Redistribution of SOD3 expression due to R213G polymorphism affects pulmonary interstitial macrophage reprogramming in response to hypoxia

Abstract

The extracellular isoform of superoxide dismutase (SOD3) is decreased in patients and animals with pulmonary hypertension (PH). The human R213G single-nucleotide polymorphism (SNP) in SOD3 causes its release from tissue extracellular matrix (ECM) into extracellular fluids, without modulating enzyme activity, increasing cardiovascular disease risk in humans and exacerbating chronic hypoxic PH in mice. Given the importance of interstitial macrophages (IMs) to PH pathogenesis, this study aimed to determine whether R213G SOD3 increases IM accumulation and alters IM reprogramming in response to hypoxia. R213G mice and wild-type (WT) controls were exposed to hypobaric hypoxia for 4 or 14 days compared with normoxia. Flow cytometry demonstrated a transient increase in IMs at day 4 in both strains. Contrary to our hypothesis, the R213G SNP did not augment IM accumulation. To determine strain differences in the IM reprogramming response to hypoxia, we performed RNAsequencing on IMs isolated at each timepoint. We found that IMs from R213G mice exposed to hypoxia activated ECM-related pathways and a combination of alternative macrophage and proinflammatory signaling. Furthermore, when compared with WT responses, IMs from R213G mice lacked metabolic remodeling and demonstrated a blunted anti-inflammatory response between the early (day 4) and later (day 14) timepoints. We confirmed metabolic responses using Agilent Seahorse assays, whereby WT, but not R213G, IMs upregulated glycolysis at day 4 that returned to baseline at day 14. Finally, we identify differential regulation of several redox-sensitive upstream regulators that could be investigated in future studies.

NEW & NOTEWORTHY Redistributed expression of SOD3 out of tissue ECM due to the human R213G SNP exacerbates chronic hypoxic PH. Highlighting the importance of macrophage phenotype, our findings reveal that the R213G SNP does not exacerbate pulmonary macrophage accumulation in response to hypoxia but influences their metabolic and phenotypic reprogramming. We demonstrate a deficiency in the metabolic response to hypoxic stress in R213G macrophages, associated with weakened inflammatory resolution and activation of profibrotic pathways implicated in PH.

INTRODUCTION

Dysregulated redox signaling has long been implicated in pulmonary hypertension (PH) (1). The extracellular isoform of superoxide dismutase (EC-SOD or SOD3) is an important redox enzyme that is abundantly expressed in the lung and vasculature, and its expression and activity are decreased in human PH (2, 3) and in animal models of PH (4–8). Multiple studies from our laboratory and others have shown that SOD3 is protective in the mouse model of chronic hypoxia-induced PH in mice, whereby loss of expression (global or smooth muscle cell-selective knockout) worsens and overexpression (transgenic or pharmacological delivery) improves PH endpoints, including small vessel muscularization, perivascular collagen deposition, right ventricular (RV) hypertrophy, and RV systolic pressure (RVSP) (9–12).

Intriguingly, a human single-nucleotide polymorphism (SNP) in SOD3 that causes an arginine to glycine substitution in the heparin binding domain (R213G; rs1799895) results in the release of SOD3 from the extracellular matrix (ECM) into extracellular fluids [i.e., circulation and bronchoalveolar lavage fluid (BALF)] and is associated with an increased risk of cardiovascular disease (i.e., ischemic heart disease) (13, 14). We previously demonstrated that mice expressing the R213G variant of SOD3 develop increased pulmonary vascular remodeling and PH in the chronic hypoxia model (13). This suggests that loss of tissue SOD3, particularly its expression in the pulmonary vasculature (11, 13), contributes to PH development that is not rescued by increased content in the serum or alveolar space. An important element of the protective effects of SOD3 includes its ability to modulate inflammation. In pulmonary disease models, we have shown that increased local SOD3 limits immune cell recruitment, the production of inflammatory chemokines and cytokines, and the activation of proinflammatory pathways, including interleukin (IL)-6, nuclear factor kappa-light-chain-enhancer of activated B cells (NfκB), tumor necrosis factor (TNF)-α, and interferon signaling (12, 13, 15–19).

Interstitial macrophage (IM) accumulation, particularly in the perivascular space, is a hallmark of PH and correlates with the severity of pulmonary vascular remodeling (20, 21). IM accumulation is observed in patients with PH and in multiple experimental PH models (20–25). In the chronic hypoxia mouse model, this accumulation occurs in the initial response to hypoxia (3–4 days) but resolves by later timepoints (≥14 days) (22). In addition, IMs undergo transcriptional reprogramming in response to hypoxia (22, 26). Pugliese et al. (22) demonstrated that following 4 days of in vivo hypoxia (peak IM accumulation), IMs upregulate genes involved in metabolism and innate immune function. However, consistent with the resolution of IM inflammation observed by flow cytometry, after 14 days of hypoxia, IMs upregulated anti-inflammatory and proreparative genes and downregulated proinflammatory pathways. More recently, a deeper look at IM responses to hypoxia utilizing single-cell RNA sequencing (scRNAseq) revealed unique hypoxia-associated IMs as well as normoxia-associated IM populations that exhibit transcriptional reprogramming in response to hypoxia. Each of these populations appears to contribute to the upregulation of pathways associated with PH pathogenesis in the total IM population (26).

Given these important data demonstrating the contribution of macrophages to PH and the fact that SOD3 expression can influence macrophage recruitment and phenotype as well as PH severity (10–13, 15), this study aimed to determine whether the R213G SOD3 variant affects the hypoxia-induced accumulation and transcriptional response of IMs. We hypothesized that loss of tissue SOD3 in R213G mice leads to exacerbated IM inflammation and upregulation of pathways that promote vascular remodeling and PH. Using a combination of flow cytometry, RNAseq, and metabolic assays, we demonstrate that IMs from R213G SOD3 mice have an altered response to hypoxia, including a lack of metabolic remodeling, mixed alternative and proinflammatory activation, and limited resolution of inflammation, that may contribute to the later development of worsened pulmonary vascular remodeling and PH.

