Endotoxin upregulates Sox17 expression in vivo

To investigate the mechanisms of EC regeneration, we performed qPCR analysis of gene expression in sorted CD31⁺ ECs obtained from mTmG-Scl mice at 6 hrs, 1 day and 2 days post-LPS challenge (12mg/kg LPS i.p.) with PBS-injected mice serving as controls. The heat map of mRNA expression depicting key genes involved in endothelial development and growth revealed that Vegfr2, Sox17, and Ccne1 were significantly upregulated at post-LPS day 2 compared to baseline (Fig. 2a). Importantly, Sox17 expression increased as early as 6h following injury whereas Vegfr2 and Ccne1 expression was increased at later time points. Because Sox17 is known to regulate vascular development³⁰, we focused on its possible role in EC regeneration. Immunoblotting of freshly isolated murine lung ECs showed that SOX17 protein expression increased 5-fold post-LPS in ECs as compared to controls (Fig. 2b, c). We also determined expression of the transcription factor ER71 linked to angiogenesis and regeneration³¹ as a potential mechanism of EC regeneration, and found its protein expression did not increase following LPS (Supplementary Fig. ²). To determine next whether Sox17 contributed to regeneration through EC proliferation, we determined the kinetics of Cyclin E1 gene and protein expression known to regulate cell cycle activity³². We observed that Ccne1 was upregulated at Day 2 post-LPS (Fig. 2a). Furthermore, immunoblotting of human lung microvascular endothelial cells (HLMVECs) showed markedly increased Cyclin E1 protein expression in the cells over-expressing Sox17 as compared to control cells (Fig. 2d, e). We also identified several putative Sox17 binding sites in the CCNE1 promoter using the Sox17 binding sequence³³ from the JASPAR database (Fig. 2f and Supplementary Fig. ³). Chromatin immunoprecipitation (Ch-IP) was carried out in HLMVECs overexpressing with Sox17. qPCR of the 4 Sox17 identified binding sites showed that Sox17 bound only to sites 3 and 4 on CCNE1 promoter (Fig. 2g). Using luciferase promoter activity reporter assay, we observed that Sox17 binding site 3 in the CCNE1 promoter increased luciferase activity by 5-fold in contrast to other binding sites (Fig. 2h). Thus, Sox17 binding site 3 in the CCNE1 promoter appears to be critical for Sox17-mediated transactivation of CCNE1. Together these results show that LPS-induced EC injury increases Sox17 expression and Sox17 induces EC proliferation via upregulating Cyclin E1.

Open in a separate window

Fig. 2

Activation of Sox17 at onset of EC regeneration and Sox17-mediated Cyclin E1 expression. a qPCR analysis of gene expression in sorted CD31⁺ cells from mTmG-Scl mice before and after injury (12mg/kg i.p.). Sox17, Vegfr2, and Ccne1 increased significantly at day 2 post-LPS compared to baseline. n=3. Color scale: the fold change increases from red to white to green color. b Western blot analysis in fresh isolated ECs from wild-type mice and quantification c showed a 5-fold increase in Sox17 protein expression within 1day following injury compared to baseline and followed by recovery within 3 days post-LPS. n=3. d, e Western blot analysis of cultured HLMVECs in which Sox17 was overexpressed showed 2.5x fold increase in Cyclin E1 protein expression relative to control cells. n=3. OE, overexpression. f Representation of the CCNE1 promoter region with Sox17 binding sites (circled numbers) and their sequences. g HLMVECs were retrovirally transduced with Sox17 or control plasmid for 3 days, and Ch-IP assay followed by qPCR was performed to amplify Sox17 binding sites in the CCNE1 promoter. n=3. h 293T cells were transfected with a Sox17 overexpression plasmid containing CCNE1 luciferase reporter constructs. Luciferase values were normalized to Renilla luciferase control reporter values. A schematic representation of corresponding deletion constructs is presented in the right panel. n=3 and duplicates per sample. **P<0.01 and ***P<0.001. Data are shown as mean±SEM. Analysis was performed using one-way ANOVA for (c) and two-way ANOVA with Bonferroni post-tests for (e, g, h)

