ARDS, a clinical syndrome defined by bilateral opacites on chest imaging denoting the presence of lung inflammation, blood hypoxemia with tissue hypoxia and the requirement for positive-pressure ventilation, has a mortality rate of up to 40%1,2. Despite decades of research, effective therapies for ARDS remain elusive, with much of the treatment research focusing on modulating the inflammatory injury, because its persistence is a poor prognostic indicator3,4. However, the effects of blood hypoxemia and the ensuing tissue hypoxia on the persistence of inflammation in ARDS have not been fully investigated.

Macrophages have a key role in driving tissue inflammation resolution. In mice, lung macrophages can be subdivided into SiglecF+ alveolar macrophages (AMs) and SiglecF interstitial macrophages (IMs), which inhabit distinct anatomical niches. Under homeostatic conditions, AMs largely self-renew5,6 whereas IMs require recruitment of circulating monocytes7,8. Evidence is emerging for key roles for IMs in inflammation regulation9,10 and repair11.

In the present study, we investigated whether hypoxemia associated with ARDS, including severe COVID-19 disease, affected the accumulation of lung monocyte-derived macrophages (MDMs) and whether this, in turn, impacted the resolution of inflammation. We found that patients with ARDS had clinical evidence of persistent hypoxemia despite ventilatory support and were profoundly monocytopenic during the first 48 h of ventilation, an observation replicated in mouse models of hypoxic acute lung injury (ALI). We further showed that systemic responses to tissue hypoxia suppressed type I interferon (IFN) signaling and fundamentally altered bone marrow hematopoiesis, with ensuing consequences for the phenotype and number of monocytes in the blood and accumulation of MDMs in the lung during ALI. This, in turn, led to the persistence of inflammation. Critically, targeting this pathway with the monocyte and macrophage growth factor colony-stimulating factor 1 (CSF-1) corrected these hypoxia-mediated changes and drove inflammation resolution.


ARDS is characterized by monocytopenia and an altered immune phenotype

Despite the heterogeneity of the etiologies leading to ARDS, one defining feature is blood hypoxemia and tissue hypoxia. We first characterized the arterial oxygen partial pressure (PaO2) in patients with moderate-to-severe ARDS. Patients were sampled within 1 week of a known insult or new or worsening respiratory symptoms in accordance with the Berlin criteria, namely, if they had bilateral opacities on a chest radiograph and evidence of respiratory failure with a PaO2/FiO2 (fraction inspired O2) < 200 mmHg and positive end-expiratory pressure >5 cmH2O. Patients with ARDS had a range of etiologies and associated pathogens (Table 1) and showed clinically important (Fig. 1a) and sustained (Fig. 1b) blood hypoxemia over a 24-h period despite supplementary oxygen therapy (Fig. 1c) and ventilatory support. In 13 of the 22 patients, this was associated with an elevated circulating lactate (Fig. 1d), indicating ongoing tissue hypoxia. To characterize the kinetics of circulating leukocyte populations, we sampled blood from ventilated patients <48 h from diagnosis of ARDS and commencement of positive-pressure ventilation (hereafter early ARDS) or 48 h to 7 d from diagnosis (late ARDS). Healthy donors were used as controls because tissue hypoxia is a common feature of critically unwell patients. Early ARDS patients had elevated circulating leukocyte counts (Fig. 1e), but significantly lower proportions and numbers of circulating monocytes (Fig. 1e) compared with controls. In late ARDS patients, circulating leukocyte numbers remained elevated compared with controls (Fig. 1f), but monocyte frequency and counts were equivalent (Fig. 1f). In early ARDS, we detected an increase in the proportion of CD14+CD16+ intermediate monocytes at the expense of classical CD14++CD16 monocytes (Extended Data Fig. 1a–c)12. ARDS monocytes, irrespective of timepoint, had a lower expression of the major histocompatibility class II (MHC-II) marker HLA-DR (human leukocyte antigen-DR) and higher expression of CD11b compared with healthy controls (Fig. 1g). Proteomic analysis of sorted blood CD14++CD16 ARDS monocytes indicated changes in the abundance of proteins with transcripts that had been reported to be sensitive to hypoxic culture in human monocytes13 when compared with healthy controls (Fig. 1h). We also observed a significant increase in secretory-granule content in ARDS CD14++CD16 monocytes (Fig. 1i,j) which was associated with altered expression of hypoxia-regulated proteins, including SLC2A314, IGFR215, PSMD416 and FTL17. NanoString analysis identified a specific transcriptional signature in ARDS monocytes (Fig. 1k), with 41 genes differentially expressed compared with healthy controls (Fig. 1l). Notably, four MHC complex genes (HLA-DMA, HLA-DMB, HLA-DQA1 and HLA-DRB3), important for antigen-presenting function, as well as genes associated with monocyte adhesion and extravasation (SELL18), transendothelial migration (CD9919) and LPS signaling (MAP2K420) and MAP3K1421) were significantly downregulated in ARDS monocytes (Fig. 1l). Thus, ARDS affected both the transcriptomic and the protein signature of blood monocyte

Table 1 ARDS patient cohort clinical characteristics and demographics
Fig. 1: Patients with ARDS are monocytopenic early in the disease with phenotypically distinct circulating monocytes.
figure 1

a,b, Lowest (a) and all (b) partial pressures of oxygen (PaO2) from clinical arterial blood samples from ARDS patients, 24 h preceding research blood sampling (green: normal range). c,d, Highest recorded FiO2 (c) and highest recorded arterial plasma lactate level within 24 h of research sampling (d) in ARDS patients (red samples: lactate ≥upper limit of normal; green: normal local reference (0.5–1.6 mmol l−1)). e, Blood leukocyte counts, monocyte proportions and monocyte counts from ARDS patients, collected within 48 h of diagnosis (early ARDS) and a healthy volunteer cohort (HC). f, Blood leukocyte count, monocyte proportions and monocyte counts from ARDS patients collected between 48 h and 7 d (late ARDS) and HC. g, Monocyte HLA-DR and CD11b expression in HC and ARDS patients. h, CD14++CD16 classical monocyte proteomic data from ARDS patients, relative to HC, for proteins associated with a human monocyte, in vitro hypoxic gene signature13. i, Classical (CD14++CD16) monocyte proteome volcano plot from HC and ARDS patients. Significantly upregulated granule-associated proteins in ARDS patients versus HC (blue), a sample of known hypoxia-regulated proteins (orange). j, Classical monocytes proteinase 3, myeloperoxidase and azurocidin 1 copy numbers in HC and ARDS patients. k,l, Pearson’s correlation (k) and heatmap (l) of differentially expressed genes from HC and ARDS patient blood monocytes. Data in ac are mean ± s.d. expressed as median (e,f) or shown as mean (g,j). In a,cg and j each datapoint represents one patient/HC; in b, each datapoint represents one independent clinical sample. Statistical testing used was: unpaired, two-tailed Student’s t-test (ef and j) and Mann–Whitney U-test (g).

