Comparative Analyses of Lung Transcriptomes in Patients with Alveolar Capillary Dysplasia with Misalignment of Pulmonary Veins and in Foxf1 Heterozygous Knockout Mice

Alveolar Capillary Dysplasia with Misalignment of Pulmonary Veins (ACDMPV) is a developmental disorder of the lungs, primarily affecting their vasculature. FOXF1 haploinsufficiency due to heterozygous genomic deletions and point mutations have been reported in most patients with ACDMPV. The majority of mice with heterozygous loss-of-function of Foxf1 exhibit neonatal lethality with evidence of pulmonary hemorrhage in some of them. By comparing transcriptomes of human ACDMPV lungs with control lungs using expression arrays, we found that several genes and pathways involved in lung development, angiogenesis, and in pulmonary hypertension development, were deregulated. Similar transcriptional changes were found in lungs of the postnatal day 0.5 Foxf1 +/− mice when compared to their wildtype littermate controls; 14 genes, COL15A1, COL18A1, COL6A2, ESM1, FSCN1, GRINA, IGFBP3, IL1B, MALL, NOS3, RASL11B, MATN2, PRKCDBP, and SIRPA, were found common to both ACDMPV and Foxf1 heterozygous lungs. Our results advance knowledge toward understanding of the molecular mechanism of ACDMPV, lung development, and its vasculature pathology. These data may also be useful for understanding etiologies of other lung disorders, e.g. pulmonary hypertension, bronchopulmonary dysplasia, or cancer.


Introduction
Alveolar Capillary Dysplasia with Misalignment of Pulmonary Veins (ACDMPV; MIM# 265380) is a rare developmental disorder of the lung [1]. ACDMPV was first described by McMahon [2] and its genetic basis emerged as more patients [3] and familial cases were reported [4]. The primary diagnostic features of ACDMPV include misalignment (malposition) of pulmonary veins, medial thickening of smooth muscles in pulmonary arteries, hyperplasia of alveolar epithelium, and drastically decreased number of capillaries and lobular underdevelopment [5]. Approximately one third of the patients also have lymphangiectasis. The disease usually presents within a few hours after birth although late presentations have been reported [6,7]. Treatment, including high pressure oxygen, nitric oxide, extra corporeal membrane oxygenation (ECMO) [8][9][10][11], and more recently Sildenafil [12], only provide temporary relief as the disease is uniformly lethal. Most patients die within the first few weeks of life although some prolonged survivals have been described [13]. The majority of the patients with ACDMPV also have extra-pulmonary anomalies, including gastrointestinal, cardiovascular, and genitourinary systems [14].
We have shown that ACDMPV results from haploinsufficiency of FOXF1 [15]. To date 42 point mutations and 18 genomic deletions involving FOXF1 or its upstream regulatory region have been identified in patients with ACDMPV [15][16][17]. Interestingly, our recent work has indicated that FOXF1 is incompletely paternally imprinted in human lungs [17,18]. Foxf1 +/2 mice exhibit neonatal mortality similar to human infants with ACDMPV and die due to defects in vasculogenesis and alveolarization [19]. They also demonstrate other foregut malformations such as narrowing of the esophagus and trachea, esophageal atresia and trachea-esophageal fistula [20]. Unlike in humans, Foxf1 was not found to be imprinted in mice (data not shown).
To better understand the molecular bases of ACDMPV and to identify major genes and pathways involved in lung development, we performed comparative analyses of lung transcriptomes in patients with ACDMPV and in Foxf1 +/2 mice using Human  and Mouse WG-6 v2 Expression BeadChip Illumina microarrays, respectively.

Ethics Statement
Patients with histopathologically-verified ACDMPV were recruited after signing consent for the study, Institutional Review Board (IRB) protocol # H-8712 ''Molecular genetics of Alveolar capillary dysplasia with misalignment of pulmonary veins (ACDMPV)'' approved by the IRB for Baylor College of Medicine (BCM) and Affiliated Hospitals. All mouse experiments were carried out under the approval of the Institutional Animal Care and Use Committee (IACUC) at BCM. Mice were housed in the Transgenic Mouse Facility (barrier level 3) under the care of the Center for Comparative Medicine (CCM), which is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

Patients
The clinical descriptions are shown in Table 1. Patients 3 and 6 were of Hispanic ethnicity and the remainder were Caucasians. Patients 2-8 had sporadic ACDMPV and patient 1 represents a familial case with neither point mutation in FOXF1 nor deletion in the 16q24.1 region identified. Patients 3 & 4 and 2, 5, & 6 were reported by Stankiewicz et al (2009) (pts D4 and M4) and by Sen et al (2013) (pts 15, 1, and 2), respectively. Patients 2 and 5 had missense mutations, patient 4 had a no stop mutation, patient 6 had a nonsense mutation and patient 3 had an approximately 1 Mb deletion involving FOXF1 as well its upstream regulatory region. Patients 2-6 had associated anomalies of the gastrointestinal system. Butterfly vertebrae were identified in the patients 3 and 6. In patient 1, autopsy was limited to heart and lung. No clinical or molecular information about patients 7 and 8 was available.