EC-specific Sox17 deletion prevents EC regeneration

We next deleted Sox17 in ECs by crossing Sox17^fl/fl mice (Jackson Laboratory) with the tamoxifen-inducible endothelial-specific endo-Scl-Cre-ERT2 mice²⁹. Tamoxifen (80mg/kg) was given daily for five consecutive days and experiments were carried out after four weeks when EC-expressed SOX17 was deleted (Fig. 3a). Immunoblotting confirmed the deletion in Sox17^EC−/− mice compared to Sox17^fl/fl mice (Fig. 3b, c). We observed that 12mg/kg LPS induced 40–50% lethality in wild-type mice. Since Sox17 regulates endothelial proliferation, we surmised that the LPS dosage of 12mg/kg in Sox17^EC−/− mice would induce even greater mortality, and therefore lowered LPS to sublethal dose of 8mg/kg in order to compare responses in Sox17^fl/fl vs. Sox17^EC−/− mice. We found that LPS increased lung vascular permeability at day 1 post-LPS with no significant difference between Sox17^EC−/− and Sox17^fl/fl mice (Fig. 3d); thus, Sox17 did not affect severity of the initial LPS-induced endothelial injury. However, measurements of lung vascular leak during the recovery phase showed the critical role for endothelial Sox17 in mediating EC regeneration following the inflammatory injury. Sox17^EC−/− mice exhibited persistently increased endothelial permeability (Fig. 3d), indicative of delayed EC regeneration response. To confirm this finding, we also generated a second EC-specific Sox17 deletion mouse model using Cre driven by Cdh5 promoter. These mice were generated by breeding Cdh5-CreERT2 mice³⁴ with Sox17^fl/fl mice. We observed in these mice the same changes seen in endo-Scl-CreERT2 mice (Fig. 3e). All Sox17^fl/fl mice survived the sublethal LPS dosage (8mg/kg i.p.) while 60% of the Sox17^EC−/− mice died (Fig. 3f). FACS analysis of freshly isolated lung ECs showed that Sox17^EC−/− and Sox17^fl/fl mice both had severe loss of EC at day 1 post-LPS. By day 5, however, the EC population was restored to baseline in Sox17^fl/fl mice but EC regeneration was severely defective in Sox17^EC−/− mice (Fig. 3g). This finding paralleled the significantly higher mortality seen in EC-specific Sox17 deleted mice. Next to assess EC proliferation, the nucleotide analog 5-bromo-2′-deoxyuridine (BrdU) was injected (i.p.) 14h before sacrifice. Immunohistochemistry and confocal microscopy showed significantly increased BrdU⁺ ECs in Sox17^fl/fl mice at day 3 post-LPS when compared to Sox17^EC−/− mice (Fig. 3h and Supplementary Fig. ⁴).