Experimental ARDS reproduces monocytopenia and phenotypic alterations

We next investigated whether the circulating monocyte profile alterations observed in ARDS patients were replicated by the induction of hypoxemia in a mouse model of ALI (Extended Data Fig. 2a). Mice exposed to 10% FiO2 demonstrated equivalent levels of hypoxemia to patients with ARDS (Extended Data Fig. 2b). Administration of nebulized lipopolysaccharide (LPS) induced an increase in circulating leukocytes and CD115+CD11b+ monocytes (Fig. 2a) in mice housed in normoxia for 24 h, relative to naive mice, yet this increase was absent in LPS-challenged mice housed in hypoxia (10% FiO2) immediately post-LPS for 24 h. A selective loss of nonclassical CD115+CD11b+Ly6Clo monocytes in LPS-challenged hypoxic mice (Fig. 2a) relative to normoxic LPS-challenged mice was observed. Blood CD115+CD11b+Ly6Chi classical monocytes from hypoxic LPS-challenged mice had decreased expression of the adhesion molecules intercellular adhesion molecule (ICAM)-1 and CD11a and increased expression of the CCR2 chemokine receptor, compared with normoxic counterparts (Fig. 2b). The absolute number of circulating leukocytes, CD115+CD11b+ monocytes (Fig. 2c) and Ly6G+CD11b+ neutrophils (Extended Data Fig. 2c) in the blood of normoxic or hypoxic LPS-treated mice normalized at day 5, although the proportion of Ly6Clo monocytes remained contracted in hypoxic LPS-challenged mice compared with normoxic LPS-challenged controls (Fig. 2c). We also observed persistent alterations in the transcriptome of Ly6Chi monocytes in hypoxic LPS-challenged mice (Fig. 2d), including decreased expression of the chemokine receptor Ccr5 and the scavenger receptor Cd36, which are markers of monocyte maturity22 and increased expression of Il1b, an inflammatory cytokine associated with poor outcomes in ARDS23. Thus, in mice with ALI, hypoxia drives alterations in monocyte numbers and phenotype that parallel those described in ARDS patients.

Fig. 2: Hypoxic acute lung injury replicates early monocytopenia in mice and alters the circulating monocyte phenotype.
figure 2

a, Blood leukocyte counts, monocyte counts and proportion of blood monocyte subgroups in naive or LPS-treated mice housed in normoxia or hypoxia for 24 h. b, Classical monocyte (CD115+CD11b+Ly6Chi) surface expression of ICAM, CD11a and CCR2 at 24 h post-LPS. c, Blood leukocyte counts, monocyte counts and proportions of monocyte sub-populations in naive or LPS-treated mice housed in normoxia or hypoxia for 5 d post-LPS. d, Differentially expressed genes in circulating classical monocytes from LPS-treated mice housed in normoxia or hypoxia for 5 d. Data represent the mean ± s.e.m. Data for a and c are pooled from two independent experiments. b is representative of 2 experiments (n=3-4/ group). Each datapoint represents an individual mouse. Statistical testing: one-way ANOVA with Tukey’s multiple comparison test (a and b).

Tissue hypoxia prevents accumulation of lung MDMs

Next we explored the effect of hypoxemia on LPS-induced inflammation in the lung. Consistent with tissue hypoxia, hypoxia-inducible factor 1α (HIF-1α) protein was most highly expressed in the lungs of hypoxic LPS-treated mice compared with normoxic LPS controls (Fig. 3a). LPS challenge significantly increased lung Ly6G+CD11b+ neutrophil numbers in hypoxia- or normoxia-housed mice compared with naive controls (Fig. 3b), with equivalent numbers of CD3+CD19MHC-II T cells and CD3CD19+MHC-II+ B cells in the lungs of these mice (Extended Data Fig. 3a,b). LPS challenge led to a significant expansion of the LinCD45+CD64hi macrophage compartment in normoxia- but not hypoxia-housed mice (Fig. 3c,d). AMs are CD64hiSiglecF+CD11c+, whereas CD64hiSiglecF macrophages almost exclusively comprise IMs in health, and include inflammation-elicited, parenchymal or alveoli-localized, CD64hiSiglecFLy6C+ MDMs after injury. The number of CD64hiSiglecF+CD11c+ macrophages was equivalent in LPS-challenged mice housed in hypoxia or normoxia (Fig. 3c,d). Hypoxia significantly blunted the LPS-mediated expansion of the CD64hiSiglecF macrophage compartment observed in normoxic mice (Fig. 3c,d), an effect that appeared to be entirely attributable to the absence of CD64hiSiglecFLy6C+ MDMs in LPS-treated hypoxic mice (Fig. 3c,d). The number of CD64hiSiglecFLy6CMHC-II+ macrophages was similar in all treatments (Fig. 3c,d). The reduction in the number of CD64hiSiglecFLy6C+ MDMs occurred despite elevated amounts of the monocyte chemoattractant CCL2 in the alveoli of LPS-treated hypoxic mice (Fig. 3e), and similar numbers of CD64loCD11b+Ly6C+ monocytes in the lungs of hypoxic and normoxic LPS-treated mice (Extended Data Fig. 3c). These observations suggested that monocytes recruited to the lung after LPS treatment did not convert to CD64hiSiglecFLy6C+ MDMs in hypoxia. The contribution from intravascular cells to lung cell counts could not be completely excluded from these data.

Fig. 3: Systemic hypoxia hampers expansion of the CD64hiSiglecF macrophage niche in ALI and S. pneumoniae infection.
figure 3

a, Representative lung immunofluorescence HIF-1α and DAPI expression from LPS-challenged mice, housed in normoxia or hypoxia for 24 h. Scale bar, 50 μm. b, Absolute numbers of live lung neutrophils in naive or LPS-challenged mice housed in normoxia (N) or hypoxia (H) for 24 h (n = 6 per group). c, Representative dot plots of the CD64hi macrophage compartment (top), CD64hiSiglecF+CD11c+ AMs (middle) and Ly6C and MHC-II expression by CD64hiSiglecF macrophages (bottom) in mice as in b. d, Absolute number of CD64hi macrophages (Mφ), CD64hiSiglecF+CD11c+ AMs, Ly6CMHC-II+ lung macrophages and CD64hiSiglecFLy6C+ MDMs as in b. e, BAL CCL2 levels from LPS-challenged mice housed in normoxia or hypoxia for 24 h. f, Frequencies of neutrophils among total lung leukocytes, of CD64hiSiglecF macrophages among lung CD64hi macrophages and MDMs among CD64hiSiglecF macrophages in mice inoculated with S. pneumoniae (n = 6 per group) or vehicle control (Veh, n = 4 per group) and housed in normoxia or hypoxia until 24 h post-inoculation. g, Representative plots of CD64hiSiglecF macrophages in mice as in f. h, Frequency of lung CD64hiSiglecF macrophages among total leukocytes, absolute numbers of lung CD64hiSiglecF macrophages, proportion of BAL MDMs and absolute numbers of BAL MDMs in naive or LPS-challenged mice housed either in normoxia or hypoxia for 48 h or for 24 h in hypoxia, followed by 24 h of normoxia (hypoxia to normoxia) (n = 3 per group). Data represent the mean ± s.e.m. Data in a represent n = 3 per group; data in be are pooled from two independent experiments; data in g represent two independent experiments. Each datapoint represents an individual mouse. Statistical testing for b,d,f and h is by one-way ANOVA with Tukey’s multiple comparison test and for e by unpaired, two-tailed Student’s t-test.

To examine the effects of hypoxia on monocyte–macrophage dynamics in a model of severe streptococcal pneumonia, C57/BL6J mice were inoculated with D39 Streptococcus pneumoniae or vehicle and housed in either hypoxia or normoxia after a 4-h recovery period. Reduced blood leukocyte and monocyte counts were detected in infected mice housed in hypoxia compared with normoxia (Extended Data Fig. 3d,e). Although the number of lung neutrophils was equivalent in hypoxic and normoxic S. pneumoniae-infected mice (Fig. 3f), the accumulation of CD64hiSiglecF macrophages was reduced (Fig. 3f), with a particular absence of CD64hiSiglecFLy6C+ MDMs (Fig. 3f,g).