MATERIALS AND METHODS

Mouse Model

Mice homozygous for the R213G SOD3 SNP (rs1799895) (13) on the C57BL/6J background were compared with C57BL/6J wild-type (WT) controls. Colonies were bred and maintained at the University of Colorado Anschutz Medical Campus (AMC). At 8–12 wk of age, male and female mice were randomly assigned to normoxia controls in Denver ambient air (5,280 ft/1,609 m) or 4 or 14 days of hypoxia exposure in hypobaric chambers simulating an approximate altitude of 18,000 ft/5,486 m above sea level, equivalent to 10% atmospheric oxygen (395 Torr), as previously described (13). All studies were approved by the University of Colorado Institutional Animal Care and Use Committee (IACUC).

Complete Blood Counts

Mice were anesthetized with 1–2% isoflurane, and blood was obtained via terminal closed-chest cardiac puncture of the right ventricle, as previously described (27). Complete blood counts (CBCs) were obtained from heparin anticoagulated blood using the hematologic analyzer Heska HT5 (Loveland, CO).

Flow Cytometry and Fluorescence-Activated Cell Sorting

Labeling of intravascular leukocytes was achieved by retro-orbitally (RO) injecting mice with 1 μg of CD45-PE antibody (BD Biosciences, Clone No. 30-F11) in 100 μL of PBS for 5 min before collecting blood and lungs. Alternatively, for identification of recruited alveolar macrophages, intratracheal administration of 1 mL of CD45-PE-Cy7 antibody per lung (BD Biosciences, Clone No. 3-F11, 1:100) was performed after mice were euthanized and incubated for 5 min before collecting the lungs. Following blood collection via open-chest cardiac puncture, lungs were flushed with 10 mL of PBS (without calcium or magnesium; Sigma) via the right ventricle to dislodge the majority of circulating leukocytes and then isolated. Lungs were processed for flow cytometric analysis and sorting, as previously described (15). In brief, collected whole lungs were digested in 1 mL of HBSS (without calcium or magnesium; Sigma) with Liberase TM (Sigma; 0.4 mg/mL) and DNAse I (Sigma; 100 U/mL) utilizing mechanical digestion with the GentleMACS system (preinstalled program m_lung_01_02 followed by 15-min incubation at 37°C with agitation and dissociation using the m_lung_02_01 program for 8 s). After filtering the cell digest through a 70-μm cell strainer, red blood cells were lysed (eBioscience 1X RBC Lysis Buffer; 3 min, room temperature), and cell suspensions were centrifuged (10 min, 300 g), washed, and incubated in FA3 buffer (PBS, 1 mM EDTA, 10 mM HEPES, 1% FBS, pH 7.4) at 4°C. For staining, FCγR blocking was first performed with anti-CD16 and anti-CD32 antibodies (BD Biosciences) for at least 20 min. Cell suspensions were then stained with Zombie Violet Fixable viability dye (BioLegend) for 15 min, followed by 30-min incubation with the antibody panel. Antibody details: anti-mouse CD45-AF700 (BioLegend, Clone No. 3-F11, 1:50), CD64-AF647 (BD Pharminogen, Clone No. X54-5/7.1, 1:50), CD11b-FITC (BioLegend, Clone No. M1/70, 1:100), CD11c-PE-Cy7 (BioLegend, Clone No. N418, 1:100), MHCII-APC-Cy7 (Clone No. M5/114.15.2, 1:100), Ly6G-BV421 (BD Biosciences, Clone No. 1A8, 1:100), CD3-BV421 (BioLegend, Clone No. 17A2, 1:100), B220-BV421 (BD Biosciences, Clone No. RA3-6B2, 1:100), Siglec-F-PE-CF594 (BD Pharminogen, Clone No. E50-2440, 1:100), and Ly6C-BV510 (BioLegend, Clone No. HK1.4, 1:100). Cell suspensions were then washed and fixed (10% PFA, 30 min at 4°C), filtered through flow tubes with 40-µm filter caps (Falcon), and run on the Gallios 561 Analyzer (Beckman Coulter) within the University of Colorado AMC Cancer Center Flow Cytometry Shared Resource Core Facility. To identify IM populations, debris and doublets were first excluded using light scatter, followed by exclusion of RO-CD45+ (intravascular) cells. The RO- population was then segregated into Dump+ (CD3+, Ly6G+, B220+, and Zombie Violet+ cells) and Dump– populations. Macrophages were selected from the Dump– gate using CD64+ expression and subsequently segregated into IM (CD45+, CD64+, CD11b+, CD11c^low/int) and resident alveolar macrophages (AMs; CD45+, CD64+, CD11b–, CD11c^hi). IMs were then separated into IM1 (CD11c^lo+, MHCII^lo+), IM2 (CD11c^lo+, MHCII^hi+), and IM3 (CD11c^int+, MHCII^hi+) subsets (28). For the identification of monocytes, we selected for Dump– CD64^int, CD11b+ before assessing Ly6C expression to determine the Ly6C^low and Ly6C^hi populations. For the assessment of alveolar macrophages in a separate panel, total AMs were first identified as Dump– IT-CD45+ CD64+ and were then separated into resident (SiglecF+CD11b–) and recruited (SiglecF–Cd11b+) populations. All cell populations were gated using fluorescence minus one (FMO) controls, and absolute counts were calculated using 123 counting beads (Invitrogen, Waltham, MA). Data were analyzed using Kaluza flow analysis software, version 2.1 (Beckman Coulter, Brea, CA). For sorting and collection of IMs for RNA and metabolic assays, the same IM antibody panel was used without IM subset gating. Sorting was performed on the MoFlo XDP Flex Cell Sorter (70 µm chip; Beckman Coulter) for all RNA extractions and on the MA900 Multi-Application Cell Sorter (70 µm chip; Sony Biotechnology Inc.) for Seahorse metabolic assays. For RNA experiments, cells were sorted at 4°C into cold FA3 buffer, and for metabolic assays, cells were sorted into 70% FBS:30% FA3 buffer at room temperature. Collected IMs were pelleted (800 g, 15 min) for downstream applications.