Open in a separate window

Fig. 3

EC specific deletion of Sox17 (Sox17^EC−/−) in mice prevents endothelial regeneration. a Schematic diagram of Sox17^fl/fl mice crossed with Scl-CreERT2 to delete Sox17 in ECs. After tamoxifen feeding for five days and rest for another four weeks, mice are challenged with LPS (sub-lethal 8mg/kg, i.p.) for analysis. b Western blot analysis of Sox17 protein expression in isolated ECs obtained from flushed lungs of Sox17^EC−/− and control mice. n=4. c Quantification of b shows 80% deletion of Sox17 in ECs of Sox17^EC−/− mice compared to control mice. n=4. d Time course of lung transvascular permeability following LPS challenge in Sox17^EC−/− and Sox17^fl/fl mice. n=4. While control mouse lungs showed increased endothelial permeability at day 1 post-LPS and then recovered to baseline by day 5, Sox17^EC−/− mice showed prolonged endothelial barrier leakiness post-LPS. e Time course of changes in lung transvascular permeability following LPS challenge was also carried out in Sox17^fl/fl mice crossed with Cdh5-CreERT2 mice. n=4. Similar as in d, these Sox17^EC−/− mice also showed persistent leakiness post-LPS while control mice fully recovered. f Survival curve of LPS challenge in Sox17^EC−/− and control mice. n=8 per group. At this sub-lethal dose, all control mice survived whereas half of Sox17^EC−/− mice died on day 2 post-LPS with increased mortality on day 3. By day 5, the death rate for control mice is 0 while for Sox17^EC−/− mice is 60%. g Flow cytometry analysis of CD31⁺CD45⁻ ECs among whole lung population in mice following injury. n=4. In contrast to control mice in which CD31⁺CD45⁻ EC population gradually recovered with day 3 post-LPS after initial loss of ECs, Sox17^EC−/− mice showed significantly delayed restoration of ECs post-LPS period. h Quantification of BrdU+ nuclei in each field of 425μm² area in flushed lung cryo-sections from mice following injury. n=4 mice per group and 6 replicates per sample. Slides are co-stained with CD31-AF488, BrdU-APC, and DAPI. At day 3 post-LPS, the control group showed a significantly higher number of BrdU+ECs compared to baseline. However, Sox17^EC−/− mice showed markedly reduced level of BrdU+ECs, indicating reduced EC proliferation. i To assess whether expression of Sox17 in ECs can restore lung endothelial integrity, studies were performed in Sox17^EC−/− mice to overexpress Sox17 protein. We used a mixture of 50μg plasmid (mouse Cdh5 promoter—Flag —Sox17) encapsulated in liposomes, which were injected i.v. 3h after LPS challenge. j At day 3 post-LPS, liposome vector-treated Sox17^EC−/− mice showed marked EC barrier leakiness as assessed by lung transvascular permeability of albumin whereas the Sox17-rescued mice showed markedly reduced endothelial permeability. n=4. OE, overexpression. *P<0.05, **P<0.01 and ***P<0.001. Data are shown as mean±SEM. Analysis was performed using two-tailed Student’s t-test for (c, j), two-way ANOVA with Bonferroni post-tests for (d, e, g, h) and Log-rank (Mantel-Cox) test for (f)

We next determined whether re-expression of Sox17 in ECs of Sox17^EC−/− mice is sufficient to restore lung endothelial integrity. These studies were conducted in Sox17^EC−/− mice in which the Sox17 protein was expressed in lung ECs via i.v. liposomal delivery^35–37. The Flag–Sox17 plasmid was driven by EC-specific mouse Cdh5 promoter to maximize its expression in ECs; and the effectiveness of the approach was determined (Supplementary Fig. ⁷), demonstrating marked augmentation of Sox17 expression in ECs isolated from the lungs during the days following i.v. delivery of liposomal Sox17 with an EC-specific promoter but no such augmentation of Sox17 expression was found in the non-EC fraction. Liposomes encapsulating Flag–Sox17 cDNA were injected i.v. 3h after LPS challenge in mice (Fig. 3i). We observed that increased Sox17 expression in Sox17^EC−/− mice significantly accelerated endothelial barrier restoration compared to control vector delivery (Fig. 3j).

We next addressed mechanisms of upstream regulation of Sox17 expression in ECs. A key feature of endotoxin-induced inflammatory lung injury is widespread influx of activated neutrophils which intensely consume oxygen and the induce relative tissue hypoxia^38,39. We thus assessed neutrophil infiltration in lungs by measuring myeloperoxidase (MPO) activity and observed 6-fold increase in MPO activity within 6h post-LPS challenge and which continued to increase up 24h (Fig. 4a). We assessed whether the hypoxia-responsive transcription factor HIF-1α^40,41 was concomitantly increased, and observed its protein upregulation in the timeframe of neutrophil infiltration (Fig. 4b, c).