To determine whether hypoxia altered the monocyte–macrophage lung compartment directly, we returned LPS-challenged hypoxic mice to normoxic conditions, for a further 24 h, after 24 h of hypoxia. Mice returned to normoxia showed a significant increase in the proportion and number of CD64hiSiglecF macrophages and bronchoalveolar lavage (BAL) MDMs, compared with mice that remained in hypoxia (Fig. 3h). Together, these data indicated that hypoxia directly induced sustained changes in the lung macrophage compartment during various inflammatory challenges.

Hypoxia directly alters hematopoiesis

To determine the mechanism by which hypoxia regulated the number of circulating monocytes, we measured bone marrow (BM) output by pulsing naive or LPS-challenged mice, housed in hypoxia or normoxia, with bromodeoxyuridine (BrdU) 12 h post-LPS12,24. Hypoxemic LPS-challenged mice had a 80% reduction in the proportion of BrdU+CD115+CD11b+Ly6Chi monocytes compared with normoxic LPS-challenged counterparts (Fig. 4a). LPS equally reduced the proportion of blood BrdU+ neutrophils in hypoxic and normoxic mice compared with naive controls (Extended Data Fig. 4a), with a similar frequency of BrdU+ lymphocytes in all samples (Extended Data Fig. 4b). Examination of the BM stem cell compartment (Extended Data Fig. 4c) 24 h post-hypoxic exposure indicated a reduction in absolute numbers of LinSca-1+Kit+ (LSK) cells in hypoxic mice compared with normoxic mice independent of LPS treatment (Fig. 4b), with a specific reduction in the CD48+CD150 HPC-1 and CD48+CD150+ HPC-2 hematopoietic progenitor cells (HPCs; Fig. 4b), which have restricted multipotency25. Irrespective of oxygenation, LPS treatment reduced the absolute number of LincKit+Sca1CD127CD16/32CD34+ common myeloid progenitor cells (CMPs) compared with naive control mice (Fig. 4b).

Fig. 4: Systemic hypoxia alters BM hematopoiesis toward increased erythropoiesis.
figure 4

a, Representative dot plots gated on live CD45+Lin(CD3/CD19/Ly6G)CD115+Ly6Chi cells and proportion of BrdU+ monocytes in naive or LPS-challenged mice housed in normoxia or hypoxia for 24 h and pulsed with BrdU for the last 12 h (naive, n = 5–6; LPS treated, n = 8 in). b, Absolute numbers of BM LSK cells, CD48+CD150 HPC-1, CD48+CD150+ HPC-2 (n = 6 per group) and LincKit+Sca1CD127CD16/32CD34+ CMPs (n = 5–6 group) in mice as in a. c, Proportion of BM pre-GMs, pre-Meg-E, pre-CFU-E and CFU-E CD41CD32/16 cells in mice as in a (n = 6 per group). d,e, Representative UMAP analysis of BM cells gated on live CD45+LineageSca1C-Kit+CD41CD32/16 cells (d) and summary data of proportions of pre-GM and CFU-E cells (e) measured in the BM of mice treated with LPS and housed in normoxia (N) or hypoxia (H) for 5 d (n = 6 per group). Data are shown as the mean ± s.e.m. Each datapoint represents an individual mouse. Data in ae are pooled from two independent experiments. Statistical testing: for ac is by one-way ANOVA with Tukey’s multiple comparison test and for e by unpaired, two-tailed Student’s t-test.

As erythrocytes and monocytes originate from CMPs26 (Extended Data Fig. 4d), we investigated whether hypoxia shifted hematopoiesis in favor of red blood cell (RBC) production by measuring the effect of hypoxia on CMP progeny (LincKit+CD41CD32/16) 24 h after LPS challenge26. Hypoxia did not affect the number of CD150CD105 pre-granulocyte/monocyte precursors (pre-GMs) (Fig. 4c), but significantly increased the proportion of CD150+CD105 megakaryocyte-erythroid precursor (pre-Meg-E) cells (Fig. 4c), and led to an increase in downstream CD150+CD105+ pre-colony-forming unit erythroid precursor (pre-CFU-E) cells (Fig. 4c). Although hematocrit values were equivalent at this 24-h timepoint (Extended Data Fig. 4e), BM CFU-E cell proportions were reduced in LPS-treated hypoxic mice compared with normoxic counterparts (Fig. 4c), suggesting increased red cell egress. The hematocrit of LPS-treated hypoxic mice was increased compared with normoxic controls at day 5 (Extended Data Fig. 4f). UMAP (Uniform Manifold Approximation and Projection) analysis of LincKit+CD41CD32/16 cells within the BM compartment post-LPS indicated that skewing of the CMPs toward erythropoiesis was achieved by day 5 (Fig. 4d,e). Taken together, these data demonstrated that hypoxia altered hematopoiesis, reducing monocyte BM output.

Hypoxia suppresses type I IFN signaling

Next, we investigated the mechanism by which hypoxia suppressed monopoiesis in mice with ALI. Erythropoietin (EPO), an RBC growth factor, was significantly increased at 24 h (Fig. 5a) and day 5 post-LPS in hypoxia-housed compared with normoxia-housed mice (Fig. 5b). Interleukin (IL)-11, a hypoxia-responsive megakaryocyte and hematopoietic growth factor27,28, was also increased at day 5 post-LPS in hypoxic compared with normoxic mice (Extended Data Fig. 5a). Type I IFN (IFN-α and IFN-β) and type II IFN (IFN-γ) are known drivers of emergency monopoeisis29,30. Although IFN-β and IFN-γ were equivalent in hypoxic and normoxic LPS-treated mice, IFN-α was markedly reduced in hypoxic mice 24 h post-LPS challenge compared with normoxic counterparts (Fig. 5c and Extended Data Fig. 5b,c). Ifnar1–/– mice, which lack the type I IFN receptor, had a contraction of the LSK compartment (Fig. 5d) and increased proportion of the megakaryocyte/erythrocyte progenitor (MEP)/pre-CFU-E/CFU-E erythroid progenitors 24 h after LPS treatment during normoxia (Fig. 5e–g). Then, 5 d post-LPS challenge, normoxic Ifnar1–/– mice had enhanced numbers of circulating RBCs and were significantly monocytopenic compared with wild-type (WT) controls (Fig. 5h). In addition, normoxic Ifnar1–/– mice had similar numbers of neutrophils and CD64hiSiglecF+ macrophages (Fig. 5i), but reduced numbers of CD64hiSiglecF macrophages and MDMs (Fig. 5i) in the lung 24 h post-LPS, compared with WT controls.

Fig. 5: Hypoxia regulates type I IFN responses, hindering lung CD64hiSiglecF macrophage expansion in response to LPS.
figure 5

a,b, Serum EPO levels in mice challenged with LPS and housed in normoxia or hypoxia for 24 h (n = 9 normoxia, n = 7 hypoxia) (a) or 5 d (n = 7 per group) (b). c, Serum IFN-α in in mice challenged with LPS and housed in normoxia or hypoxia for 24 h (n = 6 per group). d, Proportion of LSK cells in the BM of WT or Ifnar1−/− mice 24 h post-LPS challenge (n = 4 per group). e, Manual gating of erythroid precursors and granulocyte/macrophage progenitors (GMPs) in LSK cells displayed on UMAP projection in mice as in d. f, Representative expression of CD32/16 (GMP marker) and CD150/CD105 (erythroid progenitor-associated markers) in LSK cells from mice as in d using the Pronk gating strategy26 displayed on UMAP projection. g, Proportion of erythroid progenitor cells (combined MEPs, pre-CFU-E and CFU-E cells) in WT and Ifnar1−/− BM 24 h post-LPS (n = 4 per group). h, Peripheral RBCs (n = 10 WT, n = 8 knockout) and monocyte counts at day 5 post-LPS in WT and Ifnar1−/− mice. i, Neutrophils, CD64hiSiglecF+CD11c+ macrophages, CD64hiSiglecF macrophages and MDM numbers in the lungs of WT and Ifnar1−/− mice 24 h post-LPS. j, Representative HIF-1α and DAPI expression in the femoral BM in mice challenged with LPS and housed in normoxia (N LPS) or hypoxia (H LPS) for 24 h. Scale bar, 20 μm. k, IFNAR expression in the BM LSK in naive (n = 5–6 per group) or LPS-treated mice (n = 6 per group) housed in normoxia or hypoxia for 24 h. l, Fold change in Quantitative PCR of Irf8, Irf1 and Ccr5 expression (normalized to actin-β, relative quantification) in BM cells from naive mice cultured in normoxia or hypoxia for 4 h ± IFN-β (n = 3 per group) relative to untreated normoxia control. Data represent the mean ± s.e.m. All datapoints represent individual mice. Statistical testing for a–d and gi was by unpaired, two-sided Student’s t-test, for k by one-way ANOVA with Tukey’s multiple comparison test and for l by two-way ANOVA with Šídák's multiple comparison post-test. The data in ac and di represent two independent experiments, and represent n = 3 per group in j and two independent pooled experiments in l.

