Bulk RNA sequencing of human pediatric lung cell populations reveals unique transcriptomic signature associated with postnatal pulmonary development

Abstract

Postnatal lung development results in an increasingly functional organ prepared for gas exchange and pathogenic challenges. It is achieved through cellular differentiation and migration. Changes in the tissue architecture during this development process are well-documented and increasing cellular diversity associated with it are reported in recent years. Despite recent progress, transcriptomic and molecular pathways associated with human postnatal lung development are yet to be fully understood. In this study, we investigated gene expression patterns associated with healthy pediatric lung development in four major enriched cell populations (epithelial, endothelial, and nonendothelial mesenchymal cells, along with lung leukocytes) from 1-day-old to 8-yr-old organ donors with no known lung disease. For analysis, we considered the donors in four age groups [less than 30 days old neonates, 30 days to < 1 yr old infants, toddlers (1 to < 2 yr), and children 2 yr and older] and assessed differentially expressed genes (DEG). We found increasing age-associated transcriptional changes in all four major cell types in pediatric lung. Transition from neonate to infant stage showed highest number of DEG compared with the number of DEG found during infant to toddler- or toddler to older children-transitions. Profiles of differential gene expression and further pathway enrichment analyses indicate functional epithelial cell maturation and increased capability of antigen presentation and chemokine-mediated communication. Our study provides a comprehensive reference of gene expression patterns during healthy pediatric lung development that will be useful in identifying and understanding aberrant gene expression patterns associated with early life respiratory diseases.

NEW & NOTEWORTHY This study presents postnatal transcriptomic changes in major cell populations in human lung, namely endothelial, epithelial, mesenchymal cells, and leukocytes. Although human postnatal lung development continues through early adulthood, our results demonstrate that greatest transcriptional changes occur in first few months of life during neonate to infant transition. These early transcriptional changes in lung parenchyma are particularly notable for functional maturation and activation of alveolar type II cell genes.

INTRODUCTION

Human pulmonary organogenesis starts with the development of the laryngotracheal groove from embryonic anterior foregut endoderm during early weeks of gestation (1–3). From this groove forms a lung bud that continues to develop through pseudo-glandular (5–17 wk of gestation), canalicular, saccular, and alveolar (fetal age 36 wk onwards) stages to form an organ capable of gas exchange during early days of life (1–6). After birth and in the early years of life, human lung continues to mature into a more functional apparatus for gas exchange as well as a better prepared organ to act during pathogenic challenges and in clearance of inhaled particles and debris from the extensive mucosal surface. Postnatal development of human lung involves increases in gas exchange surface, maturation of microvascular architecture, and increased migration of leukocytes into the tissue (7–16). Despite the recent advances, many aspects of human postnatal lung development are yet to be adequately understood (17–20).

We investigated how postnatal lung development is associated with differential gene expression and activation of cellular/molecular pathways at different stages of this process by enriching major cellular compartments of pediatric lungs, aged 1 day to 8 yr, and examining their transcriptomic profile by bulk RNA sequencing. Differential gene expression patterns and cellular pathway activations were found with advancing age in all four enriched major cell populations with the greatest degrees of difference observed between the neonates and infants, compared with any other groups, indicating that the human lung undergoes major transcriptional changes during the neonatal phase of postnatal life despite relatively limited environmental exposures expected in that time.

MATERIALS AND METHODS

Study Population

Transplant quality donor lungs were procured through the federal United Network of Organ Sharing via the National Disease Research Interchange (NDRI) and the International Institute for Advancement of Medicine (IIAM). Lung cells from 24 pediatric subjects with no known history of respiratory disease and healthy histology, aged 1 day to 8 yr (7 female, 17 male), were examined in this study (Fig. 1A and Supplemental Table S1). Donor lungs were recovered with brief warm ischemic time, were flushed and stored immediately on wet ice in UW (University of Wisconsin) or HTK (Histidine-Tryptophan-Ketoglutarate) buffers as for organs for transplantation, and were processed as early as possible after receipt by the LungMAP Human Tissue Core at University of Rochester Medical Center (https://lungmap.net). The average warm and cold ischemic times were 0.17 ± 0.12 and 25.25 ± 1.62 h, respectively (means ± SE). Donor demographics, causes of death, and pathologists’ review of each lung are included in Supplemental Table S1. The University of Rochester IRB approved and oversees this study (RSRB00047606).

