Hypercapnia alters mitochondrial gene expression and acylcarnitine production in monocytes

CO2 is produced during aerobic respiration. Normally, levels of CO2 in the blood are tightly regulated but pCO2 can rise (hypercapnia, pCO2 > 45 mmHg) in patients with lung diseases, for example, chronic obstructive pulmonary disease (COPD). Hypercapnia is a risk factor in COPD but may be of benefit in the context of destructive inflammation. The effects of CO2 per se, on transcription, independent of pH change are poorly understood and warrant further investigation. Here we elucidate the influence of hypercapnia on monocytes and macrophages through integration of state‐of‐the‐art RNA‐sequencing, metabolic and metabolomic approaches. THP‐1 monocytes and interleukin 4–polarized primary murine macrophages were exposed to 5% CO2 versus 10% CO2 for up to 24 h in pH‐buffered conditions. In hypercapnia, we identified around 370 differentially expressed genes (DEGs) under basal and about 1889 DEGs under lipopolysaccharide‐stimulated conditions in monocytes. Transcripts relating to both mitochondrial and nuclear‐encoded gene expression were enhanced in hypercapnia in basal and lipopolysaccharide‐stimulated cells. Mitochondrial DNA content was not enhanced, but acylcarnitine species and genes associated with fatty acid metabolism were increased in hypercapnia. Primary macrophages exposed to hypercapnia also increased activation of genes associated with fatty acid metabolism and reduced activation of genes associated with glycolysis. Thus, hypercapnia elicits metabolic shifts in lipid metabolism in monocytes and macrophages under pH‐buffered conditions. These data indicate that CO2 is an important modulator of monocyte transcription that can influence immunometabolic signaling in immune cells in hypercapnia. These immunometabolic insights may be of benefit in the treatment of patients experiencing hypercapnia.


INTRODUCTION
Life on planet Earth has evolved around a background of changing CO 2 concentrations 1 and organisms have developed a range of sensing and adaptive mechanisms.
Humans have also evolved mechanisms to tightly regulate the levels of CO 2 in the blood through acute chemosensation of elevated CO 2 and adjustment of breathing, mainly via central chemoreceptors in the brain stem. 2,3 Despite the acute neuronal responses to CO 2 being well described, there is still a relative dearth of knowledge relating to the transcriptional responses elicited under conditions of elevated CO 2 which is a focus of this study. Normocapnia in humans equates to 35-45 mmHg CO 2 in the circulation; however, tissue CO 2 levels will differ depending on the physiological context. Patients with lung diseases [e.g. chronic obstructive pulmonary disease (COPD), cystic fibrosis and acute respiratory distress syndrome] can experience hypercapnia as evidenced by chronically elevated circulating pCO 2 levels > 45 mmHg. 4,5 Furthermore, local microenvironments such as the hypoxic core of solid tumors experience significantly higher than normal pCO 2 levels. 6 Hypercapnia is a risk factor with increased mortality in COPD 7 and is associated with higher intensive care unit mortality in acute respiratory distress syndrome. 4 Hypercapnia elicits a worse outcome in the context of bacterial infections, 8 impairs wound healing, 9,10 promotes muscle wasting 11 and increases airway smooth muscle contractility. 12 However, intriguingly, therapeutic hypercapnia is associated with improved prognosis in some settings (e.g. after one-lung ventilation in lung lobectomy patients). 13 Thus, hypercapnia has been likened to a "double-edged sword" where any potential beneficial effects must be balanced against known deleterious effects. 14,15 Here we contend that limitations in our current knowledge in relation to the transcriptional response of immune cells to elevated CO 2 is a major impediment to better predicting the outcome of elevated CO 2 exposure in an immune context (e.g. infection/inflammation).
Thus, the objective of this study is to use an unbiased next-generation RNA-sequencing (RNA-seq) approach to determine the impact of 10% CO 2 (akin to that observed in patients with chronic respiratory disease) on transcription in monocytes. In addition, we have buffered pH changes under these elevated CO 2 conditions to decipher the effects of CO 2 per se on monocytes as opposed to hypercapnic acidosis.
Monocytes are used as a model because their recruitment to the lung is important in COPD, 16 a disease where hypercapnia is a prominent feature. Patients with severe COPD have increased numbers of total circulating monocytes and nonclassical patrolling monocytes, compared with normal individuals and patients with less severe COPD. M2 macrophages, which we also study here, are enhanced in COPD. 17 This study is not intended to directly model COPD (which is one of several lung diseases associated with hypercapnia), but focuses on CO 2 -dependent responses in monocytes and macrophages for the reasons outlined above.
We believe that the experimental approaches taken in this study can develop our understanding of the complex milieu of cellular signaling events that predispose to an advantageous suppression of inflammation in certain contexts versus a deleterious immunosuppression in others. To our knowledge, this is the first study of its kind to use next-generation RNA-seq integrated with metabolomic analysis to examine the cellular response to elevated CO 2 in pH-buffered conditions in immune cells.

RNA-seq: effect of hypercapnia on monocytes in the basal state
We performed a principal component analysis sample similarity analysis to investigate in a nonbiased manner the robustness of our buffered hypercapnia experimental stimulus. Data in Figure 1a demonstrate separation of our normocapnia and hypercapnia monocyte replicates by PC1. 5% CO 2 -treated monocyte replicates in red clustered to the right-hand side of the graph, and the 10% CO 2 -treated monocyte replicates in turquoise blue clustered to the left-hand side of the graph. This indicates that our relatively modest 10% CO 2 -buffered hypercapnia protocol is sufficient to robustly segregate hypercapnia treated versus normocapnia control samples. We next examined the differentially expressed genes (DEGs) between normocapnia and hypercapnia using RNA-seq analysis. In Figure 1b, Log 2 -fold change and adjusted Pvalue are depicted in a volcano plot. Rather than selecting an arbitrary fold change cut-off, we included all transcripts with an adjusted P-value of ≤ 0.05. This was because several transcripts of interest [e.g. Dynein Light Chain LC8-Type 1 (DYNLL1)] were modestly changed in terms of fold change (Log 2 -fold change of 0.607) but to an extremely high degree of statistical confidence (P-adj 6.86E-06). Three hundred genes met this threshold with comparable numbers of upregulated (196 green) and downregulated (184 red) transcripts demonstrating sensitivity to hypercapnia. A list of the eight most DEGs (up and down) is shown in Figure 1c with solute carrier family 16 member 9 (SLC16A9), a pH-independent carnitine efflux transporter, 18,19 and a long noncoding RNA gene (AC048341.1; also known as MIRLET7IHG) being the most differentially upregulated and downregulated genes, respectively. Supplementary figure 1a displays the differential expression data in a different way, highlighting the top 50 most significantly affected transcripts (as determined by P-adj) ordered by expression level [log 2 of transcript per million (TPM) values]. We next performed gene ontology (GO) enrichment analysis to identify CO 2 -sensitive biological pathways/processes using the PANTHER database. The GO terms highlighted by this analysis related primarily to Figure 1. RNA-sequencing analysis of THP-1 monocytes exposed to buffered hypercapnia for 4 h. (a) Principal component analysis of THP-1 cells exposed to 5% CO 2 (red) or 10% CO 2 (blue) for 4 h. The x-axis (PC1) is the vector that displays the most variance between samples and the y-axis (PC2) displays the second most. The percentage of total variance per principal component is shown on the axis label. Data are representative of three independent experiments. (b) Volcano plot of differential expression between 5% and 10% CO 2 in the basal state (4 h). The cut-off for significance (P-adj < 0.05) is shown with a solid line on the y-axis. Significantly upregulated genes are shown in green and downregulated genes are shown in red. (c) List of the eight most significantly DEGs [up (green) and down (red)] in hypercapnia in the basal state. Includes all genes with a Log 2 FC of AE1 and P-adj of < 0.05 . (d) The top gene ontology terms associated with significant DEGs in hypercapnia in the basal state (10% CO 2 for 4 h) using a cut-off value for significance of P-adj < 0.05. Terms are ranked by fold enrichment. (a-d) Data are representative of three independent experiments in all cases. DEG, differentially expressed gene; padj, adjusted P-value; PC, principal component. mitochondrial function, protein folding and ribosomal biogenesis. The top 20 GO terms from this analysis, ranked by fold enrichment, are displayed in Figure 1d.