Protein Extraction and Western Blotting

Lung tissue was homogenized using the Fisherbrand Bead Mill 24 homogenizer (Fisher Scientific; speed = 5, 30 s × 2 cycles) in 1× RIPA lysis and extraction buffer (ThermoFisher Scientific) prepared with phosphatase inhibitor cocktails 2 and 3 and protease inhibitor cocktail, each at 1:100 dilutions (Sigma). Cell debris was removed from all lysates by centrifugation (10,000 g, 5 min) and removal of pellets. All samples were prepared in 1× XT sample buffer with 1× XT Reducing Agent (Bio-Rad). For lung, protein concentration was determined using the Pierce Rapid Gold BCA Protein Assay Kit (ThermoFisher Scientific) and loaded at 30 µg per lane in Criterion XT 4–12% Bis-Tris Precast gels (Bio-Rad) for separation via gel electrophoresis using XT-MES running buffer (Bio-Rad). For serum, 4 µL of sample was loaded into equal volumes of sample buffer. Proteins were transferred to 0.2-µm polyvinylidene difluoride (PVDF) membranes (Bio-Rad) using the Trans-Blot Turbo transfer system (Bio-Rad). Membranes were activated in methanol and blocked with 5% nonfat dry milk powder in Tris-buffered saline with 0.1% Tween 20 (TBST) for at least 1 h. After blocking, membranes were cut below the 75 kDa molecular weight marker and incubated with primary antibodies against SOD3 (goat anti-mouse; R&D Systems; 1:800) or vinculin (rabbit anti-mouse; Cell Signaling; 1:1,000) in 5% milk-TBST, incubated overnight at 4°C. Membranes were then washed with TBST and incubated with horseradish peroxidase-conjugated anti-goat (R&D Systems; 1:5,000) or anti-rabbit (Abcam; 1:10,000) secondary antibodies for 1 h at room temperature. Finally, after washing with TBST, protein expression was detected using SuperSignal Femto Chemiluminescent substrates (ThermoFisher Scientific), and densitometry analysis was performed using ImageLab Software (Bio-Rad) for relative expression versus housekeeping protein (vinculin). Fold change normalization was performed relative to the average value for the WT normoxia group.

Seahorse Metabolic Assay

Fluorescence-activated cell sorting (FACS)-isolated IMs (4 mice pooled per sample) from WT and R213G mice exposed to normoxia or 4 or 14 days of hypoxia were resuspended in warmed phenol-red free Dulbecco’s modified Eagle medium (DMEM) containing 10 mM glucose, 1 mM pyruvate, and 2 mM glutamine, pH 7.4, and plated onto Seahorse XFe 96-well culture plates (Agilent) in duplicate or triplicate (depending on IM yield) at 50,000 cells/well in 180 µL. Plated cells were degassed (37°C non-CO₂ incubator for 1 h) prior to analysis in the Seahorse XFe instrument (Agilent) for measurement of baseline oxygen consumption and extracellular acidification, according to the manufacturer’s instructions. For normalization, CyQUANT direct cell proliferation assay (ThermoFisher Scientific) was performed immediately after Seahorse assay, and signal was detected using the iD5 Multi-Mode Microplate Reader (Molecular Devices; excitation 485/20, emission 428/20, Well scan setting) and SoftMax Pro Software. Data were normalized to live cell counts via CyQUANT analysis, and duplicate/triplicate wells were averaged for one n. Each n represents data from IMs isolated from four mice (2 male and 2 female) on a separate day.

RNA Extraction and Bulk RNA Sequencing

RNA was extracted from FACS-isolated IMs (2 female and 2 male mice pooled per sample) using the Qiagen RNeasy Micro Kit according to the manufacturer’s instructions. Only RNA samples with RNA integrity number (RIN) scores of 8 or higher wereused, and RNA concentrations were determined by Qubit. RNA QC and enrichment, mRNA library preparation, and sequencing were performed by Zymo Research Corp. In brief, mRNA-Seq libraries were constructed from 20 ng of total RNA. To enrich for poly(A) RNA molecules, an oligo(dT) primer with a partial P5 adapter was used. After that, reverse transcription was performed, followed by partial P7 adapter ligation and second-strand cDNA synthesis. Finally, libraries were amplified to incorporate full-length adapters. Successful library construction was confirmed with Agilent’s D1000 ScreenTape Assay on TapeStation. mRNA-Seq libraries were sequenced on an Illumina NovaSeq to a sequencing depth of at least 30 million read pairs (150 bp paired-end sequencing) per sample.

Bioinformatic Analysis

Zymo Research performed the bioinformatic analysis using the Zymo Research RNAseq pipeline to generate raw read counts. Briefly, quality control of raw reads was carried out using FastQC v0.11.9, Unique Molecular Identifiers (UMI) were extracted using UMI-tools v1.1.1, and adapter and low-quality sequences were trimmed from raw reads using Trim Galore! v0.6.6. Trimmed reads were then aligned to the Mus musculus reference genome (Genome Reference Consortium Mouse Build 38/mm10) using STAR v2.6.1d. BAM file filtering and indexing were carried out using SAMtools v1.9 and UMI-informed read deduplication using UMI-tools. RNAseq library quality control was implemented using RSeQC v4.0.0 and QualiMap v2.2.2-dev. Library complexity was estimated using Preseq v2.0.3. Reads overlapping with exons were assigned to genes using featureCounts v2.0.1.

Differential gene expression analysis was performed using DESeq2 v1.45.0 in R using design = (∼ genotype + condition + genotype:condition) with low count (<1) genes excluded and counts normalized using lfcShrink. A gene expression heat map was generated for the top 1,000 most variable genes across all samples using pheatmap v1.0.12. Pathway analysis was performed using Ingenuity Pathway Analysis (IPA; Qiagen), gene set enrichment analysis (GSEA), and Metascape. For IPA, core analysis was performed with a false discovery rate (FDR) cutoff of <0.05, and for comparison analysis, results were filtered for P < 0.05 with Z score set to ≥|2| or ≥|2|, as indicated. For GSEA, rnk files generated using DESeq2 were used, and significant pathways were filtered for FDR <0.05 and presented using the normalized enrichment score. For Metascape analysis, results tables for selected comparisons were used, and differentially expressed genes (DEGs, FDR < 0.05) were compared between groups [e.g., Nx vs. day 4 (D4) Hpx in WT compared with Nx vs. D4 Hpx in R231G) to generate Venn diagrams. Resultant gene lists were used as input, and pathway results were presented in the order of P value [–log10(P value)].

Statistical Analysis

Data are expressed as means ± SE. Data were analyzed by two-way ANOVA followed by Tukey’s multiple-comparisons testing using Prism 10.1.2 software (GraphPad Software), with significance defined as P < 0.05.

