Nonclassical Monocytes Sense Hypoxia, Regulate Pulmonary Vascular Remodeling, and Promote Pulmonary Hypertension

Pulmonary arterial hypertension (PAH) is a heterogeneous group of diseases characterized by abnormal remodeling and, ultimately, ablation of pulmonary intralobar arteries. It is thought that in predisposed individuals, environmental and genetic factors promote endothelial cell dysfunction or damage and the recruitment and proliferation of vascular smooth muscle cells and pericytes, which leads to intimal, medial, and adventitial hypertrophy, muscularization of small arteries, and an eventual increase in pulmonary vascular resistance and pressure. The exact cause of this pulmonary vascular remodeling remains unknown, but it has traditionally been viewed as a disease of endothelial and vascular smooth muscle cells and the signaling pathways that regulate these cell types (1). However, more recently, a body of evidence has emerged suggesting that bone marrow–derived myeloid cells play an important role in PAH pathophysiology (2–4).

It has long been recognized that the development of pulmonary hypertension (PH) is associated with the infiltration of myeloid cells into the pulmonary vasculature, although the functional significance of this has not been clear (5–7). A causative role for myeloid cells in PH has been suggested by studies in which manipulations that alter myeloid cell numbers also regulate the severity of PH (8–13). It has even been suggested that PAH involves a primary bone marrow defect, as patients with PAH and their unaffected family members display abnormalities in the myeloid lineage (3, 14). This may be related to mutations in bone morphogenetic protein receptor type 2 (BMPR2), which are present in 70% of familial PAH cases and 10–25% of idiopathic PAH cases (15). In humans with PAH, BMPR2 mutations are associated with increased production of the monocyte chemoattractant GM-CSF and increased infiltration of GM-CSF receptor–expressing macrophages (MØs) in pulmonary arteries (16). In mouse models, GM-CSF infusion increased and GM-CSF blockade decreased myeloid cell filtration, small pulmonary artery muscularization, and the development of PH (16). Moreover, bone marrow transplantation with BMPR2 mutant cells caused PH in mice, whereas wild-type cells were protective in BMPR2 mutant mice (17). Together, these studies strongly suggest that myeloid cells play a causative role in the development of PH; however, a direct role for a specific disease-inciting myeloid cell type has never been demonstrated.

MØs display significant diversity and plasticity, and the functionality of specific MØ subtypes is determined by both their origins and location (18–20). Tissue MØs may arise from one of three sources: fetal MØs, CX3CR1^loLy6C^hi classical monocytes, and CX3CR1^hiLy6C^lo nonclassical monocytes (21, 22). Fetal MØs arise in the yolk sac or fetal liver and seed tissues during development. In the lung, these become resident pulmonary MØs, which, under homeostatic conditions, are maintained via proliferation through the entire lifespan of the animal (21, 23). Classical monocytes are bone marrow derived, are rapidly recruited to sites of inflammation via the activity of the chemokine receptor CCR2, and mediate acute inflammatory and fibrotic responses (24). Nonclassical monocytes are also bone marrow derived, patrol the microvasculature by crawling on the endothelium, infiltrate tissues in response to inflammatory stimuli, seemingly act as inflammation-initiating sentinel cells, and regulate the resolution of inflammation, endothelial health, and tissue repair (25–27). Previously, we demonstrated that PH is increased in CCR2-deficient mice, strongly suggesting that classical monocytes do not directly contribute to the development of PH (28, 29). In this study, using chronic hypoxia models of PH, we identify specific myeloid cell types that infiltrate small pulmonary arteries and, by sensing of hypoxia, contribute to the development of PH.

Male CD45.2 C57BL/6 mice at 6 wk of age were purchased from Charles River Laboratories (Morrisville, NC). CD45.1 C57BL/6, Cx3cr1^gfp/gfp, and Hif1α^flox/flox were purchased from Jackson Laboratories (Bar Harbor, ME). Cx3cr1-cre–transgenic animals were developed and characterized in the laboratory of Michael Gunn (30). All mice were housed in a barrier and specific pathogen-free facility at Duke University School of Medicine (Durham, NC). All procedures were approved by the Institutional Animal Care and Use Committee at Duke University. For hypoxia exposure, animals were placed at 18,000 feet altitude in an environmentally controlled hypobaric chamber. The duration of exposure is specified in each experiment. For hypoxia plus SU5416 (Sigma-Aldrich, St. Louis, MO) exposures, animals were treated with a weekly dose of s.c. 20 mg/mg SU5416 dissolved in 5% ethanol, 5% DSMO in carboxymethyl cellulose.

