The transcriptional landscape of developing human trisomy 21 lungs

Abstract

Rationale

Trisomy 21 (T21), resulting in Down syndrome (DS), is the most prevalent chromosomal abnormality worldwide. While pulmonary disease is a major cause of morbidity and mortality in DS, the ontogeny of pulmonary complications remains poorly understood. We recently demonstrated that T21 lung anomalies, including airway branching and vascular lymphatic ­abnormalities, are initiated in utero. Here, we aimed to describe molecular changes at the single-cell level in prenatal T21 lungs.

Methods

Single-cell RNA sequencing was used to generate transcriptomic profiles of individual human lung cells in tissue obtained from T21 (n = 5) and non-T21 (n = 4) prenatal lungs. Clustering of cells, marker identification, uniform manifold approximation and projection representation, and differential expression analysis were performed in Seurat. Cell type annotation and pathway analysis were annotated using ToppFun and a human fetal lung cell atlas. Spatial differences in cellular phenotypes were validated using immunofluorescence staining and fluorescent in situ hybridization.

Results

Our results detail changes in gene expression at the time of initiation of histopathological abnormalities in T21 prenatal lungs. Notably, we identify precocious differentiation of epithelial cells, widespread induction of key extracellular matrix molecules in mesenchymal cells, and hyperactivation of IFN signaling in endothelial cells.

Conclusions

This single-cell dataset of T21 lungs greatly expands our understanding of antecedents to pulmonary complications and should facilitate efforts to mitigate respiratory disease in DS.

Clinical Relevance.

This study is the first to describe changes in the distribution of cellular phenotypes, and key gene expression alterations in specific cell sub-populations in prenatal T21 lungs. Our T21 transcriptomic analysis comprehensively defines regulatory processes associated with the initiation of pulmonary anomalies observed in T21 and will help facilitate the identification of therapeutic targets for respiratory disease in DS individuals.

Introduction

Trisomy 21 (T21), resulting in Down syndrome (DS), is the most prevalent chromosomal anomaly worldwide. The partial or complete triplication of human chromosome 21 (HSA21) results in numerous phenotypic abnormalities and complications, encompassing several organ systems. Although often associated with congenital heart disease, pulmonary complications and respiratory conditions remain one of the leading causes of hospitalization and mortality in children with DS ≤3 years of age.¹ Several groups, including our own, have demonstrated that children can present with these pulmonary complications regardless of congenital heart disease.^2-4 Pulmonary complications in individuals with DS may span the entire lung, ranging from bronchomalacia in the large airways to alveolar simplification.^2,5 Considering the prevalence and severity of these complications, the ontogeny of these defects warrants a better understanding.

Lung development is achieved through cross-talk of multiple, different cell lineages.⁶ The anatomical and cellular structure of the fetal lung differs significantly from what is observed in the postnatal lung capable of gas exchange.^7,8 The developing human lung undergoes branching during the pseudoglandular stages of development, which takes place between 5 and 17 weeks of gestation, when proliferation is at its peak, and subsequently enters the canalicular stage of development between 16 and 26 weeks of gestation, when cellular differentiation takes precedence.^9-12 We recently demonstrated that the developing T21 lung presents with histopathological abnormalities starting as early as 16 weeks of gestation.¹³ Anomalies are observed in approximately 70% of those in the late pseudoglandular/early canalicular stage, with varying degrees of phenotypic severity. One of the pathways that is consistently hyperactivated in multiple T21 tissues is the type I interferon (IFN) pathway.¹⁴ In prenatal T21 lungs, we found evidence of IFN dysregulation at the time of initiation of histopathologic abnormalities associated with altered activation of both the complement and coagulation cascades and extracellular matrix (ECM) pathway.¹³ Targeted studies indicating changes in cell proliferation and differentiation suggest that cellular composition is altered in the developing T21 human lung.¹³

In this study, we aimed to define changes in the regulation of gene expression at the cellular level by performing single-cell RNA sequencing (scRNA-seq) on prenatal T21 lungs manifesting with developmental anomalies. Recent studies conducted scRNA-seq on several different prenatal human tissues, which included only a single lung (∼2000 cells) per condition, with little to no lung-­specific analyses.¹⁵ Here, we report changes in the distribution of cellular phenotypes as well as key gene expression alterations in specific cell subpopulations in fetal T21 lungs. Additionally, we note that the pathways and cellular processes being altered in the different cellular lineages are sometimes overlapping, but often distinct, highlighting the complexity of impact from triplication of HSA21. This cellular dataset comprehensively defines regulatory processes associated with the initiation of pulmonary anomalies observed in T21 and may facilitate the identification of therapeutic targets for respiratory disease in individuals with DS.

Materials and methods

Detailed methods are provided in the online data supplement.

Human tissue/ethics

As previously described,⁷ human fetal lung tissues used in this study were collected under institutional review board approval (USC-HS-13-0399 and CHLA-14-2211) provided by the University of Southern California fetal tissue biobank and the University of Washington Birth Defects Research Laboratory. Following informed consent, de-identified human fetal samples were ­collected.

Single-cell sequencing

Human fetal lung tissues were dissociated using the Miltenyi Neural Tissue Dissociation Papain-based Kit (catalog #130-092-628) according to the manufacturer’s instructions alongside additional adaptations. Single-cell capture, library production, sequencing, pathway analysis, and cell type annotation were performed as previously described by us and others.^7,16-18

Fluorescent in situ hybridization/immunofluorescence staining

Combinatorial ACD RNAScope fluorescent in situ hybridization (FISH) and immunofluorescence (IF) staining were performed on lung tissues (T21 and non-T21 age- and sex-matched controls) according to the manufacturer’s instructions and as previously described.⁷

Image analysis

Images were obtained using a Leica THUNDER inverted microscope. The different combinations of combinatorial FISH/IF were performed on 5-6 different lungs for each condition (T21 and non-T21). Sections were imaged at 40× magnification, with 10 images assessed per sample. Using HALO Image Analysis Platform version 3.5.3577 (Indica Labs), each pair was individually assessed to assure accurate quantification of RNA copy number, intensity, and protein expression. Due to the limited number of human samples, nonparametric Wilcoxon signed-rank tests were used as appropriate. A paired Student t-test was used to detect significant differences between non-T21 and T21 age- and sex-matched lungs. The results were considered significant if P ≤.05.