Increased HIF-1α stabilization was observed in the BM of hypoxic, LPS-treated, mice compared with normoxic counterparts (Fig. 5j), in keeping with tissue hypoxia. As hypoxia can alter type I IFN signaling in cancer31, we next investigated interferon-α/β receptor (IFNAR) expression in the BM. LSK cells from hypoxic mice 24 h post-LPS treatment did not upregulate IFNAR expression, compared with normoxic mice (Fig. 5k). IFNAR expression was also reduced in blood CD115+CD11b+Ly6Chi monocytes from hypoxic LPS-challenged mice compared with normoxic counterparts (Extended Data Fig. 5d). BM cells from naive mice cultured in vitro, in hypoxia (1% O2), showed significant blunting of type I IFN-mediated Irf8, Irf1 and Ccr5 expression (Fig. 5l). Taken together, these findings suggested that hypoxia directly altered the immune response through local and systemic mechanisms.

Loss of monocyte recruitment associated with persistence of inflammation

We subsequently investigated the longer-term effects of systemic hypoxia on the myeloid compartment and inflammation resolution. Infiltrating neutrophils promote vascular injury, protein leak and alveolar epithelial damage in ARDS, and drive deleterious inflammatory responses in murine models of hypoxic ALI32,33. 5 d after LPS challenge, normoxic mice had very few neutrophils within the bronchoalveolar space whereas hypoxic mice showed evidence of ongoing inflammation with significant bronchoalveolar neutrophilia (Fig. 6a). Hypoxic LPS-treated mice had a reduction in the number of bronchoalveolar CD64hiSiglecF MDMs compared with normoxic counterparts (Fig. 6b). The total number of lung Ly6G+CD11b+ neutrophils was greater in hypoxic compared with normoxic mice at day 5 post-LPS (Fig. 6c) and, although the number of CD64hiSiglecF+CD11c+ macrophages had returned to baseline (Fig. 6c), the non-AM CD64hiSiglecF macrophages remained contracted (Fig. 6c), largely as a consequence of fewer CD64hiSiglecFLy6C+ MDMs (Fig. 6c). In addition, the BAL from hypoxic mice at this timepoint had higher CXCL1 and IL-6 (Fig. 6d), parameters reported to be elevated in ARDS nonsurvivors23. Moreover, hypoxic mice showed more sustained weight loss at day 5 post-LPS, compared with the normoxia-housed controls (Fig. 6e). Collectively, these data indicated that hypoxia-induced monocytopenia was associated with persistent lung inflammation.

Fig. 6: Ongoing CD64hiSiglecF macrophage expansion failure is associated with inflammation persistence in hypoxic ALI.
figure 6

a,b, Representative dot plots and absolute numbers of BAL neutrophils (a) and CD45+Ly6GCD64hiSiglecF MDMs gated on CD45hi cells (b) in mice treated with LPS and housed in normoxia or hypoxia for 5 d. c, Lung neutrophils, CD64hiSiglecF+CD11c+ macrophages, CD64hiSiglecF macrophages and CD64hiSiglecFLy6C+ MDM numbers in LPS-challenged mice housed in normoxia or hypoxia for 5 d. d, BAL CXCL1 and IL-6 levels in mice treated with LPS and housed in normoxia (N) or hypoxia (H) for 5 d (n = 6 N LPS, n = 7 H LPS). e, Daily weight changes from baseline in LPS-challenged mice housed in normoxia or hypoxia for 5 d (n = 4 per group). Data are shown as the mean ± s.e.m. Each datapoint represents an individual mouse. Statistical testing for ac was by one-way ANOVA with Tukey’s multiple comparison test and for d by unpaired, two-tailed Student’s t-test. Data in ac were pooled from three independent experiments and in d from two independent experiments.

CSF-1 accelerates the resolution of lung inflammation in hypoxia

Finally, we tested whether increasing the number of monocytes during hypoxia facilitated inflammation resolution. LPS-challenged hypoxic mice with absent (24 h) or low (5 d) baseline levels of macrophage-colony-stimulating factor (M-CSF) (Extended Data Fig. 6a,b) were treated with four daily injections of CSF-1–Fc fusion protein34 or phosphate-buffered saline (PBS). CSF-1–Fc markedly increased the number of CD115+CD11b+ monocytes and moderately increased the number of Ly6G+CD11b+ neutrophils in the blood at day 5 compared with PBS (Fig. 7a). CSF-1–Fc-treated, hypoxic, LPS-challenged mice also had increased numbers of CD64loCD11b+Ly6C+ monocytes and CD64hiSiglecF macrophages in the lung compared with mice receiving PBS (Fig. 7b). The number of CD64hiSiglecF+CD11c+ macrophages was not affected (Fig. 7b). Importantly, the absolute number of Ly6G+CD11b+ neutrophils in the lung tissue (Fig. 7b) and BAL (Fig. 7c) was reduced in CSF-1–Fc-treated, hypoxic, LPS-challenged mice at day 5, despite equivalent levels of CXCL1 in the BAL between CSF-1–Fc-treated and PBS-treated mice (Extended Data Fig. 6c). Hypoxic LPS-challenged mice treated with CSF-1–Fc had reduced weight loss (Extended Data Fig. 6d) and reduced immunoglobulin (Ig)M levels in the BAL fluid at day 5 (Fig. 7c) compared with PBS counterparts, suggesting reduced alveolar inflammation and vascular leak. To test these observations in a model of virally-induced epithelial injury, C57/BL6J mice were inoculated with influenza A virus (PR8) and placed in hypoxia immediately after. CSF-1–Fc or PBS was administered 12 h and 36 h post-PR8 challenge. Hypoxic PR8-infected mice treated with CSF-1–Fc had increased numbers of CD64loCD11b+Ly6Chi monocytes in the lung (Extended Data Fig. 6e) and improved physiological outcomes (Extended Data Fig. 6f) compared with mice receiving PBS. This was associated with a significant reduction in BAL protein levels, a marker of lung injury (Extended Data Fig. 6g) and lactate dehydrogenase activity, as an indicator of cellular damage (Extended Data Fig. 6h).