Human lung samples were obtained from two sources. Control human lung tissues were obtained from organs declined for transplantation. All control lungs were from subjects without a history of smoking or chronic lung disease obtained from the University of North Carolina Marsico Lung Institute/Cystic Fibrosis Center Tissue Procurement and Cell Culture Core (under an approved protocol at University of North Carolina at Chapel Hill). Lung explants from patients with Group I PAH were obtained at the time of lung transplant at Duke University under a Duke University Institutional Review Board approved protocol. Fresh lung tissues were processed and digested into single-cell suspensions and prepared for multiparameter flow cytometry as described (31).

Cardiac function evaluations were performed in mice as described (28). Briefly, animals were anesthetized, intubated, and ventilated. The right jugular vein was cannulated with a fluid-filled PE-10 tube. Right ventricular (RV) pressure, heart rate, and respiratory rate were recorded using PowerLab (RRID:SCR_001620; ADInstruments, Colorado Springs, CO).

CD45.2 male animals were treated i.p. with 25 mg/kg of busulfan (Sigma-Aldrich) dissolved in a 30% DMSO/PBS solution at 72 and 48 h prior to transferring bone marrow cells. Bone marrow cells (1 × 10⁷) derived from 6- to 8-wk-old CD45.1 male animals were injected via the retro-orbital sinus into CD45.2 male recipients. Animals were maintained on sulfamethoxazole/trimethoprim supplemented water for 14 d. To assess chimerism, blood samples were collected 3 wk after cell transfer and analyzed by flow cytometry as described (32).

Phenotyping of murine and human pulmonary immune repertoires were performed using multiparameter flow cytometry and a panel of Abs previously described (32). Briefly, tissues were dissociated into single-cell suspensions and stained with a panel of Abs. Data were acquired using an LSRII flow cytometer (RRID:SCR_002159; BD Bioscience, San Jose, CA). Data were analyzed using FlowJo X (RRID:SCR_008520; BD Bioscience). Absolute numbers of each immune cell type were calculated by multiplying the total cell number recovered from each lung digest by the proportion of a particular cell type as percent of live single cells.

Murine lung tissues were fixed in 4% paraformaldehyde (Sigma-Aldrich) and washed with PBS solution. Fixed tissues were embedded in optimal cutting temperature compound. Frozen tissue sections of 6–8 μm were prepared. Immunofluorescence staining was performed using rat anti-mouse CD64 (AT152-9, RRID:AB_2687456; Bio-Rad, Hercules, CA), chicken anti-GFP (GFP-1020; Aves Labs, Tigard, OR), and rat anti-mouse CD169 (3D6, RRID:AB_10915134; BioLegend, San Diego, CA). DAPI was used for nuclear staining. CD64 and CD169 staining was performed using tyramide amplification (PerkinElmer Tyramide Plus, Waltham, MA). Small pulmonary arteries of similar sizes were selected blindly across treatment or animal groups for imaging. Confocal images were obtained with a Zeiss 710 Inverted Confocal Microscope using 20× objective (Zeiss, Cambridge, U.K.). Images were converted to JPEG file format using ImageJ (RRID:SCR_003070; National Institutes of Health).

Analyses of mRNA levels were performed using real-time RT-PCR. Total RNA was isolated from the specified immune cell subpopulations purified from the lung using a BD Aria II Sorter. Total RNA was isolated using RNAeasy Plus Micro Kit (Qiagen, Germantown, MD). Subsequently, cDNA was synthesized using QuantiTect Reverse-Transcription Kit (Qiagen). Amplification was performed using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Carlsbad, CA) on a QuantStudio Real-Time PCR System (Thermo Fisher Scientific). Primers for Hif1α are as follows: forward primer (5′-TCATCAGTTGCCACTTCCCCAC-3′) and reverse primer (5′-CCGTCATCTGTTAGCACCATCAC-3′). Gene expression was normalized to GAPDH. Using GAPDH, the primers were as follows: forward primer (5′-AGGTCGGTGTGAACGGATTTG-3′) and reverse primer (5′-TGTAGACCATGTAGTTGAGGTCA-3′). Changes in gene expression between Hif1α-sufficient and Hif1α-deficient cells were quantified using δ-δ-Ct calculation.

Seven- to eight-week-old C57BL/6 males were injected with 5 μM of PHK26PCL in Diluent B (Sigma-Aldrich) via the retro-orbital sinus. Animals were also treated with 50 mg/kg of 5-ethynyl-2’-deoxyuridine (Edu) weekly for 3 wk by i.p. injection. Lungs were harvested and prepared as single-cell suspensions, as previously described (32). Cellular proliferation was detected using Click-iT Plus EdU Proliferation Kit for flow cytometry (Thermo Fisher Scientific). Data were collected using a BD LSRII (RRID:SCR_002159; BD Bioscience) and analyzed using Flowjo X (RRID:SCR_008520; BD Bioscience).