Results

Prenatal T21 lungs display all major cell lineages

To characterize cellular heterogeneity in the developing human fetal T21 lung compared to age-matched controls, we performed scRNA-seq on cells from manually and enzymatically dissociated lungs at 17-19 weeks of gestation. We previously established that histopathological anomalies are observed in T21 lungs as early as 16 weeks of gestation, with approximately 70% of lungs at this gestational age or older presenting with anomalies.¹³ Seeking to better understand what could be contributing to these developmental defects, we sequenced 5 T21 lungs, each displaying some degree of histopathologic alteration, as determined by hematoxylin and eosin staining (Figure S1 and Table S1), and 4 non-T21 control lungs. We analyzed a total of 32 054 cells (15 345 T21 cells and 16 709 non-T21 cells), with an average detection of 2577 genes per cell. All the major cell lineages were clearly identified and validated by examining the expression of canonical marker genes associated with major cell types (Figure S2A and B). Additionally, we used He et al.⁹ as a reference to assess the fidelity of our dataset and found strong agreement in specificity of major lineage markers across the 2 datasets, with some evidence for correlated expression among sublineage marker genes. Whereas expression patterns and levels varied for the canonical lineage markers, their expression was solely observed within clusters of their respective lineage. For example, SFTPB, NXK2.1, and EPCAM are localized in the cluster classified as epithelium, whereas VWF, PECAM, and CLDN5 are expressed in the endothelial cluster (Figure S2B). For the purpose of this study, we focused primarily on the resident cell types (epithelial, mesenchymal, and endothelial), leaving the immune cells for subsequent studies.

Expression summary was assessed at both the condition (T21 vs non-T21) and sample level, demonstrating no unique T21 clusters and no sample-level bias (Figure S2C). Cell lineage proportionality was consistent when having separated and compared the T21 cells to the non-T21 cells (Figure S2D). Although no significant or distinct cell populations are observed in the developing T21 lungs when compared to non-T21 lungs, we wanted to further assess proportional changes in the distribution of specific lung cell types and several differentially expressed genes (DEGs) that could account for the structural/developmental changes observed.

Precocious epithelial cell differentiation in prenatal T21 lungs

Since the most striking histopathological anomaly observed in prenatal T21 lungs is airway dilatation,² we first focused on the epithelial lineage. After subclustering the epithelial lineage, we identified a total of 8 separate populations (Figure 1A), at clustering resolution of 0.5, with repetitive proliferating pneumocytes clusters being combined. A complete list of epithelial cluster marker genes may be found in Table S5. The annotations represented most of the major lung epithelial cell types: club cells and ciliated cells, as well as the distal AT1 and AT2 populations. However, we do appreciate that several proximal and rare epithelial cell populations, such as basal cells and neuroendocrine cells, are missing likely due to the enzymatic approach taken. Realizing their importance, but wanting to disseminate this rare and important dataset in a timely manner, we are currently using single-cell approaches on additional samples that will preserve the proximal epithelial populations for future studies. Additionally, we acknowledge that many of these fully differentiated cell types do not yet exist in the midgestation fetal lung but will refer to them using their annotated terms. We note that cluster 6, which was annotated as mixed, has many mesenchymal attributes, possibly indicative of a mixed lineage that is present during development; however, further investigation outside the scope of this study is required. A degree of redundancy was noted among the cell type annotations, primarily within the distal progenitor populations, likely representing different states of one cell type, as has been observed in other studies.^19,20 For example, cluster 1 is classified as AT1 whereas cluster 3 is annotated as late tip/AT2 and cluster 2 is considered a transitional AT2/AT1 cell. The distinctiveness of each cluster was confirmed when looking at the expression level of specific canonical lung epithelial genes via dot and ridge plot, where we see that the expression of PDPN, AGER, SFTPB, and SFTPC vary among the different clusters (Figure 1B and C). Specifically, we note that late tip/AT2 (cluster 3) differs from the transitional AT2/AT1 (cluster 2) population by having a higher average expression of SFTPC, whereas the transitional AT2/AT1 cluster expresses higher levels of SFTPB. Furthermore, comparing the AT1 population (cluster 1) to the transitional AT2/AT1 (cluster 2), we note that there is markedly higher PDPN and AGER expression in the AT1 population (cluster 1) compared to the transitional AT2/AT1 cluster (cluster 2). Trajectory analyses further validated our annotations of the distal progenitor populations, demonstrating a pseudo-time where the cells start at the mid/late tip stage (cluster 0), travel through late tip/AT2 (cluster 3), move into the transitional AT2/AT1 population (cluster 2), and end in the AT1 group (cluster 1) (Figure 1D). Having established and annotated the epithelial populations, we observed that all subclusters are present for both T21 and non-T21, confirming no unique epithelial cell populations in the T21 lungs compared to non-T21 controls (Figure S3A). Furthermore, no statistically significant proportional differences were observed in any of the subclusters between the 2 cohorts (Figure 1E and Figure S3A).

Figure 1.

Major epithelial cell types along the distal-to-proximal axis are present in T21 and non-T21 human fetal lungs. (A) UMAP displays the 8 epithelial cell subclusters. Each dot represents a single cell, and individual clusters are colored and annotated based on cell type associations identified by ToppFun and human fetal lung cell atlas,¹⁹ using marker genes for each individual cluster. (B) Dot plot of cluster markers of canonical lung epithelial cell lineage validates that these cells are of epithelial origin (CDH1, EPCAM). Expression levels of several specialized markers validate that appropriate cell type annotation; eg, FOXJ1, known to be localized to ciliated cells, is in fact only found in the ciliated subcluster (cluster 5), whereas PDPN and HOPX are most strongly expressed in cell clusters associated with AT1 cells (cluster 1 and 2). Within these populations, the percentage of cells expressing the gene (using dot size) and the average expression level based on unique molecular identifier counts (depth of color) were identified. (C) Ridge plots focusing on the distal/progenitor cell clusters (clusters 0, 1, 2, 3) of cluster markers characteristic of distal lung epithelium (SFTPC, SFTPB, NOTUM, AGER) portray each cluster as unique based on differing expression levels. The x-axis represents the normalized expression level whereas the y-axis is indicative of the relative proportion of cells expressing the gene. (D) Trajectory UMAP of distal airway cell clusters (clusters 0, 1, 2, 3) displays the lineage trajectory of mid/late tip to AT1, analyzed by Monocle 3. (E) Proportional differences in epithelial subclusters between non-T21 and T21 cohorts show no significant differences when assessed by propeller (see Supplementary Methods). Abbreviations: T21, trisomy 21; UMAP, uniform manifold approximation and projection.