RNA-seq: effect of lipopolysaccharide stimulation on monocytes
To examine the response of monocytes to hypercapnia in the context of inflammation we next exposed cells to buffered hypercapnia in the presence of the proinflammatory mediator lipopolysaccharide (LPS). LPS induced a strong transcriptional effect in both normocapnia and hypercapnia, with classical proinflammatory genes demonstrating markedly enhanced expression (e.g. ICAM1, CXCL2, IL1A, TLR4, IL6, CXCL8; Supplementary figure 2, Supplementary tables 7 and 8). Raw data relating to the LPS response in normocapnia (5% CO 2 ) have previously been deposited in the Gene Expression Omnibus repository, accession number (GSE178391). 20 These data validate our RNA-seq approach in THP-1 monocytes.

RNA-seq: effect of hypercapnia in the presence of LPS on monocytes
Similar to the basal state, buffered hypercapnia was a robust stimulus in LPS-treated monocytes as indicated by principal component analysis (Figure 2a). We next examined the DEGs between normocapnia and hypercapnia in the presence of LPS. A total of 1889 genes met this threshold, with comparable numbers of upregulated (946 green) and downregulated (943 red) transcripts demonstrating sensitivity to hypercapnia (Figure 2b). A list of the 10 most DEGs (up and down) is shown in Figure 2c with RNA component of signal recognition particle 7SL2 (RN7SL2) and Scavenger receptor class F member 2 (SCARF2) being the most differentially upregulated and downregulated genes, respectively. Supplementary figure 1b displays the differential expression data in a different way, highlighting the top 50 most significantly affected transcripts (as determined by P-adj) ordered by expression level (Log 2 of TPM) values. We next performed GO enrichment analysis, to identify CO 2sensitive biological pathways/processes in the presence of LPS using the PANTHER database ( Figure 2d). The GO terms highlighted in this analysis related largely to protein localization and telomeres. Some other terms related to glucocorticoid receptor signaling and protein folding. The top 20 GO terms from this analysis, ranked by fold enrichment, are displayed in Figure 2d. While there are clearly more differentially expressed transcripts in hypercapnia in the presence of LPS (n = 1889), compared with the basal state (n = 370), we found a high proportion of genes (n = 254) that were sensitive to elevated CO 2 regardless of LPS (Figure 2e). We propose that this subset of transcripts represents particularly robustly CO 2 -sensitive genes, for example, mitochondrially encoded cytochrome C oxidase 1 (MT-CO1), ATPase sarcoplasmic/endoplasmic reticulum Ca 2+ transporting 1 (ATP2A1) pyrroline-5 carboxylate reductase 1 (PYCR1) and receptor interacting serine/ threonine kinase I (RIPK1). In addition, most genes that are regulated by CO 2 in both the basal and stimulated states respond to CO 2 in the same direction (Figure 2f).

RNA-seq: effect of hypercapnia on mitochondrial gene expression in monocytes
Based on the data in Figures 1 and 2, there was a clear signature for CO 2 -dependent changes in genes related to the mitochondria and oxidative phosphorylation (OXPHOS) in both the basal and LPS-stimulated states. To better understand the effect of hypercapnia on mitochondria, we examined our RNA-seq data for mitochondrially encoded genes which were differentially expressed in 10% CO 2 compared with 5% CO 2 in both the basal and LPS-stimulated states. Of the 37 mitochondrially encoded genes, 35 were detected in our experiments. A heatmap of the expression (in TPM) of the 35 detected mitochondrial genes is shown in Figure 3a. In the basal state, nine mitochondrial genes were significantly differentially expressed in hypercapnia, seven OXPHOS protein subunits and two transfer (t) RNAs. In the LPS-stimulated state, 18 mitochondrial genes were significantly differentially expressed, 10 OXPHOS protein subunits and 8 transfer RNAs. These significant DEGs are highlighted in Figure 3b. These data are consistent with information presented in Figure 2f, highlighting concordant responses of genes in both the basal and the LPS-stimulated states. In both the basal and LPS-stimulated states, several (but not all) mitochondrial genes, which produce subunits of complex I of the OXPHOS pathway, were significantly upregulated by hypercapnia.
Mitochondrially encoded NADH: ubiquinone oxidoreductase core subunit 4 (MT-ND4;  . RNA-sequencing analysis of THP-1 monocytes exposed to buffered hypercapnia for 4 h in the presence of LPS. (a) Principal component analysis of THP-1 cells exposed to 5% CO 2 (red) or 10% CO 2 (blue; 4 h) in an LPS-stimulated state (2.5 lg mL À1 for 2 h). The x-axis (PC1) is the vector that displays the most variance between samples and the y-axis (PC2) displays the second most. The percentage of total variance per principal component is shown on the axis label. (b) Volcano plot of differential expression between 5% and 10% CO 2 (4 h) in an LPS-stimulated state (2.5 lg mL À1 for 2 h). The cut-off for significance (P-adj < 0.05) is shown with a solid line on the y-axis. Significantly upregulated genes are shown in green and downregulated genes are shown in red (c) List of significant DEGs in hypercapnia in the LPS-stimulated state. Includes the top 10 most DEGs up and down ranked by Log 2 FC with a P-adj of < 0.05. (d) The top gene ontology terms associated with significant DEGs in hypercapnia (4 h) in an LPS-stimulated state (2.5 lg mL À1 for 2 h) using a cut-off value for significance of P-adj < 0.05. Terms are ranked by fold enrichment. (e) Venn diagram representing the number of genes that are commonly and exclusively sensitive to CO 2 in THP-1 cells in the basal (blue) and LPS-stimulated (yellow) states. Selected representative examples are included for each group. (f) Heatmap of the top 50 common DEGs between THP-1 cells exposed to 5% and 10% CO 2 in the basal and LPS-stimulated states. The top 50 genes were selected by significance (P-adj) in the basal state and then ranked by differential expression level (Log 2 FC) in the basal state for visualization. All genes had a P-adj < 0.05. Data are representative of three independent experiments in all cases (a-f). DEG, differentially expressed gene; LPS, lipopolysaccharide; padj, adjusted P-value; PC, principal component.
were upregulated in 10% CO 2 compared with 5% CO 2 in both the basal and LPS stimulated states.
Thus, we observed that several mitochondrially encoded genes were upregulated in hypercapnia. However, there are also many nuclear-encoded genes whose protein product is localized in the mitochondria. To gain a more global view of the effect of short-term hypercapnia on mitochondria in monocytes, we used MitoCarta3.0, a database of proteins localized to the mitochondria. We analyzed our expanded DEG lists of . Taken together, these data support the idea that 4 h of hypercapnia exposure is a robust cell stimulus capable of selectively increasing the transcription of mitochondrial genes, with both mitochondrial and nuclear-encoded transcripts significantly increased.