Fig. 7: CSF-1 rescues the hypoxic monocytopenia driving inflammation resolution.
figure 7

ac, Monocyte and neutrophil counts in the blood (a), monocyte, CD64hiSiglecF macrophage, CD64hiSiglecF+ CD11c+ macrophage and neutrophil counts in the lung (b), and absolute numbers of neutrophils and IgM titers in the BAL (c) in hypoxic LPS-challenged mice treated with four daily injections of PBS or CSF-1–Fc. d,e, Representative lung CD64hiSiglecFLy6C+ MDM CD45.2 and CD45.1 expression (d) and CD64hiSiglecFLy6C+ MDM:blood monocyte chimerism and CD64hiSiglecFLy6C macrophage:blood monocyte chimerism (e) in lung-protected, naive or LPS-challenged mice that are normoxia or hypoxia housed and treated with PBS or CSF-1–Fc. f, Differentially expressed genes in PBS- or CSF-1–Fc-treated, LPS-challenged mice in Ly6Chi blood monocytes at day 5 post-LPS challenge. g, Overlap between differentially downregulated genes in ARDS blood monocytes and genes upregulated in CSF-1–Fc-treated mice relative to PBS-treated mice. h, Comparison of Ly6Chi blood monocyte proteomes from hypoxic, LPS-challenged, CSF-1–Fc-treated mice relative to the PBS-treated counterparts (granule-associated proteins identified). i, Representative Lyve1 expression and number of lung CD64hiSiglecFLy6CLyve1+MHC-II macrophages in naive or LPS-challenged mice housed in normoxia or hypoxia and treated with PBS or CSF-1–Fc for 5 d. j, Chimerism of CD64hiSiglecFLy6CLyve1+MHC-II macrophages to blood monocytes in lung-protected, LPS-challenged chimeras housed in normoxia or hypoxia, treated with PBS or CSF-1–Fc. k,l, Serum IL-10 (k) and representative tiled immunofluorescence of lung sections stained for F4/80, IL-10 and DAPI (l), and F4/80, Lyve1 and DAPI in LPS-challenged mice housed in hypoxia for 5 d and treated with CSF-1–Fc or PBS. Scale bar, 200 μm. m, Differentially regulated genes in CD64hiSiglecFLyve1+MHC-II macrophages from LPS-challenged, CSF-1–Fc-treated mice relative to CD64hiSiglecFMHC-II macrophages from PBS-treated counterparts, housed in hypoxia for 5 d. np, Total lung CD64hiSiglecF macrophages and CD64hiSiglecFLy6CLyve1+MHC-II macrophages (n), BAL MDMs (o) and body-weight change (relative to baseline) (p) in LPS-challenged Ifnar1−/− mice treated with PBS (Ifnar1−/− PBS) or CSF-1–Fc (Ifnar1−/− CSF-1) for 5 d. Data represent the mean ± s.e.m. Each datapoint represents an individual mouse. Statistics for b,c,k and np were by unpaired, two-sided Student’s t-test, for i and j by a one-way ANOVA with Tukey’s post-test and for e by a two-tailed Mann–Whitney U-test following D’Agostino and Pearson’s normality test. For b and c data are pooled from three independent experiments, and for f,k,j and np data are pooled from two independent experiments. In g all genes have a fold-change >1, except H2-DMa, H2-DMb2, IL-17Ra and Nlrp3 where the fold-change is >0.5 and P < 0.05.

To dissect the mechanism by which CSF-1–Fc accelerated inflammation resolution, we tracked the ontogeny of CD64hiSiglecF macrophages in CD45.1+CD45.2+ C57BL6 mice reconstituted with CD45.2+ BM cells after lung-protected, single-dose irradiation. Then 8 weeks after BM reconstitution (Extended Data Fig. 6i), the mice were challenged with LPS, housed in normoxia or hypoxia for 5 d and treated with PBS or CSF-1–Fc daily. In keeping with increased recruitment from the blood, the proportion of CD64hiSiglecFLy6C+ MDMs derived from donor CD45.2+ BM cells in LPS-treated mice paralleled that seen in the blood (Fig. 7d,e). In the absence of LPS challenge, the proportion of CD45.2+CD64hiSiglecFLy6C macrophages relative to the CD45.2+CD115+CD11b+ blood monocyte pool was ~20%, indicating that maintenance of this subset was dependent on blood monocytes (Fig. 7e)35. In LPS-treated mice housed in normoxia, the proportion of CD45.2+CD64hiSiglecFLy6C macrophages relative to CD45.2+CD115+CD11b+ blood monocytes was ~80% (Fig. 7e), suggesting that the expansion of this population during inflammation was predominantly through recruitment of blood monocytes. The chimerism of this population was ~60% in hypoxic, LPS-challenged, PBS-treated mice and ~80% in hypoxic, LPS-challenged mice treated with CSF-1–Fc (Fig. 7e), indicating that CSF-1–Fc replenished the number of CD64hiSiglecF lung macrophages in LPS-treated hypoxic mice predominantly through increased recruitment of circulating CD115+CD11b+ monocytes.

In LPS-treated hypoxic mice, CSF-1–Fc treatment also normalized the hypoxic suppression of the type I IFN-associated gene Ccr5, without enhancing expression of Il1b (Fig. 7f). CSF-1–Fc significantly induced the expression of a number of genes that were suppressed in the ARDS patient samples, such as Itga5, Cd99, Sell and Anxa36 (Fig. 7g and Supplementary Table 1). A proteomic survey showed reduced abundance of secretory-granule proteins in circulating Ly6Chi monocytes of LPS-challenged, hypoxic, CSF-1-treated compared with PBS-treated counterparts (Fig. 7h), suggesting that CSF-1 altered their phenotype toward a less inflammatory profile. Increased numbers of lung MHC-IILyve-1+ macrophages, a subset of repair-associated macrophages reported in bleomycin-induced lung injury11, were observed in CSF-1–Fc-treated, hypoxic, LPS-challenged mice compared with their PBS counterparts (Fig. 7i and Extended Data Fig. 6j), with increased chimerism noted in the CSF-1–Fc-treated mice in the lung-protected chimera model (Fig. 7j)

To explore how CSF-1–Fc-mediated expansion of CD64hiSiglecF macrophages facilitated neutrophil clearance, we investigated the expression of known mediators of efferocytosis in the blood and lung37. IL-10 was elevated in the serum of CSF-1–Fc-treated, hypoxic, LPS-challenged mice compared with the PBS-treated counterparts (Fig. 7k). Immunofluorescence staining showed increased numbers of IL-10-expressing interstitial F4/80+ cells (Fig. 7l and Extended Data Fig. 7a), and F4/80+Lyve1+ macrophages (Fig. 7l and Extended Data Fig. 7b) in the lung of hypoxic, LPS-challenged, CSF-1–Fc-treated mice compared with their PBS-treated counterparts. Sorted CD64hiSiglecFMHC-IILyve1+ macrophages from these CSF-1–Fc-treated mice expressed the Il-10 transcript (Extended Data Fig. 8a). Transcriptional profiling indicated that CD64hiSiglecFMHC-IILyve1+ macrophages from CSF-1–Fc-treated, LPS-challenged, hypoxic mice had lower expression of archetypal inflammatory genes (Il1b, Il6, Tnf and Il18 and S100a11) and genes associated with lung fibrosis (Mmp8, Mmp12, Col14a1, Fpr1, Pdgfa and Pdgfb) relative to CD64hiSiglecFMHC-II from PBS-treated counterparts. Furthermore, genes reported to be enriched in severe SARS-CoV2 infection in BAL monocytes or macrophages, such as CD14, CCL3 and S100A8 (ref. 38), were suppressed in the CD64hiSiglecFMHC-IILyve1+ macrophages from the CSF-1–Fc-treated hypoxic mice compared to CD64hiSiglecFMHC-II macrophages from the PBS-treated counterparts (Fig. 7m), indicating that CSF-1 expanded the number of IL-10-producing macrophages and enabled resolution of hypoxia-driven inflammation.