Figure 1.

Pediatric lungs showed more structural complexity and different cellular compositions with increasing age. Lungs from 24 term-born pediatric subjects with no known lung disease, aged 1 day up to 8.5 yr, were assessed for alveolar development by pathological examination and cellular complexity by flowcytometry. A: plot showing donor ages in months. B: regression plot showing increasing radial alveolar counts (RAC) obtained by counting the numbers of alveoli from the center of a respiratory bronchiole to the nearest septal division or pleural margin (shown as average of 10). C: microscopic images of H&E-stained lung sections showing a 1-day-old neonate (top left), a 6-mo-old infant (top right), a 20-mo-old toddler (bottom left), and a 5-yr-old male donor lung (bottom right). D: regression plots showing changing proportions of four major pulmonary cell populations with increasing age determined by flowcytometry. Dissociated lung cells were stained with 7AAD and anti-CD235a antibody to identify and exclude the dead cells and erythrocytes, respectively. From remaining live cells, leukocytes (MIC) were identified using an anti-CD45 antibody. Percentages of CD45+ cells shown are percent of cells positive for CD45 among the live nonerythrocytes (% of live) as detected by the flow cytometer during sorting. Endothelial cells (END) were identified from live nonleukocytes as CD31+CD144+ cells. The CD326+ cells were identified as epithelial cells (EPI), and cells not positive for those membrane proteins were identified as nonendothelial mesenchymal cells (MES). Since the percentages of CD45+ cells (leukocytes) were significantly increased with age, the percentages of END, EPI, and MES are shown here as percent of CD45-negative nonleukocytes (shown as % of CD45−) rather that percent of all live cells. The R and two-tailed P values were determined by Spearman r test, solid lines represent linear regression lines, dashed lines showing 95 percentile error bands (B and D).

Lung Tissue Sectioning, Hematoxylin and Eosin Staining, and Determination of Radial Alveolar Counts

In brief, the right lower lobe (RLL) of each lung was inflated with 10% buffered formalin to 20 cm water pressure for 24–48 h, the lobe was then sliced and fixed for a further 24 h, then dehydrated as ∼1 cm × 0.5 cm tissue blocks in 70% ethanol for 24 h before processing into paraffin as formalin-inflated paraffin-embedded (FFPE) tissues. Five-micron thick FFPE sections were deparaffinized, rehydrated, and stained with hematoxylin and eosin. Details of our laboratory standardized protocols are available online at protocols.io (21–24). Radial alveolar counts (RAC) were assessed by G.D. during histological examination of hematoxylin-eosin (H&E)-stained slides obtained from the RLL (average 6–7 blocks/case). The RAC was calculated as the mean of the number of alveolar septa transected by a perpendicular line drawn from the center of a respiratory bronchiole to the nearest septal division or pleural margin, with a minimum of 10 measurements/case. These measurements were performed on images (magnification ×10) visualized with a digital camera mounted on a Nikon Eclipse 80i microscope using NIS-Elements Advanced Research Software v4.13 (Nikon Instruments Inc., Melville, NY). Digital whole slide images were also created on a Keyence BZ-X810 microscope (Keyence, Itasca, IL).

