Acute and chronic hypoxia differentially predispose lungs for metastases

Oscillations in oxygen levels affect malignant cell growth, survival, and metastasis, but also somatic cell behaviour. In this work, we studied the effect of the differential expression of the two primary hypoxia inducible transcription factor isoforms, HIF-1α and HIF-2α, and pulmonary hypoxia to investigate how the hypoxia response of the vascular endothelium remodels the lung pre-metastatic niche. Molecular responses to acute versus chronic tissue hypoxia have been proposed to involve dynamic HIF stabilization, but the downstream consequences and the extent to which differential lengths of exposure to hypoxia can affect HIF-isoform activation and secondary organ pre-disposition for metastasis is unknown. We used primary pulmonary endothelial cells and mouse models with pulmonary endothelium-specific deletion of HIF-1α or HIF-2α, to characterise their roles in vascular integrity, inflammation and metastatic take after acute and chronic hypoxia. We found that acute hypoxic response results in increased lung metastatic tumours, caused by HIF-1α-dependent endothelial cell death and increased microvascular permeability, in turn facilitating extravasation. This is potentiated by the recruitment and retention of specific myeloid cells that further support a pro-metastatic environment. We also found that chronic hypoxia delays tumour growth to levels similar to those seen in normoxia, and in a HIF-2α-specific fashion, correlating with increased endothelial cell viability and vascular integrity. Deletion of endothelial HIF-2α rendered the lung environment more vulnerable to tumour cell seeding and growth. These results demonstrate that the nature of the hypoxic challenge strongly influences the nature of the endothelial cell response, and affects critical parameters of the pulmonary microenvironment, significantly impacting metastatic burden. Additionally, this work establishes endothelial cells as important players in lung remodelling and metastatic progression.

(A) eletion of HIF-1 in primary C carrying double-floxed HIF-1 allele and exposed to Cre-expressing Adenovirus is consistently above 0% (WT control cells were treated with adenovirus expressing β-galactosidase); (B) epresentative IF images of HIF-isoform signal in all treatments and all genotypes; Animals expressing lung endothelium specific L1 Cre show visibly less HIF signal for the respective isoform. Original whole membrane images of WB using nuclear extracts of normoxic and hypoxic primary cells probed for HIF-1 , cropped and shown in main text. PV F membranes were cut at 5 a (using Bio ad Precision Plus dual collor ladder). Bottom halft was used to probe for -actin or TATA-binding protein, and upper half for HIF-1 . Multiple membranes were occasionally scanned at once when target signal intensity was comparable. ections framed in blue represent for HIF-1 and green -actin control.  Whole scans of western blot of nuclear extracts of primary cells probed for HIF-2 , cropped and shown in main text (biological replicate 2). PV F membranes were cut at 5 a (using Bio ad Precision Plus dual colour ladder). Bottom half was used to probe for -actin and upper half for HIF-2 . Multiple membranes were occasionally scanned at once when target signal intensity was comparable. ections framed in orange represent for HIF-2 and green -actin control, in the whole scan images. α α β− cans of western blot membranes of whole lung protein probed for HIF-1 , HIF-2 and -actin control, cropped and shown in the main text. PV F membranes were cut at 5 a (using Bio ad Precision Plus dual colour ladder). amples were prepared in one large mix and same volume (corresponding to 15ug of protein) were loaded in multiple wells, to probe for multiple targets. The upper half of the membranes was used for targets running above 5 a, and the bottom half was used to probe for -actin and other targets smaller than 5 a. Cropped sections used in main gure are framed in blue for HIF-1 , orange for HIF-2 and green for -actin control. Images of the membranes are provided. These were superimposed to the luminescence images to locate the targets in relation to the standards. The membrane used for HIF-2 probe was also used for i O (shown in upplementary g. 2 ).
Extended methods -Reiterer et al., Acute and chronic hypoxia differentially predispose lungs for metastases Animal models: Deletion of HIF-1α or HIF-2α in lung EC was obtained by crossing female animals homozygous for the floxed alleles of HIF-1α 1 or HIF-2α 2 (double-floxed, DF) to HIF-1α df /L1Cre + or HIF-2α df /L1Cre + males 3 . Experimental cohorts including L1Cre + were perfomed using Cre-littermate controls (DF animals with no deletion, physiologically wildtype, as previously shown 1,2 ). We have previously shown that L1-driven Cre expression is confined to the pulmonary endothelium 4 . C57Bl/6 WT mice were purchased from the Charles River laboratories (UK).