Finally, we asked whether CSF-1 was sufficient to overcome the loss of CD64hiSiglecFLy6C+ MDMs and CD64hiSiglecF macrophages observed in the absence of type I IFN signaling. Normoxic LPS-challenged Ifnar1 / mice treated with CSF-1–Fc (four daily injections) had increased numbers of CD115+CD11b+ monocytes in the blood, Ly6Chi monocytes (Extended Data Fig. 8b,c) and CD64hiSiglecF macrophages in the lung (Fig. 7n) compared with those receiving PBS. In addition, these mice had increased lung CD64hiSiglecFLyve1+ macrophages (Fig. 7n) and BAL-recovered CD64hiSiglecF MDM numbers (Fig. 7o), with enhanced return to baseline body weight at day 5 post-LPS challenge (Fig. 7p) compared with LPS-challenged, PBS-treated Ifnar1−/− mice.


In the present study, we showed that monocyte recruitment and conversion into lung macrophages are required to drive inflammation resolution in hypoxic ALI. Hypoxemic mice with ALI demonstrated an increase in erythropoiesis, with an associated reduction in monopoiesis, monocytopenia and failure to expand the MDM and non-AM CD64hiSiglecFmacrophage compartment in the lung. In the context of prioritizing the preservation of tissue oxygen delivery, increased erythropoiesis makes physiological sense, such as in adaptation to altitude, where monocytopenia had been reported as early as 1969 (ref. 39). However, when engagement of an effective innate immune response is also required, our data demonstrated that hypoxia-induced immune changes observed in early disease have long-term consequences for inflammation resolution, such as persistence of neutrophilic inflammation, a well-known poor prognostic feature of ARDS3.

Monocytes are professional phagocytes and key mediators of the restoration of homeostasis. There is increasing appreciation that the phenotype of circulating monocytes can be predetermined by systemic cues affecting their BM progenitors40. The presence of a specific phenotypic profile in ARDS patients’ circulating monocytes, irrespective of the sampling timepoint, would be in keeping with alterations within their progenitors. It will be important, in future work, to explore whether the absence of systemic IFN-α, and suppressed IFNAR expression on BM LSK and circulating monocytes, as observed in our mouse model of hypoxic ALI and in hypoxemic patients with severe SARS-Cov2 infection41,42, may be sufficient to drive the phenotypic and functional changes observed in the circulating monocytes of a hypoxic ARDS cohort.

Expanded numbers of airway monocytes and macrophages in the BAL of patients with severe SARS-Cov2 infection have been reported38, although it is unclear whether this expansion is promoting or limiting disease pathogenesis. An important limitation of our present study is the inability to sample the lung macrophage compartment in patients with ARDS. Transcriptomic survey of the airway monocytes and macrophages in patients with severe SARS-Cov2 infection identified an enrichment of markers of immaturity, inflammatory proteins and cytokines38. We show that treatment with CSF-1 increased the number of CD64hiSiglecF macrophages and monocyte differentiation toward CD64hiSiglecFMHCLyve1+ macrophages, a cell type that has tissue-repair roles in various disease contexts43, including in the lung11. In addition, CSF-1–Fc suppressed the expression of several genes reported to be enriched in the BAL of patients with severe SARS-Cov2 infection38. These findings, compounded by the effect of CSF-1–Fc on accelerating inflammation resolution and reducing lung injury, underscored the therapeutic potential of CSF-1 in ARDS. The use of other growth factors has been trialed in ARDS without success, with systemic-44 or lung-delivered45 granulocyte–macrophage colony-stimulating factor (GM-CSF) failing to improve ARDS mortality. Although GM-CSF plays a key role in AM homeostasis, its pleotropic nature means that it can also act as a neutrophil chemoattractant and growth factor46, with neutrophils pathogenic in the context of ARDS and hypoxia promoting neutrophil survival and proinflammatory function32,33. On the other hand, monocytes and MDMs can directly inhibit neutrophil-mediated damage to the host47. In our system, CSF-1–Fc treatment expanded the CD64hiSiglecF macrophage compartment and led to an increase in lung IL-10+ macrophages, with a concomitant increase in systemic IL-10, and reduced numbers of neutrophils in the airspace and the lung. Macrophages are known to release IL-10 on efferocytosis, which drives the resolution of inflammation37. Furthermore, IL-10-producing lung IMs are key regulators of both allergic8,9 and endotoxin-mediated lung injury10. These findings strengthen the case of the therapeutic potential of CSF-1 in human ARDS.


Resources availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, A.M. (

Human healthy control blood donors

Patients with ARDS were recruited and informed consent obtained directly or by proxy under the ‘META-CYTE’ study (17/SS/0136/AM01) and ‘ARDS-NEUT’ study (20/SS/0002), as approved by the Scotland A Research Ethics Committee. Samples were also obtained under the ‘Effects of Critical Illness on the Innate Immune System’ study as approved by Health Research Authority (REC no. 18/NE/0036).

All healthy participants gave written informed consent in accordance with the Declaration of Helsinki principles, with ACCORD Medical Ethics Research Committee approval for the study of healthy human volunteers through the University of Edinburgh Centre for Inflammation Research blood resource (15-HV-013).

Up to 20–40 ml of whole blood was collected into citrate tubes and up to 10 million cells were stained for flow cytometry assessment and sorting. Briefly, the whole blood was treated with red cell lysis buffer (Invitrogen) and cells counted before staining for flow cytometry. Cells were incubated with anti-CD16/32 Fc block (2:50) for 30 min, followed by staining for 30 min with antibodies (Table 2) followed by a wash with FACS buffer (PBS + 2% fetal calf serum (FCS)). DAPI (1:1,000) was added before flow cytometry to determine live cells. Monocytes were identified as Singles DapiCD45+ nongranulocyte Lin(CD3/CD56/CD19±CD66b) HLADR+ CD14+ and/or CD16+ cells.

Table 2 List of antibodies

Samples obtained from April 2020 were fixed before acquisition given the potential for SARS-Cov2 dissemination. Briefly, 1 μl of Zombie Aqua fixable viability dye (stock 1:20 dilution) was added to 100 μl of whole blood for 15 min at room temperature in the dark. Then, 2 μl of Fc Block was added for a further 30 min, on ice. Samples were then stained as above and fixed/lysed using BD FACS Lyse for 10 min at room temperature. The sample was then resuspended in 300 μl of FACS buffer and 50 μl of Countbright beads added (Thermo Fisher Scientific) before acquisition.


Male C57/BL6J mice aged 8–15 weeks were purchased from Envigo or Charles River. Ifnar1/ (ifnar tm/agt) mice were obtained from J.S. who purchased them originally from the Jackson Laboratory. Animal experiments were conducted in accordance with the UK Home Office Animals (Scientific Procedures) Act of 1986 with local ethical approval.

Mouse LPS ALI model

Mice were treated with nebulized LPS (3 mg) and then housed in normoxia or hypoxia (10% O2) immediately thereafter for up to 5 d. Mice were treated daily (days 1–4 post-LPS), by subcutaneous injection, with PBS or 0.75 mg kg−1 of porcine CSF-1 fused to the Fc region of porcine IgG1a (generated by David Hume), prior to cull on day 5.

D39 S. pneumoniae infection

Mice were anesthetized and 107 colony-forming units (c.f.u.) (or vehicle) was delivered in 50 μl of PBS via intratracheal intubation. After reversal of the anesthetic and a period of recovery, the mice remained in normoxia or were placed in hypoxia.

Influenza A (PR8) virally induced ALI model

Mice were lightly anesthetized using isoflurane and 20 plaque-forming units (p.f.u.) of PR8 influenza A virus in Dulbecco’s modified Eagle’s medium (DMEM) was inoculated intranasally. After 1 h of recovery time, mice were placed in hypoxia for 48 h. Subcutaneous PBS or CSF-1–Fc injections (as above) at 12 h and 36 h were administered. Sickness scores were determined using methods described previously32.