Animals
Foxf1 +/2 mice with one Foxf1 allele disrupted by deletion of the forkhead binding domain were generated ( Figure S1) and maintained on a mixed 129-C57BL6/J background. The midday of vaginal plug identification was considered as embryonic day 0.5 and the mid-day when pups were born was considered as postnatal day (P) 0.5.

RNA isolation
Total RNA was isolated from approximately 10 mg frozen lung samples from patients and normal control humans using TRIzol (Invitrogen). Controls C1-C4 are age-matched lung samples from individuals that had no lung problems and died of other reasons. Control samples C1-C3 are from females and control sample C4 is a male sample. Control sample C5 is a commercially available human total lung RNA (Life Technologies Cat. No# AM7968). The tissue was homogenized in microfuge tubes with a hand held homogenizer in the presence of 200 ml of TRIzol. The aqueous phase was collected after spinning at 10,000 rpm for 5 min. This step was repeated and the pooled aqueous phase was loaded onto an RNeasy column (Qiagen). RNA was eluted from the columns following the manufacturer's instruction. The RNA was quantified using Nanodrop and its quality was assessed on a BioRad Bioanalyzer. In spite of the differences in their process of collection and preservation, the RNA isolated from the lung tissues of patients with ACDMPV were of good quality.
Mouse RNA was extracted from frozen whole lungs of P0.5 Foxf1 +/2 mice and their wildtype littermate controls. Whole lungs were homogenized using TissueLyser LT (Qiagen) in QIAzol lysis reagent (Qiagen) according to manufacturer's instructions. Total RNA was isolated from lung homogenates using the miRNeasy mini kit (Qiagen). RNAs were tested for quality on the Agilent 2100 bioanalyzer and quantified on the Nanodrop Spectrophotometer by the Genomic and RNA Profiling Core (GARP) facility at BCM.

Microarray analysis
500 ng of total human RNA was converted into cDNA and then cRNA following the manufacturer's instructions. The cRNA was biotinylated and hybridized to the Human Ref8.V3 microarray (Illumina) contains probes representing ,22,000 curated genes and ESTs conjugated with streptavidin-Cy3. Quality standards for hybridization, labeling, staining, background signal, and basal level of housekeeping gene expression for each chip were verified. The scanned images of the chip were analyzed using GenomeStudio software (Illumina). Gene expression analyses of RNA from patients' lungs (n = 8) was compared to those from the control lungs (n = 5).
Transcriptomes of six Foxf1 +/2 and six wildtype littermate samples were assessed in two separate microarray experiments using MouseWG-6 v2.0 Expression BeadChip kit (Illumina) containing probes for over 45,200 transcripts. 500 ng of total RNA was labeled using Illumina TotalPrep RNA Amplification Kit (Ambion) and hybridized as per manufacturer's instructions. The array data were analyzed using the lumi bioconductor package [21], normalized by robust spline normalization (RST), and transformed using variance stabilization transformation (VST) [22]. ComBAT batch correction [23] was applied to correct for differences in expression values between the two separate microarray experiments. Two-sample t-test was applied to determine differentially expressed genes between Foxf1 +/2 and wild type lung groups. Differential expression p-values were adjusted for false discovery rates (FDR). Fold changes were calculated using reverse VST.
Database for Annotation, Visualization, and Integrated Discovery (DAVID), Ingenuity pathway analysis (IPA), and Gene set enrichment analysis (GSEA) softwares were used for gene ontology, pathway, and network analyses.

qRT-PCR
Confirmation of the differential gene expression was done by qRT-PCR using an ABI 7900HT Fast real time PCR System. Universal probe library and Fast start probe master mix from Roche Applied Science was used following manufacturer's instructions.
For the mouse microarray analysis validation, RNA was reverse transcribed to cDNA using the SuperScript III First-Strand Synthesis System (Life Technologies) and amplified using the Power SYBR Green PCR Master Mix (Life Technologies). Primers used for qRT-PCR validation of the mouse microarray are listed in table S5.

Immunohistochemistry
Antibody against COL1A1 was obtained from Sigma-Aldrich. The respective dilutions of antibodies and other experimental conditions used were following manufacturer's instructions.