Animals were exposed to 18,000 feet altitude plus 20 mg/kg SU5416 weekly for 3 wk. Lungs were harvested and prepared for sorting as described above. For RNA sequencing (RNAseq), 250 nonclassical monocytes or pulmonary interstitial MØs (IMØ) were directly sorted into 1× lysis buffer containing RNase inhibitor (Takara Bio). Libraries were prepared using SMARTer@ Stranded Total RNA-Seq Kit - Pico Input Mammalian Library Prep Kit, per the manufacturer’s protocol. Library qualities and quantities were verified using Agilent Bioanalyzer System (Agilent, Santa Clara, CA) and Qubit fluorometric quantification (Thermo Fisher Scientific). Sequencing was performed using an Illumina Hiseq 2500 (Illumina, San Diego, CA), with 126 pair-ended runs and an approximate depth of 50 million reads per library. The sequencing reads were trimmed to remove the Illumina adapters and any low-quality base at the ends (the Phred quality score <30) by using Cutadapt (v1.12) (33). Subsequently, concordant pair-end reads for each sample were successfully aligned to the mouse reference transcriptome, GRCm38, using Tophat (v2.1.1) (34). Finally, the read count–per-gene measurements for each sample were performed by htseq-count (HTSeq v0.6.0) (35). The read counts were then filtered and normalized by estimated size factors by using the R package “DESeq2” (36). The gene expression differences across the treatment groups were then evaluated using the default generalized linear model in DESeq2. Genes passing the threshold, a false discovery rate of <5%, were considered as significantly differentially expressed.

Murine lung tissues were perfused with PBS, followed by fixation in 4% paraformaldehyde. Harvested tissues were washed extensively with PBS. All lobes of the lungs were sliced in cross-section (perpendicular to main pulmonary arteries), at ∼0.5 cm thickness. All lung sections were embedded as a single block in paraffin. Tissue sections of 5 μm were obtained, stained with anti-human von Willebrand factor and alkaline phosphatase conjugated mouse anti-human α–smooth muscle cell actin (SMA) (1A4; Sigma-Aldrich), and developed with VectorRed Alkaline Phosphatase (VectorLabs, Burlingame, CA). Tissues were counterstained with Hematoxylin QS (VectorLabs). The entire slide was imaged and stitched with Zeiss Axio Imager Widefield Fluorescence Microscope (Zeiss). In a blinded fashion, all vessels under 50 μm on the entire slide were categorized as no muscularization (absence of SMA staining around the vessel), partially muscularized (SMA staining in parts of the vessel), and fully muscularized (SMA staining encircles the entire circumference of the vessel). Vessel muscularization was expressed as the percentage of total vessels enumerated.

All data are expressed as mean (±SEM) across experimental repeats, as stated. Group comparisons were performed in GraphPad PRISM 6 (RRID:SCR_002798; La Jolla, CA) using Student t test, one-way ANOVA, or two-way ANOVA. Any p values <0.05 were considered statistically significant. All experiments were repeated at least three times.

To identify the myeloid cell types that accumulate in lungs during hypoxia, we used flow cytometric analysis to quantify myeloid populations in the lungs of hypoxic and normoxic mice (Supplemental Fig. 1) (32, 37). Based on our prior finding that myeloid cell infiltration peaked at 3 wk of hypobaric hypoxia exposure, C57BL/6 animals were exposed to normoxia or hypoxia, with or without Su5416, for 3 wk (28). Su5416 is a tyrosine kinase inhibitor that is thought to induce endothelial cell dysfunction, leading to the development of severe PH in mice (28, 38). Relative to normoxic controls, mice exposed to either hypoxia or hypoxia plus Su5146 exhibit a 2- to 3-fold increase in the number of CD11b⁺CD64⁺ pulmonary IMØ (Fig. 1A). No significant changes in the numbers CD11b⁻CD64⁺ alveolar MØs (AMØ) or CD11c⁺MHCII⁺CD24⁺CD64⁻ dendritic cells (DCs) were observed (Fig. 1A). Consistent with our previous report, hypoxia-exposed animals displayed a significant reduction in the numbers of both classical (Ly6C^hi) and nonclassical (Ly6C^lo) monocytes in the lungs (Fig. 1A) (28). However, in hypoxia plus Su5416–treated animals, compared with normoxia plus Su5416–treated controls, monocyte reduction was only observed in the nonclassical (Ly6C^lo), but not classical (Ly6C^hi), subset (Fig. 1A). No significant changes in granulocyte populations, including neutrophils and eosinophils, were observed (data not shown). Overall, these findings demonstrate that IMØ are the only myeloid cell type to increase in the lungs of mice exposed to hypoxia.