In vivo Hypoxia treatments: Eight-week old male animals were exposed to 10% O2 for either 24 h (acute exposure, optimised for preferential activation of HIF-1 in lung tissue) or 10 d Metastasis Assay: A total of 5 × 10 5 LLC cells in 200 μl sterile PBS were injected into tail veins of animals pre-conditioned in hypoxia or normal air controls. Injections were performed inside the hypoxia chamber for the hypoxic pre-conditioned animals, and animals were that all animals were exactly the same age at the time of injection (hypoxia treatments started at day -10 for chronic treatments, day -1 for acute treatments, and at day 0 all animals received tumour cells from the same culture batch). Animals were transferred to normal air after tumour cell injection. Lung tissue was collected either 24h (to monitor immediate responses) or 14 days post-injection (to quantify metastatic take). Lung tissue was collected directly into OCT, immediately processed for flow cytometry, snap frozen in liquid N2 (for protein and RNA extractions), or perfused with 10 mg/mL heparin in PBS, and fixed in 4% PFA for paraffin embedding and H&E staining. Lung tumors were counted in H&E-stained 10 μm evenly spaced sections across whole lungs (≥ 20 sections per animal).
Tumor area was quantified using Image J software 5 .   Click-iT reaction buffer and additive, according to manufacturer's protocol. Secondary immunostaining, when performed, was done after this step. All slides were mounted in Vectashield anti-fade mounting medium with DAPI (Vector) and imaged the following day.
Microscopy and image analysis: Images were taken with Leica fluorescence microscope using a x20, x40 or x63 (oil immersion) objective. Gain and offset were set to negative controls, and used to standardize image acquisition for each experiment (tissue: three images per section, a ≥ 8 randomly chosen sections from each animal; cells: ≥ 4 images per slide chamber, three slides per treatment). Quantification of fluorescence was done using ImageJ. After setting the threshold to a duplicate image, stained areas are identified and selected. "Dark background" is selected for fluorescence. "Analyze-Analyze particles" was used for measurement of various parameters; background values were acquired by selecting three random areas in gray-scale image using the circle tool and "Analyze-Measure" function. Quantification in tissue samples was performed as follows: The mean fluorescence for each image was calculated: Extended methods -Reiterer et al., Acute and chronic hypoxia differentially predispose lungs for metastases To correct for uneven fluorescence the following formula was used: The number of TUNEL + cells was identified by DAPI co-staining.
Confocal Imaging and co-localization: Images were acquired in a Leica SP5 confocal microscope using x20, x40 or x63 (Oil Immersion) objective. Images were acquired through LAS (Leica) and analysed using ImageJ 5 . Pearson's correlation was calculated to assess co-localization through the Coloc 2 plugin. Cell culture and hypoxia treatments: Primary EC were isolated and cultured from lungs of HIF-1α df , HIF-2α df , iNOS df male animals between 6 and 8 weeks of age, as previously described 12 . Cre recombinase-mediated gene deletion was performed ex-vivo by overnight Western blotting: Equal amounts of protein (15 μg) were obtained from either cells or whole lung tissue lysed in RIPA buffer, resolved in 3-8% acrylamide Tris-Acetate gels (Life Sciences, EA0375BOX) and transferred to PVDF membrane using BioRad Transblot Turbo Transfer System. Primary antibodies were used at 1:1,000 (iNOS, sc-651; HIF-1α, NB-100-049) or 1:500 dilutions (HIF-2α, R&D AF2997). Membranes were cut at 75KDa band and targets were probed on the upper half of the membrane, whereas -actin (Sigma, A1978, running at ~50KDa), used as normalization control, was probed on the bottom half of the membrane. Membranes were not re-probed for multiple antibodies. Protein signals were detected following secondary incubation with HRP-conjugated antibodies for 1h at room temperature, and ECL Plus chemiluminescence detection kit (Amersham, Cat. # RPN2232), according to manufacturer's protocol. Image capture and quantification were performed