Hypercapnia Alters Expression of Immune Response, Nucleosome Assembly and Lipid Metabolism Genes in Differentiated Human Bronchial Epithelial Cells

Hypercapnia, the elevation of CO2 in blood and tissues, commonly occurs in severe acute and chronic respiratory diseases, and is associated with increased risk of mortality. Recent studies have shown that hypercapnia adversely affects innate immunity, host defense, lung edema clearance and cell proliferation. Airway epithelial dysfunction is a feature of advanced lung disease, but the effect of hypercapnia on airway epithelium is unknown. Thus, in the current study we examined the effect of normoxic hypercapnia (20% CO2 for 24 h) vs normocapnia (5% CO2), on global gene expression in differentiated normal human airway epithelial cells. Gene expression was assessed on Affymetrix microarrays, and subjected to gene ontology analysis for biological process and cluster-network representation. We found that hypercapnia downregulated the expression of 183 genes and upregulated 126. Among these, major gene clusters linked to immune responses and nucleosome assembly were largely downregulated, while lipid metabolism genes were largely upregulated. The overwhelming majority of these genes were not previously known to be regulated by CO2. These changes in gene expression indicate the potential for hypercapnia to impact bronchial epithelial cell function in ways that may contribute to poor clinical outcomes in patients with severe acute or advanced chronic lung diseases.

GO biological process-associated gene clusters targeted by hypercapnia. Major clusters from hypercapnia-downregulated genes are linked to immune response, nucleosome assembly, cell differentiation, oxidation reduction, and ion and lipid transport (Fig. 2). Clusters from upregulated genes induced by high CO 2 (Fig. 3) involve biological processes related to lipid metabolism, cholesterol biosynthesis, signal transduction, and transport. A number of these important clusters, labelled A-E in Figs 2 and 3, are analyzed in more detail in the following sections. Their corresponding gene lists are depicted in Figs 4,5,and Supplementary Fig. 3. Hypercapnia differentially regulates genes associated with innate immunity and nucleosome assembly. Cluster A, represented in Fig. 4a, includes hypercapnia-regulated genes involved in signal transduction, immune and inflammatory responses, and leukocyte chemotaxis. Notably, TLR4, multiple chemokines (CCL28, CXCL1, CXCL2, CXCL6, and CXCL14) and the IL-6 receptor gene (IL6R) were all downregulated by elevated CO 2 . On the other hand, the IL-1 receptor like 1 gene (IL1RL1) was upregulated by hypercapnia. To validate the microarray results related to CO 2 -induced changes in key immunoregulatory genes, expression of CXCL1, CXCL14, CCL28, ILR6 and TLR4 was also assessed by qPCR. We found that these genes were all downregulated at levels similar to those in the microarray analysis (Fig. 4c). Indeed, the degree of CO 2 -induced downregulation of these transcripts assessed by qPCR and microarray was highly correlated (r 2 = 0.7981).  In addition, to determine whether downregulation of a key immunoregulatory transcript by hypercapnia was accompanied by a similar change in protein expression, we assessed expression of TLR4 protein in differentiated NHBE cells. Immunofluorescence microscopy ( Fig. 4d) and immunoblotting (Fig. 4e) both showed that exposure to 20% CO 2 for 24 h decreased NHBE cell expression of TLR4 protein. Full-length blots are included in Supplementary Fig. 4. Taken together, these results suggest that hypercapnia would suppress airway epithelial innate immune response to microbial pathogens and other inflammatory stimuli.
