The study was approved by the local ethic committee (EA4/025/18), and healthy blood donors provided written informed consent. Neutrophils and monocytes were isolated from heparinized venous whole-blood samples using density-gradient centrifugation as described previously (30, 31). In brief, 1% dextran was added for red blood cell sedimentation followed by Ficoll-Hypaque density gradient centrifugation of the resulting supernatant. Monocytes were isolated from the interphase and remaining erythrocytes in the pellet containing neutrophils were hypotonically lysed for 15 s before normal osmolality was achieved through the addition of 3.6% sodium chloride solution. Neutrophils were centrifuged (10 min at 1,050 rpm), and isolated monocytes and neutrophils were resuspended in HBSS⁺⁺ (Gibco, Waltham, USA) and counted using a Beckman Coulter system with a purity of >95%.

Roxadustat (TargetMol, Wellesley Hills, USA), YC1, LPS serotype O111:B4, Actinomycin D, rapamycin, 4EGI-1, and cycloheximide (Merck, Darmstadt, Germany) were diluted in DMSO. Human recombinant TNFα, GM-CSF, IL6, IL8 (Bio-Techne GmbH, Wiesbaden, Germany), and G-CSF (PeproTech, Waltham, USA) were diluted in PBS (Gibco, Waltham, USA) with 0.1% bovine serum albumin (BSA, Merck). fMLP (Merck) was diluted in water. Monoclonal antibodies blocking integrin activation were mouse anti-human IgG CD11b clone 2LPM19c (#sc-20050, RRID:AB_626883, Santa Cruz Biotechnology, Heidelberg, Germany) and mouse anti-human IgG CD18 clone 7E4 (#IM1567, RRID:AB_131640, Beckman Coulter, Krefeld, Germany). The integrin activating antibodies were mouse anti-human IgG CD11b clone Bear1 (#IM0190, RRID:AB_3095685, Beckman Coulter) and mouse anti-human IgG CD18 clone MEM-148 (#MCA2086XZ, RRID:AB_323901, Bio-Rad, Feldkirchen, Germany). The murine IgG₁ isotype control was from Sigma-Aldrich (#M5284, RRID:AB_1163685, Merck). The blocking PECAM-1 antibody was mouse monoclonal anti-human IgG CD31 clone Gi18 (#ALX-805-003A-C100, RRID:AB_2051038, Enzo Life Sciences, Lörrach, Germany). The exclusion of reagent cytotoxic side effects is shown in Supplementary Figure 1 .

For the preparation of whole-cell lysates, 1.25×10⁶ monocytes or 2.5×10⁶ neutrophils were resuspended in 50 µl of sonication buffer (150 mM Tris, pH 7.8, 1.5 mM EDTA, and 10 mM KCl) supplemented with protease inhibitors (Protease Inhibitor Cocktail III, 2 mM PMSF, 1 mM Na₃VO₄, 0.5 mM DTT, 2 mM levamisole, cOmplete™ 25×), sonicated in a Bioruptor® Plus sonication device (Diagenode, Seraing, Belgium) for 10 min, and centrifuged at 13,000 g for 10 min at 4°C. Whole-cell lysates in the supernatants were collected and boiled with the appropriate volume of reducing loading buffer (4× ROTI® LOAD, Carl Roth GmbH+Co.KG, Karlsruhe, Germany) at 95°C for 5 min after photometric protein quantification using ROTI® Nanoquant 5× reagent (Carl Roth GmbH+Co.KG) and a VERSAmax™ microplate reader (Molecular Devices, Sunnyvale, USA).