Due to the recognized importance of surfactant during lung development, we interrogated relevant genes of interest that were differentially expressed including SFTPC, SFTPB, SFTPD, and NAPSA (Figure 2A and Figure S3C). Each of these markers was significantly upregulated in the distal progenitor populations (clusters 0, 1, 2, 3) of the T21 lungs compared to the non-T21 lungs. Furthermore, our data showed significantly increased expression of ABCA3 in several clusters, including the late tip/AT2 cluster (cluster 3; Figure S3C and D), a gene recently implicated as a marker of mature AT2 cells.^21,22 It was also noted that the progenitor cell markers SOX2 and SOX9 were significantly downregulated in several T21 clusters when compared to the non-T21 clusters (Figure 2A), corroborating our previous studies.¹³ We next interrogated pathway activation in the epithelial cell clusters from developing T21 lungs. For each cluster, we identified DEGs and used those to assess dysregulated pathways. A complete list of DEGs for each epithelial cluster is presented in Table S6. From these analyses we noted upregulation of the surfactant metabolism pathway (Figure 2B), as indicated by the differentially increased expression of genes previously listed such as SFTPB, SFTPC, SFTPD, NAPSA, ABCA3, and SLC3A2. A complete list of upregulated and downregulated pathway analyses for each epithelial cluster is presented in Tables S7 and S8, respectively.

Figure 2.

Trisomy 21 (T21) epithelial subclusters suggest accelerated differentiation compared to controls. (A) Violin plots of differentially expressed genes in T21 lungs compared to non-T21 controls. SFTPB, SFTPC, NAPSA, and ICAM1 are increased in several epithelial subclusters of T21 compared to non-T21, whereas SOX2 and SOX9 are decreased in several of the T21 epithelial subclusters when compared to non-T21. Significance based on adjusted P value (P <.05) is indicated by * above a bar linking the groups being compared. (B) Pathway analysis comparing the T21 epithelial cells to non-T21 cells shows upregulation (red) and downregulation (green) of several pathways. Surfactant metabolism is shown to be upregulated in several of the T21 epithelial subclusters, suggestive of accelerated differentiation. Alternatively, other pathways, such as the translation initiation/elongation pathway, are downregulated. (C) Fluorescent in situ hybridization of non-T21 control and T21 age- and sex-matched fetal lungs clearly demonstrate increased expression of both SFTPB and SFTPC within the epithelium of T21 lungs compared to non-T21 lungs, with several cells colocalized for both genes in the T21 lung (indicated by arrow; Figure S3F). SFTPC: copy number P = .0711; intensity P = .0298; SFTPB: copy number P = .0265, intensity P = .0259 (n = 6 for each condition; scale bar: 64.7µm). Box plots: center line, mean; box limits, upper and lower quartiles; whiskers, minimum to maximum.

We used combinatorial FISH-IF to validate differential expression with context to cellular location. Specifically, we observed a trend toward an increase in the percentage of epithelial cells positive for SFTPD (P = .0535) as well as the number of SFTPD transcripts per cell (P = .1981) in the T21 lungs compared to the non-T21 lungs (Figure S3E). However, we found higher expression of SFTPC within the distal epithelium of the T21 lungs when compared to non-T21 sex- and age-matched controls, and stronger presence of SFTPB in the proximal epithelium, both of which were determined by the quantification of average transcription copy number per cell as well as expression intensity (SFTPC: copy number P = .0711; intensity P = .0298; SFTPB: copy number P = .0265, intensity P = .0259; Figure 2C). Interestingly, the T21 lungs have a larger number of double-positive SFTPC/SFTPB cells within the proximal epithelium (Figure 2C), which was suggested in the single-cell data (P = .1101; Figure S3F). Furthermore, although the percentage of epithelial cells positive for ABCA3 is comparable between T21 and non-T21 lungs, there appears to be on average more transcripts per cell (P = .0535) in the T21 lungs, specifically within the epithelium (P = .0223), with a stronger intensity (P = .0208), also primarily within the epithelial cells (P = .0128), validating that ABCA3 expression is in fact greater within the epithelium of developing T21 lungs (Figure S3D). Taken together, the increased differential expression of several surfactant proteins, as well as the increase of ABCA3 in the AT2 cluster along with the downregulation of the progenitor markers, strongly suggests premature differentiation and/or maturation of distal epithelial cells in T21 lungs. Separated color panels of all epithelial FISH/IF stains may be found in Figure S4 to help better discern areas of co-localization and expression pattern.

Through pathway analyses it was also noted that T21 lungs display an upregulation of the integrin pathway and downregulation of the translation initiation/elongation pathway (Figure 2B). Having observed abnormal epithelial architecture in the histopathology of the T21 lungs, dysregulation of the integrin pathway required some further investigation. This was achieved by performing gene set enrichment analysis (GSEA) and interrogating the Molecular Signatures Database (MSigDB) hallmark Apical Junction gene set, which proved to be upregulated in several of the epithelial subclusters. This was confirmed through the dysregulation of genes contributing to the Apical Junction hallmark gene set, MDK and ICAM1 (Figure 2A and Figure S3G), which are differentially expressed in several epithelial subclusters.