Lung Cell Processing, Fluorescence-Activated Cell Sorting, and RNA Extraction

Cell dissociation from distal lung tissues, isolation/enrichment of four major cell populations (leukocytes, epithelial, endothelial, and nonendothelial mesenchymal cells) from mixed lung cells by fluorescence-activated cell sorting (FACS), and RNA extraction from those enriched cell populations were performed as described previously (25). Tissues were processed immediately after receiving the organ in our repository. In brief, lung tissues were first cut into small pieces by scissors and then placed into a c-tube (Miltenyi Biotech) with 10 mL digestive enzyme cocktail containing 2 mg/mL collagenase type A (Roche), 1 mg/mL dispase II (Gibco), 0.5 mg/mL porcine pancreas elastase (Worthington Biochemical), and 2 mg/mL bovine pancreas deoxyribonuclease-I (DNase-I, Sigma). Tissues in c-tubes were further disrupted mechanically using a Miltenyi Biotech GentleMACS Octo Tissue Dissociator using the program “mouse tumor implant program-01.01.” Cells were then incubated with digestion cocktail at 37°C for 1 h. After removing the extra digestive enzymes by centrifugation, any contaminating red blood cells were lysed and removed by ammonium-chloride-potassium (ACK) lysis buffer (Biowhittaker) treatment for 5 min at room temperature (20 mL buffer/tube). After the removal of ACK buffer by centrifugation, cells were counted and were kept frozen in 90% fetal bovine serum (Atlanta Biologicals), 10% DMSO (Sigma) in the vapor phase of liquid nitrogen until further processing (∼50 × 10⁶ cells/freezing vial). An illustrated flowchart describing the dissociation process is shown in Supplemental Fig. S1. To enrich lung epithelial cells (EPI), endothelial cells (END), nonendothelial mesenchymal cells (MES), and mixed leukocyte population (MIC) by FACS, thawed lung cells were stained with these conjugated antibodies—CD45 V450 (clone HI30), CD31 BV605 (clone WM59), CD144 FITC (clone 55-7H1), CD235a PE-Cy5 (clone GA-R2) (BD Bioscience, San Jose, CA), CD326/EpCAM PE (clone 1B7, eBioscience, Waltham, MA), and Podoplanin AF647 (clone NC-08, Biolegend, San Diego, CA) as previously described (25). The antihuman Podoplanin AF647 antibody was included in the panel for phenotyping purpose only and was not used for sorting. Cells were also stained with viability marker 7AAD (BD Biosciences). After identifying and removing 7AAD⁺ dead cells and any contaminating CD235a⁺ erythrocytes by FACS sorting, MIC, END, and EPI were sequentially identified and collected by their CD45, CD31 plus CD144, and CD326 expressions, respectively. Remaining cells, negative for these membrane markers, were also collected as MES. A pictorial description of our cell-type identifying and gating strategy for fluorescence-activated cell sorting is shown in Supplemental Fig. S2. RNA was extracted from enriched cell populations immediately after sorting using Qiagen RNeasy kit (Qiagen, Valencia, CA) following manufacturer-provided protocol. RNA quality, indicated by RNA integrity number (RIN) and gel analysis, and quantity were determined by Agilent Bioanalyzer 2100 (Agilent, Santa Clara, CA). Extracted RNA were stored at −80°C until sequenced.

RNA Sequencing

Twenty-four human pediatric lung samples, ranging from 1-day full-term infant to 8 yr old, were digested into single cell suspensions and sorted into major cell types as previously described (25). Bulk RNA sequencing was performed with 1 ng of total RNA per sample amplified by SMARter ultra low amplification kit (Clontech, Mountain View, CA) as described previously (25, 26). Sorted cell gene expression was assessed by RNAseq using the Ilumina NovaSeq6000. Reads were aligned to human reference genome 38 (Hgr38) + gencode 28 using Splice Transcript Alignment to a Reference (STAR_2.6.0c) algorithm (27). Then high-throughput sequence analysis in Python (HTSeq) was used to summarize the expression values (28). Detected transcripts have been deposited on LungMAP.net and the LGEA Web Portal (29).

Analysis of Bulk RNA Sequencing Data

Differential expression.

Differential gene expression comparisons were performed pairwise between samples in adjacent age groups within major cell lineages (i.e., endothelial samples from neonates were compared with endothelial samples from infants). DESeq2 (1.38.3) was used to perform differential expression using ASHR (2.2–63) to estimate shrinkage. R v. 4.2.2 (www.R-project.org) was used within Rmarkdown 2.24 in Rstudio 2022.12.0 + 353. EnhancedVolcano 1.16.0 (github.com/kevinblighe/EnhancedVolcano), ggplot2 3.4.3, plotly 4.10.2, pcaExplorer 2.24.0, pheatmap 1.0.12, and ComplexHeatmap 2.15.4 were used in generating figures (30–36).

Pathway enrichment analysis.

EnrichR (3.2) was used in R (4.2.2) for functional enrichment analysis against the Gene Ontology Biological Process Library (http://geneontology.org) on genes that were differentially expressed above an adjusted P value level of less than 0.05 (37).

Statistical analysis.

In addition to gene expression analyses, RAC and lung cellularity determined by FACS were assessed by Spearman’s rank correlation performed using GraphPad Prism software (GraphPad, La Jolla, CA). Significance is marked when P value was <0.05.