Open in a separate window

Fig. 4

soHIF-1α signaling induces Sox17 expression. a MPO activity of flushed lung sample from mice challenged with LPS (12mg/kg i.p.) for 6 and 24h. n=3. b Western blot analysis of wild-type mice lung before and after LPS-induced injury (12mg/kg i.p.) and its quantification c showed that HIF-1α protein expression increased within 6h post-LPS and remained increased until day 2. n=3. d Western blot analysis in freshly isolated ECs from Hif1a^fl/fl mice and quantification e showed significantly increased Sox17 protein expression after injury when compared to baseline levels. n=3. f Western blot analysis in freshly isolated ECs from Hif1a^EC−/− mice and quantification g showed no significant difference in Sox17 protein expression before and after injury. n=3. **P<0.01 and ***P<0.001. Data are shown as mean±SEM. Analysis was performed using one-way ANOVA for (b, c, e, g)

To study the relationship between HIF-1α and Sox17 in vivo, we deleted HIF-1α in an EC-specific manner by crossing Hif1a^fl/fl mice with Tie2-CreERT2 mice. At baseline EC-specific HIF-1α deletion in mice did not affect SOX17 expression (Supplementary Fig. ⁸). Immunoblotting showed that while SOX17 expression significantly increased in ECs post-LPS in Hif1a^fl/fl mice (Fig. 4d, e), SOX17 expression did not change either before or after LPS in Hif1a^EC−/− mice (Fig. 4f, g); thus, EC specific deletion of HIF-1α prevented the increase in endothelial Sox17 expression following LPS induced EC injury.

HIF-1α signaling in ECs activates Sox17 transcription and expression

To examine the role of HIF-activation in regulating Sox17 expression, we designed specific guide RNAs to delete either HIF-1α (Hif1a^KO) or HIF-2α (Hif2a^KO) in HLMVECs using CRISPR/Cas9⁴². Dimethyloxalylglycine (DMOG), a prolyl-4-hydroxylase inhibitor^43,44, was added to HLMVECs to stabilize both HIFs. We observed that HIF-1α, HIF-2α, and SOX17 protein expression was significantly increased after DMOG treatment as compared to control (Fig. 5a, b). Importantly, CRISPR/Cas9 mediated deletion of HIF-1α prevented the increase of SOX17 induced by DMOG whereas HIF-2α deletion did not (Fig. 5a, b), indicating the essential role of HIF-1α in mediating Sox17 expression.

Open in a separate window

Fig. 5

HIF-1α activates transcription of Sox17. a Western blot analysis in control HLMVECs and HLMVECs for which CRISPR/Cas9 was used to delete HIFs. ECs were treated with the HIF prolyl hydroxylase inhibitor DMOG to induce HIF expression. DMOG (1mM) increased HIF-1α and HIF-2α protein expression in control ECs but not in ECs lacking HIF-1α or HIF-2α. Induction of HIF expression was coupled to Sox17 upregulation. n=3. b Quantification of a showed that protein expression of HIF-1α and HIF-2α in ECs was significantly increased by DMOG treatment. Sox17 showed a 2.5-fold increase in Sox17 expression control DMSO treated ECs. The increase in Sox17 was significantly reduced in HIF-1α-deleted ECs but preserved in HIF-2α-deleted ECs, indicating the importance of HIF-1α in mediating Sox17 expression. n=3. c Representation of the SOX17 promoter region with 3 HREs indicated by circled numbers, and their respective sequences are displayed. d HLMVECs were exposed to normoxia or 1% O2 (hypoxia) for 8h. Ch-IP assay followed by qPCR was performed to amplify the HRE regions in the SOX17 promoter. Studies were performed in ECs exposed to either normoxia or hypoxia. n=3. e 293T cells were transfected with a HIF-1α overexpression plasmid containing SOX17 luciferase reporter constructs. Luciferase values were normalized to Renilla luciferase control reporter values. A schematic representation of corresponding deletion constructs is presented in the right panel. n=3 and duplicates per sample. Results show that hypoxia activation of SOX17 HRE3 was required for Sox17 expression. **P<0.01 and ***P<0.001. Data are shown as mean±SEM. Analysis was performed using two-way ANOVA with Bonferroni post-tests for (b, d, e)