Acknowledging that certain proximal epithelial cell types were missing from our single-cell analyses, we investigated several of these populations in situ to understand how they compared in prenatal T21 lungs to age- and sex-matched non-T21 controls. FISH for TP63 displayed a trend toward a decrease in the number of TP63 transcripts per cell (P = .0532) in T21 lungs compared to non-T21 lungs (Figure S5A–D), which was confirmed by qRT-PCR (Figure S5E; P = .0589), suggesting fewer basal cells in the developing T21 lung. Assessing the rare neuroendocrine population, which has been described to be the first specialized cell type to appear in the lung,²³ a bombesin immunohistochemical stain demonstrated significantly fewer positive cells in the T21 lungs when compared to the non-T21 controls (Figure S5F–I; P = .0311). Alternatively, IF stains for SCGB1A1 and SCGB3A2, interrogating the club cell population, were significantly increased in the T21 lungs (Figure S5J–M and O–R; SCGB1A1: P = .0296; SCGB3A2: P = .0138), which was also confirmed at the gene expression level by qRT-PCR (Figure S5N and S; SCGB1A1: P = .0169; SCGB3A2: P = .0468). This suggests that precocious epithelial cell differentiation may also be observed in the proximal epithelial cells of T21 lungs, as it has been previously described that club cell differentiation begins in the canalicular stage of development, with many club cell proteins increasing with advancing gestation.¹⁰ However, as previously mentioned, we appreciate that this does not fully account for the full transcriptomic profile of these proximal epithelial cell types and are therefore continuing subsequent studies, with modified isolation techniques, to preserve these populations and conduct sequencing in the future.

Dysregulated mesenchymal ECM production in T21 lungs

Proper lung branching morphogenesis is a dynamic process that requires cross-talk between the epithelium and mesenchyme and can be driven by spatiotemporal deposition of ECM. Therefore, to better understand the branching defects observed in the developing T21 lung, we next investigated the mesenchymal cell lineage. Initial subclustering of the mesenchyme revealed that several subclusters presented as nonspecific. After investigating this further, we realized it was in result to several mesenchymal cells being in either the S or M phase of the cell cycle. Although our first instinct was to regress out proliferation, the results continued to be inconclusive. Therefore, being that these cells were clustering on their own in the total object, we decided to separate the mesenchyme into 2 groups: mesenchyme and proliferating mesenchyme (Figure S2A). For the mesenchyme, we determined that the fetal human lung has 16 separate subclusters (Figure 3A) at a clustering resolution of 0.4. A complete list of mesenchymal cluster marker genes can be found in Table S9. There were 2 nonspecific mesenchymal clusters, clusters 8 and 13, both categorized as mixed, suggesting the presence of possibly several “multipotent” cell populations in the developing lung that remain poorly described. Alternatively, these may represent cells that were misclassified during initial steps of lineage assignment.

Figure 3.

Diverse mesenchymal cell types are present in T21 and non-T21 human fetal lungs. (A) UMAP displays the 16 mesenchymal cell subclusters. Each dot represents a single cell, and individual clusters are colored and annotated based on cell type associations identified by ToppFun and the human fetal lung cell atlas,¹⁹ using marker genes for each individual cluster. (B) Dot plot of cluster markers of canonical lung mesenchymal cell lineage (MEST, COL1A1) validates that these cells are of mesenchymal origin. Expression levels of several specialized markers validate appropriate cell type annotation; eg, PDGFRB, which is a known marker of pericytes, is primarily found in the pericyte subcluster (cluster 6), whereas ACTA2 is primarily localized in the smooth muscle cell clusters (clusters 3, 5, 14). Within these populations the percentage of cells expressing the gene (using dot size) and the average expression level based on unique molecular identifier counts (depth of color) were identified. (C) Several mesenchymal subclusters demonstrated a degree of redundancy, having annotated as some form of an alveolar fibroblast. Therefore, ridge plots of cluster markers (FN1, COL6A1, A2M, STAT3) were used to help distinguish the differences between these clusters (clusters 0, 1, 2, 4, 10, 12). The x-axis represents the normalized expression level whereas the y-axis is indicative of the relative proportion of cells expressing the gene. (D) Trajectory UMAP of alveolar fibroblast–associated cell clusters (clusters 0, 1, 2, 4, 10, 12) displays the different lineage trajectories of immature lung fibroblasts (cluster 0) to lipofibroblasts (cluster 4), FABP7 alveolar fibroblasts (cluster 10), and alveolar fibroblasts (cluster 12), analyzed by Monocle 3. (E) Proportional differences in mesenchymal subclusters between non-T21 and T21 cohorts show that T21 lungs have a reduced population of airway fibroblasts (cluster 11; P = .1005). They also have a trend toward a larger population of pericytes (cluster 6; P = .0582) and myofibroblasts (P = .0436) (*P <.05). Assessed by propeller (see Supplementary Methods). Abbreviations: ASMC, airway smooth muscle cell; T21, trisomy 21; UMAP, uniform manifold approximation and projection; VSMC, vascular smooth muscle cell.

While some mesenchymal clusters display redundant cell type annotations, the expression levels of several canonical lung mesenchymal genes in each cluster differ, suggesting a different cell state (Figures 3B and C). Whereas clusters 0, 2, 4, 10, and 12 are all partially classified as alveolar fibroblasts, each cluster has differing levels of expression of both COL6A1 and FN1 (Figure 3C). Furthermore, some of these clusters express very specific markers, making them quite unique. In cluster 4, A2M is highly expressed, providing it the co-annotation of a lipofibroblast, whereas cluster 10 is the only one expressing FABP7 (Figure 3B and C). To further investigate the concept of different cell states in the alveolar fibroblast populations, we performed trajectory analyses (Figure 3D). From this we noted that the cells go from immature lung fibroblasts (cluster 1), through alveolar/intermediate fibroblasts (cluster 0) and branch off into the 3 distinct “mature” alveolar fibroblast populations: alveolar/lipofibroblasts (cluster 4), FABP7 alveolar fibroblasts (cluster 10), or through the alveolar/mid fibroblasts (cluster 2) to ultimately reach alveolar fibroblasts (cluster 12). Another set of mesenchymal subclusters that likely represent different cell states is clusters 3, 5, and 7. Whereas cluster 3 is labeled as airway smooth muscle cell (ASMC)/myofibroblasts, cluster 5 is labeled solely as ASMC and cluster 7 solely as myofibroblasts. When comparing the 3 clusters to one another, it is evident that cluster 3 shares markers with the other 2, yet when cluster 7 and 5 are compared to one another there is no overlap of gene expression. For example, cluster 3 and 5 both express ACTA2, which is not expressed in cluster 7 (Figure 3B). Alternatively, cluster 3 and 7 both express CXCL14, which is lacking in cluster 5 (Figure 3B).