To determine if IMØ are also the cells that infiltrate small pulmonary arteries in response to hypoxia, we exposed Cx3cr^wt/gfp mice to normoxia or 3 wk of hypoxia in the presence or absence of Su5416 and examined the localization of IMØ in lung sections by confocal microscopy. Compared with hypoxia alone, addition of Su5416 treatment induces increased pulmonary vascular remodeling and severity of PH (38). In Cx3cr^wt/gfp mice, all pulmonary MØs express CD64, but only IMØ express GFP⁺, leaving AMØ GFP⁻ (32). In both models of hypoxia-induced PH, lungs of normoxic mice displayed no muscularization of small pulmonary arteries, as demonstrated by a lack of SMA staining and no accumulation of GFP⁺ CD64⁺ IMØ around the small pulmonary arteries (Fig. 1B, columns 1 and 2). In contrast, the lungs of hypoxic mice displayed robust muscularization of small pulmonary arteries (SMA⁺) and infiltration of these remodeled small pulmonary arteries by GFP⁺ CD64⁺ IMØ (Fig. 1B, columns 3 and 4). To confirm that the small pulmonary artery–infiltrating cells were MØs, we also examined expression of CD169, a marker expressed on IMØ and AMØ but not monocytes or DCs (32). The lungs of hypoxic mice displayed infiltration of SMA⁺ small pulmonary arteries by GFP⁺CD169⁺ IMØ (Supplemental Fig. 2A). Moreover, the infiltration of small pulmonary arteries by IMØ appeared to occur to a greater extent in smaller vessels (<50 μm) (Supplemental Fig. 2B). These findings demonstrate that IMØ are the myeloid cell type that infiltrate small pulmonary arteries during the development of hypoxic PH.

Because IMØ may arise from either resident fetal-derived MØs or circulating bone marrow–derived monocytes, we determined the origin of the IMØ that accumulate in lungs in response to hypoxia. We transplanted congenic CD45.1⁺ bone marrow into CD45.2⁺ recipients after myeloid cell ablation with busulfan. The recipients were exposed to normoxia or hypoxia, and the origin of IMØ was examined (Fig. 2A). Analyses of peripheral blood 3 wk after transplant demonstrated effective chimerism in which >85% of circulating monocytes, neutrophils, and eosinophils were of donor (CD45.1⁺) origin (Fig. 2B). Consistent with prior reports, busulfan treatment generally preserves the CD45.2⁺ recipient origin of lung resident tissue MØs but allows the replacement of bone marrow–derived circulating monocytes with those of CD45.1⁺ donor origin (Fig. 2C, 2D) (39). As with Cx3cr^wt/gfp mice, recipients exposed to hypoxia displayed a 2- to 3-fold increase in the frequency of lung IMØ (Fig. 2E, 2F). In normoxia, ∼25% of IMØ were CD45.1⁺ (Fig. 2G, 2H). In contrast, in the lungs of hypoxic mice, over 55% of IMØ were CD45.1⁺ donor origin (Fig. 2G, 2H). This enrichment of CD45.1⁺ donor cells in IMØ with hypoxia exposure demonstrates that the vast majority of newly arriving IMØ are bone marrow derived.

Bone marrow–derived IMØ may arise from either classical or nonclassical monocytes. Classical monocytes (Ly6c^hi) are dependent on CCR2 signaling to traffic into the tissue (40, 41). Because MØ numbers increased to the same extent in hypoxic wild-type and CCR2^−/− mice, this strongly suggested that hypoxia-induced IMØ do not arise from classical monocytes (28). To determine if hypoxia-induced IMØ arise from nonclassical monocytes, we initially considered tracking the fate of these cells after adoptive transfer; however, mice do not contain sufficient numbers of nonclassical monocytes in blood, spleen, or bone marrow to allow effective adoptive transfer. We, therefore, performed in vivo cell labeling using PKH26PCL, a dye that is retained by tissue MØs and nonclassical (Ly6C^lo) monocytes (Fig. 3A) (42). To optimize and validate cell type labeling by PKH26PCL, we examined dye uptake and retention by immune cells in the lung and blood. At 24 h after PKH26PCL administration, all circulating and tissue-immune cells were labeled with PKH26PCL (data not shown). However, 4 d after PKH26PCL administration, as expected, there was selective dye retention by tissue IMØ and AMØ (Fig. 3C). In addition to tissue MØs, in both blood and lung, PKH26PCL was retained in nonclassical (Ly6C^lo) monocytes but not classical (Ly6C^hi) monocytes (Fig. 3B, 3C). Other immune cells, including DCs and neutrophils, did not retain the PKH26PCL dye (Fig. 3C). To determine cellular origin of infiltrating IMØ in response to hypoxia, mice were injected i.v. with PKH26PCL. Four days after administration of PKH26PCL, animals were exposed to normoxic versus hypoxic condition, with or without Su5416, for 21 d (Fig. 3A). Additionally, to determine the extent that expanded IMØ population was due to the resident MØ proliferation, mice were also injected with Edu at weekly intervals (Fig. 3A, 3D, 3E). Lung tissues were then harvested, and IMØ labeling by PKH26PCL and Edu were examined (Fig. 3D–G). Similar to findings in Figs. 1, 2, in animals exposed to hypoxia alone or hypoxia plus Su5416, compared with controls, pulmonary IMØ were increased 2- to 3-fold (data not shown). With or without Su5416 exposure, IMØ derived from the lungs of normoxic mice demonstrated a low level (∼7%) of proliferation (Fig. 3D, 3E). In mice exposed to 21 d of hypoxia, IMØ proliferation, as measured by Edu, remained low and similar to that of normoxic animals (∼7%) (Fig. 3D, 3E). Consistent with the above-mentioned finding that the majority of expanding IMØ derive from circulating bone marrow cells, this Edu staining further demonstrates that expansion of IMØ was not due to proliferation of PHK26PCL⁺ resident IMØ. The staining also demonstrated that, similar to normoxic animals, nearly all IMØ from hypoxic animals are PKH26PCL⁺ (Fig. 3F, 3G). Thus, the expanding IMØ population in hypoxic animal must arise from the circulating PKH26PCL⁺ cell type: nonclassical monocytes. Taken together, these findings demonstrate that circulating nonclassical monocytes serve as the major precursors to hypoxia-induced pulmonary IMØ.