Next, we analyzed cluster B, which includes hypercapnia-regulated genes that codify proteins involved in nucleosome assembly (Fig. 5a). The heat map in Fig. 5b shows that hypercapnia downregulates genes encoding multiple family members of the core histones H2A and H2B 27 , as well as the nucleosome assembly protein 1-like 1 (NAP1L1), which regulates protein complex assembly, chromosome organization and DNA metabolism. The only upregulated gene in cluster B is H1F0, encoding histone H1, which is normally expressed in terminally differentiated and slowly dividing cells. To validate the microarray data from cluster B, we performed qPCR for selected transcripts whose expression was significantly altered in the microarray analysis. Figure 5c shows that expression of the histone genes HIST1H2AC, HIST1H2BD, and HIST1H2BK was downregulated by hypercapnia as assessed by qPCR, again similar to the microarray results. and others), cell surface receptor signaling (EGFR, IFNAR1, IL6R and others) and apoptosis (BCL2L15, DAPL1, SEMA6A and others). The impact of elevated CO 2 on expression of these genes would be expected to alter epithelial metabolism and barrier function, as well as innate immune and inflammatory responses.

Discussion
To our knowledge, the present study is the first to investigate the impact of hypercapnia on global gene expression in airway epithelial cells. Of importance, we utilized primary NHBE cells cultured at ALI to achieve a differentiated state closely resembling normal human bronchial epithelium. Our principal finding was that hypercapnia altered expression of a small number of specific genes (309 out of 20,390 transcripts assayed, or 1.5%) in differentiated NHBE cells. Of these, 183 genes (59%) were downregulated, while 126 (41%) were upregulated. Thus, the effects of elevated CO 2 are highly selective, involving both differential repression and differential activation of specific gene subsets. The overwhelming majority of these genes were not previously known to be regulated by CO 2 . Furthermore, gene ontogeny analysis showed enrichment of hypercapnia-regulated genes involved in a variety of fundamentally important cellular processes. Altering expression of genes related to these processes would be expected to impart functional changes in bronchial epithelial cells, which could in turn influence the pathophysiology and outcomes of many respiratory diseases.
Our data show that hypercapnia alters expression of multiple components of the innate immune system, including downregulation of the IL-6 receptor (IL6R); the neutrophil chemokines CXCL1, CXCL2 and CXCL6 28 ; the mucosal-associated chemokines CCL28 and CXCL14 [29][30][31] and importantly TLR4. Hypercapnia also upregulated CD55 and CD86, which bind virus at the cell surface 32,33 . While hypercapnia downregulated TLR4, it increased the expression of IL1RL1, which has been shown to inhibit TLR4 activation 34   defense against multiple respiratory pathogens [36][37][38][39][40] . Interestingly, airway epithelial TLR4 expression was reduced in patients with severe COPD as compared to those with less severe COPD 41 , possibly due to hypercapnia in patients with advanced disease. Reduced expression of immune response genes was also seen in the lungs of newborn mice exposed to moderate hypercapnia (8% CO 2 ) for the first two weeks of life 42 . While the immune genes downregulated by hypercapnia in the newborn mice differed from those we found in NHBE cells, the mucosal immunity chemokine CXCL14 43 was commonly downregulated in both systems. Taken together, these observations indicate that the airway epithelium is an important target for hypercapnic suppression of innate immune gene expression. This, along with the suppressive effects of elevated CO 2 on macrophage, neutrophil, alveolar epithelial cell functions [13][14][15][17][18][19] likely contributes to the deleterious impact of elevated CO 2 on lung injury and host defense.
Another cluster impacted by hypercapnia includes genes related to nucleosome assembly, which also have antibacterial properties. The nucleosome consists of 145-147 base-pair-segments of DNA wrapped around a histone octamer containing one (H3-H4) 2 tetramer, two H2A-H2B dimers, and histone chaperones or linkers that facilitate nucleosome assembly 44 . Regulation of nucleosome assembly following DNA replication, DNA repair and gene transcription is critical for the maintenance of genome stability and epigenetic information 44 . Within this gene cluster, hypercapnia downregulated transcripts for the core histones H2A and H2B 27 , the histone chaperone NAP1L1 45 , and the linker histone H1 27 . Downregulation of histone gene expression can be triggered by DNA-damage or indirect inhibition of DNA synthesis 46 and might lead to alterations of chromatin structure that would influence transcriptional regulation of many genes and even genome stability 47 . Exchange of core histones with histone variants might also alter the chemical nature and physical properties of the nucleosome, thereby affecting distinct cellular processes 48 . In addition, histones H2A and H2B also can inactivate endotoxin and function as antimicrobial proteins 49,50 .