We confirmed that no unique mesenchymal cell populations are present in the T21 lungs (Figure S6A). Proportional differences for some mesenchymal subpopulations were observed (Figure 3E) and were consistent with our prior observations indicating reduced branching in T21 fetal lungs.^3,13 Although not significant, we note a reduced population of airway fibroblasts (cluster 11), accompanied by a larger myofibroblast population (cluster 7). Furthermore, non-T21 lungs have a larger population of cluster 2 alveolar/mid fibroblasts, whose cluster markers are indicative of a more naive alveolar fibroblast. Altogether the reduction of airway fibroblasts, which normally support epithelial branching, and increase of myofibroblasts, which cause excessive ECM deposition, could structurally impair the branching observed in the T21 lung.

We next interrogated DEGs in each of the 16 mesenchymal subclusters. A complete list of DEGs for each mesenchymal cluster is found in Table S10. Considering how large the mesenchyme is, and the diverse populations it entails, we explored DEGs in 2 groups: fibroblasts (clusters 0, 1, 2, 4, 9, 10, 11, 12) and myogenic-associated clusters (clusters 3, 5, 6, 7, 14). We observed an upregulation of several interstitial matrix–associated fibrillar collagens, whose role is to contribute to the mechanical strength and framework of the tissue. COL6A1 and COL6A2 were consistently upregulated in the mesenchymal subclusters (Figure 4A and Figure S6C). Additionally, dysregulation of fibronecting (FN1), which has been shown to be essential for proper lung branching morphogenesis,²⁴ was also observed in several clusters (Figure S6C). Pathway analysis of DEGs confirmed upregulation of ECM and collagen pathways in several mesenchymal subclusters. Interestingly, this analysis also identified downregulation of the smooth muscle contraction pathway (Figure 4B) resulting from the downregulation of several DEGs, including ACTA2, ACTG1, and IGF1 (Figure 4A and Figure S6C). A complete list of upregulated and downregulated pathway analyses for each mesenchymal cluster is presented in Tables S11 and S12, respectively.

Figure 4.

Dysregulated ECM is observed in prenatal T21 lungs. (A) Violin plots of differentially expressed genes in T21 lungs compared to non-T21 controls. Mesenchymal violin plots have been separated based on association. Subclusters annotated as fibroblasts are grouped into one violin plot (clusters 0, 1, 2, 4, 9, 10, 11, 12), whereas myogenic-associated subclusters are grouped into another (clusters 3, 5, 6, 7, 14). COL6A1, COL6A2, and COL1A1 are increased in several mesenchymal fibroblast subclusters of T21 when compared to non-T21, whereas ACTA2, ACTG1, MYH11, and IGF1 are decreased in several of the T21 mesenchymal myogenic-associated subclusters compared to non-T21. Significance based on adjusted P value (P <.05) is indicated by * above a bar linking the groups being compared. (B) Pathway analysis comparing the T21 mesenchymal cells to non-T21 shows upregulation (red) and downregulation (green) of several pathways. Both the collagen and the ECM pathways are upregulated in several of the T21 mesenchymal subclusters, with overlap between the 2 pathways. Alternatively, the SMC pathway is shown to be downregulated in in several of the T21 mesenchymal subclusters. (C) Fluorescent in situ hybridization of non-T21 control and T21 age- and sex-matched fetal lungs clearly demonstrate increased expression of both COL6A1 and COL1A1 within the mesenchyme of T21 tissues compared to non-T21, with several cells colocalized for both genes in T21 (Figure S6E). COL6A1: copy number P = .0337, intensity P = .0381; COL1A1: copy number P = .2866, intensity P = .2979 (n = 6 for each condition; scale bar: 64.7 µm). Box plots: center line, mean; box limits, upper and lower quartiles; whiskers, minimum to maximum. Abbreviations: ASMC, airway smooth muscle cell; ECM, extracellular matrix; SMC, smooth muscle cell; T21, trisomy 21; VSMC, vasculat smooth muscle cell.

We quantified the expression of different ECM genes in context to their localization using combinatorial FISH-IF. COL6A1 and COL6A2, located on HSA21, were significantly increased in both the copy number per cell as well as signal intensity in T21 lungs when compared to non-T21 lungs (COL6A1: copy number P = .0337, intensity P = .0381; COL6A2: copy number P = .0346, intensity P = .0322) (Figure 4C and Figure S6D). Conversely, there was no significant increase in either copy number or signal intensity for COL1A1 (copy number P = .2866, intensity P = .2979). However, we did note that there was a significantly larger population of double-positive COL6A1/COL1A1-expressing cells in the T21 lungs compared to the non-T21 lungs (P = .0289; Figure S6E). Separated color panels of all mesenchymal FISH/IF stains may be found in Figure S7 to help better discern areas of co-localization and expression pattern.

Comparable proliferative mesenchyme population between T21 and non-T21 lungs

As previously mentioned, when assessing the mesenchyme, there was a subset of cells clustering based on their proliferative signature. Therefore, these cells were treated as a separate “lineage” from the rest of the mesenchyme and labeled proliferative mesenchyme. We determined that the fetal human lung has 7 separate proliferative mesenchymal subclusters (Figure S8A). A complete list of proliferative mesenchyme cluster marker genes can be found in Table S13. Assessing cluster markers, it was evident that these populations segregated based on S and M cell cycle phase (Figure S8B and C). Clusters 0, 2, and 3 primarily expressed PCNA, which is more indicative of the S phase, whereas clusters 1, 4, and 5 expressed more TOP2A and MKI67, suggesting the M phase. Post-subclustering, the proliferative cluster markers were regressed out and the expression levels of several canonical lung mesenchymal genes were used to help determine the specific mesenchymal cell types, such as CXCL14, which is only expressed in the myofibroblast cluster (cluster 3; Figure S8B). We confirmed that no unique proliferative mesenchymal cell populations exist in the T21 lungs (Figure S8D). However, given the sparsity of the “lineage,” totaling 1759 cells and comprising the smallest population compared to the total number of cells, minimal DEGs were determined. Therefore, the only differences assessed between T21 and non-T21 was proportionality, where based on unadjusted P values a decrease of proliferating mid/intermediate fibroblasts (cluster 0, P = .0330) and an increase of early/mid fibroblasts (cluster 4, P = .0494) in the T21 lungs were noted (Figure S8F).