For HIF1α and HIF2α, 10 µg (monocytes) or 30 µg of protein (neutrophils) per sample were loaded onto 8% SDS-polyacrylamide gels and transferred to PVDF membranes (pore size 0.2 µM, Thermo Scientific, Waltham, USA). For other proteins, we used 15% SDS-polyacrylamide gels for gel electrophoresis. Membranes were blocked with 5% skimmed milk for 1 h before overnight incubation with the appropriate dilution of primary antibody in 5% BSA-containing buffer at 4°C. The following primary antibodies were used: rabbit polyclonal anti-human IgG HIF1α (1:500, #3716, RRID:AB_2116962, Cell Signaling Technologies, Leiden, The Netherlands), monoclonal mouse anti-human IgG HIF2α clone ep190b (1:500, #NB100-132, RRID:AB_10000898, Novus Biologicals, Wiesbaden, Germany), polyclonal rabbit anti-human IgG HIF2α (1:500, #NB100-122, RRID:AB_535687, Novus Biologicals), polyclonal rabbit anti-human IgG HIF2α (1:500, # PA1-16510, RRID:AB_2098236, Thermo Fisher Scientific), monoclonal rabbit anti-human IgG β-actin clone 13E5 (1:2000, #4970, RRID:AB_2223172), monoclonal rabbit anti-human IgG phospho-STAT3 (Tyr705) clone D3A7 (1:1000, #9145, RRID:AB_2491009), monoclonal mouse anti-human IgG STAT3 clone 124H6 (1:1000, #9139, RRID:AB_331757), polyclonal rabbit anti-human IgG phospho-eIF4E (Ser209) (1:1000, #9741, RRID:AB_331677), monoclonal rabbit anti-human IgG phospho-4EBP1 (Ser65) clone D9G1Q (1:1000, #9451, RRID:AB_330947), monoclonal rabbit anti-human IgG mTOR clone 7C10 (1:1000, #2983, RRID:AB_2105622), and polyclonal rabbit anti-human IgG phospho-mTOR (Ser2448) (1:1000, #2971, RRID:AB_330970, all from Cell Signaling Technologies). The secondary HRP-conjugated antibodies were diluted in 5% skimmed milk 1:1000-1:5000 (rabbit anti-mouse IgG from Agilent Technologies Denmark, #P0260, RRID:AB_2636929, and donkey anti-rabbit IgG from Cytiva, #NA934V, RRID:AB_772206). Chemiluminescence was detected on a VWR ECL reader using SuperSignal™ West Dura Extended Duration Substrate (Thermo Fisher Scientific). All bands presented were at the predicted molecular weight for the protein of interest. Densitometric analysis was performed using ImageJ (RRID: SCR_003070). Fold changes of protein abundance were calculated from optical density ratios of the target protein and the loading control was normalized to the ROX-treated sample (or as specified in the figure legends).

PolyHema (poly 2-hydroxyl-ethyl methacrylate, Merck) was dissolved in 95% ethanol at 80 mg/ml and 500 µl per well was pipetted on a 12-well tissue culture plate (TPP Techno Plastic Products AG, Trasadingen, Switzerland) and left to dry overnight. Fibronectin (Roche, Mannheim, Germany) was dissolved at 20 µg/ml in HBSS⁺⁺ before the wells were coated for 1 h at room temperature.

Fibronectin-coated sterile transwells (3 µm pore diameter, Sarstedt, Germany) were inserted into a tissue culture plate containing HBSS⁺⁺, chemoattractant, and ROX in the appropriate experimental conditions. Neutrophils were pipetted in the transwell and were allowed to migrate to the lower well at 37°C with 5% CO₂. After 4 h, migrated neutrophils were either collected for protein extraction (see immunoblotting) or lysed using 0.5% Triton X-100 for the measurement of MPO activity using a 2,2′-Azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)-based (Merck) colorimetric assay on a VERSAmax™ microplate reader (Molecular Devices).

Neutrophils were resuspended in Roswell Park Memorial Institute (RPMI) medium (Gibco) supplemented with autologous serum and cultured for 4 h/20 h at 37°C with 5% CO₂ with 2.5×10⁶ cells per well. Neutrophil adhesion was induced by fibronectin-coated wells. Apoptotic neutrophils were washed, transferred to Annexin binding buffer (BD Biosciences, Heidelberg, Germany), and co-stained with Annexin V-APC (BD Biosciences) and 7-AAD (Merck). FACS analysis was performed immediately after staining on a BD CANTO II cytometer. FACS data were analyzed on FlowJo version 7 (Treestar, USA).