We also found that elevated CO 2 upregulated NHBE cell expression of cholesterol and fatty acid biosynthesis genes, while downregulating ATP-binding cassette (ABC) transporters, which promote the efflux of cholesterol and phospholipids from the cell 51 . Interestingly, enveloped viruses subvert preexisting lipids for viral entry and trafficking and also reprogram lipid synthesis and lipid distribution in lipid rafts to establish an optimal environment for their replication, assembly and egress 52 . Furthermore, host defense against viral infection involves interferon-mediated downregulation of sterol biosynthesis 53 . Thus, hypercapnia-induced cholesterol accumulation might contribute to the entry, replication, and shedding of respiratory viruses in the airways.
As noted above, in a previous study, we showed that hypercapnia downregulates the TCA cycle enzyme IDH2, resulting in mitochondrial dysfunction and impaired proliferation of fibroblasts and A549 lung epithelial cells 20 . However, in the current study, hypercapnia did not alter IDH2 expression in NHBE cells, indicating that CO 2 -mediated regulation of gene expression is cell-type-specific. On the other hand, a number of genes involved in mitochondrial function were regulated by hypercapnia in NHBE cells. Among these, upregulated genes included acyl-CoA dehydrogenase short/branched chain (ACADSB) and acyl-CoA synthetase short chain family member 2 (ACSS2), which encode enzymes involved in fatty acid synthesis and oxidation 54 . Genes downregulated by elevated CO 2 included gamma-butyretaine hydroxlase 1 (BBOX 1), which catalyzes synthesis of L-carnitine, an essential co-factor in beta-oxidation 55 ; kynurenine 3-monooxygenase (KMO), an outer mitochondrial membrane protein that hydroxylates tryptophan to form kynurenine 56 ; BCL2 interacting protein 3 (BNIP3), a BH3 domain protein with pro-apoptotic activity 57 ; and mitochondrial assembly of ribosomal large subunit 1 (MALSU1), an inhibitor of translation at the mitochondrial ribosome 58 . The diverse activities of these genes indicate the potential for hypercapnia to disrupt multiple mitochondrial functions in NHBE cells.
While the current study does not reveal the molecular mechanism(s) underlying hypercapnia-induced changes in gene transcription, other recent work suggests a path to elucidating components of a putative CO 2 -induced signaling pathway leading to inhibition of innate immune gene expression and impaired host defense. We previously reported that elevated CO 2 inhibits expression of antimicrobial peptide genes and suppresses antibacterial host defense in Drosophila 59 , suggesting that the immunosuppressive effect of hypercapnia is evolutionarily conserved. Using a genome-wide RNAi screen, we identified a small number of genes whose expression is required for CO 2 -induced immunosuppression in Drosophila cells, and which are conserved in mammalian systems 60 . Flies deficient in of one these genes, a zinc finger homeodomain transcription factor known as zfh2, were protected from CO 2 -induced mortality associated with bacterial infection 60 . This opens up the opportunity to test whether orthologues of zfh2 and other genes identified in the Drosophila screen mediate hypercapnic immunosuppression in mice and ultimately in humans.