Effect of hypercapnia on mitochondrial content and mitochondrial function in monocytes
To investigate the impact of hypercapnia on mitochondrial function and metabolism in monocytes we performed a range of cellular assays. To determine the impact of elevated CO 2 on mitochondrial mass we used a quantitative PCR (qPCR)-based approach to compare the ratio of mitochondrial DNA (mtDNA)-associated MT-ND4 to the DNA associated with a nuclear-encoded gene [beta-2-microglobulin (B2M)]. In principle, if there is an increase in the ratio of MT-ND4 DNA:B2M DNA, this is an indication of increased mitochondrial number or increased mitochondrial mass per cell. Figure 4a demonstrates that there was no significant change in the ratio of MT-ND4:B2M DNA at 2, 4 or 24 h of exposure to 10% CO 2 AE LPS. In addition, using a different approach, we measured the expression of a mitochondrial membrane lipid, cardiolipin, using the fluorescent dye nonyl-acridine orange (NAO). In response to 10% CO 2 over 24 h, THP-1 cells demonstrated significantly less NAO-dependent fluorescence compared with 5% CO 2 controls (Figure 4b).
To determine mitochondrial superoxide production, we employed the superoxide fluorescent indicator MitoSOX. There was no difference in the relative fluorescence between 5% and 10% CO 2 at 4 h; however, rotenone treatment did result in a statistically significant increase in fluorescence regardless of the CO 2 treatment (Figure 4c). We next used Amplex Red to measure extracellular and intracellular hydrogen peroxide, as well as intracellular peroxidase activity in THP-1 cells in 5% and 10% CO 2 from 0.5 to 24 h. Extracellular peroxide was measured in the culture media. We observed a time-dependent change in extracellular peroxide, increasing up to 24 h; however, there was no apparent CO 2 -dependent effect (Figure 4d, i). Intracellular peroxide level was measured in cell lysates. There was an increase in intracellular peroxide at 24 h; however, there was no difference between 5% and 10% CO 2 (Figure 4d, ii). Intracellular peroxidase activity was also measured in cell lysates; however, again there was no difference between 5% and 10% CO 2 (Figure 4d, iii). Taken together, we observed an increase in extracellular peroxide and intracellular peroxide up to 24 h of exposure, with no effect of elevated CO 2 evident. We next exposed THP-1 cells to 5% and 10% CO 2 and performed mitochondrial enrichment. Using an antibody capable of measuring OXPHOS proteins, we observed an increased expression of the nuclear-encoded mitochondrial ATP-synthase subunit ATP5A in our mitochondrial enriched fractions (Figure 4e and quantified in Supplementary figure 4e).
To determine the level of reductase activity in our monocytes exposed to normocapnia and hypercapnia, we employed a 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT) assay where MTT is converted to formazan by intracellular reductases. Monocytes were exposed to normocapnia or hypercapnia for 22 h followed by addition of MTT to the media for a further 2 h. Formazan detection at 570 nm was significantly higher in cells exposed to elevated levels of CO 2 , indicative of enhanced cellular reductase activity in hypercapnia ( Figure 4f). We also performed real-time cell metabolic analysis using Seahorse technology but observed no significant CO 2 -dependent difference in basal oxygen consumption rate, maximal respiration, basal extracellular acidification or glycolysis (Supplementary figure 5). However, because of the nature of this assay, the metabolic analysis must take  in 5% (pink) or 10% (orange) CO 2 . mtDNA content was determined as a ratio of MT-ND4 to B2M. Data shown as mean AE standard error of the mean for three independent experiments. (b) Nonyl acridine orange (NAO) fluorescence readings were obtained using a CLARIOstar plate reader (Excitation 438 nm/ Emission 535 nm) from THP-1 monocytes exposed to 5% or 10% CO 2 for 24 h. Data are representative of five independent experiments. Data shown are the mean values for each individual n-number. Statistical analysis was performed using a paired t-test. **P < 0.01. (c) Relative MitoSOX fluorescence at 675-715 nm in THP-1 cells exposed to 5% or 10% CO 2 for 4 h with or without rotenone (1 lM) for 2 h. Fluorescence is normalized to an unstimulated control at 5% CO 2 . Data are shown as mean AE standard error of the mean and representative of three individual experiments. Statistical analysis was performed using a two-way ANOVA followed by Sid ak's multiple comparison's test. *P < 0.05. (d) Measurement of (i) extracellular peroxide, (ii) intracellular peroxide and (iii) peroxidase activity in THP-1 cells cultured in buffered phenol-free DMEM media at 5% and 10% CO 2 for up to 24 h. Concentrations were normalized to a standard curve. Data are shown as mean AE standard error of the mean and are representative of three or four individual experiments, with all conditions containing at least two replicates. (e) Western blot analysis of mitochondrial lysates from THP-1 cells exposed to 5% or 10% CO 2 for 24 h. Lysates were probed using a revert total protein stain and imaged in the 700-nm channel or incubated with a mitochondrial cocktail primary antibody followed by a fluorescent secondary mouse antibody and imaged in the 800-nm channel on an Li-COR imaging system. Image is representative of four independent experiments. (f) MTT assay was performed on THP-1 monocytes exposed to 5% or 10% CO 2 for 24 h with MTT addition to the media for the final 2 h. Formazan production was determined on a CLARIOstar plate reader at 570 nm (Ref 690 nm). Data are representative of five independent experiments. Data shown are the mean value for each individual n-number. Statistical analysis was performed using a paired t-test. **P < 0.01. DMEM, Dulbecco's modified Eagle medium; LPS, lipopolysaccharide; mtDNA, mitochondrial DNA; MT-ND4, mitochondrially encoded NADH: ubiquinone oxidoreductase core subunit 4; MTT, 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide. place at ambient CO 2 levels (0.04%), which is a limitation when seeking to compare the effect of elevated CO 2 on metabolic activity. Taken together, the data from Figure 4 suggest that despite consistent CO 2 -dependent changes in mitochondrial gene expression (after 4 h), this does not appear to be associated with increased mitochondrial content as measured by mtDNA:nuclear DNA ratio or NAO assay. Mitochondrial oxidant activity was also not significantly different in hypercapnia using Amplex Red and MitoSOX assays; however, we did observe a CO 2 -dependent increase in ATP5A protein expression in mitochondrial fractions and a significant increase in cellular reductase activity using the MTT assay.

Effect of hypercapnia on acylcarnitine expression and regulation in monocytes
We next took a liquid chromatography (LC)-mass spectrometry approach to examine the functional metabolic consequences of elevated levels of CO 2 on monocytes. Given a reported role for CO 2 in lipid metabolism and b-oxidation in skeletal muscle 21 and the important role of b-oxidation in cells of leukemic origin, 22 we first focused our analysis on a targeted metabolomic screen of 40 acylcarnitine species following 4 h of exposure to 5% or 10% CO 2 with or without LPS. Intriguingly, of the detected acylcarnitine species in our cell lysates, several were statistically significantly increased under conditions of elevated CO 2 (regardless of LPS treatment). At 4 h C3, C4, C5, C14, C16 and C16:1 were significantly different in hypercapnia in both the basal and LPS-stimulated states (Figure 5a, Supplementary figure 6). In general, the longer acylcarnitine species were increased at 10% CO 2 compared with controls. Reexamination of our RNA-seq data revealed several transcripts associated with b-oxidation and carnitine shuttling to be significantly differentially expressed at 10% CO 2 with LPS, for example, acyl-coenzyme A synthetase (ACSL) genes 1, 4 and 5, which are involved in the addition of acyl groups to long-chain fatty acids at the outer mitochondrial membrane; and short-chain acylcoenzyme A dehydrogenase (ACADS), a key enzyme in the b-oxidation cascade. qPCR analysis of ACADS over a longer time course (up to 24 h) revealed it to be differentially expressed at 10% CO 2 with and without LPS (Figure 5b, c), most notably at 24 h of exposure. Taken together, the data in Figure 5 and Supplementary figure 6 reveal a marked change in monocyte lipid metabolism in response to elevated levels of CO 2 that is evidenced by significant differences in several acylcarnitine species and associated changes in specific genes related to lipid metabolism. These functional data are concordant with changes in the expression of the acylcarnitine transporter SLC16A9 in Figures 1b, c and 2b, c. Given that the data above linked CO 2 -sensitive genes associated with fatty acid metabolism (RNA-seq and qPCR) with functional changes in metabolites [liquid chromatography with tandem mass spectrometry (LC-MS/MS)], we re-examined our RNA-seq data to identify robustly CO 2 -sensitive metabolic genes that are rate limiting in the regulation of amino acids on our LC-MS/ MS screen. PYCR1 is a mitochondrial enzyme that catalyzes the NAD(P)H-dependent conversion of pyrroline-5 carboxylate to proline. PYCR1 mRNA was significantly increased in response to elevated CO 2 in both the basal and LPS-stimulated states in our RNA-seq experiment (Figure 2e, Supplementary table 9). qPCR validation of PYCR1 expression was performed in separate samples where again there is evidence of enhanced PYCR1 expression in hypercapnia in both the basal and LPS-stimulated states (Figure 5d, e). Intriguingly, our LC-MS/MS data revealed a specific and corresponding CO 2 -dependent decrease in the cellular levels of the amino acid proline in both the basal and LPS-stimulated states (Figure 5f, Supplementary figure 6). Taken together, these data reveal CO 2 -dependent modulation of lipid metabolism and amino acid metabolism in terms of actual metabolite concentration and the expression of key transcripts involved in the expression of those metabolites.