For qPCR, RNA was transcribed into cDNA after deoxyribonuclease I-treatment (Qiagen, Venlo, The Netherlands) using hexanucleotide primers and the RevertAid First Strand cDNA Synthesis Kit following the manufacturer’s protocol (Thermo Fisher Scientific). qPCR was performed with the Fast SYBR Green Master Mix or the TaqMan Fast Universal PCR Master Mix (Applied Biosystems, Waltham, USA) and run on a QuantStudio plus machine (Applied Biosystems). Primers for qPCR were designed with Primer 3 and were as follows: HIF1α (forward, 5′-CATAAAGTCTGCAACATG GAAGGT-3’; reverse, 5′-ATTTGATGGGTGAGGAATGGGTT-3′), 18S (forward, 5′-ACATCCAAGGAAGGCAGCAG-3′; reverse, 5′-TTTTCGTCACTACCTCCCCG-3′, 5′-6-FAM-CGCGCAAATTACCCACTCCCGAC-TAMRA-3′), and VEGFA (forward, 5′-GAGGAGGGCAGA ATCATCAC-3′; reverse, 5′-ACACAGGATGGCTTGAAGATG-3′). Quantitation was performed using the ΔΔCT-method using the ROX-treated sample as the reference sample.

Results are given as means ±standard deviation. Statistical comparisons were made using the ratio-paired t-test for 2 groups and one-way ANOVA for experimental groups of >2 using GraphPad Prism9 software. Multiple testing correction was performed using Šidák’s method. Differences were considered significant at p<0.05 (*), p<0.01 (**), p<0.001 (***), or p<0.0001 (****).

We first assessed HIF1α in neutrophils both resting and exposed to a variety of inflammatory mediators during culture in test tubes. Under normoxia, we did not detect HIF1α protein in the former and observed a marginal HIF1α signal at most in the latter ( Figure 1A ). Under pseudohypoxia with ROX-dependent PHD inhibition, HIF1α protein significantly increased in resting neutrophils. Moreover, IL8 acted additively, and GM-CSF and LPS even synergistically to pseudohypoxia in HIF1α protein upregulation. We selected 15 µM of ROX because reported blood concentrations in patients range from 2.8–28 µM (32–34) and determined a 4 h incubation period based on a time course study ( Supplementary Figure 2A ).

When we investigated isolated human blood monocytes in test tubes, we found that pseudohypoxia, similar to neutrophils, reliably led to HIF1α protein accumulation in resting and stimulated cells but, in contrast to neutrophils, none of the inflammatory mediators showed additive or synergistic effects with ROX ( Figure 1B ).

We did not detect HIF2α protein in resting and stimulated neutrophils as well as monocytes under pseudohypoxia using three different antibodies that detected HIF2α in appropriate control lysates ( Supplementary Figures 2B, C ).

Thus, inflammatory mediators were weak myeloid cell HIF1α activators at most but acted additively or even synergistically with PHD inhibition in human neutrophils. Based on these findings, we focused on HIF1α regulation in neutrophils and selected GM-CSF as a synergistic inflammatory stimulus for further experiments.

Neutrophils encounter their inflammatory challenges either when circulating in the bloodstream or in inflamed tissues where they interact with extracellular matrix proteins. We tested the hypothesis that adhesion provides an essential factor for HIF1α protein induction limiting HIF1α effects to emigrating neutrophils. We applied two distinct conditions, namely neutrophil incubation on fibronectin (FN)-coated plates to promote extracellular matrix interactions, and PolyHema-coated plates (S) to prevent adhesion and reduce cell-cell contacts (35) mimicking bloodstream neutrophils. We confirmed that the incubation of GM-CSF-treated neutrophils on FN resulted in cell adhesion and spreading that was not observed on PolyHema ( Figures 2A, B ). We then compared these two biologically relevant conditions with our initially used assays in test tubes (T) that explicitly allow for plastic and cell-cell contact. Culturing neutrophils on FN did not result in HIF1α protein per se neither in resting nor in GM-CSF-treated cells. However, we detected HIF1α protein in resting and, fourfold stronger, in GM-CSF-treated neutrophils on FN under pseudohypoxia. Notably, the synergistic ROX/GM-CSF HIF1α effect was even more pronounced on FN than with tubes (T). In sharp contrast, HIF1α protein was not induced by pseudohypoxia under stringent suspension conditions on PolyHema (S), neither in resting nor in GM-CSF-stimulated neutrophils ( Figure 2B ). Based on these findings, we explored the role of β₂-integrins in HIF1α induction. Blocking antibodies to CD11b or CD18 prevented HIF1α induction in GM-CSF-stimulated neutrophils incubated under pseudohypoxic conditions in tubes (T) and on FN ( Figures 2C, D ). Conversely, activating β₂-integrin antibodies enabled HIF1α protein expression in ROX/GM-CSF-treated suspension neutrophils on PolyHema (S, Figure 2E ). Because VEGFA is a target gene for the HIF1α transcription factor, we validated the β₂-integrin dependent HIF1α pathway activation by measuring VEGFA mRNA. We observed that transcription of the HIF1α target gene VEGFA was only upregulated in ROX- and GM-CSF/ROX-treated adherent neutrophils on FN, whereas no significant regulation was found in suspended neutrophils on PolyHema that lacked integrin-elicited HIF1α protein expression ( Supplementary Figure 3 ).