To determine if IMØ derived from nonclassical monocytes play a causative role in the development of hypoxia-induced PH, we examined mice in which hypoxia-inducible factor-1α (Hif1α) was selectively eliminated in the nonclassical monocyte lineage using Hif1α^flox/flox mice. Hypoxia-inducible factors are the critical oxygen-sensing molecules in mammalian cells, and animals heterozygous for Hif1α are protected from hypoxia-induced PH (43, 44). Because CX3CR1 is expressed most highly in Ly6C^lo nonclassical monocyte, we used Cx3cr1-cre mice to target the nonclassical monocyte lineage (30, 32). The specificity of Cre expression in these mice was examined by crossing them to Rosa26-fGFP reporter mice. Cx3cr1cre;Rosa26-fgfp mice displayed expression of farnesylated GFP on Ly6C^lo, but not Ly6C^hi, monocytes in both the blood and lung (Supplemental Fig. 3A). GFP expression was also detected in ∼50% of lung MØs and ∼60% of DCs (Supplemental Fig. 3A). There was no GFP expression in CD45⁻ nonimmune cells in the lung, demonstrating that there was no off-target gene deletion in nonimmune cells (data not shown). When crossed to Rosa26-dtr mice, Ly6^lo, but not Ly6C^hi, monocytes were depleted when the resulting Cx3cr1cre;Rosa26-dtr mice were treated with diphtheria toxin (Supplemental Fig. 3B).

Cx3cr1cre;Hif1α^Δ/Δ mice displayed a selective deletion of Hif1α in nonclassical monocytes. By semiquantitative PCR, there was an ∼90% reduction of Hif1α expression in Ly6C^lo monocytes but normal expression levels in Ly6C^hi monocytes (Fig. 4A). Interestingly, whereas farnesylated GFP expression in Cx3cr1cre;Rosa26-fgfp mice was detected in subpopulations of CD64⁺ MØs and IA/IE⁺CD11c⁺CD64⁻CD24⁺ DCs by semiquantitative PCR, Hif1α was not effectively deleted in these cell types (Fig. 4B, 4C, Supplemental Fig. 3A). Although Hif1α was not deleted in MØs under steady-state conditions (Fig. 4A), Hif1α deletion was observed in IMØ derived from hypoxic animals (Supplemental Fig. 3C). Hif1α deletion had no significant effect on pulmonary immune cell composition (Supplemental Fig. 3D), baseline body weight, heart rate, or hemoglobin levels (Supplemental Table I). Relative to Hif1α^flox/flox controls, normoxic Cx3cr1cre;Hif1α^Δ/Δ mice displayed no abnormalities in RV systolic pressures (RVSPs), right ventricular hypertrophy (as assessed by Fulton index [RV/(LV+S)]), or pulmonary vessel muscularization in either the absence or presence of SU5416 (Fig. 4D–F). However, after 4 wk of hypobaric hypoxia, Cx3cr1cre;Hif1α^Δ/Δ mice displayed a significant reduction in RVSP compared with Hif1α^Δ/Δ littermates (Fig. 4D). Similarly, when exposed to hypoxia plus SU5416, Cx3cr1cre;Hif1α^Δ/Δ mice displayed significant reductions in RVSP, right ventricular hypertrophy, and pulmonary vessel muscularization compared with Hif1α^Δ/Δ littermates (Fig. 4B–E). Associated with the decreased muscularization of small pulmonary arteries in hypoxia-exposed Cx3cr1cre;Hif1α^Δ/Δ mice, there was a decreased accumulation of CD64⁺ pulmonary IMØ around small pulmonary arteries (Fig. 4G). These findings demonstrate that nonclassical monocytes are the myeloid cell type that senses hypoxia, infiltrates small pulmonary arteries, differentiates into IMØ, and directly contributes to the development and/or progression of PH.