Effect of hypercapnia on polarized primary murine macrophages
To further investigate the immunometabolic consequences of hypercapnia in a primary immune cell background we generated primary bone marrow-derived macrophages (BMDMs) and differentiated them toward an M2 phenotype using interleukin IL4. Here we focus on the expression of inflammatory and metabolic genes in IL4-stimulated BMDMs under conditions of buffered hypercapnia. The polarizing effect of IL4 stimulation on BMDMs is illustrated in Supplementary figure 7. Using targeted transcriptomic analysis, we identified 192 genes that were differentially expressed when IL4 polarized BMDMs were exposed to hypercapnia versus normocapnia for 24 h in pH-buffered media. Gene set analysis revealed pathways that were upregulated and downregulated in response to elevated CO 2 in primary IL4-stimulated BMDMs (Figure 6a). Absolute mRNA levels for a selection of CO 2 -sensitive exemplar genes have been presented. These exemplar genes correspond to several upregulated pathways (red)-lysosomal degradation [hexosaminidase subunit beta (HEXB)], fatty acid synthesis [fatty acid binding protein 5 (FABP5)], Figure 5. LC-MS/MS analysis of acylcarnitine species in monocytes exposed to hypercapnia. (a) Detected acylcarnitine species from THP-1 cells exposed to 5% or 10% CO 2 for 4 h. Data shown are median values and individual data points for 10 independent experiments. Statistical analysis was performed using one-way ANOVA followed by a Fisher's LSD post-hoc test. *An FDR of < 0.05. RT-qPCR of cDNA generated from THP-1 cells exposed to 5% or 10% CO 2 for 4, 6 or 24 h in the basal state (b, d) or in the presence of LPS (2.5 lg mL À1 ) (c, e) using SYBR green primers targeted to ACADS and PYCR1. Data shown are mean AE standard deviation for four independent experiments. Statistical analysis was performed using two-way ANOVA followed by Sid ak's multiple comparisons test. *P < 0.05, **P < 0.01. (f) Proline metabolite concentration from THP-1 cells exposed to 5% or 10% CO 2 for 4 h. Data shown are median values and individual data points for 10 independent experiments. Statistical analysis was performed using one-way ANOVA followed by a Fisher's LSD post-hoc test. *An FDR of < 0.05. ACADS, short-chain acylcoenzyme A dehydrogenase; cDNA, complementary DNA; FDR, false discovery rate; LC-MS/MS, liquid chromatography with tandem mass spectrometry; LSD, least significant difference; PYCR1, pyrroline-5 carboxylate reductase 1; RQ, relative quantification; RT-qPCR, quantitative reverse transcription polymerase chain reaction. Figure 6. IL4-polarized BMDM response to hypercapnia. BMDMs were exposed to 5% (pink) or 10% (orange) CO 2 for 6 h followed by IL4 (100 ng mL À1 ) stimulation for an additional 18 h. Absolute mRNA levels and differential gene expression were assessed by a metabolic and inflammatory gene panel on the NanoString nCounter platform.