RESULTS

Pediatric Lung Showed Increased Tissue Complexity, Cell Heterogeneity, and Altered Proportions with Increasing Age

To investigate the changes in tissue architecture and cellular heterogenicity in developing human lung, we examined 24 pediatric lung samples from neonates (less than a month old, n = 4), infants (aged 1 mo to 12 mo, n = 9), toddlers (1 to 2 yr old, n = 6), and children (older than 2 yr up to 8 yr old, n = 5) (Fig. 1A). We quantified distal pulmonary structural growth and maturation by radial alveolar count (RAC). RAC is considered a reliable microscopy parameter for evaluating lung complexity and growth when, as described, standardized inflation and multiple measurements are preformed (38). As expected, we found surface area available for gas exchange, as determined by increased RAC (Fig. 1B) and microscope observation of tissue sections by an experienced pediatric lung pathologist (Fig. 1C), increased with advancing age. Increased age also showed changes in proportions of lung cell populations, with significantly increased proportions of leukocytes among live cells as detected during sorting flow cytometry accompanied by decreased endothelial cell percentages among nonleukocytes (Fig. 1D).

RNA-Seq Data Quality

To study differential gene expression by major pulmonary cell populations in pediatric lung, we enriched endothelial (END), epithelial (EPI), nonendothelial mesenchymal (MES) cells, and mixed leukocyte populations (MIC) by FACS as described previously (25). Upon assessment by flowcytometry, dissociated cells showed a median viability of 97.3% (Fig. 2A) and all sorted cell types showed high enrichment (Fig. 2B). Post-sort median enrichment of END, EPI, MES, and MIC were 92.4%, 90.5%, 99.6%, and 91.5%, respectively. We assessed the quality of RNA obtained from those sorted cells by RNA integrity number (RIN). RNA extracted from END, EPI, MES, and MIC showed median RIN of 8.5, 7.8, 8.7, and 8.9, respectively (Fig. 2C). RNA quality was assessed again by high throughput RNA sequencing. Libraries generated 22.8 ± 0.5 million (means ± SE) total reads at a targeted sequencing depth of 10 million reads across all cell types (END 22.5 ± 1.1, EPI 22.1 ± 0.7, MES 23.2 ± 1.0, and MIC 23.6 ± 1.0, all values means ± SE) (Fig. 2D). The percentage of uniquely mapped sequences (mapping rates) among all cell types was 85.9% (mean) with slight variation among these four cell types [EPI 82.8 ± 1.4, END 86.9 ± 0.8, MES 86.2 ± 0.6, and MIC 87.6 ± 2.9, means ± SE (Fig. 2E)]. The transcript coverage for all cell types was 59.5 ± 0.3% (EPI 59.3 ± 0.5%, END 60.1 ± 0.6%, MES 57.6 ± 0.7%, MIC 61.1 ± 1.0%, means ± SE) (Fig. 2F).

Figure 2.

RNA sequencing of enriched human pediatric lung cells showed maximum number of differentially expressed genes during neonate to infant transition. Dissociated lung cells from 24 pediatric subjects aged between 1 day to 8 yr were stained with fluorochrome tagged antibodies against CD235a, CD45, CD31, CD144, and CD362 as well as viability determining dye 7AAD. After identifying and excluding 7AAD+ dead cells and CD235a+ RBCs, CD45+ cells were identified and collected as mixed immune cells (MICs). Among CD45-cells, endothelial (END) and epithelial (EPI) cells were identified as CD31+CD144+ and CD326+, respectively, and enriched. Viable cells not expressing any of the above-mentioned membrane proteins were also collected as nonendothelial mesenchymal cells (MES). A: antibody-stained dissociated cells showed high viability during sorting. B: all sorted cell types showed high enrichment compared with their presort percentages. C: RNA extracted from those sorted cell populations showed high RNA integrity number showing excellent RNA quality. D–F: plots showing number of total reads, transcript coverage (percent), and mapping rates bulk RNA during sequencing. G: principal component analysis scatter plot showing clustering of four enriched lung cell populations. H: genes with top loadings in first and second principal components. I: differential gene expression analysis showing the number of differentially expressed genes (DEG) comparing 1) neonates to infants 2) infants to toddlers, and 3) toddlers to children 2 years and older. #P < 0.001 compared with END, MES, and MIC by unpaired t test.