ACDMPV lung gene expression analysis
Statistical analysis revealed 337 genes deregulated in ACDMPV lungs compared to control lungs (adjusted p-value ,0.05, foldchange approximately ./ = 2 or ,/ = 22), of which 205 (60.83%) were down-regulated and 132 (39.17%) were upregulated ( Figure 1A, Table S1). DAVID analysis revealed gene ontologies (GO) related to inflammation and immune responses, vasculature development, cell adhesion and chemotaxis associated with the deregulated genes ( Figure 1B). The microarray data were validated using qRT-PCR for eight differentially expressed genes: IL8, SPP1, IGFBP3, COL1A1, SOD2, GLI2, ANGPTL4, and HES4 (data not shown). Further, differences in protein quantity were assessed by immunohistochemistry (IHC) using antibodies against a significantly differentially expressed (COL1A1) product ( Figure 1C). The amount of COL1A1 was significantly higher in the patient samples as compared to the control lung.
GSEA and DAVID analyses showed enrichment of pathways related to signaling in the immune system, hemostasis, cell surface interactions at the vascular wall, and the lysosome (Table S2). Further analyses by GSEA allowed determining gene sets associated with chemical and genetic perturbations enriched in the 337 genes deregulated in the ACDMPV lungs (Table S3) (Table S4). GSEA and DAVID analyses revealed GO terms related to response to cytokine stimulus, cell division, blood circulation, regulation of cell proliferation and cell adhesion associated with deregulated genes ( Figure 3B).
The microarray data were confirmed using qRT-PCR for five differentially expressed genes: Dll4, Hgf, Sema3C, Cma1, and Ren1 ( Figure 3C). Canonical pathways affected include genes involved in focal adhesion, homeostasis, p53 signaling, as well as in Vegf, Endothelin, and Notch signaling (Table S6). Additional GSEA analysis identified genes down-regulated by E-cadherin and genes up-regulated in vascular smooth muscle cells by MAPK8 to be enriched in the list of down-regulated genes. Genes regulated by hypoxia were found to be enriched in the list of up-regulated genes (Table S7). Network analysis by IPA identified interactions amongst genes involved in cellular movement, inflammatory response, and hematological and cardiovascular system development ( Figure 4).