RESULTS

Interstitial Macrophage Accumulation in Response to Hypoxia Is Less Robust in R213G Mice

We first examined how the R213G genotype impacted IM accumulation in response to hypoxia. Consistent with prior studies (21, 23), in WT mice, total IMs significantly increased at 4 days of hypoxia and returned to baseline numbers at day 14 (Fig. 1B). Hypoxia also significantly increased total IMs at day 4 in R213G mice; however, although the R213G mice have been shown to have worsened chronic hypoxic PH, they did not have increased IM accumulation compared with WT mice at 4 days. In contrast, the R213G mice tended to have fewer total IMs at 4 days and significantly fewer IMs at day 14 (Fig. 1B). We then examined three IM subsets (IM1, IM2, and IM3; Fig. 1, C–E) described by Gibbings et al. (27). There were no baseline differences in the total number of IMs or any of the studied IM subsets in R213G mice. All three IM subsets transiently increased in the WT mice at 4 days, with the greatest increase observed for subsets IM1 and IM2. The accumulation of IMs in the R213G mice did not reach statistical significance for any of the three subsets, with only the IM2 subset tending to increase in R213G mice at day 4 (Fig. 1D; R213G Nx vs. R213G 4 D Hpx, P = 0.1). When comparing the response of the two strains, specific IM subsets were lower in the R213G mice, with reduced IM1 and IM2 subpopulations at day 4 and day 14, respectively (Fig. 1, C and D). To confirm no contribution of recruited alveolar macrophages to our IM populations, we included intratracheal CD45 antibody (IT-CD45) administration in a new cohort of mice using a separate flow cytometry panel to exclude IT-CD45-labeled alveolar macrophages and observed no effect of hypoxia or R213G SOD3 on recruited macrophages in the alveolar compartment (IT-CD45+Dump–CD64+ CD11b+ SiglecF–) (Fig. 2).

Figure 1.

Interstitial macrophage accumulation in response to hypoxia is less robust in mice expressing R213G SOD3. Flow cytometric analysis of lung interstitial macrophage (IM) numbers in wild-type (WT) and R213G SOD3 polymorphism (R213G) mice at baseline (normoxia; Nx) and following 4 or 14 days of hypoxia exposure (Hpx) is shown. Intravenous (IV) CD45 antibody administration was used to exclude intravascular cells as part of the “Dump” gate. A: gating strategy to separate IM (Dump–, CD64+, CD11b^hiCD11c^low/int) from resident alveolar macrophages (AMs; Dump–, CD64+, CD11b–, CD11c^hi) and identify IM1 (CD11c^lowMHCII^low), IM2 (CD11c^lowMHCII^hi), and IM3 (CD11c^intMHCII^hi) subsets. B–E: total IM (B), IM1 (C), IM2 (D), and IM3 (E) counts in the whole lung over the hypoxia time course. Two-way ANOVA, Tukey’s post hoc tests for selected comparisons; *P < 0.05, **P < 0.01, and ***P < 0.001. P values indicated where trends (P < 0.15) toward a significant difference were observed. n = 6–13 mice. All data are expressed as means ± SE.

Figure 2.

Monocyte-derived macrophages are not recruited to the alveolar compartment in response to hypoxia. Flow cytometric analysis of resident vs. recruited alveolar macrophage (AM) numbers in wild-type (WT) and R213G SOD3 polymorphism (R213G) mice at baseline (normoxia; Nx) and following 4 days of hypoxia exposure (Hpx) is shown. Intratracheal (IT) CD45 antibody administration was used to identify cells in the alveolar space. A: gating strategy to identify all AMs (Dump–, IT CD45+, CD64+) and separate resident (SiglecF+, CD11b–) and recruited (SiglecF–, CD11b+) AMs. B–D: total AM (B), resident AM (C), and recruited AM (D) counts in the whole lung. Two-way ANOVA = ns. n = 7 or 8 mice. All data are expressed as means ± SE.

Circulating and Lung Monocytes Are Unaffected by R213G SOD3

Pulmonary IMs can arise from monocytes recruited from the circulation (20, 24, 25, 29, 30), and consistent with our prior report (13), we observed that R213G mice have elevated serum but low lung expression of SOD3 (Supplemental Fig. S1). Therefore, we measured blood and lung monocyte populations following hypoxia exposure to investigate whether changes in local SOD3 content due to the R213G variant would impact the recruitment of monocytes. As determined by both flow cytometry (Fig. 3) and complete blood cell counts (Supplemental Fig. S2), there were no significant differences in circulating monocyte populations in R213G compared with WT mice. There was, however, a small but significant effect of hypoxia (2-way ANOVA, hypoxia effect P < 0.05) on monocyte numbers across the lung and vascular compartments in both strains of mice. We observed a decrease in circulating (Fig. 3B) and intravascular (IV-CD45+) lung total and Ly6C^hi monocytes on day 4 and a subsequent rebound increase on day 14 (Fig. 3E). The reverse patterns were observed for total and Ly6C^low monocyte populations in the IV- lung compartment, with a significant decrease in total and Ly6C^low monocytes between days 4 and 14 in R213G mice (Fig. 3, G–I).

Figure 3.

Hypoxia-driven changes in circulating and migrating monocyte numbers are not affected by R213G SOD3 variant. Flow cytometric analysis of blood and lung monocyte numbers in wild-type (WT) and R213G SOD3 polymorphism (R213G) mice at baseline (normoxia; Nx) and following 4 or 14 days of hypoxia exposure (Hpx) is shown. Intravenous (IV) CD45 antibody administration was used to identify intravascular cells within the lung. A–C: total (A), Ly6C^hi (B), and Ly6C^low (C) circulating monocyte numbers per 1 mL of blood. D–F: total (D), Ly6C^hi (E), and Ly6C^low IV+ (F) monocyte numbers in the whole lung. G–I: total (G), Ly6C^hi (H), and Ly6C^low IV- (I) monocyte numbers in the whole lung. Two-way ANOVA; #P < 0.05 and ##P < 0.01 for hypoxia effect; Tukey’s post hoc tests for selected comparisons, *P < 0.05. P values indicated where trends (P < 0.15) toward a significant difference were observed. n = 7–13 mice. All data are expressed as means ± SE.