Gene Expression Patterns of Enriched Lung Cell Populations Shifted the Most during Neonate to Infant Transition Period

Principal component analysis (PCA) on the unabridged dataset showed that ∼73% of variance is captured within the first two principal components, with the first principal component (PC1) being driven primarily by the mixed immune cell population (Fig. 2G). The samples in the epithelial, endothelial, and mesenchymal lineages are well separated with the second principal component (PC2). This is also reflected by the loadings present in the first two principal components (Fig. 2H). Epithelial cell expressed genes largely account for higher PC2 score, whereas endothelial cell expressed genes were top contributors to lower PC2 scores (Fig. 2H). The comparisons between the neonate and infant groups yielded the largest number of differentially expressed genes in all cell types compared with that between infant to toddler and toddler to older children group (Fig. 2I). In END, during neonate to infant transition, there were 1,688 and 1,237 genes (out of 56,870 transcript reads) that were significantly upregulated and downregulated, respectively (adjusted P value < 0.05). In contrast, fewer genes were found to be differentially expressed during infant to toddler (68 upregulated and 66 downregulated) and toddler to older children transition (12 upregulated and 30 downregulated) in END (Fig. 2I). The EPI showed the greatest number of DEGs during neonate to infant transition (3,601 upregulated, 2,932 downregulated) compared with infant to toddler- (56 upregulated, 19 downregulated) and toddler to older child (1 upregulated, 0 downregulated) transitions. The MES and MIC showed similar trends (Fig. 2I). Since comparisons between neonate and infant groups yielded the largest number of differentially expressed genes, we focused subsequent analysis primarily on the differences between these two groups.

Differential Gene Expression between Neonatal and Infant Enriched Lung Cell Populations Point to Differentiation, Maturation of Cellular Function, and Increased Pathogen Response Capacity

To further assess the transcriptomic changes during neonate to infant transition most likely to affect cellular function, we: 1) performed PCA analysis of each sorted cell type to identify age-cohort specific clustering, 2) examined the list of differentially expressed transcripts to identify activation/deactivation of genes associated with lung parenchymal function, tissue/cellular maintenance, antigen presenting molecules, cytokine/chemokine signaling, and 3) performed pathway enrichment analysis using GO: Biological Pathway_2021 library (37).