Altered T21 lung vascular development is associated with aberrant type I IFN signaling in endothelial cells

We and others have demonstrated that the T21 lung often presents with abnormal vascular growth in utero.^13,25 Whereas an upregulation of certain anti-angiogenic factors has been reported in these lungs, there is limited information describing the ontogeny of these defects. Focusing specifically on the endothelial cell population, we determined a total of 10 distinct endothelial subpopulations (Figure 5A). A resolution of 1.0 was selected for the endothelium, with repetitive general capillary clusters being combined. A complete list of endothelial cluster marker genes is shown in Table S14. Much like the other cell lineages, the annotations represented most of the major lung endothelial cell types, including the newly described aerocytes (cluster 7). Interestingly, there were 2 clusters (clusters 2 and 8) that annotated as pericytes, a cell type often associated with mesenchymal cells. When ­observing where these cells localized in the total object, it was noted that the mesenchymal pericytes clustered on their own next to the mesenchymal population, whereas the endothelial pericytes clustered within the endothelium (Figure S9A). Furthermore, when assessing the expression of canonical endothelial markers, mesenchymal markers, and pericyte-specific markers within the mesenchymal pericyte cluster, endothelial pericyte clusters, and other cells, it was noted that the endothelial pericyte clusters expressed PDGFRB and HIGD1B as well as PECAM1 and CLDN5, thus establishing their endothelial nature (Figure S9B). Alternatively, the mesenchymal pericytes also expressed pericyte markers, like PDGFRB and HIGD1B, alongside mesenchymal markers PLAC9 and ACTA2 with no endothelial marker expression. Having exhausted the appropriate decontamination protocols, and clearly demonstrating cells with coexpression of endothelial and pericyte-­specific markers in both the single-cell data (Figure S9C) and by performing a FISH-IF of HIGD1B and CD31 in lung sections (Figure S9D), these clusters thus reveal the cellular diversity observed in the developing human lung and would require further investigation in the future.

Figure 5.

All major lung endothelial cell types are present in prenatal non-T21 and T21 lungs. (A) UMAP displays 10 endothelial cell subclusters. Each dot represents a single cell, and individual clusters are colored and annotated based on cell type associations identified by ToppFun and the human fetal lung cell atlas,¹⁹ using marker genes for each individual cluster. (B) Dot plot of cluster markers of canonical lung endothelial cell lineage validates that these cells are of endothelial origin. PECAM1 and CLDN5, universal developmental lung endothelial cell markers, are expressed in each of the subclusters. Alternatively, expression levels of several specialized markers validate appropriate cell type annotation; eg, PROX1, a known lymphatic marker, is solely found in the lymphatic subcluster (cluster 6) and ACKR3, a venous marker, is strongly expressed in the venous subcluster (cluster 5). Within these populations, the percentage of cells expressing the gene (using dot size) and the average expression level based on unique molecular identifier counts (depth of color) are identified. (C) The different microvasculature cells of the developing lung display a similar gene signature. Therefore, ridge plots of cluster markers (SOX7, CD93, THBD, EDNRB) for the microvasculature (clusters 0, 1, 3, 4, 7) were used to help distinguish differentiation (naïve cap to late gCap) and phenotypic (gCap vs aerocyte) differences. The x-axis represents the normalized expression level whereas the y-axis is indicative of the relative proportion of cells expressing the gene. (D) Trajectory UMAP of microvasculature-associated cell clusters (clusters 0, 1, 3, 7) displays the different lineage trajectories of naïve capillaries (cluster 1) branching into either late gCaps (cluster 3) or aerocytes (cluster 7), analyzed by Monocle 3. (E) Proportional differences in endothelial subclusters between the non-T21 and T21 cohorts were minimal when assessed by propeller (see Supplementary Methods). The only subcluster to show a trend toward a larger population in the T21 lungs is the aerocytes (cluster 7; P = .0625). Abbreviations: gCap, general capillary; T21, trisomy 21; UMAP, uniform manifold approximation and projection.

Similar to what was observed in both the epithelium and mesenchyme, there was a slight degree of redundancy in the endothelial clusters’ annotations, primarily within the microvasculature. This again suggests a different cell state, as indicated by cluster 1 annotating as a naïve capillary, cluster 0 as a mid/late general capillary (gCap), cluster 3 as a late gCap, and cluster 7 as an aerocyte capillary. However, when assessing the expression of well-established endothelial markers, they are evidently separate populations (Figure 5B and C). Specifically, cluster 3 expresses more SOX7, SOX11, and CD93, markers known to be expressed in more established gCaps, when compared to an embryonic microvascular endothelial cell phenotype (cluster 1, naïve cell)^26,27 (Figure 5C and Figure S10A). Alternatively, we note that VIM is highly expressed in the naïve (cluster 1) and mid/late gCap (cluster 0) clusters, with lower levels of expression in the late gCap (cluster 3), and SPARCL1 being most strongly expressed in the mid/late gCap (cluster 0) (Figure 5C and Figure S10A). Furthermore, cluster 7 is the only one expressing the known aerocyte markers TBX2 and EDNRB (Figure 5C and Figure S10A). To establish the phenomenon of different cell states, we assessed all microvasculature clusters via pseudo-time (Figure 5D). From this it was observed that the cells initiate in the naïve capillary (cluster 1) stage, travel through the mid/late gCap (cluster 1), and bifurcate into becoming late gCaps (cluster 3) or aerocytes (cluster 7). We confirmed that no unique endothelial cell populations exist in the T21 lungs (Figure S10B). When assessing proportionality of the endothelial subclusters, it was interesting to note that, although not statistically significant, the T21 lungs appear to have a larger population of aerocytes compared to the non-T21 lungs (Figure 5E).