The above results raise questions concerning the mechanisms by which hypoxia-stimulated nonclassical monocytes and IMØ may stimulate the development of PH. To identify functional pathways in these cell types that may contribute to PH pathogenesis, we performed RNAseq of nonclassical monocytes and IMØ purified from the lungs of Cx3cr1cre;Hif1α^Δ/Δ and control Hif1α^flox/flox mice after 3 wk of exposure to hypoxia plus SU5416 (Fig. 5A). Approximately 20–∼30 million 126-bp paired-end reads were generated for each sample. Subsequently, 75% (15–23 million) quality concordant pair-end reads for each sample were successfully aligned to the mouse reference transcriptome, GRCm38. Finally, the read count–per-gene measurements for each sample were performed to convert the mapped reads to read counts for a total of 26,608 genes. The read counts were then filtered and normalized by estimated size factors by using the R package “DESeq2” (36). The gene expression differences across the treatment groups were then evaluated using the default generalized linear model in DESeq2. Genes passing the threshold, a false-discovery rate of <5%, were considered to be significantly differentially expressed. Comparing Hif1α-deficient to Hif1α-sufficient nonclassical monocytes exposed to hypoxia plus Su5416, 314 differentially expressed genes were identified, with 260 downregulated and 54 upregulated genes. For Hif1α-deficient IMØ, 1577 differentially expressed genes were identified (1003 downregulated and 574 upregulated).

Relative to controls, Cx3cr1cre;Hif1α^Δ/Δ IMØ expressed lower levels of MØ maturation markers, including CD64 (Fcɣr1), F4/80 (Emr1), CD88 (C5ar1), and LysM (Lyz2) (Fig. 5B). In addition to these conventional markers for MØ identification, Cx3cr1cre;Hif1α^Δ/Δ IMØ also downregulated pathways associated with classical MØ functions, including phagocytosis and complement activation (Fig. 5C). These MØs also downregulated pathways associated with Ag presentation, which is essential for subsequent activation of adaptive immune response. In Cx3cr1cre;Hif1α^Δ/Δ IMØ, all processes required for efficient Ag presentation including phagosomes, lysosomes, and protein processing in the endoplasmic reticulum and Ag presentation onto the cell surface were reduced (Fig. 5D). Overall, these findings suggest an impaired maturation of Cx3cr1cre;Hif1α^Δ/Δ nonclassical monocytes into disease-promoting IMØ, with a reduced capacity to mediate phagocytosis, complement activation, Ag presentation, and subsequent activation of innate and adaptive immune responses.

Both IMØ and nonclassical monocytes from Cx3cr1cre;Hif1α^Δ/Δ animals expressed significantly lower levels of cytokines that have been associated with PH, including IL-1, IL-6, TNF and TGF-β (Fig. 6). These findings suggest that nonclassical monocytes and nonclassical monocyte–derived infiltrating IMØ sense hypoxia and promote PH.

The above findings suggest that, in hypoxic mice, the activity of nonclassical monocytes in the lungs plays a key role in the development of PH. To determine if cells of the nonclassical monocyte lineage accumulate in human PAH lungs, we examined monocyte populations in peripheral lung tissues from control and PAH patients using flow cytometry and the gating strategy shown in Fig. 7C (31). Control lungs were derived from human donors who were declined at the time of transplant. PAH lung tissues were obtained at the time of transplant. In control lungs, the vast majority of monocytes were CD14^hiCD16^lo, the human equivalent of murine classical monocytes (Fig. 7A, 7B). In comparison, PAH lungs displayed a 5- to 6-fold increase in CD14^loCD16^hi nonclassical monocytes, the human equivalent of Ly6C^lo nonclassical monocytes (Fig. 7A, 7B). These findings demonstrate that nonclassical monocytes specifically accumulate in the lungs of human PAH patients, suggesting that these cells contribute to the pathogenesis of PAH in humans.

Myeloid cell infiltration of the pulmonary vasculature is a common feature of PH in humans and all animal models; however, the specific infiltrating cell type has not been identified, and the causal relationship between this inflammatory response and disease pathogenesis has remained unclear. In this study, using mouse models of hypoxic PH, we can draw three novel conclusions about the myeloid cells that infiltrate the pulmonary vasculature during the development of PH. First, we show that the vasculature-infiltrating cells are IMØ. Second, we show that these cells arise from circulating nonclassical monocytes. Third, we show that these nonclassical monocyte–derived vascular-infiltrating MØs sense hypoxia and directly contribute to PH pathogenesis.