Principal component analysis of epithelial cell dataset from all donors showed that neonatal epithelial cells were clustered at the lower side of the PC1 that captured 61.22% variance (Fig. 3A). Top and bottom loading plots of PC1 and PC2 showed genes driving most variation on PC1 (such as SFTPA, SFTPB, SFTPC, LAMP3) were alveolar type-2 cell (AT-2)-associated genes (Figs. 3, B and C). Genes associated with parenchymal function and lineage (such as surfactant protein related genes, as well as MUC5, AGER, SCGB1A1), antigen presentation (HLA—A, B, C, and D genes), leukocyte recruitment and activation (several cytokines, chemokines and their receptors), and cell adhesion molecules were significantly upregulated in infant EPIs compared with neonatal EPIs (Supplemental Fig. S3). Biological pathways related to more neutrophil recruitment and cytokine signaling were among the most enriched in infant EPIs compared with that of neonates along PC-1 that explained 33.29% variance (Fig. 3D). The neonatal ENDs were scattered on the lower side of the PC-1 compared with the END from rest of the subjects. Among the 1,688 upregulated and 1,237 downregulated END genes, top genes that explained a higher PC-1 were IL1R, CCL23, and IL18R, indicating more involvement in immune function (Fig. 4, B and C). Genes upregulated in END cells in infants compared with neonates were associated with antigen presentation (HLA), scavenger receptor CD163, complement system proteins (C2, C6, C7), complement decay accelerating factor (CD55), as well as CD59, the inhibitor of complement-mediated membrane complex (Supplemental Fig. S4). Infant END also showed upregulated VWF, PECAM1, ICAM1, as well as several cytokine receptors and chemokines. Interestingly, neonatal END had upregulated IL12A and CXCL14 expression (Supplementary Fig. S4). Biological pathway enrichment analysis of DEGs revealed that pathways linked to the regulation of cell migration/motility and response to cytokines were among top more enriched, while cell division cycle-associated pathways were among least enriched in infant END compared with neonatal END (Fig. 4D). Similarly, the MES dataset from this 24-donor study showed a neonatal MES cluster separate from the older age group MES (Fig. 5A). PC-1 captured 44.87% variance, and the neonatal MES clustered on the lower side of PC-1 axis. Top 10 determinant genes of PC-1 variation included immune function-associated genes C1Q, FCER1G, and CCL18 (Fig. 5, B and C). Infant MES showed significant upregulation of genes associated with antigen presentation (HLA), adhesion molecule ICAM1, complement proteins (C3, C5, C8), several lymphokines (such as IL16, IL18, CXCL5), and lymphokine receptors (e.g., IL1R1, IL6R, CXCR1) compared with neonatal MES (Supplemental Fig. S5). In contrast, neonatal MES showed upregulated IL12A, IL17B, TGFB1, CXCL-12 and 14, and CX3CL1 compared with that of infants, indicating their immunomodulatory capability (Supplemental Fig. S5). The top enriched biological pathways in infants compared with neonates were linked to neutrophil recruitment and activation, response to cytokines, and modulation of lymphocyte function (Fig. 5D). Although we noticed increased percentages of leukocytes in dissociated cells with advancing age, PCA of MIC dataset did not demonstrate clear separation of age-based clusters (Fig. 6A). Neonate MIC did trend toward the lower side of PC1, which overall accounted for 34.69% variance. The top genes that drove PC1 variance included CCL18, B-cell expressed IGHA1, and often macrophage-associated MARCO and CD163 (Fig. 6, B and C). Differentially expressed genes between infant and neonate MIC did not indicate maturation but rather pointed to a change in balance in cytokine response. Infant MIC had upregulated Th2 cytokine IL10, while simultaneously downregulated IL13, another Th2 cytokine (Supplemental Fig. S6). The neonatal MIC population had comparatively upregulated B-cell marker CD19 and dendritic cell (DC)-associated membrane molecules CD1a, CD1c, CD209 indicating that B cells and DC migrated early in neonatal lung compared with other leukocytes. Infant MIC also had increased expressions of proinflammatory IL6, neutrophil chemo-attractants CXCL2, 5 and 8, as well as CSF1 and TGFB1 (Supplemental Fig. S6). Biological pathways associated with neutrophil recruitment, ubiquitination and endoplasmic-reticulum-associated protein degradation (ERAD), and cytokine response dominated the top 10 enriched biological pathways in infant MIC compared with neonatal MIC (Fig. 6D). Differential gene expression analysis showed, 1) increased CCL18, CCL20, and CCL23 expression in all four enriched cell types, 2) upregulated CXCL2, CXCL5, and CXCL8 in EPI and MIC, and 3) increased CXCL2 in END (Supplemental Figs. S3–S6). Concurrently, in infants, 1) decreased CXC14 expression in all cell types, 2) reduced CXCL12 in EPI, MES, and MIC along with, 3) decreased END expression of CXCR4, receptor for both CXCL12 and CXCL14, were found compared with neonates (Supplemental Figs. S3–S6). Our results also demonstrated increased CX3CL1, IL-16, and IL-18 in infant EPI and MES compared with neonates (Supplemental Figs. S3 and S5). We also observed a simultaneous decrease in infant MIC IL-18BP, the protein that binds and blocks IL-18, compared with neonates (Supplemental Fig. S6). These differential gene expression patterns indicate an apparent shift in cytokine gradient-dependent cell migration and activation during neonate to infant transition.

Figure 3.

Human pediatric lung epithelial cells (EPI) undergo dramatic transcriptional changes during neonate to infant transition. Bulk RNA sequencing was performed on enriched human lung cell populations from subjects 1 day to 8 yr old. Enrolled subjects were divided into four age cohorts, neonates (1–30 day old, n = 4), infants (1–12 mo, n = 9), toddlers (1–2 yr old, n = 6), and children (> 2 yr–8 yr, n = 5). The epithelial cell dataset was further assessed for differential gene expression and pathway enrichments. A: principal component analysis of 24 epithelial cell bulk RNA sequencing dataset showing distinct position of neonatal EPI on the negative side of PC plot. B: top 10 genes driving the top and bottom loadings of PC1 and PC2 shown in the loading plot. C: volcano plot showing differentially expressed genes between infant and neonatal epithelial cells. Genes that are top drivers of PC1 and PC2 are highlighted. D: GO biological pathway analysis showing 10 most enriched biological pathways with respect to time in epithelial cells. PC1, principal component 1, PC2, principal component 2.

Figure 4.