IM RNAseq Reveals Strong Hypoxia Response and Genotype Separation at Day 14

After demonstrating that expression of the R213G SOD3 variant did not exacerbate the accumulation of IMs in response to hypoxia, we proposed that it may instead alter IM transcriptional responses. To investigate this, we performed next-generation RNAseq on FACS-isolated IMs (total IM population) from each genotype at baseline (Nx) compared with 4 and 14 days of hypoxia (Hpx) exposure (Figs. 4, 5, and 6). Principal component analysis (PCA) and the gene expression heatmap demonstrate strong separation of normoxia groups from hypoxia groups and genotype separation in hypoxia, particularly between the day 14 groups (Fig. 4, A and B). Hypoxia-driven differences on the PC1 axis explain the most variance (35%), whereas PC2 (19% variance) strongly separates the two genotypes at the day 14 timepoint.

Figure 4.

Overview of gene regulation in response to hypoxia in interstitial macrophages (IMs) from WT and R213G mice. Bulk-RNAseq data were generated from flow-sorted IMs at each timepoint from each genotype and analyzed using DESeq2 in R. Differential gene expression analysis was performed using the design “genotype + condition + genotype: condition,” and main (WT) effects of hypoxia (i.e., condition) as well as hypoxia effects in R213G mice were determined and filtered for FDR < 0.05. A: principal component analysis (PCA) of IMs collected at baseline (Nx; circles) and after 4 (triangles) and 14 (squares) days of hypoxia exposure from WT (blue) and R213G (red) mice. Each shape represents data generated from IMs collected from four mice. B: heatmap depicting sample clustering based on the top 1,000 most variable genes in the RNAseq dataset. C: Venn diagram showing the total number of unique and overlapping differentially expressed genes (DEGs) regulated in response to 4-day (4D) hypoxia in each genotype. D: Venn diagram showing the total number of unique and overlapping DEGs regulated in response to 14-day (14D) hypoxia in each genotype. All data were analyzed from n = 3 or 4 IM samples per experimental group, with each n representing IMs collected from four mice (2 male and 2 female). FDR, false discovery rate; WT, wild-type.

Figure 5.

Pathway analysis identifies differences between WT and R213G IM phenotypic reprogramming in response to 4 or 14 days of hypoxia exposure. Pathway analysis was performed on the RNAseq data using gene set enrichment analysis (GSEA) and Ingenuity Pathway Analysis (IPA; Qiagen) software. Pathways and regulators related to metabolism that were upregulated in WT IMs are highlighted with blue boxes, whereas those related to macrophage activation and extracellular matrix interactions were enhanced in R213G and are highlighted with red boxes. Pathways related to regulation of translation were differentially regulated in R213G vs. WT at day 14 and are highlighted in green. A–D: GSEA was used to determine significantly upregulated and downregulated canonical pathways (m2.cp.v2023.Mm) for each comparison (FDR < 0.05) from preranked gene lists generated using DESeq2. Bar graphs depict the top 20 regulated pathways ordered by significance and depicting the normalized enrichment score for WT 4-day (4D) Hx vs. Nx IMs (A), R213G 4D Hx vs. Nx IMs (B), WT 14-day (14D) Hx vs. Nx IMs (C), and R213G 14D Hx vs. Nx IMs (D). E and F: core IPA was performed on the entire gene list for all hypoxia groups compared with normoxia, with a FDR cutoff of 0.05. Comparison analysis was then performed for 4D hypoxia (E) and 14D hypoxia (F), and the upstream regulators were filtered for significance (P < 0.05) and an absolute Z score of ≥3 in at least one group. Upstream regulators that were the same in both genotypes were excluded from the figure but are presented in the Supplemental Material. Color intensity indicates Z score level, with blue depicting inhibition and orange depicting activation. Gray dots indicate a nonsignificant Z score (Z < 2). All data were analyzed from n = 3 or 4 IM samples per experimental group, with each n representing IMs collected from four mice (2 male and 2 female). FDR, false discovery rate; IM, interstitial macrophage; WT, wild-type.

Figure 6.

Blunted anti-inflammatory response in R213G IMs between 4 and 14 days of hypoxia exposure. A: Venn diagram showing the total number of unique and overlapping differentially expressed genes (DEGs) regulated in each genotype between 14 days (14D) and 4 days (4D) of hypoxia exposure. B and C: core Ingenuity Pathway Analysis (IPA; Qiagen) was performed on the entire gene list for each genotype’s 14D vs. 4D comparison with a FDR cutoff of 0.1. Comparison analysis was then performed and the top canonical pathways (B) and upstream regulators (C) were filtered for significance (P < 0.05). Results were filtered for an absolute Z score of ≥2 or ≥3 in at least one group for B and C, respectively. Color intensity indicates Z score level, with blue depicting inhibition and orange depicting activation. Gray dots indicate a nonsignificant Z score (Z < 2). Pathways and regulators related to metabolism that were regulated in WT IMs are highlighted with blue boxes, whereas those relating to the regulation of inflammation that were unchanged or differentially regulated in R213G are highlighted with red boxes. Pathways and regulators related to regulation of translation are highlighted in green. All data were analyzed from n = 3 or 4 IM samples per experimental group, with each n representing IMs collected from four mice (2 male and 2 female). D: diagram depicting potential interacting pathways and regulators that were differentially expressed at day 14 in WT (blue arrows) vs. R213G (red arrows) IMs highlighting potential redox-sensitive targets. FDR, false discovery rate; IMs, interstitial macrophages; WT, wild-type.

We next examined differential gene expression between experimental groups using DESeq2 in R. We found no significant differences in gene expression between the two genotypes in normoxia (see data repository) but substantial differences in their responses to hypoxia. To illustrate the extent and genotype differences in gene regulation in response to hypoxia (D4 vs. NMX and D14 vs. NMX), Venn diagrams were generated from differentially expressed gene (DEG; FDR < 0.05) lists. Most strikingly, R213G IMs exhibited a stronger transcriptional response to 4 days of hypoxia with over threefold more DEGs compared with WT mice (Fig. 4C; 211 in WT vs. 748 in R213G). Consistent with the PCA plot, although each genotype had a similar number of DEGs at day 14 compared with normoxia, the majority of these were uniquely regulated in each genotype, with >500 unique genes for each genotype and almost 300 overlapping (Fig. 4E). This suggests that the IM phenotypes in WT and R213G mice at day 14 are quite distinct.