The endothelial population made up a small percentage of the total number of cells, resulting in their subclusters being relatively small in each condition (non-T21 vs T21), thus limiting the number of DEGs observed. For a complete list of DEGs for each endothelial cluster, please see Table S15. Therefore, we focused on using GSEA, a rank-based method, to perform pathway analysis and assess endothelial cell differences in developing T21 lungs.²⁸ Specifically, we interrogated the MSigDB hallmark gene sets relevant to endothelial cell development.²⁹ These analyses revealed that the hallmark IFN-α response (type I IFN) and the hallmark IFN-γ response (type III IFN) were upregulated in T21 lungs compared to the non-T21 lungs (Figure 6A). Several genes (BST2, IFI27, IFITM2, IFITM3) contributing to these 2 hallmark gene sets were upregulated in many of the endothelial subclusters (Figure 6B). This upregulation is of interest given that high type I IFN inhibits vascular endothelial growth factor–induced development of capillary-like structures^30,31 and is associated with a decrease of endothelial progenitor cells.³² Thus, the upregulation of these pathways may possibly be implicated in the vascular “simplification” phenotype observed in the T21 lung.^13,25

Figure 6.

Upregulated type I IFN signaling is associated with disrupted endothelium in prenatal T21 lungs. (A) Performing gene set enrichment analysis and interrogating the Molecular Signatures Database hallmark gene set collection demonstrated upregulation of the type I IFN-α response and type III IFN-γ response gene set in numerous endothelial subclusters. Several genes (BST2, IFI27, IFITM2, IFITM3) contributing to these 2 hallmark gene sets were upregulated in many of the endothelial subclusters, as represented by violin plots. (B) Violin plots of differentially expressed genes in T21 lungs compared to non-T21 controls. BST2, IFI27, IFITM2, IFITM3, and DUSP1 are increased in several endothelial subclusters of T21 when compared to non-T21, whereas LYVE1 is decreased in several of the T21 endothelial subclusters compared to non-T21. Significance based on adjusted P value (P < .05) is indicated by * above a bar linking the groups being compared. (C) Fluorescent in situ hybridization of non-T21 control and T21 age- and sex-matched fetal lungs clearly demonstrate increased expression of both type I IFN markers IFI27 (red) and BST2 (green) within the endothelium (CD31, white) of T21 lung compared to non-T21, with several cells colocalized for both genes in T21 (Figure S10D). IFI27: copy number P = .0470, intensity P = .0163; BST2: copy number P = .1471, intensity P = .0777 (n = 6 for each condition; scale bar: 64.7 µm). Box plots: center line, mean; box limits, upper and lower quartiles; whiskers, minimum to maximum. Abbreviations: gCap, general capillary; T21, trisomy 21.

We validated these results using combinatorial FISH-IF to quantify the expression of different type I IFN–associated genes in context to their localization. In both the non-T21 and T21 lungs, BST2 and IFI27 are primarily localized in the endothelium. However, the average transcript number and intensity per endothelial cell of IFI27 are significantly greater in T21 compared to non-T21 lungs (IFI27: copy number: P = .0470; intensity: P = .0163). Additionally, there is an upward trend of BST2 transcript number and intensity in the T21 lungs compared to the non-T21 lungs (BST2: copy number: P = .1471; intensity: P = .0777) (Figure 6C). When assessing the coexpression of these 2 genes, we see a significantly larger population of double-positive IFI27/BST2 expressing endothelial cells in T21 lungs when compared to non-T21 lungs (P = .0230; Figure S10D). Separated color panels of endothelial FISH/IF stains may be found in Figure S10E to help better discern areas or co-localization and expression pattern.

Discussion

Adult and pediatric respiratory morbidity has long been recognized as a complication of Down syndrome. Overall, the gross structure of the mature T21 lung is quite comparable to that of a non-T21 lung, but may present with mild alveolar, airway, and vascular changes. We recently described a fetal origin for lung histopathological abnormalities in T21.¹³ The goal of the studies reported here was to comprehensively catalog cellular and molecular alterations in the fetal T21 lung, coincident with the onset of histological changes. It comes as no surprise that there is a similar composition of cell types at the lineage level (epithelial, mesenchymal, proliferative mesenchymal, endothelial, immune) in T21 and non-T21 lungs (Figure S2). However, it is interesting to note in our data presented herein, that multiple differences in gene expression are observed dependent upon individual lineages. Our data uncover an early and complex effect of HSA21 triplication on lung morphogenesis and provide a foundational description of molecular changes at the cellular level. Furthermore, the complexity of the system becomes more apparent when considering that dysregulation occurs not only in genes localized on HSA21 (ie, IFNAR1, COL6A1) but also in several genes located elsewhere (ie, IFI27, COL1A1). Understanding the interplay and regulation between these 2 gene sets will require further investigation to fully elucidate the underlying altered mechanisms.

In the epithelial lineage we noted strong evidence of precocious cellular differentiation in the T21 lungs compared to the non-T21 controls.¹⁹ These observations are of interest given that DS has been described as a progeroid syndrome, characterized by accelerated maturation or aging.^33-35 Individuals with DS start aging prematurely, and present with conditions more characteristic of the geriatric population.^35,36 These signs of early aging have been described to affect primarily the neurological system, with high prevalence of neurodegenerative disorders (ie, dementia and Alzheimer disease) as well as the immune system.³⁷ This accelerated aging may be associated to genome-wide perturbation in epigenetic regulation.^38,39 When conducting pathway analysis on our 8 epithelial subclusters, we determined an upregulation of surfactant metabolism in the T21 lungs, which was confirmed by increased differential expression of several surfactant proteins (SFTPB, SFTPC, SFTPD) (Figure 2A and B;Figure S3C). It is also interesting to note that the decreased translation initiation/elongation pathway observed in several epithelial clusters of the T21 lungs could be indicative of a regulatory process toward the advanced cell differentiation taking place (Figure 2B). It has been previously demonstrated that ribosomal proteins regulate cell growth, proliferation, and differentiation and attempt to maintain an appropriate stoichiometry to assure a functional proteome.⁴⁰ This dysregulation of ribosomal protein expression is a phenomenon that has been previously described in other studies assessing prenatal mouse and human T21 tissues,¹⁵ and requires further investigation. Furthermore, our single-cell data showed significantly increased expression of ABCA3, a gene recently shown to be a marker of mature AT2 cells.²¹ ABCA3 is expressed in the human fetal lung normally after 28 weeks of gestation^21,22; however, we observe a robust expression at 18 weeks of gestation in T21 cluster 3 late tip/AT2 (Figure S3C and D). This concept of premature maturation may further be extrapolated from our previous findings where we demonstrated significantly decreased proliferation in the human fetal T21 lungs when compared to age- and sex-matched non-T21 lungs in conjunction with decreased proximal airway SOX2 expression, which we further validated in our single-cell dataset (Figure 2A), and abnormal ASMC patterning.¹³ Additionally, assessing several of the proximal epithelial lung cell types, we noted fewer basal and neuroendocrine cells in the T21 lungs, and more club cells, which are known to start differentiating during the canalicular stage of development¹⁰ (Figure S5). Altogether these data support that premature differentiation and precocious development is occurring in the epithelium of T21 fetal lungs.