Differential gene expression patterns in human lung endothelial cells between neonates and infants. Bulk RNA sequencing was performed on enriched human lung cell populations from subjects 1 day to 8 yr old [four neonates (1–30 days old), nine infants (1–12 mo), six toddlers (1–2 yr), and five older children (>2–8 yr)]. A: principal component analysis of 24 endothelial cell bulk RNA sequencing dataset showing distinct position of neonatal END on the negative side of PC plot. B: top 10 genes driving the top and bottom loadings of PC1 and PC2 shown in the loading plot. C: volcano plot showing differentially expressed genes between neonatal and infant END with top PC loading genes highlighted. D: plot showing 10 most enriched biological pathways with respect to time determined by GO Biological Pathway analysis. PC1, principal component 1, PC2, principal component 2.

Figure 5.

Differences in transcriptomic signature in human infant and neonatal lung mesenchymal cells. Bulk-sequencing was performed on RNA extracted from enriched human lung nonendothelial mesenchymal cells (MES) from subjects 1 day to 8 yr old [four neonates (1–30 days old), nine infants (1–12 mo), six toddlers (1–2 yr), and five older children (>2–8 yr)]. A: principal component analysis of 24 mesenchymal cell bulk RNA sequencing dataset showing distinct position of neonatal MES on the negative side of PC plot. B: loading plot showing top 10 gene expression driving top and bottom loadings of PC1 and PC2. C: MES differentially expressed genes between neonates and infants with top and bottom PC1 and PC2 loading genes highlighted. D: pathway enrichment analysis plot showing top 10 most enriched biological pathways with respect to time. PC1, principal component 1, PC2, principal component 2.

Figure 6.

Comparison of gene expression and biological pathway enrichment in neonatal and infant human lung enriched leukocytes. Mixed immune cell (MIC) bulk-RNA sequencing dataset from 24 pediatric donors [four neonates (1–30 days old), nine infants (1–12 mo), six toddlers (1–2 yr), and five older children (>2–8 yr)] were analyzed for differential gene expression and relevant biological pathway enrichment. A: principal component analysis of 24 mixed immune cell bulk RNA sequencing dataset showing distinct position of neonatal MIC on the negative side of PC plot. B: top 10 genes driving the top and bottom loadings of PC1 and PC2 shown in the loading plot. C: MIC differentially expressed genes between neonates and infants with top and bottom PC1 and PC2 loading genes highlighted. D: plot showing top 10 most enriched biological pathways with respect to time. PC1, principal component 1, PC2, principal component 2.

DISCUSSION

Our knowledge of differential gene expression patterns and enrichment of associated cellular pathways in developing human pediatric lung cells is limited despite recent advances in evaluating transcriptomes in both adult and fetal lungs (39–42). Current understanding of postnatal lung cell transcriptional profiles is based on studies done on murine lung (43–45). This study assessed the transcriptomic profiles of all four major pulmonary cell populations to identify gene expression patterns linked to their functional maturation and developmental states. We included donor lungs aged 1 day to 8 yr old to cover the age groups that experience dynamic postnatal lung development. To investigate transcriptomic signatures that are unique to a specific cell population, we first enriched four major lung cell types by identifying them based on their membrane protein expression and then performed high throughput RNA sequencing on each population. Instead of identifying unique cell types associated with different developmental stages by performing RNA sequencing at single cell level involving fewer subjects, we preferred to perform high throughput bulk-cell RNA sequencing enrolling 24 subjects for greater depth of transcript identification and statistical power to provide better insight into age-dependent transcriptomic profiles of known major lung cell lineages. In this study, we report profound changes in gene expression patterns linked to advancing stages of postnatal human lung development in all major cell types with most dynamic transcriptomic changes occurring during first few months of life.

Human lung undergoes enormous structural development and functional maturations during the first few years of life that continues till adulthood (4, 8, 40, 46). In this study, we assessed the transcriptional changes associated with postnatal lung development. Not surprisingly, we confirm advancing age to be linked with increased alveolar formation and complexity, as demonstrated by increasing alveolar counts and tissue architecture. An altered cellular diversity also accompanied development (Fig. 1). Bulk-cell RNA-sequencing resulted in clear separation between four enriched cell types by PCA. Top genes determining PC clustering included leukocyte-exclusive PTPRC, EPI-exclusive surfactant protein subunits, and END-expressed VWF, further validating our previously published FACS-based lung cell enrichment process (Fig. 2H) (25). Comparison of differentially expressed genes (DEG), both upregulated and downregulated, between two adjacent age cohorts (neonate vs. infant, infant vs. toddler and toddler vs. children 2 yr and older) showed that the highest level of differential gene expression, in all four major cell populations in lung was found between the neonates and infants, despite ongoing structural development throughout childhood (Fig. 2I). The assessment of individual gene expression along with unbiased pathway enrichment analysis yields insights into cellular differentiation and maturation during early stages of postnatal lung development.