We next investigated whether normobaric hypoxia produced similar results as pharmacological PHD inhibition. Culturing neutrophils in 1% O₂ on FN, but not in normoxia (21% O₂) on FN, induced HIF1α protein in resting neutrophils and approximately 8.5-fold more in GM-CSF-treated neutrophils. Similar to pseudohypoxia, HIF1α protein in GM-CSF-treated hypoxia-exposed neutrophils on FN was prevented by blocking antibodies to β₂-integrins, which was not observed in suspension neutrophils cultured on PolyHema (S), and was strongly induced by activating β₂-integrins on PolyHema ( Figures 2F, G ). Our data establish an indispensable role for β₂-integrins in neutrophil HIF1α induction under both pharmacological pseudohypoxic and normobaric hypoxic conditions.

During emigration from the blood into hypoxic inflammatory sites, neutrophils move against a chemotactic gradient and interact with extracellular matrix proteins in a β₂-integrin-dependent manner. To mimic these conditions, we performed a migration assay using FN-coated transwells and two chemoattractants ( Figure 3A ), namely GM-CSF that had, and fMLP that did not have, a synergistic HIF1α-increasing effect with ROX ( Figure 1A ). Either chemoattractant was pipetted into the lower well together with buffer control or ROX to mimic hypoxia at the inflammatory site. Neutrophil migration toward the two chemoattractants was similar and was not affected by HIF1α activation with ROX ( Figure 3B ). However, neutrophils from the lower well that had migrated towards GM-CSF/ROX increased HIF1α significantly more than those that had migrated toward fMLP/ROX ( Figure 3C ). In addition, HIF inhibition with YC1 ( Figure 3D ) also did not affect neutrophil transmigration ( Figure 3E ) but abrogated HIF1α protein accumulation in transmigrated neutrophils ( Figure 3F ). These data further support the notion that HIF1α is strongly induced in neutrophils migrating into inflammatory sites where hypoxia and specific cytokines synergistically increase HIF1α. Next, we investigated mechanisms by which GM-CSF and β₂-integrins cooperate in HIF1α protein induction under pseudohypoxia.

We hypothesized that GM-CSF upregulates HIF1α transcription leading to increased HIF1α protein when PHDs are inhibited. To evaluate whether neutrophil HIF1α transcription contributes to neutrophil HIF1α protein expression, we employed the dsDNA intercalating agent actinomycin D (ActD) to inhibit the transcription of RNA (36). We detected constitutive HIF1α mRNA in resting neutrophils on FN that was not affected by ROX but significantly upregulated by GM-CSF ( Figure 4A ), resulting in increased HIF1α protein when PHDs were inhibited by ROX ( Figure 4B ). ActD treatment for 4 h inhibited HIF1α transcription in ROX- and GM-CSF-treated neutrophils ( Figure 4A ) and decreased HIF1α protein in GM-CSF/ROX-treated neutrophils, indicating the need for continuously active HIF1α transcription ( Figure 4B ).

We used the JAK2 inhibitor AZD1480 to suppress GM-CSF-induced STAT3 phosphorylation ( Figure 4C ). AZD1480 abrogated the synergistic GM-CSF effect on HIF1α protein in neutrophils on FN ( Figure 4D ) by inhibiting GM-CSF-enhanced HIF1α transcription ( Figure 4E ). These data indicate that GM-CSF increases HIF1α mRNA, at least in part, using the JAK2-STAT3 pathway.