Discussion
To date, we have recruited greater than 100 families with children affected with ACDMPV. This is the largest collection of ACDMPV samples worldwide. To identify transcriptional changes in the neonate ACDMPV lungs, we compared the whole genome expression in the lungs of eight patients with ACDMPV with transcriptional changes in P0.5 Foxf1 +/2 mouse lungs, and appropriate control lungs. This approach enabled us to identify gene ontologies and pathways associated with ACDMPV and gain a better understanding of the disease.
The Foxf1 +/2 mouse lungs showed down-regulation of Notch signaling genes Dll4 and Notch1 and slightly lower but significant (fold change: 21.46) down-regulation of the Notch target gene Hey1 (data not shown). Notch signaling has been reported to be affected in Foxf1 +/2 lungs [24]. Genes involved in pulmonary vasculature development, Sema3c and lung branching morphogen-esis, Hgf and Lama1, were also down-regulated. Sema3c, a semaphorin gene expressed in the lung endothelium, interacts with neuropillin-1 during vasculature development; loss of semaphorin-neuropilin-1 signaling showed lung pathology reminiscent of ACDMPV [25]. Sema3c was down-regulated in our study and Foxf1 was down-regulated in lungs of Nrp1 Semamice, suggesting a possible interaction between Foxf1 and Sema3C in murine lung pathology. These transcriptional changes potentially lead to underdeveloped lungs causing respiratory distress at birth.
Transcriptional changes found in Foxf1 +/2 lungs, potentially resulting from the respiratory distress, include deregulation of pathways associated with hypoxia-induced vascular remodeling. There is increased expression of a-smooth muscle actin Acta2, indicating greater muscularization of pulmonary vessels and possible pulmonary hypertension. Greater expression of Acta2 could also be due to endothelium-to-mesenchyme transition that is found to accompany vascular remodeling in pulmonary hypertension [26]. Additionally, members of the Endothelin-1 signaling pathway were deregulated. Endothelin-1 (ET-1) is a potent vasoconstrictor regulating vascular tone and is implicated in pulmonary hypertension. Edn1 over-expression in mice leads to development of systemic hypertension with altered vascular reactivity [27].
Fibronectin and several members of the collagen gene family are also found to be up-regulated, likely leading to increased production of extracellular matrix (ECM) proteins. Accumulation of ECM proteins such as collagen, elastin, and fibronectin leads to thickening of pulmonary blood vessels, likely contributing to pulmonary hypertension. Interestingly, endothelin induced collagen remodeling has been reported in a mouse model of pulmonary hypertension [28], indicating that the increased endothelin-1 signaling in Foxf1 +/2 lungs may be causing an increase in collagen production. Integrins involved in cell surface interactions at the vascular wall were also found to be down-regulated. Similar transcriptional changes have been observed to be associated with matrix and vasculature remodeling in TGF-a induced pulmonary fibrosis [29]. Foxf1 was shown to play an essential role in mesenchymal migration by transcriptionally regulating integrin-b3 (Itgb3) [30]. Integrin-ECM interactions also contribute to vascular remodeling. Altered integrin-ECM interactions have been documented in chronic hypoxia and monocrotaline-induced pulmonary hypertension [31].
Mast cell chymase and tryptase genes (Mcpt4, Mcpt6, Cma1, and Cpa3) are highly up-regulated i n Foxf1 +/2 lungs, showing evidence of pulmonary mastocytosis. Pulmonary mastocytosis and enhanced lung inflammation in Foxf1 +/2 mice after chemically induced or allergen-mediated lung injury was reported [32]. These inflammatory molecules could contribute to the vascular remodeling through their effects on vascular cell wall permeability, matrix protein production, and by inducing recruitment of mesenchymal precursor cells [33]. Similar accumulation of mast cell mediators was found in lungs of Fgfr3/4 mutant mice that exhibit Bronchopulmonary Dysplasia (BPD) like pathology and in lungs  of children with BPD [34], suggesting that pulmonary mastocytosis could be a common feature found in diffuse lung developmental conditions. Ren1 which along with Cma1 and Cpa3 is a component of the blood pressure regulating renin angiotensin system is also up-regulated.
Other transcriptional changes include up-regulation of Mki67, Ccnd1, and Foxm1, showing an increase in proliferation in the lung. Foxf1 has been reported to be down-regulated in mice with endothelial cell-specific deletion of Foxm1 [35], indicating a reciprocal feedback loop between Foxf1 and Foxm1. Finally, upregulation of Hif3a and Nos3 is also observed, suggesting a compensatory mechanism to counteract the effects of hypoxiainduced pulmonary hypertension.
Microarray analysis of ACDMPV human lung transcriptomes primarily identified deregulation of genes involved in inflammation and cell surface interactions at the vascular wall. Members of the collagen gene family are highly up-regulated similar to the upregulation seen in Foxf1 +/2 lungs, potentially leading to increased ECM collagen production, causing thickening of pulmonary blood vessels, and resulting in pulmonary hypertension, a typical feature found in neonates with ACDMPV.
IHC for COL1A1 confirmed the increased protein levels of COL1A1 in ACDMPV patient lungs ( Figure 1C). Structural changes in ACDMPV lungs that could be attributed to hypoxiainduced vascular remodeling include medial thickening of smooth muscles in pulmonary arteries. Transcriptional changes which could be described as compensatory mechanisms to counteract the pulmonary hypertension include down-regulation of genes involved in TREM1 signaling (TREM1, TLR4, IL1B, and ITGAX) and up-regulation of NOS3. Pulmonary inflammatory responses play an important role in hypoxia-induced vascular remodeling [33] and down-regulation of TREM1 signaling could be a compensatory response to attenuate pulmonary inflammatory responses. Tlr-4 deficient mice were shown to be resistant to chronic hypoxia-induced pulmonary hypertension by attenuating the pulmonary vascular inflammatory response to hypoxia [36]. Up-regulation of NOS3 also could be a compensatory response to counteract the hypertension.
Other affected pathways include signaling related to the lysosome function, carbohydrate metabolism, cell cycle, and axon guidance which are also deregulated in Foxf1 +/2 lungs. Interestingly, we found targets of E-cadherin enriched in the deregulated genes.
Comparison of the deregulated genes in both experiments  enabled identification of 14 genes, COL15A1, COL18A1, COL6A2,  ESM1, FSCN1, GRINA, IGFBP3, IL1B, MALL, NOS3, RASL11B,  MATN2, PRKCDBP, and SIRPA, similarly altered in both ACDMPV and Foxf1 heterozygous lung transcriptomes. ESM1 is mainly expressed in endothelial cells and its expression is upregulated by cytokines like TNFa and IL-1b suggesting that it plays a role in endothelium-dependent pathological disorders [46]. ESM1 as well as IL1B are downregulated in both the microarray datasets. ESM1 has also been computationally predicted to have a FOX: ETS binding motif in its vicinity suggesting that it could be a direct target of FOXF1 in endothelial cells [47]. Another common downregulated gene in both datasets, SIRPA has been shown to interact with surfactant proteins A & D in the absence of inflammation to suppress alveolar macrophage phagocytosis [48]. Whereas there is greater variation for Sirpa in the mouse data, the average fold change difference for this gene correlates with the human data.
In aggregate, we present the results of genome-wide analyses of transcriptional changes in lungs from patients with ACDMPV due to point mutations or genomic deletions in FOXF1 and in mice heterozygous for Foxf1 loss. The transcriptional changes should help better define the molecular mechanisms involved in the pathogenesis of ACDMPV and may also be a valuable source for future research on lung vasculature development and pulmonary disease conditions such as pulmonary hypertension, bronchopulmonary dysplasia, and lung cancer.