Pulmonary alveolar epithelial cells perform the crucial task of gas exchange (47–50). This unique task places those cells in direct contact with inhaled particles and pathogens. We assessed if transcriptional changes during neonate to infant transition reflect gain of parenchymal function as well as maturing of immune surveillance capacity. Compared with neonates, infant EPI had upregulated LAMP3, AGER, PDPN, SCGB1A1, FOXJ1, CEACAM-5/6, and expression of surfactant proteins, indicating differentiation of functionally specialized EPI subtypes (Supplemental Fig. S3). These infant EPI also had increased gene expression of antigen presentation-associated molecules and complement proteins, suggesting a greater role in immune surveillance. Endothelial cells form the thin barrier between alveolar epithelial cells and blood stream to facilitate gas exchange (51–53). They are also critical regulators of immune response and angiogenesis (54–56). Like EPI, neonatal END clustered separately on negative side of PC1 demonstrating inter-donor homogeneity but age-dependent modulation in transcription signature (Fig. 4). When we examined the extent of transcriptional changes happening during the neonate to infant transition, we found increased complement protein, cytokine, and chemokine transcripts. Similarly, neonatal MES clustered separately in PC plot and infant MES showed upregulated antigen presentation and complement pathway-related gene expression compared with neonatal MES. In contrast, PCA of MIC did not show distinct clustering of neonatal cells although MIC had the largest numbers of differentially expressed genes among neonates and infants. The list of differentially expressed genes demonstrated increased macrophage-associated scavenger receptor CD163 (57, 58) and decreased dendritic cell [CD1 molecules (59–61), CD209 (59, 62)] and B-cell- [CD19 (63, 64)] associated gene expression in infants compared with neonates, indicating a possible subpopulation shift in lung-resident MIC (Supplemental Fig. S6).

All enriched cell types in infants showed significantly upregulated expressions of chemokines CCL18, CCL20, and CCL23 compared with neonatal lung cell populations. These chemokines play critical roles in driving adaptive immunity by attracting Th2 cells (CCL18), by directing migration of dendritic cells (DC), effector/memory T cells (T_E/M), and memory B cells (CCL20), and by recruiting resting T cells and monocytes (65–69). Neonate to infant transition also witnessed increases in CXCL2, CXCL5, and CXCL8 transcripts by many enriched cell types. These chemokines are potent neutrophil chemo-attractants and regulator of neutrophil-mediated immune response (70–81). DEG study also showed relatively increased neonatal expressions of CXCL14 in all cell types, and increased CXCL12 transcripts in EPI, MES, and MIC compared with infants. Notably, relative increases in CXCL-14 and 12 occurred simultaneously with increased neonatal END expression of their common receptor, CXCR4 (82, 83).

This transcriptomic study of bulk-FACS sorted human lung cell lineages comparing healthy, full-term born, at postnatal day of life 1–8 yr of age provides analysis of a valuable dataset evaluating human postnatal lung development. Limitations of the study include relatively small sample numbers at each age, concern for which is reduced by the consistency of the data within each age cohort. Histopathology also demonstrates inflammatory infiltration in several cases, likely the effect of terminal mechanical ventilation, which was not significantly different between age groups. In summary, the transcriptomic data support maturation of lung epithelial cell secretory function and enhanced antigen presentation and complement activation by nonimmune cells with age, as well as in the immune cell compartment, in enhanced T-cell and macrophage activation, with rise in chemokines and cognate receptors.

DATA AVAILABILITY

Transcript data are available on https://lungmap.net and the Lung Gene Expression Analysis Web Portal (https://research.cchmc.org/pbge/lunggens/mainportal.html).

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.24759453.

Supplemental Figs. S1–S6: https://doi.org/10.6084/m9.figshare.25135190.

GRANTS

This study was supported by National Heart, Lung, and Blood Institute (NHLBI) Molecular Atlas of Lung Development Program Human Tissue Core Grants U01HL122700 and U01HL148861 (to G.S.P).

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

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