Lung and alveolar cell sampling

Mice were culled with an overdose of intraperitoneal anesthetic (Euthetal) followed by blood collection from the inferior vena cava. Alveolar leukocytes were collected by BAL, then mice were perfused lightly with PBS through the heart, before harvesting the lung tissue. On occasion, the lower limbs were harvested for BM leukocyte assessment (see below).

Tissue leukocytes were extracted from surgically dissociated lung tissue by enzymatic digestion with 2 ml of enzyme mix (RPMI with 0.625 mg ml−1 of Collagenase D (Roche), 0.85 mg ml−1 of Collagenase V (Sigma-Aldrich), 1 mg ml−1 of dispase (Gibco, Invitrogen) and 30 U ml−1 of DNase (Roche Diagnostics GmbH)) for 45 min at 37 °C in a shaking incubator. The digest material was passed through a 100-μm cell strainer with the addition of FACS buffer (PBS with 0.5% BSA/2% FCS and 0.02 mM EDTA). Cell pellets were treated with red cell lysis buffer (Sigma-Aldrich) and washed in FACS buffer. The resulting cell suspension was subsequently passed through a 40-μm strainer before cell counting using a Casey TT counter (Roche). Single-cell suspensions (5 million cells per sample) were then stained for flow cytometry. BAL samples were counted before staining for flow cytometry.

Blood and BM sampling

Mouse blood and BM were treated with RBC lysis buffer (BioLegend) before counting and staining for flow cytometry (Table 2).

Hematopoietic cell assessment was performed using both hind legs, which were crushed using a pestle and mortar until a homogeneous cell suspension was achieved, or flushed though using a 32G needle. Cells were collected in cold FACS buffer and filtered through a 70-μm nylon strainer (BD Falcon, catalog no. 352340). Cells were treated with RBC lysis buffer (BioLegend) before staining.

Tissue-protected chimeras

C57BL/6J CD45.1+CD45.2+ mice aged 6–8 weeks were anesthetized and irradiated with a single dose of 9.5-Gy γ-irradiation, with all but the hind legs and lower abdomen protected by a 5cm lead shield. The next day, the mice received 2 × 106–5 × 106 BM cells from CD45.2+ C57BL/6J by intravenous injection. The chimerism of blood monocytes (proportion of donor cells) was determined by flow cytometry in each individual mouse at day 5 and the chimerism in the lung macrophage populations (as described in the figures) was divided by this reference value, thereby determining the proportion of the cells that were of blood ontogeny.

Flow cytometry

Mouse cells were treated with α-CD16/32 Fc block (eBioscience) (1:100) before staining with antibodies (Table 2). Relevant fluorescence − 1 samples were used as controls. Zombie Aqua fixable viability dye (BioLegend) was used before Fc Block to exclude dead cells from digest samples or DAPI for single-cell suspensions.

Cells were acquired on the LSRFortessa (Becton Dickinson) or sorted on an Aria II or Fusion machine (Becton Dickinson). Compensation was performed using BD FACSDiva software and data analyzed in FlowJo v.10 or FCS Express 7 for t-distributed stochastic neighbor embedding analysis.

Gating strategies

  • Human monocytes: Singles DapiCD45+non-granulocyte Lin(CD3/CD56/CD19/CD66b) HLADR+ CD14+ and/or CD16+ cells.

  • Mouse blood monocytes: Singles DapiCD45+ Lin(CD3/CD19/ Ly6G)CD115+CD11b+Ly6Chi, Ly6Cint or Ly6C.

  • Mouse blood neutrophils: Singles, DapiCD45+Ly6G+CD11b+Ly6Cint.

  • Mouse lung/BAL alveolar macrophages: Singles, Zombie AquaCD45+Lin (CD3/CD19/Ly6G)CD64hiSiglecF+CD11c+.

  • Mouse lung interstitial /BAL inflammatory macrophages: Singles, Zombie AquaCD45+Lin (CD3/CD19/Ly6G)CD64hiSiglecFCD11c+/− then Ly6C+/−MHC-II+/−.

  • Lung classical monocytes: Singles, Zombie AquaSinglesCD45+Lin (CD3/CD19/Ly6G)CD64lo CD11b+Ly6C+.

  • Lung/BAL neutrophils: Singles, Aqua or Dapi, CD45+CD11b+Ly6G+.

  • Lung cDC1 subset: Zombie Aqua-Singles CD45+CD11chi,CD103+, CD64MHC-II+.

  • BM HSPC SLAM analysis Alive: Singles LK (Lin-cKit+) and LSK (Lin-cKit+Sca-1+) cells. LSK cells were further sub-gated on hematopoietic stem cells (HSCs: LSK CD48CD150+), multipotent progenitors (MPPs: LSK CD48CD150), HPC-1 (LSK CD48+CD150) and HPC-2 (LSK CD48+CD150+).

  • BM erythroid progenitors based on Pronk analysis26: Singles, Dapi or Aqua, Lin, CD11b, cKit+, Sca1, CD32/16, CD41, CD105+ or CD150+ (pre-Meg-E CD150+CD105, pre-CFU-E CD150+CD105+, CFU-E CD150CD105+).

Further gating strategy information can be made available on request.

BAL/serum cytokine/chemokine quantification

BAL and serum supernatants were collected and stored at −80 °C until use. Cytokine and chemokine levels were measured using an MSD V-plex plate per the manufacturer’s instructions.

Lung injury measurements

IgM BAL levels were measured using the Ab133047 Abcam kit as per the manufacturer’s instructions.

BAL lactate dehydrogenase activity (measured as colorimetric reduction of NAD to NADH) was performed using Ab102526 (Abcam) as per the manufacturer’s instructions.

BAL total protein was measured using Pierce BCA Assay (Thermo Fisher Scientific) as per the manufacturer’s instructions.

In vitro BM culture

Naive WT C57BL/6 BM was obtained by flushing the femoral and tibial bones and RBCs were lysed. Cells were cultured in hypoxia (FiO2 1%) or normoxia (FiO2 21%) with conditioned DMEM for 1 h before the addition of IFN-β 10 ng ml−1 (RnD 8234-MB-010) for a further 3 h. Cell pellets were collected and QIAGEN RLT buffer added (containing 10 μl ml−1 of 2-mercaptoethanol). Pellets were snap-frozen and stored at −80 °C for RNA extraction.

RNA isolation and relative quantification

RNA was isolated from BM cells using the genomic DNA eliminator solution for purification of total RNA (RNeasy Plus Mini Kit, QIAGEN). Complementary DNA was synthesized using AMV reverse transcriptase with random primers (Promega). TaqMan gene expression assays (Applied Biosystems, Thermo Fisher Scientific) and PrimeTime qPCR Probe Assays (IDT) were used for relative quantification of cDNA using SDS 2.4 (Thermo Fisher Scientific) and normalized to ACTB expression.


Murine paraffin-embedded blocks were prepared from lungs fixed via the trachea with 10% buffered formalin. The lung sections were stained with anti-IL-10 (catalog no. ab189392, Abcam), anti-F4/80 (catalog no. ab6640, Abcam) or isotype control after deparaffinization and antigen retrieval. Antigen retrieval was performed by microwave heating in citric acid-based, antigen-unmasking solution (Vector, catalog no. H-3300-250). The following were used: tyramide signal amplification (TSA) plus system amplification (catalog no. NEL744B001KT, Perkin Elmer) and autofluorescence quenching with TrueView (Vector, catalog no. SP-8400). The nuclei were stained with DAPI (catalog no. 422801, Sigma-Aldrich). Images were obtained using EVOS FL Auto 2 (Invitrogen). All image acquisition and processing steps were performed using the same settings for both sample groups.