We then assessed HIF1α transcription in neutrophils incubated in PolyHema, tube, and FN conditions. GM-CSF caused a similar HIF1α mRNA increase under all three conditions ( Figure 4F ) but resulted in detectable HIF1α protein only when β₂-integrins were engaged (see Figure 4 ). In addition, blocking β₂-integrins in neutrophils on FN did not decrease, and activating β₂-integrins in suspended neutrophils on PolyHema (S) did not increase, HIF1α transcription ( Supplementary Figure 4 ). Thus, basal and GM-CSF-upregulated HIF1α transcription are necessary but not sufficient to explain the corresponding HIF1α protein expression in adherent neutrophils under pseudohypoxia and normobaric hypoxia. Consequently, we hypothesized that GM-CSF increased HIF1α transcription but that β₂-integrins are essential for translation to yield HIF1α protein.

To assess the contribution of translation to basal and GM-CSF-upregulated HIF1α protein in neutrophils on FN, we treated neutrophils on FN with the translation elongation inhibitor cycloheximide (CHX, 2.5 µg/ml) (37). CHX prevented HIF1α protein in resting and GM-CSF-treated neutrophils on FN under pseudohypoxia without affecting HIF1α transcription ( Figure 5A ). Together with our findings on PolyHema, namely substantial HIF1α transcription without translation into the corresponding protein, we considered β₂-integrin-controlled HIF1α translation in human neutrophils on FN.

Translation initiation factor elF4E and 4EBP1 phosphorylation (eIF4E: Ser209, 4EBP1: Ser65) and their subsequent dissociation are rate-limiting steps for 5′-cap-dependent mRNA translation (38). The dissociation of phosphorylated 4EBP1 and eIF4E can be specifically suppressed by the compound 4EGI-1 (39, 40). We found basal phosphorylation of both co-factors in neutrophils on FN that increased with GM-CSF treatment over time with a maximum at approximately 60 min ( Figure 5B ). Selecting the 60-min timepoint, GM-CSF-induced 4EBP1 and elF4E phosphorylation was significantly reduced by the blocking CD18 antibody ( Figure 5C ). Preventing eIF4E and 4EBP1 dissociation in neutrophils on FN by 25 µM of 4EGI-1 (39), HIF1α protein levels were completely abrogated in resting and GM-CSF-stimulated neutrophils under pseudohypoxia and normobaric hypoxia ( Figure 5D ).

Next, we explored mammalian target of rapamycin (mTOR) pathway-dependent HIF1α protein accumulation in adherent neutrophils, as the mTOR pathway is a canonical regulator of translation in numerous cell types (41). Using phospho-specific antibodies, we found basal mTOR phosphorylation (Ser2448) in freshly isolated neutrophils adhering to FN that was augmented by GM-CSF treatment and reduced by the mTOR inhibitor rapamycin (RAP) ( Figure 6A ), indicating the RAP-mediated inhibition of mTOR autophosphorylation (42). Consequently, RAP partly reduced HIF1α protein accumulation in GM-CSF/ROX-stimulated neutrophils adherent to FN ( Figure 6B ) without affecting HIF1α transcription ( Figure 6C ). Together, these data support the notion that GM-CSF-induced HIF1α transcription and β₂-integrin-mediated HIF1α mRNA translation as a conditio sine qua non cooperate in HIF1α protein upregulation. HIF1α translation but not transcription is, at least in part, controlled by the mTOR pathway.

Neutrophils that migrate from the circulation toward the inflammatory site leave the serum-rich blood, engage β₂-integrins by interacting with extracellular matrix, and encounter hypoxia. We mimicked these conditions by incubating neutrophils with high (10% v/v) and low (0.1% v/v) autologous serum (AS) concentration on FN either under pseudohypoxia (ROX) or normobaric hypoxia (1% O₂). Apoptosis was measured after 20 h in the absence or presence of the HIF1α inhibitor YC1 to determine whether HIF1α has a role in neutrophil survival under these conditions ( Figures 7A, B ). In low serum concentration (0.1% AS), we observed significantly increased HIF1α protein ( Figure 7C ) together with delayed neutrophil apoptosis under pseudohypoxia (ROX) and hypoxia. The HIF1α inhibitor YC1 not only abrogated HIF1α protein but also reversed delayed neutrophil apoptosis under pseudohypoxia and hypoxia. The apoptosis-delaying effects of pseudohypoxia and hypoxia were less pronounced in high serum concentration (10% AS) but were nevertheless counteracted by YC1. Our data suggest that HIF1α activation promotes neutrophil survival, especially after the cells left the serum-rich blood circulation migrating to their destination in the tissue.