On analyzing the promoter region of SOX17 for hypoxia response elements (HREs), we found 3 sites (Fig. 5c and Supplementary Fig. ^5A), one of which is conserved across multiple species (HRE1, −294~−285) (Supplementary Fig. ^5B). Chromatin immunoprecipitation (Ch-IP) was carried out to determine binding of HIF-1α to SOX17 promoter using HLMVECs collected after 8h of hypoxia (1% O₂) with 21% O₂ as the normoxic control. qPCR of amplified HRE regions showed that HIF-1α bound to HREs in the SOX17 promoter (Fig. 5d and Supplementary Fig. ^5C). We observed using a luciferase promoter activity reporter assay that only HRE3 significantly increased luciferase activity  upon co-expression with a degradation-resistant HIF-1α construct (Fig. 5e).

2-photon microscopy

The tissue live imaging experiment followed a previously described protocol⁷⁷. Generally, the mice were sacrificed, and the lung was flushed with Phosphate-buffered saline (PBS, Sigma #P3813). Then 1ml 37°C 2% low-melting-temperature agarose was slowly instilled through the trachea into the lung. Once the lungs were fully inflated, we let the lungs cool by pouring 4°C PBS over the lung and removed the lungs from the mice. We sectioned the largest lobe of the lung into 300-μm slices and placed the slices on plastic coverslips in the imaging chamber. Images were taken with a Prairie Technologies Ultima In Vivo Multiphoton Microscopy System (Bruker) and analyzed by Imaris software (Bitplane).

Mouse cell isolation

Flushed mouse lungs were  minced and digested with 5ml Type 1 collagenase I (2mg/ml in PBS) at 37°C water bath for 1h. Mixtures were titrated with #18 needles and then pipetted through a 40μm disposable cell strainer. After centrifuging at 300×g for 5min and washing with PBS, the isolated cells were treated with red blood cell lysis buffer (eBioscience) for 5min on ice to lyse red blood cells.

Flow cytometry

After isolation, remaining cells were incubated with anti-mouse CD16/CD32 (1:50, BD Pharmingen #553142) to block endogenous Fc for 10min on ice. After this, cells were stained with antibodies including CD45-EF450 (1:2000, eBioscience #48–0451–82) and CD31-APC (1:100, eBioscience #17–0311–82) for 45min at 4°C. After washing, the cells were resuspended in 500μl buffer and analyzed on an LSRFortessa (BD Pharmingen) cell analyzer. Obtained data were analyzed by Summit software (Beckman Coulter).

Mouse EC pull down assay

After isolation, cells were incubated with 5µg CD31 antibody (10µl antibody in 1ml buffer, BD #553370) at 4°C for 20min with gentle tilting and rotation. After washing, cells were then incubated with pre-washed Dynabeads (25µl beads in 1ml buffer, Invitrogen #11035) at 4°C for 25min with gentle tilting and rotation. ECs were harvested by magnetic separation.

RNA isolation and qPCR

RNA was extracted from the harvested cells using TRIzol™ Reagent (ThermoFisher #15596026) according to the manufacturer’s protocol. Then RNA was quantified by Nanodrop 1000 (ThermoFisher) and reverse transcribed into cDNA using High-Capacity cDNA Reverse Transcription Kit (ThermoFisher #4368814). FastStart Universal SYBR Green Master (ThermoFisher #4913850001) was used for relative quantification of cDNA on the ViiA 7 Real-Time PCR System (ThermoFisher) (Primer information included in Supplementary Table ²).

qPCR heatmap

The qPCR relative fold change values for a set of genes were plotted using the heatmap.2 function in the R “gplots” package. Upregulation of genes is shown in red, white, and green, respectively. Red corresponds to minimal increase while green corresponds to dramatic increase relative to the untreated control.

