSAC at St. Jude to Paul Thomas and Tim Flerlage, AI154470 to Paul Thomas and Adrienne Randolph, and PICFlu Study NIH AI084011 to Dr. Adrienne Randolph. The funders had no role in the design or analysis of the study.

Author information

Authors and Affiliations

Department of Infectious Diseases, St. Jude Children’s Research Hospital, Memphis, TN, USA
Tim Flerlage, Shaoyuan Tan & Sherri Surman
Division of Pediatric Critical Care Medicine, Department of Pediatrics, University of Tennessee Health Science Center, Memphis, TN, USA
Tim Flerlage & Alina N. West
Department of Immunology, St. Jude Children’s Research Hospital, Memphis, TN, USA
Jeremy Chase Crawford, E. Kaitlynn Allen & Paul G. Thomas
Department of Pediatrics, University of Tennessee Health Science Center, Memphis, TN, USA
Danielle Severns
Hartwell Center for Biotechnology, St. Jude Children’s Research Hospital, Memphis, TN, USA
Granger Ridout
Department of Anesthesiology, Critical Care, and Pain Medicine, Boston Children’s Hospital, Boston, MA, USA
Tanya Novak & Adrienne Randolph
Department of Anaesthesia, Harvard Medical School, Boston, MA, USA
Tanya Novak & Adrienne Randolph
Department of Pediatrics, Harvard Medical School, Boston, MA, USA
Adrienne Randolph

Authors

Tim Flerlage
View author publications
You can also search for this author in PubMed Google Scholar
Jeremy Chase Crawford
View author publications
You can also search for this author in PubMed Google Scholar
E. Kaitlynn Allen
View author publications
You can also search for this author in PubMed Google Scholar
Danielle Severns
View author publications
You can also search for this author in PubMed Google Scholar
Shaoyuan Tan
View author publications
You can also search for this author in PubMed Google Scholar
Sherri Surman
View author publications
You can also search for this author in PubMed Google Scholar
Granger Ridout
View author publications
You can also search for this author in PubMed Google Scholar
Tanya Novak
View author publications
You can also search for this author in PubMed Google Scholar
Adrienne Randolph
View author publications
You can also search for this author in PubMed Google Scholar
Alina N. West
View author publications
You can also search for this author in PubMed Google Scholar
Paul G. Thomas
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

T.F. wrote and designed the study, enrolled the patients, processed samples, performed single cell assays, analyzed the data and prepared the manuscript. J.C.C performed data analysis and contributed to manuscript and figure preparation. E.K.A. contributed to single cell and ViroCap assays and manuscript preparation. D.S. contributed to data collection and manuscript preparation. S.T. contributed to ViroCap data analysis. S.S. performed protein concentration measurements in tracheal aspirate samples provided by collaborators at Boston Children’s Hospital (T.N. and A.R.). G.R. contributed to ViroCap assays and manuscript preparation. T.N. provided samples and associated clinical data for external validation cohort from Boston Children’s Hospital and contributed to manuscript preparation. A.R. provided samples and clinical metadata for external validation cohort from Boston Children’s Hospital and contributed to manuscript preparation. A.N.W. contributed to study design, patient identification and enrollment, and manuscript preparation. P.G.T. oversaw the study, contributed to data analysis, manuscript preparation, and figure preparation.

Corresponding author

Correspondence to Paul G. Thomas.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Andrew Kau and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Flerlage, T., Crawford, J.C., Allen, E.K. et al. Single cell transcriptomics identifies distinct profiles in pediatric acute respiratory distress syndrome. Nat Commun 14, 3870 (2023). https://doi.org/10.1038/s41467-023-39593-0

Download citation

Received: 24 August 2022
Accepted: 21 June 2023
Published: 30 June 2023
DOI: https://doi.org/10.1038/s41467-023-39593-0

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.