DISCUSSION
This study has for the first time used next-generation RNA-seq to investigate the transcriptional response to CO 2 in immune cells. Using pH-buffered conditions and modest degrees of hypercapnia, we observe significant transcriptional changes in monocytes exposed to 10% CO 2 , resulting in comprehensive changes to mitochondrial associated genes and mitochondrial metabolism.
CO 2 -dependent changes in immunometabolism observed in monocytes are subsequently supported and validated in primary BMDMs polarized with IL4.
Monocyte infiltration/activation and elevated levels of CO 2 are key components in the pathophysiology of hypercapnic COPD. To directly interrogate the influence of CO 2 levels on monocyte function under pH-buffered conditions, we initially exposed monocytes to 5% CO 2 or 10% CO 2 for 4 h. Previous studies have used higher concentrations of CO 2 , up to 20%, 23 and/or a much longer duration of exposure, 3-6 days. 24,25 However, we chose to utilize a modest and more clinically relevant concentration of CO 2 26 to determine a physiological role for CO 2 -dependent transcriptional regulation in a focused fashion. It is technically challenging to separate hypercapnia from pH under human physiologic conditions, so the purpose of this study is to dissect the specific role of CO 2 in monocyte gene expression and function. To our knowledge, this is the first study of its kind to use next-generation RNA-seq to examine the cellular response to elevated CO 2 in pH-buffered conditions in immune cells.
Principal component analysis (Figures 1a and 2a) shows clear separation of the 5% and 10% CO 2 samples in relation to PC1. The volcano plot (Figure 1b) further reveals that 10% CO 2 is a modest but robust stimulus in monocytes and that hypercapnia with LPS results in a much more pronounced transcriptional response (Figure 2b). These data are consistent with the idea that hypercapnia can serve as a microenvironmental modulator capable of modifying immune responses. Enriched GO terms in the basal state relate to muscle contraction (e.g. ATP2A1), mitochondrial function (e.g. MT-CO1), calcium signaling and transmembrane transport (Supplementary table 5). These data align well with current literature as these pathways and processes have all been previously identified as showing some sensitivity to CO 2 . For example, high CO 2 has been shown to downregulate skeletal muscle protein anabolism 27 and cause mitochondrial dysfunction. 24 In the presence of LPS, GO terms associated with the most DEGs (Log 2 FC > 1 or < À1) again revealed transcripts related to mitochondrial activity, angiogenesis and developmental processes (Supplementary table 6), which is also consistent with the current literature. Taken together, there is evidence for a core CO 2 -sensitive cohort of genes that is regulated by CO 2 in both the basal and stimulated states and respond to CO 2 in a congruent and consistent manner (Figure 2f). This core cohort of CO 2sensitive targets is enriched in genes associated with mitochondrial metabolism, calcium signaling and muscle contraction.
While several inflammatory genes [e.g. Toll-like receptor 4 (TLR4)] were differentially expressed (Figure 2e, Supplementary figure 8) in hypercapnia, immune signaling was not among the most prominent GO terms (Figure 1d and 2d) as has been reported in similar models. 28 There was, however, a particularly strong signature for alterations in mitochondrial function, mainly in relation to OXPHOS. Mitochondrially encoded genes were among the top upregulated genes in both the basal and LPS-stimulated states (Figures 1c and 2c). Furthermore, GO terms related to mitochondrial function, specifically the electron transport chain, were among the top terms for both comparisons. From the literature, chronic exposure to hypercapnia causes mitochondrial dysfunction in lung epithelium and fibroblasts. 24 In rats, acute 20% CO 2 inhaled for 10 min improved mitochondrial function and upregulated the expression of the mitochondrial biogenesis regulators peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1a) and TFAM. 29 Interestingly in this study, TFAM was increased in hypercapnia in the presence of LPS at 4 h (Supplementary figure 3b,  Supplementary table 4). To our knowledge, this study is the first to demonstrate a distinct upregulation of mitochondrially encoded genes in hypercapnia. Unbiased RNA-seq revealed that almost all the mitochondrial mRNA and several mitochondrial transfer RNA transcript levels were higher in 10% CO 2 compared with 5% CO 2 , suggesting an increase in mitochondrial transcription in hypercapnia. Interestingly, in both the basal and LPS states, nuclear-encoded mitochondrial transfer RNA synthetases (Supplementary figure 4a, b) and mitoribosome proteins (Supplementary figure 4c, d) were significantly upregulated. As both sets of these proteins are involved in the translation of mitochondrial mRNA, these data provide further evidence that mitochondrial transcription and transcripts relating to the translation of mitochondrial proteins are increased in a CO 2 -dependent manner.
Mitochondria play a central bioenergetic role in cells. Regulation of mitochondria occurs not only at a transcriptional level, but also by autophagy, with dynamic regulation of cellular mitochondrial content critical to the maintenance of functional mitochondria. 30 The data in Figures 1-3 and Supplementary figure 4 indicating enhanced mitochondrial gene expression suggest that mitophagy is unlikely under these conditions. With our observations of increased mitochondrial gene expression in hypercapnia, we investigated the effect of hypercapnia on the mitochondrial content of monocytes for up to 24 h. We observed no CO 2 -dependent difference between the mtDNA content of monocytes using a qPCR-based technique (Figure 4a). Using the fluorescent dye NAO to stain the mitochondrial membrane lipid cardiolipin, we observed a statistically significant decrease in NAOdependent fluorescence following 24 h of exposure to 10% CO 2 (Figure 4b). Altered cardiolipin levels in response to cellular stress have the potential to affect multiple pathways, 31 thus reduced cardiolipin levels following exposure to 10% CO 2 is further suggestive of a CO 2 -dependent shift in mitochondrial homeostasis. The reactive oxygen species O 2 À and subsequently H 2 O 2 are produced by electron leakage during OXPHOS, primarily from complex I and complex III. Our data suggested that mitochondrial reactive oxygen species production may be altered in our model. However, the data (Figure 4c, d) demonstrate that hypercapnia does not significantly affect mitochondrial reactive oxygen species production in monocytes. It is possible; however, that there are small differences in oxidant production below the threshold of sensitivity of our assays. Looking at mitochondrial protein expression, ATP5A was relatively abundant in our mitochondrial extracts in comparison to other proteins involved in OXPHOS and was enhanced in response to hypercapnia (Figure 4e). We next investigated cellular reductase activity in monocytes exposed to elevated CO 2 using MTT yellow tetrazole dye. MTT conversion is proposed to occur intracellularly and likely reflects changes in several cytosolic subcompartments including the mitochondria. 32 Interestingly, we observed an increase in MTT conversion to formazan in monocytes exposed to elevated CO 2 for 24 h (Figure 4f). Taken together, the data in Figure 4 do not support the concept of wholesale increases in mitochondrial mass under conditions of elevated CO 2 . Thus, we believe that our observed transcriptional increase in mitochondrial subunit expression represents an early adaptive response to the metabolic stress of elevated CO 2 , which subsequently becomes modified following more prolonged exposure. For this reason, we next used an LC-MS/MS-targeted screen of acylcarnitine species and amino acids to probe specific changes in metabolite levels in hypercapnia.
Metabolomic analysis of acylcarnitines permits a quantitative evaluation of mitochondrial lipid metabolism under conditions of elevated CO 2 . Immunometabolism is central to immune cell phenotype and function, with M1 "proinflammatory" and M2 "anti-inflammatory" macrophages having distinct metabolic profiles. In general, M1 macrophages are considered to be more glycolytic, while M2 macrophages use more OXPHOS and b-oxidation. 33 This is, however, an oversimplification of a complex topic, 34 but the metabolic context of the cell is clearly intrinsically linked to its physiological role and inflammatory phenotype. Notably, we observed several statistically significantly increased acylcarnitine species from hypercapnic cells compared with normocapnic controls (Figure 5a and Supplementary figure 6a). Acylcarnitines are intermediates in the transport of long-chain acyl (coenzyme A) into the mitochondria for b-oxidation. They can also be exported and released into the plasma of organisms 35 by acylcarnitine efflux proteins (e.g. SLC16A9). 18 In the context of this cell-based experiment we are examining intracellular acylcarnitine species. Altered acylcarnitine levels are associated with several factors including inherited genetic defects, exercise, fasting, insulin resistance, obesity and cardiovascular disease. 36 Skeletal muscle cells exposed to chronically high levels of CO 2 demonstrated increased levels of specific acylcarnitine species. 21 Altered lipid metabolism is also associated with COPD and lung health. 37 The shuttling of fatty acids into the mitochondria involves several key steps 38 and incomplete b-oxidation is associated with elevated serum acylcarnitine levels. 39 A number of transporters are linked to acylcarnitine transport across the plasma membrane, including organic cation/carnitine transporter family members 40 and SLC16A9. 18 Interestingly, in addition to observing a CO 2dependent shift in acylcarnitine species and a key transporter (SLC16A9) in our model, we observed changes in gene expression in specific steps within the fatty acid metabolism pathway. for example, ACSL1, 4 and 5 were all significantly increased in THP-1s in hypercapnia in the presence of LPS and transcripts associated with CPT1a were enhanced in IL4-polarized BMDMs (Figure 6d). ACADS was significantly different in response to CO 2 in the presence of LPS (Supplementary table 4), and a significant increase in the transcript levels of this key b-oxidation enzyme was determined by qPCR over a longer time course (24 h; Figure 5b, c). Furthermore, the acylcarnitine transporter SLC16A9 is among the most differentially enhanced transcripts in response to elevated CO 2 in both the basal and LPS-stimulated states (Figures 1b, c and 2b, c) suggesting a highly coordinated CO 2 -dependent shift in acylcarnitine processing. Notably, SLC16A9 is described as a pH-independent transporter, 18 which further supports our contention that our responses are elicited by CO 2 per se in our buffered hypercapnia system.
We also observe CO 2 -dependent transcriptional changes and altered metabolite levels with respect to amino acid metabolism. PYCR1 gene expression was enhanced in both our RNA-seq and qPCR experiments and this was associated with a highly significant CO 2dependent decrease in the cellular levels of proline in both the basal and LPS stimulated states (Figure 5f,  Supplementary figure 6b). Proline can be synthesized intracellularly using a number of substrates including arginine and glutamine. PYCR1 catalyzes the final conversion of pyrroline-5 carboxylate to proline. Proline is an important structural amino acid linked with collagen deposition and also has roles as an antioxidant and in immune responses. 41 Proline metabolism is also linked to the development of "trained" macrophages that predispose to allergic asthma. 42 The full consequence of reduced proline levels in monocytes exposed to hypercapnia remains to be elucidated, but may be relevant in wound healing and in determining immune cell phenotype. Our data relating to PYCR1 and proline, in addition to our data relating to ACADs and acylcarnitine, clearly suggest a coordinated transcriptomic and metabolomic response to acute hypercapnia under pH-buffered conditions.
There are certain limitations associated with using a cancer-derived THP-1 monocyte model in our study. To verify the main findings from our THP-1 monocyte experiments in a primary macrophage model, we generated BMDMs and polarized them with IL4 to generate "M2-like" macrophages. M2 macrophages or "alternative macrophages" are generally characterized as producing low levels of inflammatory cytokines and elevated levels of b-oxidation. M2-related genes are enhanced in alveolar macrophages of smokers with COPD compared with healthy smoker controls 43 and the ratio of M2/M1 phenotype macrophages is increased in a mouse COPD model. 17 Here, the IL4-polarized BMDMs elicited a marked transcriptional response to elevated CO 2 (24 h) in pH-buffered media ( Figure 6). Interestingly, key pathways (e.g. fatty acid metabolism) that were identified in our unbiased THP-1 RNA-seq experiments were also enhanced in IL4-polarized BMDMs (Figure 6a, c, d). Genes associated with glycolysis (ALDOA) and cytokine signaling (TNF) were reduced (Figure 6f, g), supporting the notion of a CO 2 -dependent shift in immunometabolism in IL4-polarized BMDMs. These data from primary macrophages validate several key findings from our THP-1 experiments. Data in Supplementary figure 8 directly compare RNA-seq TPM data from THP-1 monocytes with absolute mRNA counts from IL4-polarized BMDMs for selected genes. TLR4 and SLC16A3 (monocarboxylate transporter 3) are comparably decreased in hypercapnia in both models, while NFKB1 is CO 2 sensitive in both models but is different in pattern (which could be explained by different duration of CO 2 exposure). In addition to TLR4 and SLC16A3, another 10 genes were CO 2 sensitive (in the same direction) in both models [ALDOA, GLUL, STK11, TK2 (downregulated), ASNS, BCL2L1, FNIP2, ITGB1, CCND1 and CBR4 (upregulated)]. Taken together, the IL4-polarized BMDM data are highly supportive and validate key CO 2 -sensitive pathways (e.g. lipid metabolism) and transcripts (e.g. TLR4) identified in our THP-1 experiments. To validate a CO 2 -sensitive transcript from our NanoString experiment, NFKB2 (which encodes the nuclear factorkappa B family member p100), we next examined p100 protein expression in THP-1 cells exposed to hypercapnia under the same experimental conditions (buffered hypercapnia for 24 h). NFKB2 mRNA expression is significantly decreased in IL4-primed BMDMs exposed to 10% CO 2 for 24 h under buffered conditions and p100 protein expression is also significantly decreased under the same conditions in THP-1 cells (Supplementary figure 9). These data further validate our parallel experimental approach using THP-1 monocytes and IL4polarized BMDMs. Further validation of our RNA-seq data is provided in Supplementary figure 10, where RIPK1 mRNA and RIPK1 protein are reduced in response to elevated CO 2 in THP-1 monocytes. RIPK1 regulates cellular decisions between prosurvival/cell death signaling in the nuclear factor-kappa B pathway and is emerging as a potential therapeutic target in multiple inflammatory diseases. 44 Cellular metabolic changes in immune cells are fluid to allow cellular metabolism to dynamically change in response to the emerging needs of the cell in response to a changing environment. These changes can represent alterations in cell fate decisions as well as effector function. 45 A classic example of this is the differentiation of M1 macrophages to a more glycolytic phenotype in response to LPS. This primes the cells for proinflammatory effector function and ensures that this is possible through glycolysis-dependent macromolecule generation via the pentose phosphate pathway. Thus, immune cells are capable of dramatic reprogramming of cellular metabolism in response to environmental signals (e.g. bacterial products, cytokines, hypoxia) and here we provide evidence of metabolic shifts in response to elevated CO 2 levels. The position within the immune cell's life cycle where the environmental signal occurs is also of importance, as cell differentiation can be affected in uncommitted cells while effector function can be modified in mature differentiated cells. Given that monocyte infiltration, M2 macrophage polarization and hypercapnia are co-incident in COPD, our data provide novel insights into the immunometabolic landscape of hypercapnic lung disease. Hypercapnia and acidosis have previously been reported to attenuate M1 macrophage differentiation and migration, 46 which is indicative of an attenuated inflammatory immune response. Here we build on these studies to demonstrate that CO 2 per se causes a marked immunometabolic shift in monocytes and macrophages that further attenuates proinflammatory signaling. A CO 2 -dependent suppression of transcripts associated with glycolysis in both monocytes and IL4polarized macrophages suggests a further deviation from proinflammatory M1-like "classic" signaling. Suppression of TLR4 transcripts in both cell types is congruent with this. Enhancement of transcripts associated with fatty acid oxidation and lipid metabolism furthermore points away from proinflammatory-type signaling and more toward that of an immunosuppressive phenotype, where fatty acid synthesis can fuel OXPHOS in M2 macrophages. 47 Furthermore, in monocytes there is evidence of a potential CO 2 -dependent shift away from expression of chemotactic factors (e.g. CX3CR1 and CXCR3), while promoting the expression of more secretory factors (e.g. IL12B; Supplementary table 4). This microenvironmental shift in cellular signaling capacity could be of benefit in the context of a destructive inflammatory milieu, but may be less responsive to pathogenic challenges through reduced innate signaling (e.g. reduced TLR4; Supplementary figure 8). These scenarios are highly relevant to lung pathologies including COPD. Furthermore, an enhanced understanding of how CO 2 affects metabolism and immune signaling in monocytes and macrophages may reveal new opportunities for therapeutic hypercapnia, where the anti-inflammatory effects of CO 2 can be of benefit. 13 CONCLUSIONS Taken together, this study has for the first time used next-generation RNA-seq to investigate the transcriptional response to CO 2 under pH-buffered conditions in immune cells. Hypercapnia alone elicits a modest but robust and significant transcriptional response that is enhanced in the presence of the proinflammatory stimulus LPS. The transcriptional response of monocytes to hypercapnia revealed a novel and marked change in gene expression in relation to mitochondria and mitochondrial associated genes, suggesting that the elevated level of CO 2 in this model is a metabolic stressor. Despite widespread changes in mitochondrial gene expression in our RNA-seq experiments, our data indicate that significant changes in mitochondrial reactive oxygen species production or mitochondrial mass are not associated with the CO 2 response under these conditions. Mitochondrial function/ activity; however, was significantly altered. Cellular reductase activity was enhanced under hypercapnic conditions, and we observed a marked change in acylcarnitine species as early as 4 h after CO 2 exposure, indicating a CO 2 -dependent change in lipid metabolism. Proline levels were also significantly reduced under these conditions. These changes in lipid metabolism and amino acid metabolism were also associated with altered transcription of specific genes associated with mitochondrial lipid processing (e.g. ACADS) and proline biosynthesis (e.g. PYCR1). Importantly, many of the key observations observed in THP-1 monocytes were also evident in IL4-polarized primary BMDMs (e.g. increased expression of genes associated with fatty acid metabolism; Supplementary figure 11). These data help us to understand the immunomodulatory effects of hypercapnia on monocytes and macrophages and may provide insight into future therapeutics for patients experiencing hypercapnia.