Immunoblotting

Harvested cells or tissues were lysed in RIPA buffer (Sigma #R0278) with protease (Sigma #P8340) and phosphatase inhibitor (Sigma #P5726 and #P0044). Protein concentration was measured using DC™ Protein Assay Kit II (Bio-Rad #5000112) in Epoch Microplate Spectrophotometer (BioTek). Immunoblotting was performed with Sox17 (1:250, Origene #AM32707PU-N for mouse Sox17; 1:200, R&D #AF1924 for human Sox17), HIF-1α (1:200, Cayman Chemical #10006421), HIF-2α (1:500, Novus Biologicals #NB100–122), Flag M2 (1:1000, Sigma #F1804), Cyclin E1 (1:500, R&D #AF6810) primary antibody using Bio-Rad Protein Electrophoresis and Blotting system. Sample loading was confirmed by GAPDH expression (1:5000, Proteintech #60004). Western blot pictures were taken using ImageQuant LAS 4000 (GE Healthcare). Uncropped versions of all Western blots are available as a Source Data file.

Cell culture

HLMVECs (Lonza #CC-2527) were maintained in culture in a humidified 5% CO₂ atmosphere at 37°C in complete Clonetics EGM-2MV BulletKit medium (Lonza #CC-3202) and 15% fetal bovine serum (Atlanta Biologicals). Cells were thawed at passage 3 and transduced at the same passage. For hypoxic signaling activation studies, HLMVECs were treated with the HIF1α activator DMOG (Calbiochem #400091). 293T cells (ATCC #CRL-3216; Clontech #632180) were maintained in culture in a humidified 5% CO₂ atmosphere at 37°C in DMEM medium (Gibco #11995–065) with 10% fetal bovine serum (Atlanta Biologicals).

CRISPR/Cas9 studies

Guide RNA targeting human HIF-1α and HIF-2α gene for Cas9-mediated CRISPR disruption were identified using online design software at http://crispr.mit.edu. The two targeting sequences for HIF-1α were: 5′-TTCTTTACTTCGCCGAGATC-3′ and 5′-CCTCACACGCAAATAGCTGA-3′. The two targeting sequences for HIF-2α were: 5′-GCTGATTGCCAGTCGCATGA-3′ and 5′-CAAGGCCTCCATCATGCGAC-3. The transduction of lentiviral sgRNA and adenoviral Cas9 followed a previous protocol⁴². Generally, DNA oligos were synthesized by IDT and the overlapping PCR products were cloned into pLX-single sgRNA lentiviral vector (a gift from Eric Lander and David Sabatini⁷⁸ and obtained through Addgene as plasmid #50662). Lentivirus was prepared by co-transfection of lentiviral plasmids with psPAX2 (produced by Didier Trono and obtained through Addgene as plasmid #12260) and pMD2.G (produced by Didier Trono and obtained through Addgene as plasmid #12259) packaging plasmids into 90% confluent human 293T cells using Lipofectamine 2000 as per manufacturer’s protocol. Lentiviral supernatant was collected at 48 and 72h post-transfection, concentrated by Lenti-X concentrator (Clontech #631232) as per manufacturer’s protocol, and used to transduce HLMVECs at a Multiplicity of infection (MOI) of 10 (measured by Global UltraRapid Lentiviral Titer Kit, System Biosciences #LV961A-1) in the presence of 8μg/ml polybrene (Sigma #TR-1003). At 8h later, the EGFP-tagged Cas9 adenovirus (Ad-GFP-Cas9, Vector Biolabs #1901) was also added at a MOI of 10 to transduce HLMVECs. After 2–3 days when HLMVECs reached confluency, the cells were trypsinized and split 1:3 and received a second batch of sgRNA lentivirus and EGFP-Cas9 adenovirus. Two to three days later, the cells were ready for experimental studies.