It has been previously thought that DCs are the main accumulating myeloid cell type in PH (6, 8). In contrast, we found that IMØ are the only myeloid cell type that accumulates in lungs in hypoxia-induced PH. Our conclusion is based on the finding that the only cells that accumulate in PH lungs are CD45⁺, CD11c^int/hi, CD11b^hi, CD64^hi, CX3CR1⁺ and CD169⁺, while being Ly6G⁻, Ly6C⁻, CD24⁻, and Siglec F⁻, and that the vasculature-infiltrating cells are CD64⁺, CX3CR1⁺ and CD169⁺. This is the phenotype of pulmonary IMØ as we and others have described (32, 37). The finding of expanded pulmonary IMØ in PH is consistent with our preliminary findings and other recent reports (45–47). Previous reports of DC accumulation likely stem from the fact that pulmonary IMØ also express CD11c and MHC class II, which were commonly used to identify DCs but are now known to be expressed by MØs. Thus, the myeloid cells that accumulate in PH are pulmonary IMØ.

Our second conclusion is that vasculature-infiltrating IMØ arise primarily from circulating bone marrow–derived nonclassical monocytes. This conclusion is based on three findings: 1) the cells that accumulate in hypoxic lungs are of donor origin after bone marrow transplantation; 2) after labeling circulating bone marrow–derived nonclassical, but not classical, monocytes with PKH26PCL, the vast majority of IMØ in hypoxic lungs are PKH26PCL⁺; and 3) IMØ in hypoxic lungs display only minimal proliferation. This conclusion is consistent with our prior finding that classical monocytes, which depend on CCR2 for trafficking into tissue, are not major precursors to the accumulating pulmonary IMØ (28). Also, the decrease in Ly6C^lo nonclassical monocyte numbers in the lungs of animals exposed to hypoxia or hypoxia + Su5416 likely reflects maturation of monocytes into monocyte-derived MØs (28). Mature MØs are specialized cells with enhanced phagocytic capacity. PKH26PCL was designed to label highly phagocytic cells and has been used to differentiate infiltrating Ly6c^hi classical monocyte–derived MØs from resident MØs. Labeling of monocyte subsets for in vivo imaging and tracking has been challenging (25). In this study, we demonstrate that PKH26PCL administration and timing can be optimized to label monocyte subsets differentially. Prior studies suggest that Ly6C^lo nonclassical monocytes represent a group of more mature and specialized monocytes that patrol the vasculature (25). The enhanced ability of Ly6C^lo nonclassical monocytes to take up and retain PKH26PCL, compared with Ly6C^hi classical monocytes, is consistent with the view that Ly6C^lo nonclassical monocytes represent a more mature population along the monocyte-MØ continuum (25). Additionally, our findings that cells of Ly6C^lo nonclassical monocyte lineage accumulate in small pulmonary arteries in PH further support a specialized role for these cells in regulating vessel homeostasis and disease (27, 48).

The accumulation of myeloid cells has been described extensively in humans and animal models of PH. The extent to which these cells play a direct causal role in PH pathogenesis has remained controversial because of the fact that evidence for such a role for these cells has been limited. Our third and key conclusion is that cells of the nonclassical monocyte lineage play a direct causal role in PH development. We find that the sensing of hypoxia by this lineage promotes vascular remodeling and contributes significantly to the development of hypoxia-induced PH. Using Cx3cr1cre to target Hif1α deletion in Ly6C^lo nonclassical lineage, we show that Hif1α-deficient animals have a reduced severity of PH. Based on the expression of Cx3cr1cre:fGFP, we had expected that Cx3cr1cre would be expressed and reduce Hif1α transcripts in subpopulations of pulmonary MØs and DCs (Fig. 4B, 4C, Supplemental Fig. 3). However, by semiquantitative PCR, Hif1α transcripts remained intact in pulmonary MØs and DCs. This finding demonstrates that specific gene deletions are affected by factors beyond presence or absence of cre recombinase. Potential determinants include timing and the expression level of cre recombinase in relation to targeted gene locus accessibility (49, 50). Monocytes are derived from a distinct precursor lineage, compared with majority of resident pulmonary MØs and classical DCs (19). This distinct cell differentiation and maturation program may affect the efficiency of Hif1α deletion. Thus, the efficiency and specificity of targeted gene deletion should not be inferred based on reporter expression but should be confirmed empirically (51). Using Cx3cr1cre;Hif1α, we are not able to examine the contribution of resident MØs to PH pathogenesis. However, the finding of Hif1a gene deletion in IMØ derived from hypoxic, but not normoxic, animals is consistent with the view that, in response to hypoxia, nonclassical monocytes serve as progenitors for infiltrating IMØ (Fig. 4A, Supplemental Fig. 3C). Thus, our findings suggest a direct causal role of infiltrating Ly6C^lo nonclassical monocytes to PH development. This conclusion is supported by previous studies in which interventions that increased cell numbers that would include nonclassical monocytes, their precursors, or their derivatives increased the severity of PH. In contrast. interventions that decreased the number of such cells also decreased PH severity (4, 8, 10, 11, 16). Our findings suggest a model in which hypoxia, working via hypoxia-inducible factor activation, stimulates nonclassical monocytes to infiltrate small pulmonary arteries, differentiate into IMØ, and produce factors that stimulate pulmonary vascular remodeling.