Alterations in expression of innate immune and other genes in airway epithelial cells may be of central importance in the CO 2 -induced increase in mortality of Pseudomonas pneumonia we previously observed in mice 17 . In addition, the suppressive effect of elevated CO 2 on immune gene expression in the airway epithelium, along with similar effects on immune cells, suggest a reason why severe COPD and other lung disease associated with hypercapnia all carry a high risk of pulmonary infection. Bacterial and viral infections are a principal cause of acute COPD exacerbations [61][62][63][64] , which are linked to the need for hospitalization and to mortality 65,66 . Thus, CO 2 -induced alterations in airway epithelial gene expression may underlie the increase in mortality associated with hypercapnia in advanced COPD, as well as community-acquired pneumonia 9 , adenoviral lung infections 10 and cystic fibrosis 11 . It is notable in this regard that reducing hypercapnia with noninvasive ventilatory support has been shown to decrease hospital readmissions and mortality in patients with severe COPD 67,68 . Further investigation of the molecular mechanisms and mediators of CO 2 effects on gene expression may reveal targets for pharmacologic intervention to prevent hypercapnic immune suppression in patients with advanced respiratory disease.

Primary Normal Human Bronchial Epithelial Cells. Primary NHBE cells isolated from airways of
humans without known lung disease were obtained from a commercial source (Lonza). The cells were plated on collagen-coated plastic dishes, grown to confluence in BEGM TM Bronchial Epithelial Cell Growth Medium (Lonza), and passaged after enzyme dissociation with trypsin 69 . Cells from passage-3 were seeded onto 24-mm, 0.4 μm pore size, polyester, transwell inserts (Corning) at 0.5 × 10 6 cells per insert (4.67 cm 2 ) and cultured in a serum-free medium 70 , comprised of 1:1 mixture of BEBM (Lonza): DMEM (Mediatech), supplemented with hydrocortisone (0.5 μg/ml), insulin (5 μg/ml), transferrin (10 μg/ml), epinephrine (0.5 μg/ml), triiodothyronine (6.5 ng/ml), epidermal growth factor (0.5 ng/ml), retinoic acid (50 nM), bovine pituitary extract (0.4%), gentamycin (50 μg/ml), and amphotericin B (50 ng/ml). After the cells reached confluence in submersion culture, the medium above the inserts was removed and the cells were maintained in ALI culture for two more weeks, at which point differentiation to a pseudostratified mucociliary epithelium with characteristics of airway epithelium in vivo was established 69,71 . Differentiation after ∼2 wk on ALI culture was confirmed by the presence of beating cilia and mucus production, as previously described 72 . Culture of NHBE cells up to the point of full differentiation was carried out in an atmosphere of humidified 5% CO 2 /95% air at 37 °C. Hypercapnia Exposure. After differentiation, NHBE cells were cultured in ALI for an additional 24 h in humidified 20% CO 2 /21% O 2 /59% N 2 (hypercapnia) or maintained in humidified 5% CO 2 /95% air (5% CO 2 /20% O 2 /75% N 2 ; normocapnia), as control. The growth medium was pre-saturated with appropriate CO 2 concentration for 4 h prior to the addition to the cells. The PCO 2 and pH of the pre-saturated media were measured using a pHOx Plus Blood Gas Analyzer (Nova Biomedical Corp). For the normocapnia-and hypercapnia-equilibrated media, the PCO 2 s were 44 and 112 mmHg, and the corresponding pH values were 7.4 and 7.1 respectively.

Cytotoxicity Assay.
To determine whether hypercapnia induces cytotoxicity, lactate dehydrogenase (LDH) release to the apical and basolateral compartments was assessed using a colorimetric Cytotoxicity Detection Kit (Roche) according to the manufacturer's instructions. Absorbance at 490 nm was measured using a VersaMax Tunable Microplate Reader (Molecular Devices). Percent LDH release was calculated as the amount of LDH measured in the basolateral supernatant or apical wash divided by the total amount of LDH in the culture (LDH in cell lysates plus that measured in apical and basolateral compartments) times 100.

RNA Isolation and Affymetrix GeneChip Hybridization. Total RNA was isolated using the RNeasy
Mini kit (Qiagen). Quality and quantity of each RNA sample were assessed using a 2100 BioAnalyzer (Agilent).