The lung sections were stained with anti-mouse LYVE-1 (catalog no. 103-PA50AG, ReliaTech GmbH) and anti-mouse F4/80 (catalog no. ab6640, Abcam) overnight at 4 °C after deparaffinization and antigen retrieval. Antigen retrieval was performed by microwave heating in citric acid-based, antigen-unmasking solution (Vector, catalog no. H-3300-250). The following were used: TSA plus system amplification (catalog no. NEL744B001KT, Perkin Elmer) and autofluorescence quenching with TrueView (Vector, catalog no. SP-8400) according to the manufacturer’s instructions. The nuclei were stained with DAPI (catalog no. 422801, Sigma-Aldrich). Images were acquired using a EVOS FL Auto 2 (Invitrogen).

All image acquisition and processing steps were performed using the same settings for both sample groups.

The nCounter NanoString platform analysis

For human monocytes, 5,000 HLADR++ cells were sorted using the aforementioned human monocyte gating strategy directly into 2 μl of RLT buffer using a BD Fusion Sorter (patients 4–8 were sampled). 5,000 mouse classical monocytes were sorted from mice treated with LPS and housed in normoxia, hypoxia and hypoxia + CSF-1 gating on single DAPICD45+LinCD115+Ly6Chi cells into 2 μl of RLT. Cell pellets were vortexed and centrifuged before immediate freezing until ready for processing. NanoString gene expression plates of human and mouse myeloid inflammation were run as per the manufacturer’s instructions at the University of Edinburgh HTPU Centre within the MRC Institute of Genetics and Molecular Medicine/Cancer Research UK Edinburgh Centre.

Proteomic analysis

Sorted classical monocytes were processed for proteomics using the ‘in-cell digest’, as described by Kelly et al.48, resuspended in digestion buffer (0.1 M triethylammonium bicarbonate + 1 mM MgCl2) and digested with benzonase (>99%, Millipore) for 30 min at 37 °C, followed by trypsin (Thermo Fisher Scientific, 1:50 w:w protein) overnight at 37 °C. A second aliquot of trypsin (1:50) was subsequently added and incubated at 37 °C for 4 h. A minimum of 25 ng of trypsin was added. Digests were acidified and desalted using StageTips49 and subjected to either tip-based fractionation or direct analysis by liquid chromatography–tandem mass spectroscopy (LC–MS/MS).

After digestion, and to generate the reference spectral library, peptides were subjected to reverse-phase, high pH, tip fractionation following the general guidelines described by Rappsilber et al.49. In brief, tips for fractionation were made using three SDB-XC disks (Merck) per tip. The tip was cleaned and conditioned using, sequentially, methanol, 80% acetonitrile (MeCN) (Thermo Fisher Scientific) in 0.1% NH4OH (v:v), and 0.1% NH4OH (52 mM) (v:v). Peptides, also resuspended in 0.1% NH4OH, pH 10, were spun through the SDB-XC disks and the flow-through was collected, acidified and concentrated on C-18 StageTips before being subjected to MS analysis. Fractionation was then achieved by sequential elution with 7%, 14%, 21%, 28%, 35%, 55% and 80% MeCN in 0.1% NH4OH. Fractions were then dried at ambient temperature (Concentrator 5301, Eppendorf) and prepared for MS analysis by resuspension in 6 μl of 0.1% trifluoroacetic acid (TFA).

Data-dependent acquisition LC–MS analyses were performed on an Orbitrap Fusion Lumos Tribrid Mass Spectrometer (Thermo Fisher Scientific) coupled, on-line, to an Ultimate 3000 HPLC (Dionex, Thermo Fisher Scientific). Peptides were separated on a 50-cm (2-µm particle size) EASY-Spray column (Thermo Fisher Scientific), which was assembled on an EASY-Spray source (Thermo Fisher Scientific) and operated constantly at 50 °C. Mobile phase A consisted of 0.1% formic acid in LC–MS-grade water and mobile phase B consisted of 80% acetonitrile and 0.1% formic acid. Peptides were loaded on to the column at a flow rate of 0.3 μl min−1 and eluted at a flow rate of 0.25 μl min−1 according to the following gradient: 2–40% mobile phase B in 120 min and then to 95% in 11 min. Mobile phase B was retained at 95% for 5 min and returned back to 2% a minute after until the end of the run (160 min in total).

The spray voltage was set at 2.2 kV and the ion capillary temperature at 280 °C. Survey scans were recorded at 60,000 resolution (scan range 400–1,600 m/z) with an ion target of 1.0 × 106 and injection time of 50 ms. MS2 was performed in the orbitrap (resolution at 15,000), with an ion target of 5.0 × 104 and higher-energy C-trap dissociation (HCD) fragmentation50 with a normalized collision energy of 27. The isolation window in the quadrupole was 1.4 Thomson. Only ions with a charge between 2 and 6 were selected for MS2. Dynamic exclusion was set at 60 s. The cycle time was set at 3 s.

Samples subjected to data-independent acquisition (DIA) were prepared for MS analysis by resuspension in 0.1% TFA. MS analyses were performed on an Orbitrap Fusion Lumos Tribrid Mass Spectrometer (Thermo Fisher Scientific). LC conditions (instrumentation, column and gradient) were the same as described above.

Survey scans were performed at 15,000 resolution, with a scan range of 350–1,500 m/z, maximum injection time 50 ms and AGC target 4.5 × 105. MS/MS DIA was performed in the orbitrap at 30,000 resolution with a scan range of 200–2,000 m/z. The mass range was set to ‘normal’, the maximum injection time to 54 ms and the AGC target to 2.0 × 105. The inclusion mass list with the corresponding isolation windows is shown in Table 3. Data for both survey and MS/MS scans were acquired in profile mode. A blank sample (0.1% TFA, 80% MeCN, 1:1 v:v) was run between each sample to avoid carryover.

Table 3 List of masses, default charge states and isolation windows used for data independent acquisition

MS raw data files were processed using Spectronaught v.14.7.201007.47784 with either a human or a mouse reference FASTA sequence from UniProt, using default search parameters. The resulting protein-level data were analyzed using R v.3.5.0. Protein parts per million (p.p.m.) intensities were calculated by dividing the mean p.p.m. intensities between conditions (for example, for the human monocyte samples, ARDS patients and healthy controls), the P values were calculated using a Student’s t-test on log(transformed p.p.m. intensities). Proteins were designated as significantly changing if they showed P values <0.05 and fold-changes exceeding 1.96 s.d. away from the mean (that is, z-score >1.96). Only proteins that were quantified in all samples are shown in the volcano plot.

Gene expression analysis

Normalization of data was carried out using the geNorm selection of housekeeping genes function on NanoString nCounter analysis software. The resulting log2(normalized values) were used in subsequent analyses. Differential genes (‘DE genes’) were defined as genes with log2(fold-change) > 1, P < 0.05 across sample groups. Hierarchical clustering of sets of DE genes was carried out using Euclidian and Ward methods based on Pearson’s correlation values across transcriptional scores. The z-score scalar normalization of data was applied to the data before plotting as heatmaps. Analyses, including the drawing of heatmaps and volcano plots, were carried out in R using the package ggplot2 ( Analysis of datasets was carried out by Thomson Bioinformatics, Edinburgh, UK.

Quantification, statistical analysis and reproducibility

Statistical tests were performed using Prism 8.00 and 9.0.2 software (GraphPad Software Inc.) (specific tests detailed in figure legends). Significance was defined as a P < 0.05 (after correction for multiple comparisons where applicable). Sample sizes (with each n number representing a different blood donor for human cells or an individual mouse for animal experiments) are shown in each figure.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.