BrdU assay of EC proliferation

Mice were injected i.p. with BrdU (Sigma #B5002) 150mg/kg 14h before sacrifice. After the mice being sacrificed, mice lung was flushed with PBS, infused with 70% OCT intratracheally and then excised in the cassette and put on dry ice. After storage at  −80°C for at least 48h, mouse lungs were sectioned into 5µm frozen sections.

Immunofluorescence and confocal microscopy

The frozen sections were fixed with 4% paraformaldehyde (Sigma #P6148) and permeabilized with 0.25% Triton X-100 (Fisher Scientific #BP151–100). After a wash with PBS/Tween-20 (Fisher Scientific #BP337–100), slides were treated with 3M hydrochloric acid (Fisher Scientific #A144S-500) for 10min to open the nucleus structure if needed. Slides were blocked with 10% donkey serum (Sigma #D9663) in 2% BSA (Sigma) in PBS/Tw-20 for 1h at room temperature. Afterwards, slides were probed with primary antibodies (CD31 1:25, BD #550274; BrdU-APC 1:50, BD; Alexa Fluor 488 Mouse anti-BrdU 1:10, BD #558599; Flag M2 1:500, Sigma #F1804) and incubated overnight at 4°C. The next day, slides were washed and incubated with the fluorescence-conjugated secondary antibody (AF488 donkey anti-rat 1:300, Invitrogen #A-21208; AF594 donkey anti-mouse 1:300, Invitrogen #A-21203; AF594 donkey anti-rat 1:300, Invitrogen #A-21209). Images were taken with a confocal microscope LSM880 (Zeiss) and analyzed by Zen software (Zeiss).

Sox17 binding sites in CCNE1 promoter

We identified human CCNE1 promoter region using the UCSC genome browser. For the prediction of human Sox17 binding sites in the promoter of CCNE1, we searched for the presence of ATTGT motifs in the JASPAR database. The binding sites were found to be at −1090~−1082 bp, −1335~−1327 bp, −2022~−2014 bp, and −2092~−2084 bp upstream of the CCNE1 ATG start site.

Identification of HREs in Sox17 promoter

We performed a sequence alignment of the human SOX17 promoter region using ClustalW2. For the prediction of human HIF-1α binding sites in the promoter of SOX17, we searched for the presence of RCGTG motifs. In addition, we only considered RCGTG motifs that were conserved in, at least, four species, including the mouse. The HREs were found to be at −294~−285 bp, −318~−311 bp, and −338~−330 bp upstream of the SOX17 ATG start site.

Chromatin-immunoprecipitation assay

The Ch-IP protocol is based on the protocol of Bryan Dynlacht and Richard Young laboratories. For the Sox17—CCNE1 experiment, HLMVECs transduced with pMXs-ms-Sox17 (a gift from S. Yamanaka⁷⁹, Addgene plasmid #50781) were crosslinked by using 1% formaldehyde (Fisher Scientific #F79–1), washed three times with cold PBS, and resuspended in cell lysis buffer. The nuclei portion was resuspended in nuclear lysis buffer and sonicated to break down the genomic DNA using S220 Focused-ultrasonicator (Covaris). After centrifugation, the supernatant was immunoprecipitated with 5μg anti-Sox17 (R&D #AF1924) or an equal amount of goat IgG (Santa Cruz #sc-2028). The DNA obtained from the IP was amplified by qPCR with primers specifically recognizing different binding sites in CCNE1 promoters (Primer information included in Supplementary Table ²). For the HIF-1α-SOX17 experiment, HLMVECs cultured in normoxic or 1% O₂ hypoxic conditions were processed as above and then immunoprecipitated with 5μg anti-HIF1α (Santa Cruz #sc-10790) or an equal amount of rabbit IgG (Santa Cruz #sc-2027). The DNA obtained from the IP was amplified by qPCR with primers specifically recognizing different HRE sites in SOX17 promoters (Primer information included in Supplementary Table ²).