Metabolic Remodeling Response to Hypoxia Observed in IMs From WT but Not R213G Mice

We used three different methods of enrichment analysis to identify specific pathways that could differentiate the R213G from WT IMs in hypoxia (IPA, gene set enrichment analysis of ranked gene lists, and Metascape analysis). Multiple pathways were regulated in response to hypoxia in both genotypes, including complement and platelet activation, ECM remodeling, and regulation of immune responses (Fig. 5 and Supplemental Figs. S3 and S4). There were certain pathways, however, that predominated in specific genotypes. In IMs isolated from WT mice, we observed a strong metabolic response to hypoxia, consistent with prior studies in hypoxia (22, 26). Remarkably, however, activation of metabolic pathways was not observed in IMs from mice expressing the R213G variant of SOD3. Activation of multiple metabolic pathways, including mitochondrial dysfunction, oxidative phosphorylation, electron transport chain, ATP formation, and NADH generation, was observed in response to hypoxia in WT IMs, particularly at day 4, but was absent from the R213G analysis (Fig. 5 and Supplemental Figs. S3 and S4, highlighted with blue boxes). Glycolysis was also significantly regulated in WT IMs at day 4 (Supplemental Fig. S6) but not in the IMs from the R213G mice. Furthermore, a comparison of day 14 versus day 4 IMs revealed subsequent inhibition of glycolysis in IMs from WT but not from R213G mice (Fig. 6B).

Although many upstream regulators (URs) identified in the response to hypoxia by IPA were similar, there was a subset of URs that were either differentially regulated or unique to one strain of mice. Relevant to metabolic differences, MLX Interacting Protein Like (MLXIPL), also known as carbohydrate response element binding protein (ChREBP), and rapamycin-insensitive companion of mammalian target of rapamycin (RICTOR) stood out as being differentially regulated. At day 14, strong activation of MLXIPL (Z score = 3.2) and inhibition of RICTOR (Z score = –3.7) were evident in WT IMs compared with MLXIPL inhibition (Z score = –3.7) and activation of RICTOR (Z score = 3.3) between days 4 and 14 in IMs from R213G mice (Figs. 5F and 6C). Further investigation of the UR analysis revealed several additional metabolic regulators that were activated in hypoxia in WT but not in R213G IMs. This included mitochondrial uncoupling protein 1 (UCP-1) and phosphoinositide 3-kinase (PI3K) complex at day 4 and phosphatase and tensin homolog deleted on chromosome 10 (PTEN), AMP-activated protein kinase (AMPK), and sirtuin 1 (SIRT1) at day 14 (P < 0.05, Z ≥ |2|; Supplemental Figs. S7 and S8).

To validate our gene-level observations of differences in metabolic remodeling in IMs across the time course of hypoxia, we performed bioenergetic assays using the Agilent Seahorse Bioanalyzer to measure baseline IM metabolism. Consistent with the RNAseq analysis, the extracellular acidification rate (ECAR), a proxy for glycolysis, was significantly increased in WT IMs after 4 days of hypoxia and returned to baseline by day 14 (Fig. 7A). Importantly, this response did not occur in IMs isolated from R213G mice. Despite our observations of gene regulation of oxidative phosphorylation and ETC pathways in WT, we did not see any alterations in baseline oxygen consumption rates (OCR) (Fig. 7B).

Figure 7.

Hypoxic regulation of glycolysis in WT but not R213G IMs. Seahorse assay was performed on FACS-isolated IMs to determine baseline extracellular acidification rate (A; ECAR), a proxy for glycolysis, and oxidative phosphorylation (B; oxygen consumption rates). Results were normalized to live cell counts using CyQUANT direct cell stain following each Seahorse assay. Two-way ANOVA, *P < 0.05, Tukey’s post hoc tests for selected comparisons. All data are expressed as means ± SE; n = 4 or 5 IM samples per experimental group, with each n representing IMs collected from four mice (2 male and 2 female). FACS, fluorescence-activated cell sorting; IMs, interstitial macrophages; WT, wild-type.

Enhanced Activation of Inflammation, Adhesion, and ECM Interaction Pathways in IMs From R213G Mice in Hypoxia

In contrast to metabolic pathways regulated in WT IMs, many pathways were more strongly or uniquely activated in R213G IMs in response to hypoxia. When comparing the top 20 pathways by GSEA, these pathways were predominantly related to chemotaxis and ECM interactions (Fig. 5, highlighted with red boxes). IPA canonical pathway analysis also showed distinct activation of cell cycle control in R213G, supported by Metascape analysis that suggested this was to downregulate proliferation (Supplemental Fig. S3). Enrichment of several migration and adhesion pathways, as well as alternative macrophage activation, was also unique to R213G IMs in hypoxia (Supplemental Fig. S3). Although this “M2” phenotype can be anti-inflammatory in some contexts, the inflammatory response was also enriched in R213G IMs (Supplemental Fig. S3), and alternatively activated macrophages and the M2 stimuli IL-4 and IL-13 are implicated in PH (31–33). A proinflammatory phenotype is also supported by the UR analysis that showed robust activation of colony stimulating factor 2 (CSF2), signal transducer and activator of transcription 6 (STAT6), C-C motif chemokine receptor 2 (CCR2), and tumor necrosis factor ligand superfamily member 12 (TNFSF12), as well as activation (P < 0.05, Z ≥ |2|) of IL-4, IL-13, and transforming growth factor-β1 (TGF-β1) (Fig. 5E and Supplemental Fig. S7). Overall, these data suggest a proinflammatory, profibrotic macrophage phenotype in R213G mice in hypoxia at both day 4 and day 14.