RNA was hybridized to GeneChip ® Human Genome U133 2.0 Plus Array (Affymetrix). A total of 6 chips, each hybridized to a cRNA from different normocapnic (n = 3) or hypercapnic (n = 3) NHBE cell cultures were used in this study. The U133 2.0 Plus Arrays contain probes for approximately 56,921 transcripts and variants, including over 45,000 well characterized human genes. Fluorescent images were detected in a GeneChip ® Scanner 3000 and expression data were extracted using the GeneChip Operating System v 1.2 (Affymetrix).

Microarray Data Analysis. Differential gene expression between normocapnia and hypercapnia was
assessed by a statistical linear model analysis using the BioConductor package limma 73,74 (https://www.bioconductor.org/help/faq/), in which an empirical Bayes method is used to moderate the standard errors of the estimated log-fold changes of gene expression. The moderated t-statistic p-values derived from the limma analysis were further adjusted for multiple testing by Benjamini and Hochberg's method 75 to control false discovery rate SCIENTIFIC REPORTS | (2018) 8:13508 | DOI:10.1038/s41598-018-32008-x (FDR). Many genes whose expression signals were below background were defined as "absent". Transcripts absent in all samples were filtered out, leaving 54,675 probes corresponding to 20,390 genes in the downstream analysis. The lists of differentially expressed genes were obtained by the FDR criteria of <0.05 and fold-change cutoff of >1.4. Differential gene expression in hypercapnia versus normocapnia was depicted in a pie chart, volcano plot of statistical significance (−log 10 P value) plotted against log 2 fold change, and hierarchical clustering by Pearson correlation represented as heat maps generated using Heatmapper 76 and Gene-E (https://software.broadinstitute. org/GENE-E/).
Over representation analysis (ORA) of gene ontology (GO) terms from biological processes of all genes downregulated or upregulated by hypercapnia were separately analyzed using the Gene Ontology Analysis InnateDB tool 77 which utilizes a manually-curated knowledgebase of genes, proteins, interactions and signaling pathways involved in mammalian innate immune responses. Results from the Innate DB analysis were confirmed using GeneGo Metacore (Thomson Reuter), a separately curated database and pathway analysis tool. Microarray data have been deposited to the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/projects/geo) complied with MIAME standards (accession number GSE110362).
Network Ontology Analysis. Subsequent analysis of global expression changes and ontology network assessment on the differentially selected genes was performed using Mathematica ® v11.2 (Wolfram Research, Inc., Mathematica, Version 11.2, Champaign, IL (2017)). Ontology groups were generated using inbuilt GenomeData, matching the annotated genes with pre-defined processes and intracellular functions. Two approaches were used for analysis of genome wide expression changes: unbiased measurements of intra-network gene expression and fold-change ranked segmentation. Unbiased intra-network changes were assessed for cellular processes that contained at least five genes in the post-screen data. Mean-fold change, the variance of the fold-change, and Pearson correlation of expression were measured for each process. Intra-network heterogeneity of relative expression was measured by calculating the standard deviation of the relative expression for genes within any given ontological process. For instance, if a gene was classified as belonging to both "Nucleosome Assembly" and "Signal Transduction", it was assigned to both groups and a connection between these processes was indicated. To further understand the impact of hypercapnia-induced differential gene expression, cluster domains of GO biological processes containing 5 or more genes and with at least 4 connections were also generated using Mathematica ® v11.2. These processes were broadly grouped based on gene function and by their connections.
Quantitative TaqMan Real-time RT-PCR. Total RNA was isolated from NHBE cells and first-strand cDNA was generated using MultiScribe ™ MuLV reverse transcriptase (Applied Biosystems). The first-strand cDNA was used to quantitate the mRNA levels by TaqMan real-time PCR system (Applied Biosystems). The level of expression of eukaryotic translation elongation factor 1 alpha 1 (EEF1A1) was used as reference, and fold change of target genes was calculated by the ∆∆ CT method 78 .
Statistical analysis. Data are presented as means ± SE. Differences between two groups were assessed using Student's t test. Levene's test was used to analyze the homogeneity of variances. Significance was accepted at p < 0.05.