THP-1 monocytes
The immortalized cell line THP-1 was chosen as a suitable substitute for primary monocytic cells. THP-1 cells originated from a 1-year-old male with acute monocytic leukemia. They have been well described and demonstrate characteristic monocyte functions such as phagocytosis, 48 inflammatory cytokine production and differentiation capacity. In addition, THP-1 cells respond to LPS 49-51 stimulation and differentiated THP-1 macrophages have shown an altered immune response to elevated CO 2.

28,46,52
We have recently extensively characterized the shNT version of these cells described by Phelan et al. 20 where monocytes have been stably transduced with a nontarget short hairpin RNA. In this study both wildtype and shNT-THP-1s were employed as some of these studies were carried out as part of a wider study. 20 We have previously published that shNT-THP-1s phenocopy the wildtype THP-1s. 53,54 THP-1 cells were maintained in Roswell Park Memorial Institute 1640 medium (61870-010, Thermo Fisher, Waltham, MA, USA) supplemented with 10% fetal bovine serum (10270-106, Thermo Fisher) and 1% penicillinstreptomycin (15140122, Thermo Fisher). Cells were maintained at a density of 2 9 10 5 -1 9 10 6 cells mL À1 and subcultured two times per week. For subculture, cells were counted and media containing the appropriate number of cells was transferred to a T75 tissue culture flask containing fresh media, to a total of 20-30 mL. All procedures and treatments prior to cell lysis were performed in a class II biological safety cabinet or a CO 2 chamber. shNT-THP-1 monocytes were used in Figures 1-3, 4a, c, d and 5, as well as in Supplementary  figures 1, 2, 4a-d, 6, 8a-c and 10a. Untransformed THP-1 monocytes were used in Figure 4b, e, f and Supplementary figures 4e, 5, 9a, b and 10b, c.

Primary murine macrophages
Primary BMDMs were isolated and identified as previously described 46,55 from healthy wild-type black 6 mice at about 15-20 weeks of age. In brief, mice were killed by cervical dislocation by trained and licensed animal handlers in the University College Dublin Biomedical Facility. We received the mice postmortem. Both femurs and tibias were in toto excised and washed in sterile phosphate-buffered saline (PBS; (D8537; Merck, Darmstadt, Germany). Epiphyses were removed and medullary cavity was flushed with cold sterile PBS using a 27G needle. The flow-through was collected into a 50-mL tube (210270; Greiner Bio-One International). The cell suspension was centrifuged at 500g for 10 min, and the supernatant was discarded. After lysing of erythrocytes and repeated centrifugation, cells were resuspended in Roswell Park Memorial Institute 1640 (Merck) with 10% fetal calf serum, 1% penicillin-streptomycin, 2 mM L-glutamine and 10 ng mL À1 macrophage colony-stimulating factor. Pooled cells of one mouse were seeded in one T-75 flask (660175-120G; Greiner Bio-One International, Frickenhausen, Germany) and split after 48 h. After 7 days in culture, cells were harvested and identified by F4/80 immunostaining as described previously. 46 Primary murine macrophage polarization After isolation, differentiation and identification of BMDMs, murine macrophages were further polarized to analyze the effect of CO 2 on macrophage polarization. To induce macrophage polarization, 5 9 10 5 cells per well were seeded in 12-well plates (665180; Greiner Bio-One International) and placed within gas chambers at different CO 2 concentrations (5% versus 10% CO 2 ). After 6 h of CO 2 treatment primary murine macrophages were polarized into immunomodulatory (M2)-like macrophages using IL4 (100 ng mL À1 ; I1020-5UG Merck) for additional 18 h. At the end of the experiments, cells were harvested for RNA isolation.