Proper lung branching morphogenesis is a dynamic process that requires cross-talk between the epithelium and mesenchyme⁶ and can be driven by spatiotemporal deposition of ECM. The significant upregulation of both COL6A1 and COL6A2 (Figure 4A and Figure S6C) in the T21 lungs is of particular interest given that deletion of COL6 has been shown to increase airway branching.⁴¹ It has also been described that inhibition of FN1 activity results in diminished cleft formation and branching in the lung,²⁴ making the dysregulated FN1 in the T21 lungs also noteworthy. Additionally, downregulation of the smooth muscle contraction pathway was observed in the T21 lungs, which was confirmed by the dysregulation of several muscle contraction–associated genes in several mesenchymal clusters (ACTA2, ACTG1, etc) (Figure 4A and Figure S6C). It has been previously demonstrated that smooth muscle cells play an important role in branching during fetal lung development, where disrupted ACTA2 results in epithelial dilatation.^12,42 Furthermore, T21 fetal lungs are known to present with a thinner and interrupted layer of smooth muscle cells around the dilated airways.¹³ Taken together, the upregulation of both the collagen and ECM pathways in T21 lungs, as well as the decreased smooth muscle cell contraction, suggest that dysregulated ECM components of the fetal T21 likely contribute to the compromised branching observed.

We and others have demonstrated that the T21 lung often presents with abnormal vascular growth in utero.^13,25 Whereas an upregulation of certain anti-angiogenic factors has been reported in these lungs, there is limited information describing the ontogeny of these defects. The type I IFN pathway is one of the major signaling cascades consistently activated in T21.^14,43 This was further confirmed in T21 fetal lungs, where bulk sequencing demonstrated an appreciable upregulation of IFN pathway genes.¹³ Here, we demonstrate that this aberrant IFN signaling is highly localized in the endothelium in the fetal T21 lung (Figure 6A and C), with several subclusters demonstrating altered expression of several downstream type I IFN targets: IFI27, IFITM2, BST2, and IFITM3 (Figure 6B). While multiple groups have reported congested capillaries, lymphatic dilatation, and muscularized arteries in the developing T21 lung,^13,44 the nature of these defects and the different endothelial cell populations contributing to such defects are yet to be understood. Upregulation of type I IFN signaling has been previously shown to inhibit endothelial cell proliferation and block angiogenesis,³⁰ therefore likely contributing to the observed compromised vasculature. Interestingly, altered vascular development has been shown to also affect fetal lung ­branching.⁴⁵

In this study we present how the transcriptional landscape differs in T21 lungs. Specifically, we describe changes in gene regulatory networks at the single-cell level in T21 fetal lungs presenting with developmental anomalies. We demonstrate many cell populations displaying unique patterns of differential gene expression. These data provide critical insight as to how specific pathways and cellular processes are being altered in different cellular lineages, and how these may lead to advanced cellular differentiation and defective lung development in T21. While we recognize there are some limitations to our study, including the small number of T21 fetal lung samples and the absence of proximal epithelial cell populations, we believe the insights gained from these rare specimens are invaluable. These findings advance our understanding of developmental T21 lung biology. However, we acknowledge the significance of understanding these cell types more deeply, and plan to investigate them further in future studies. This benchmark cellular atlas of T21 lungs should greatly facilitate discoveries toward ameliorating respiratory morbidity in patients with Down syndrome.

Author contributions

S.B., T.J.M., D.A.A., and S.D. designed the research studies. S.B., A.F., C.C., and S.D. conducted the experiments. S.B., A.F., C.C., G.D., and S.D. acquired data. S.B., A.F., T.J.M., D.A.A., and S.D. analyzed the data. S.B., A.F., T.J.M., D.A.A., and S.D. accessed and verified the underlying data. G.D. and I.A.G. provided human tissues and input on data. S.B., A.F., C.C., G.D., I.A.G., T.J.M., D.A.A., and S.D. wrote, edited, and approved the final manuscript. No honorarium or other form of payment was given to anyone to produce the manuscript.

Supplementary material

Supplementary material is available at American Journal of Respiratory Cell and Molecular Biology online.

Conflicts of interest

Please see the ICMJE disclosure forms, which have been provided as supplementary material.

Funding

This work was supported by the National Heart, Lung, and Blood Institute (NHLBI), Office of the Director, National Institutes of Health (NIH) (R01HL155104 to S.D.); the NIH/NHLBI (R21HL165411 to S.D. and D.A.A. and R01HL141856 to D.A.A.); and the NIH/Eunice Kennedy Shriver National Institute of Child Health and Human Development (R24HD000836 to I.A.G.).

Data availability

The datasets generated during and/or analyzed during the current study are available in the National Center for Biotechnology Information Sequence Read Archive/Gene Expression Omnibus repository (accession number GSE272470). Additionally, code used in analyses may be found at https://github.com/sdanopoulos/git_Danopoulos_T21_lung_sc. This article has a data supplement, which is accessible at the Supplements tab.

Artificial intelligence disclaimer

No artificial intelligence tools were used in writing this manuscript.