At present, the specific factors that stimulate pulmonary vascular remodeling remain to be identified. In our transcriptome analyses, by principal component analysis scatter, there was not complete separation between IMØ derived from Hif1α-sufficient and Hif1α-deficient animals. This may be because our sorted IMØ population would have included Hif1α-sufficient resident IMØ (Fig. 5A). Consistent with the mixture of Hif1α-sufficient resident IMØ and Hif1α-deficient IMØ in our RNAseq analyses, and the increased absolute number of pulmonary IMØ (∼2-fold) in hypoxic animals, our transcriptomic analyses displayed an ∼50% reduction in Hif1α expression between IMØ derived from Hif1α-deficient versus Hif1α-sufficient hypoxic animals (Fig. 1A, Supplemental Fig. 3C). Even with this mixture of infiltrating and resident IMØ populations, we identified significant alterations in IMØ derived from nonclassical monocyte–specific Hif1α-deficient animals. Hif1α-deficient IMØ express lower levels of MØ-specific transcripts (Fcɣr1/CD64, Emr1/F4-80, CD68, C5ar1/CD88, and Lyz2), suggesting these MØs are immature. Consistent with this immature phenotype, many pathways associated with quintessential MØ functions are decreased, including phagocytosis, complement activation, and processes leading to Ag presentation. Ag presentation by MØs are important for regulating both T and B cell functions. Thus, Hif1α-deficient IMØ may have a reduced capacity to activate or perpetuate subsequent innate and adaptive immune response. In addition, nonclassical monocytes and IMØ derived from targeted Hif1α-deleted animals display reduced transcripts for both M1 and M2 cytokines and chemokines that have been implicated in PH development and progression, including IL-6, TNF, and TGF-β (11, 15, 52). Thus, rather than skewing MØ M1 versus M2 profiles, deletion of Hif1α in nonclassical monocytes prohibits maturation of these monocytes into disease-promoting MØs. Thus, nonclassical monocyte–derived cells orchestrate a vascular microenvironment, through immune activation and cytokine production, that promotes vascular remodeling.

With respect to human PH, our finding that cells of the nonclassical monocyte lineage are uniquely increased in lung explants from patients with Group I PAH supports the view that cells of this lineage may play a role in human PH. One significant difference between our murine and human results was that in the murine studies, we demonstrated an accumulation of pulmonary IMØ, whereas in Group I PAH lung explants, we found an accumulation of nonclassical monocytes. This difference may be explained by the disparate disease stages. Mouse models of PH generally produce relatively mild and reversible pathology that is similar to early stages of disease. Human PAH lung explants are derived from patients with nonreversible, end-stage disease. It is likely that cellular differentiation kinetics differ between disease stages and species. Despite this limitation, our findings support a role for the nonclassical monocyte lineage in PH pathogenesis. This view is consistent with studies showing that humans with mutations in BMPR2 display increased production of GM-CSF, a potent monocyte chemoattractant, in small pulmonary arteries and infiltration of vessels by GM-CSFRa⁺CD68⁺ cells, the phenotype of monocyte-MØs (16). Our results support the view that BMPR2 mutations promote PAH, at least in part, by stimulating the accumulation of IMØ lineage cells in small pulmonary arteries. This would represent one “hit” in the “multiple-hit” model of PAH pathogenesis (53).

It is known that multiple distinct cell types contribute to the development of PH through the sensing of hypoxia. The specific deletion of Hif2α in endothelial cells reduces the severity of PH in hypoxic models (54, 55). Pulmonary smooth muscle cell–specific HIF1α has also been shown to drive disease development (56, 57). Until now, direct evidence that immune cells also contribute to PH through the sensing of hypoxia has been lacking (58). In this study, we demonstrated that nonclassical monocyte–derived pulmonary IMØ sense hypoxia and directly contribute to PH pathogenesis. This suggests that in multiple types of PH, infiltration of IMØ into small pulmonary arteries may be stimulated by hypoxia itself in the context of alveolar hypoventilation, pulmonary emboli, systemic inflammation, or local infection (59, 60). If such IMØ infiltration is a common phenomenon, it would predispose those affected to the development of PH. As such, specific factors produced by IMØ that stimulate pulmonary vascular remodeling may represent attractive therapeutic targets for multiple types of PH in patients.

We thank Dr. Scott Randell and the staff of the University of North Carolina Marsico Lung Institute/Cystic Fibrosis Center Tissue Procurement and Cell Culture Core for providing human lung tissues. Flow cytometry and sorting was performed in the Duke Human Vaccine Institute Research Flow Share Resources Facility (Durham, NC).

The authors have no financial conflicts of interest.