Hypercapnic exposures
All hypercapnic exposures were performed in humidified environmental chambers (Coy Laboratories, Grass Lake, MI, USA) at 37°C, at 5% or 10% CO 2 . All experiments were performed in buffered, pre-equilibrated Dulbecco's modified Eagle medium (DMEM) base media (D1152, Merck) + supplements 56 or phenol-free DMEM base media (D2902, Merck) + supplements (MitoSOX experiments). For most THP-1 experiments, cells were used at a density of 750 000 cells mL À1 of media. We have recently demonstrated that intracellular pH does not change significantly in Phorbol 12myristate 13-acetate (PMA)-treated THP-1s cultured under these buffered hypercapnia conditions. 46 Representative media pH, pCO 2 and pO 2 measurements are included in Supplementary table 1.

RNA-seq
Total RNA was extracted from cultured cells using the E.Z.N.A. Total RNA Kit I. All lysates were homogenized by pipetting 109 up and down using a 21-gauge needle and a 1-mL syringe prior to column extraction. RNA concentration and purity were determined using a NanoDrop 2000 spectrometer. RNA clean-up was performed (if required) using ethanol precipitation. Subsequent Qubit results determined an RNA integrity number > 9.5 for all samples. cDNA library preparation was performed with polyA selection using Illumina HiSeq, 2 9 150-bp configuration, single index, per lane (~350 millionraw paired-end reads per lane). Library preparation and sequencing were performed by GeneWiz (Germany). Raw data quality was evaluated with FastQC. Sequence reads were trimmed using Trimmomatic version 0.36 and subsequently mapped to the Homo sapiens GRCh38 reference genome using the STAR aligner version 2.5.2b. Unique gene hit counts were calculated using featureCounts from the Subread package version 1.5.2. Downstream differential expression analysis was performed using DESeq2. 57 P-values and log 2 fold changes were generated using the Wald test. Significant DEGs were called as genes with an adjusted Pvalue (P-adj) < 0.05 (Supplementary tables 3, 4 and 7). Initial stringent GO analysis was performed using GeneSCF version 1.1-p2 to generate GO graphs where a Fisher's exact test was used to determine P-values (Supplementary tables 5, 6 and 8, Supplementary figure 2d). Principal component analysis and read count distribution analysis were performed by GeneWiz. Volcano plots of DEGs were generated using GraphPad Prism (version 8, San Diego, CA, USA), with log 2 FC on the x-axis and adjusted P-value on the y-axis. Significant DEGs were determined as any genes from the DESeq2 workflow, which had an adjusted P-value (P-adj) < 0.05. Nested comparisons for multiple groups were completed by cross-referencing lists of DEGs from pairwise comparisons to generate secondary lists of either common or specific DEGs. GO analysis was performed on each list of DEGs. This analysis was completed by the PANTHER Classification system (pantherdb.org) using statistical overrepresentation tests. The background reference list for these comparisons was generated by compilation of all genes expressed to any degree in any of the experimental conditions (raw TPM values). Raw data from the RNA-seq experiment are displayed as mean TPM (AE standard error of the mean) to best visualize expression changes between several groups. Statistical comparisons between individual groups used DESeq2 analysis of normalized counts. The bulk RNA-seq data presented in the study are deposited in the Gene Expression Omnibus repository, accession number GSE206333.

Measurement of mtDNA content by qPCR
Cells were collected in 15-mL centrifuge tubes using a pipette and cell scraper as required and centrifuged for 5 min at 300g. DNA extraction was performed using the QIAamp DNA Mini Kit (Qiagen, Hulsterweg, Netherlands) according to the manufacturer's instructions. DNA concentrations were measured using a nanodrop spectrophotometer. A second elution was performed using 100 lL Buffer AE to maximize DNA yield. DNA was stored at À20°C.
DNA was normalized to 20 ng lL À1 and measured again on the nanodrop. A serial dilution was performed to dilute the DNA to 0.5 ng mL À1 . SYBR Green qPCR was performed on the extracted DNA as described above using specific primers for the mitochondrially encoded (MT-ND4) and the nuclear-encoded B2M and cycle threshold (CT) values were determined. Primer details are included in Supplementary table 2. DCT was calculated as CTnucDNA -CTmtDNA or CTB2M -CTMT-ND4. mtDNA copy number per cell was calculated as 2 9 2 DCT . This method assumes that there are two copies of the nuclear gene B2M per (diploid) cell and one copy of the mitochondrial gene MT-ND4 per molecule of mtDNA. The calculation used gives the number of copies of MT-ND4 relative to B2M and can therefore be used to infer the amount of mtDNA per cell. Similar methods have been used to assess mtDNA copy number in multiple species, including the use of B2M to represent nucDNA in humans. 58,59 The use of this qPCR method to assess mitochondrial biogenesis in THP-1 cells has also been demonstrated previously. 49

NAO assay
The NAO assay uses the chemical dye acridine orange 10-nonyl bromide (A1372, Merck) to measure mitochondrial mass. NAO binds to cardiolipin, a phospholipid specifically present on the mitochondrial membrane. 60 In brief, approximately 100 000 THP-1 monocyte cells/well in 200 lL of media were seeded on flat-bottomed 96-well plates in either 5% or 10% CO 2 -buffered media. Treatments were carried out in sextuplicate. Cells were incubated for 24 h. Subsequently, the suspension of THP-1 cells was centrifuged at 1500g for 5 min (brake 1) and media was carefully aspirated; 5 lM of NAO dissolved in PBS and 100 lL was added to each well in a 96-well plate. Cells were incubated at 37°C and 5% CO 2 for 30 min in the dark. The plate was centrifuged again, and the dye removed. Cells were resuspended in 100 lL PBS and fluorescence measured at (Excitation/ Emission 485/538 nm) on a CLARIOstar plate reader (BMG Labtech). Background was removed by subtracting negative control (cells without NAO dye) value from all samples.

Western blot analysis
Whole-cell protein lysates were prepared using either whole cell lysis buffer (150 mM NaCl, 25 mM Tris pH8, 1 mM ethylenediaminetetraacetic acid, 1% NP-40) supplemented with protease inhibitor cocktail (P2714, Merck) or precipitated from RNA extraction flow-through and suspended in 1 M HEPES (N-2-hydroxyethylpiperazine-N-2-ethane sulfonic acid) and 1% sodium dodecyl sulfate (Merck). Mitochondrial lysates were prepared using a mitochondrial isolation kit (Thermo Scientific) according to the manufacturer's instructions, with isolated mitochondria lysed in 2% CHAPs (Merck) in Tris Buffer Saline (TBS). Lysates were quantified using the DC Protein Assay kit (Biorad, Hercules, CA, USA) before sodium dodecyl sulfate-polyacrylamide gel electrophoresis on the Bio-Rad mini-protean system using TGX precast gels. Wet transfer was performed onto nitrocellulose membranes and reversibly stained with Revert 700 total protein stain (Li-COR, Lincoln, NE, USA). Membranes were imaged at 700 nm, washed in Revert 700 wash solution (Li-COR) and destained in Revert 700 destaining solution (Li-Cor)  MitoSOX superoxide assay THP-1 cells were suspended in 1 lM MitoSOX (Thermo Fisher) working solution or dimethyl sulfoxide vehicle (for unstained cells), at a concentration of 1 million cells mL À1 . Cells were incubated in a 37°C water bath for 20 min, protected from light, with gentle agitation every 5 min. Cells were centrifuged at 300g and pellets were washed with warm PBS three times (with centrifugation between washes), before dividing cells evenly between tubes for resuspension (~750 k cells mL À1 ) in pre-equilibrated low glucose phenol-free (D2902; Merck) 5% CO 2 media or 10% CO 2 supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin in the environmental chambers. A volume of 1 mL of the cell suspension was added per well of a 24-well tissue culture plate, protected from light, with wells of stained and unstained cells immediately transferred to 1.5-mL microcentrifuge tubes to confirm successful staining. Plates were incubated for 4 h with or without rotenone (Merck) (1 lM) for 2 h and protected from the light. Following incubation, the cell suspension was analyzed in a CytoFLEX S flow cytometer (Beckman Coulter, Brea, CA, USA). The blue laser (488 nm) was used to excite the fluorophore and emission was measured from 675 to 715 nm. Cells were selected for analysis based on the forward and side scatter and subsequently single cells were selected based on forward scatter area and forward scatter height. Cell loading with MitoSOX was confirmed by an increase in the median fluorescence detected in the 675-715-nm range. Treatment or CO 2 -dependent changes in fluorescence were calculated by relative fluorescence with respect to the untreated 5% CO 2 control at each time point.