IMs from R213G Mice Have Attenuated Resolution of Inflammation at Day 14

In WT mice, IMs shift to a less inflammatory, more reparative phenotype after 14 days of hypoxia (22). Since R213G mice develop worse PH (13) and local SOD3 can promote anti-inflammatory signaling (15–17, 19), we next investigated whether the redistribution of SOD3 in R213G mice leads to a blunted anti-inflammatory IM response at day 14 when compared with the phenotype at day 4. Supporting previous studies, we demonstrate a robust inhibition of inflammatory pathways in WT IMs at day 14 (Figs. 5 and 6 and Supplemental Figs. S3 and S4). In general, we observed a decreased transcriptional response in R213G IMs compared with WT in this analysis with 400 fewer DEGs in the IMs from R213G mice (574 in WT D14 vs. D4 and 174 in R213G D14 vs. D4; Fig. 6A). This suggests that in R213G mice, IMs do not change as substantially at this later timepoint, sustaining much of their early hypoxic response. Indeed, pathway analysis revealed less robust inhibition of inflammatory pathways compared with WT. Between days 4 and 14, we observed inhibition of IL-3, C-X-C chemokine receptor 4 (CXCR4), and NO/reactive oxygen species (ROS) production pathways in IMs from WT mice, that were all activated in R213G mice. In addition, hypoxia-inducible factor 1α (HIF1α), thrombin (trend for activation in R213G Z score = 1.9), interleukin (IL-1), inducible nitric oxide synthase (iNOS), integrin, high mobility group box 1 (HMGB1), epidermal growth factor (EGF), IL-7, and IL-15 signaling were all inhibited in IMs from WT mice but unchanged in R213G mice (Fig. 6A, highlighted with red boxes). Finally, immune regulation through inhibition of IL-4 and interferon regulatory factor 5 (IRF5) and activation of suppressor of cytokine signaling 1 (SOCS1) and prostaglandin E₂ receptor 4 (PTGER4) was observed in IMs from WT but not from R213G mice at day 14. Of note, PTGER4 has been shown to suppress proinflammatory myocyte enhancer factor 2 (MEF2A) via activation of protein kinase A (PKA) signaling (34)and MEF2A inhibition alongside PKA activation was observed in day 14 WT IMs (Fig. 6 and Supplemental Fig. S8).

Interestingly, we also observed differential regulation of pathways related to translation (cytoplasmic ribosomal units, formation of a pool of 40S subunits, nonsense-mediated decay, translation) in R213G (inhibition) compared with WT (activation) IMs at day 14 (Fig. 5, highlighted with green boxes). This was consistent with their differential regulation of eukaryotic initiation factor 2 (EIF2) signaling and the inhibition of MYC observed in day 14 IMs from R213G mice (Fig. 6).

DISCUSSION

Despite recent advances in the field, there remains a critical need to better understand the pathophysiology of PH to support the development of novel treatment strategies and improve clinical outcomes. Our laboratory and others have convincingly established a protective role for the extracellular redox enzyme SOD3 in this progressive and ultimately fatal disease (2–13). Low SOD3 promotes inflammation, and there is strong evidence that the accumulation and reprogramming of interstitial macrophages contributes to vascular remodeling and PH (20–23). It was unknown, however, how alterations in SOD3 impact macrophage accumulation or programming. To address this question, we began by considering a human R213G SOD3 polymorphism known to increase the risk of human cardiovascular disease (14) and worsen chronic hypoxic PH in mice expressing the R213G variant (13). This SOD3 variant lowers tissue binding affinity, thus redistributing SOD3 from tissue ECM into the bronchoalveolar space and circulation without changing enzyme activity. We hypothesized that the R213G SNP in SOD3 exacerbates pulmonary IM accumulation and pathogenic IM reprogramming in response to hypoxia. We show that although R213G SOD3 does not augment the previously described early IM accumulation in response to hypoxia, it does lead to important differences in the transcriptional reprogramming of these cells at early (day 4) and intermediate (day 14) timepoints. Most strikingly, IMs from R213G mice exposed to hypoxia demonstrate 1) exacerbated inflammatory responses, including alternative macrophage activation; 2) limited inflammatory resolution at day 14; and 3) a lack of metabolic remodeling.

Our first key observation was that although hypoxia increased IMs in both WT and R213G mice at day 4, this hypoxia-induced IM accumulation was less robust in R213G mice. Given the important role IMs play in PH development and that R213G mice develop worse disease, this was contrary to our initial hypothesis. This emphasized to us that the phenotype of these cells and reprogramming in response to hypoxic stress may be a more crucial determinant of disease progression than simply the number recruited. It is now recognized that IMs do not represent a single uniform population but can be subdivided into several distinct phenotypes based on surface markers, ontogeny, and transcriptional programs (26, 28, 35–38). We used one such method of separating IM subpopulations based on differential cell surface expression of CD11c and MHCII (28) and found less IM1s (CD11c^–MHCII^low) at day 4 and less IM2s (CD11c^–MHCII^hi) at day 14 in R213G mice when compared with WT. Although their roles in PH remain uncertain, IM1 and IM2 appear closely related, only distinguished by their expression of MHCII and antigen presentation genes and, during homeostasis, both have higher cell surface CD206, CD169, Lyve1, and “reparative” gene expression compared with IM3 (CD11c⁺MHCII^hi), which expresses higher levels of cell surface CCR2 (28). Altered proportions of IM subsets in R213G mice prompted us to next examine the IM response to hypoxia more deeply using RNAseq, where we observed important strain differences in hypoxia-induced reprogramming.

Our analysis of WT IM reprogramming showed significant regulation of metabolic pathways. Key changes included early activation and subsequent inhibition of HIF1α and inflammation, robust activation of EIF2, and inhibition of RICTOR at day 14, all in agreement with prior studies of this model (22, 26, 39). Some differences in our pathway analysis results for WT mice as compared with these studies included less regulation of Rho pathways but enhanced activation of complement, platelet, and ECM-related pathways. These subtle differences could be due to both differences in baseline controls (Denver altitude in our study vs. sea level) and analysis methods (comparison with alveolar macrophages vs. R213G genotype, singe-cell- vs. bulk-RNAseq). Importantly, macrophage interactions with complement (39), platelets (40), and ECM (41–43) are implicated in PH, suggesting that this reprogramming could contribute to the pathogenesis of this disease.

The R213G IM RNAseq analysis demonstrated upregulation of multiple leukocyte migration and adhesion pathways that were not identified in WT. In addition, a subset of pathways related to ECM interactions were uniquely activated in R213G IMs in hypoxia, including degradation of ECM, collagen degradation, and collagen formation at day 4 and hyaluronan metabolism at day 14. These observations were associated with unique activation of alternatively activated macrophage (AAM) signaling via STAT6, IL-4, IL-13, and TGF-β1 URs, combined with typical proinflammatory activation including CSF2, CCR2, and TNF. Interestingly, although the chronic hypoxia model is not typically associated with type 2 inflammation, this is consistent with existing human data demonstrating increased IL-4 and IL-13 in BALF of patients with PH and in additional animal models of pulmonary fibrosis and PH (44, 45). Th2 and AAM-driven responses are well-established contributors to Schistosoma-induced pulmonary vascular remodeling and PH (31, 44–46). Furthermore, macrophage TGF-β1 signaling leads to the production of thrombospondin-1 to promote chronic hypoxic PH (24, 46). Overall, our RNAseq data from R213G IMs in hypoxia are indicative of a proinflammatory/profibrotic IM phenotype that is implicated in PH (33). We therefore suggest that the loss of SOD3 from tissue ECM (including within the pulmonary vasculature) in R213G mice exposed to hypoxia leads to IM reprogramming that exacerbates vascular remodeling and PH through enhanced ECM remodeling and inflammation.