Amplex red H 2 O 2 assay
Following CO 2 exposure, cells were centrifuged at 300g for 5 min and the media was removed. The media is used to measure extracellular peroxide. Cells were lysed in 19 reaction buffer, 0.1% Triton (Merck), briefly pulsing samples on a vortex every 3 min for 15 min on ice. Homogenates were centrifuged for 15 min at max speed (21 000g

MTT assay
The MTT assay exploits the chemical reduction of a yellow tetrazole dye [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazol ium bromide] to insoluble formazan by intracellular reductases and NADH in particular. The conversion of MTT to formazan mainly occurs intracellularly in mitochondria as well as other intracellular organelles. 32 The MTT assay is occasionally used as an assay of cellular viability as intracellul ar reductase activity can be proportional to cell viability. However, the assay can also determine augmented or reduced reductase activity linked to enhanced enzymatic activity and/or mitochondrial mass.
In brief, approximately 100000 THP-1 monocyte cells/well in 200 lL of media were seeded on flat-bottomed 96-well plates in either 5% or 10% CO 2 -buffered DMEM media. Treatments were carried out in sextuplicate. Cells were incubated for 22 h with or without the addition of MTT stock solution (5 mg mL À1 ; Merck) at a ratio of media:dye of 3:1 for a further 2 h in the dark. Subsequently, the suspension of THP-1 cells was centrifuged (Eppendorf 5810R, Eppendorf, Hamburg, Germany) at 1500g for 5 min (brake 1) and the media was carefully aspirated. Following this, 100 lL of dimethyl sulfoxide (Merck) was added to each well to dissolve the produced formazan and incubated at 37°C and 5% CO 2 for 30 min. The absorbance was measured at a wavelength of 570 nm (Ref 690 nm) on a CLARIOstar plate reader (BMG Labtech). Background was removed by subtracting negative control (cells without MTT dye) value from all samples.

LC-MS/MS assay
THP-1 monocytes (~750 k cells mL À1 ) were incubated at 37°C with either 5% or 10% CO 2 -buffered DMEM (high glucose; Merck D1152 with 10% fetal bovine serum and 1% penicillinstreptomycin) for 4 h, with or without LPS (Invivogen, San Diego, CA, USA) (2.5 lg mL À1 ) for the final 2 h. The cell suspension was transferred to a 15-mL tube and any attached cells were collected from the plate with a cell scraper and cold PBS. Cells were pelleted at 1000g for 5 min and washed two times with ice-cold PBS. The final cell pellet was snap-frozen in liquid nitrogen before extraction. Metabolite extractions were performed to isolate a broad range of metabolites from cell pellets (10 technical replicates per condition). In brief, cells were extracted with ethanol/phosphate buffer (85%v/v). The metabolites were measured using LC-MS/MS analysis and flow injection analysis-MS/MS analysis based on the p180 biocrates assay. The data were acquired on a SCIEX QTRAP 6500plus mass spectrometer coupled to SCIEX ExionLC Series UHPLC capability as previously described. 61 Liquid chromatography was performed using a custom ultra-highperformance liquid chromatography column with acetonitrile and water with 0.2% formic acid as mobile phase. The targeted LC-mass spectrometry platform captures over 120 metabolites giving a comprehensive coverage of multiple metabolites. For the purposes of this study we are focusing on the acylcarnitine species and amino acids. Analysis was performed using a one-way ANOVA followed by a Fisher's least significant difference post-hoc test. A false discover rate (FDR) of < 0.05 was deemed significant using the total data set.
Quantitative reverse transcription polymerase chain reaction RNA was extracted using the EZNA total RNA Kit I (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer's instructions. DNase I digestions were performed on all RNA samples intended for use in qPCR. Samples were incubated for 15 min at 20°C in a thermocycler before addition of 1 lL of 25 mM ethylenediaminetetraacetic acid solution to inactivate the DNase I (Thermo Fisher), and samples were heated to 65°C for 10 min. Complementary DNA (cDNA) was synthesized using Moloney murine leukemia virus.
Moloney Murine Leukemia Virus Reverse Transcriptase (Promega, Madison, WI, USA). Master mixes were prepared using diluted RNA, Moloney murine leukemia virus, reaction buffer, dNTPs random primers (Thermo Fisher) and nucleasefree water. Control samples lacking the reverse transcriptase (no reverse transcriptase) were used to detect any genomic DNA contamination.
The samples were incubated for 60 min at 37°C for the cDNA synthesis reaction. cDNA samples were used immediately or stored at À20°C.
For qPCR, cDNA samples were diluted 1 in 4 in nucleasefree water. Master mixes for SYBR green reactions (Thermo Fisher) were prepared using master mix solutions, forward and reverse primers (Eurofins, Val Fleurie, Luxembourg) and nuclease-free water for 10-lL reactions in a 384-well plate. The plate was centrifuged briefly to collect samples and remove bubbles. qPCR was performed in an Applied Biosystems RT-PCR machine with the appropriate Applied Biosystems Quant Studio 7 software. For SYBR Green primers, melt curves were generated for each sample for assessment of primer performance. Primer details are included in Supplementary table 2.

Targeted transcriptomic analysis in primary macrophages
Targeted transcriptomic analysis was performed on RNA derived from primary BMDMs. Total RNA isolation was achieved using TRIzol (TRI Reagent; Merck). RNA samples were subsequently analyzed on the NanoString nCounter. The Metabolic Pathways Panel (XT-CSO-MMP1-12; NanoString Technologies, Seattle, WA, USA) was utilized for analysis and encompassed 768 inflammatory and metabolic genes and 20 internal reference genes for data normalization. Normalization of absolute mRNA counts was accomplished using the nSolver software (advanced analysis module version 2.0.134). The log 2transformed output data were analyzed using R (version 3.3.2). Genes with normalized expression values below 20 were removed. The remaining genes were utilized for global gene set analysis, pathway score analysis and individual gene expression analysis.

Real-time cell metabolic analysis
THP-1 cells were exposed to 5% or 10% CO 2 for 24 h in buffered DMEM as per previous experiments. About 1 h prior to termination of experiment cells were counted, removed from the chambers and washed with phenol-free, bicarbonatefree serum free seahorse XF DMEM supplemented with glutamine (2 mM), glucose (10 mM) and pyruvate (1 mM; Agilent, Santa Clara, CA, USA). Cells were resuspended in seahorse DMEM and seeded onto Cell-Tak (Corning, Tewksbury, MA, USA) adhesive-coated seahorse 96-well plates at~150 000 cells per well, with eight technical replicates per treatment. Cells were centrifuged on the 96-well plate (500g) for 5 min to promote adhesion. The plate was then incubated for 90 min at 37°C in a non-CO 2 incubator prior to running the assay. A seahorse XF Cell Mito Stress Test cartridge that had previously been prepared with water and subsequently assay calibrant was loaded with oligomycin (15 lM), Carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone (FCCP) (10 lM), antimycin A/rotenone (5 lM) and 2-deoxy D-glucose (500 mM; Sigma Aldrich) in ports A-D, respectively. Assay was run on a Seahorse XF 96 well Analyzer (Agilent). Basal oxygen consumption rate (OCR) was calculated as follows: OCR preinjection (3rd measurement) minus OCR postrotenone injection. Maximal respiration was calculated as follows: maximal OCR post-FCCP injection minus OCR postrotenone/antimycin A injection (third measurement). Basal extracellular acidification rate (ECAR) was calculated as ECAR preinjection (third measurement). Glycolysis was calculated as ECAR preinjection (third injection) minus ECAR post-2-deoxy-glucose (third measurement).

Statistical analysis
Statistical analysis was performed for the RNA-seq and targeted transcriptomic analysis as described above. ANOVA or t-tests were applied as indicated in the figure captions and figures prepared using GraphPad Prism (version 8

ETHICAL APPROVALS
Ethical review and approval was not required for the animal study because animal tissue was acquired postmortem from animals that were killed in the University College Dublin Biomedical Facility in accordance with the University Animal Research Ethics Committee's approval. The policy to make use of postmortem tissue is in alignment with Clause 27 of Directive 2010/63/EU which states "To promote the principle of reduction, Member States should, where appropriate, facilitate the establishment of programmes for sharing the organs and tissue of animals that are killed".