In addition to differences in macrophage activation, another striking strain difference in hypoxia-induced transcriptional reprogramming of IMs was robust metabolic remodeling in WT IMs that was absent in R213G IMs and correlated with less prominent anti-inflammatory signaling in R213G at day 14. HIF1α signaling plays a central role in regulating macrophage metabolic and inflammatory pathways in response to hypoxia, including upregulation of glycolysis (47, 48), and contributes to pulmonary vascular remodeling in this model (49, 50). Importantly, HIF1α signaling was inhibited as part of the resolving response in WT, but not in R213G, and coincided with the inhibition of proinflammation and glycolytic pathways. We confirmed this pattern of glycolytic regulation in WT IMs using bioenergetic analysis and observed no functional metabolic changes in R213G IMs in response to hypoxia. This suggests that the metabolic switch that occurs between days 4 and 14 in WT IMs may be required for the resolution of the hypoxia response, which was lacking in R213G IMs. The lack of increased glycolysis in R213G IMs at day 4 could also reflect their profibrotic AAM phenotype that is less reliant on glycolysis than a typical proinflammatory macrophage. Differential regulation of translation factor EIF2 at day 14 (activated in WT but inhibited in R213G) was also striking. Given that EIF2 can inhibit NFκB and proinflammatory signaling (51), these observations may also be linked. Overall, our data suggest that R213G IMs do not respond metabolically and do not recover as effectively to hypoxic stress as WT IMs, with sustained activation in longer term hypoxia.

Finally, we identified a group of interacting URs involved in inflammation and metabolism that were differentially regulated in the day 14 hypoxia response in R213G compared with WT IMs. mTOR signaling, in agreement with that reported by Pugliese et al. (22), appeared to be a central regulator of the hypoxia response. Our analysis showed that WT IMs activated MLXIPL, a factor that can regulate mTOR (52–55), and inhibited RICTOR, a component of the mTORC2 complex, at day 14. By contrast, R213G IMs activate RICTOR and inhibit MLXIPL between days 4 and 14. Macrophage MLXIPL is required for metabolic reprogramming and suppresses inflammatory responses (56), suggesting that it could also be central to the day 14 resolution and glycolysis regulation in WT IMs. Macrophage mTOR signaling can be activated by a variety of stimuli, including IL-4 and TLR agonists (57), and plays a complex role in macrophage function. Central to cross talk between metabolism and inflammation, mTOR complex 1 (mTORC1) is considered predominantly proinflammatory and upregulates glycolysis through HIF1α (58, 59). Furthermore, mTORC1 is negatively regulated by AMPK, PKA, PTEN, and SIRT1 (60–64) that were all activated in WT, but not R213G, IMs at day 14. Many of these regulators can be influenced by the redox environment, making the role of SOD3 in their regulation plausible. Redox-dependent regulation of mTOR, HIF1α, NfκB, and EIF2 is well established (65–67), while SOD3-dependent AMPK activation has been demonstrated in several studies (68–70). A summary of the potential interactions between these regulatory molecules, redox-sensitive targets, and their differing responses at day 14 in WT versus R213G IMs is depicted in Fig. 6D.

There are some important limitations and knowledge gaps that should be highlighted. Our selected time course and use of RNAseq on the whole IM population could miss important differences in immune cell recruitment, subpopulations, and reprogramming responses. We also assume a predominantly perivascular accumulation of IMs based on prior reports (22, 23, 30) but did not investigate the localization of these cells, and, despite our focus on IMs in this study, the R213G SNP is likely to affect the function of many cell types that could drive vessel remodeling and PH. Excitingly, new methodologies for spatial transcriptomic studies at a single-cell level will allow the determination and analysis of the location of recruited IMs and nearby cell types and the identification of important cell-cell interactions that are driving metabolic and phenotypic differences in R213G IMs in hypoxia. Although our bioenergetic analysis of isolated IMs demonstrated important strain differences in extracellular acidification, we did not detect changes in oxidative phosphorylation despite the robust enrichment of these pathways in WT IMs after 4 days of hypoxia. It is important to note, however, that OCR is very difficult to quantify in flow-sorted cells that are no longer responsive to electron transport chain inhibitors and uncouplers, and therefore, we may have missed some critical changes in metabolic function that were present in vivo. Finally, much of our conclusions are based on gene expression data, and therefore, studies that assess redox modifications using redox proteomics will be important to verify protein-level changes and to determine targetable pathways. To confirm direct effects of altered IM reprogramming on PH outcomes in R213G mice, future investigations should target these identified pathways in macrophages.

In conclusion, we show that the R213G polymorphism of the vascular antioxidant enzyme SOD3, associated with worsened PH, does not exacerbate IM accumulation but alters IM transcriptional reprogramming in response to hypoxia exposure. This reprogramming enhances proinflammatory and profibrotic signaling that fails to resolve. Furthermore, R213G SOD3 prevents the hypoxic metabolic remodeling that occurs in WT macrophages in this model. Taken together, we propose that the redistribution of SOD3 expression and altered local redox environment due to the R213G SNP leads to a defect in IM metabolic and immune regulation in response to hypoxic stress that exacerbates processes involved in vascular remodeling and sustains proinflammatory responses to worsen PH outcomes.

DATA AVAILABILITY

All raw and processed RNAsequencing data files have been uploaded to the NCBI Gene Expression Omnibus (Accession No. GSE269892). Additional data files will be made available from the authors upon request.

SUPPLEMENTAL MATERIAL

Supplemental Figs. 1–8: https://doi.org/10.6084/m9.figshare.26612233.v1.

GRANTS

This work was supported by American Heart Association Grant CDA 23CDA1045594 (to C.V.L.), National Institutes of Health (NIH) Grant R35HL139726 (to E.S.N.), NIH Grant K12HD047349 (to C.S.), and NIH Grant P01HL152961-01 (to C.D.). This work was supported in part by NIH Grant P30CA046934-funded University of Colorado Cancer Center Flow Cytometry (RRID:SCR_022035) and Genomics (RRID: SCR_021987) shared resources.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

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