Myeloperoxidase activity

Mice lungs were flushed with PBS and homogenized in RIPA buffer (Sigma #R0278). 50μg extracted sample was mixed with 1ml of 50mM phosphate buffer (Sigma #S3264 and #S3139) containing 0.167mg/ml of o-Dianisidine dihydrochloride (Sigma #D3252) and 0.0005% H₂O₂ (Sigma #216763). The reaction proceeded for 3–5min at room temperature, and then was stopped by adding 3M Hydrochloric acid (Sigma #H1758). Absorption was determined by OD460 using UV/vis spectrophotometer (Beckman #DU530) and normalized by protein concentration measured using DC™ Protein Assay Kit II (Bio-Rad #5000112) in Epoch Microplate Spectrophotometer (BioTek).

Evans blue-albumin tracer measurement of pulmonary transvascular permeability

Evans blue–albumin (EBA) (40mg/ml BSA with 1% Evans blue dye Sigma #E2129) was injected into the right jugular vein of the mice (anesthetized by ketamine-xylazine mix approved from pharmacy) at a dose of 8µl per mouse body weight gram and allowed to circulate in the blood vessels for 45min. Intravascular Evans blue was washed by PBS perfusion from the right ventricle for 2min. Mouse lungs were excised, homogenized in 1ml PBS, and extracted in 2ml Formamide (Fisher Chemical #F84–1) overnight at 60°C. Evans blue content was determined by OD620 using UV/vis spectrophotometer (Beckman #DU530) of the formamide extract and normalized by body weight. The body weights of mice for the Evans blue experiments are provided in Supplementary Table ¹.

Plasmid constructs

The 2.5kb Cdh5 promoter was cloned from mouse genomic DNA and then inserted into the pCDNA 3.0 vector. The original CMV promoter was destroyed. Then the mouse Sox17 protein coding region was amplified from pMXs-ms-Sox17 plasmid and inserted into the pCDNA-Cdh5 vector. A flag tag (GACTACAAAGACGATGACGACAAG) was added to the N-terminus of the Sox17 coding region. For the liposome-gene complex biodistribution experiment, the EGFP protein coding region was amplified from a pWPXL plasmid (a gift from Didier Trono, Addgene plasmid #12257) and inserted into the pCDNA-Cdh5 vector. Full-length human SOX17 and CCNE1 promoter plasmids (GenoCopeia #HPRM50295-PG04 and #CS-HPRM38973-PG04–01) and the truncation mutants (PCR amplified) were subcloned into a pGL3-basic vector (Promega #E1751). The HIF-1α constitutively active mutant HA-HIF1alpha P402A/P564A-pBabe-puro was a gift from William Kaelin (Addgene plasmid #19005)⁸⁰. Control pRL-TK plasmid was purchased from Promega (#2241).

Luciferase reporter assay

293T cells were seeded in 24-well cell culture plates and transfected with the indicated plasmids using Lipofectamine 2000 (Invitrogen #11668). One day after transfection, cells were lysed, and luciferase activity was measured using the dual-luciferase reporter assay system (Promega #E1910) on a GloMax luminometer (Promaga #SA3030). Relative luciferase units were determined by dividing the luciferase activity of the Firefly target pGL3 reporters by that of the pRL-TK Renilla luciferase control reporter.

Liposome gene delivery in lung ECs

Dimethyldioctadecylammonium bromide (Sigma #D2779) was dissolved in Chloroform (Fisher Chemical #C606SK-1) in a 1:1 molar ratio with Cholesterol (Millipore #228111). Chloroform was evaporated in a Rotavapor (Buchi #R-300) at 37°C, 100rpm for 20min. Afterwards, the dried lipid film was resuspended in 5% dextrose (Fisher Scientific #D16–500) in water and then sonicated in Branson 2510 Ultrasonic Cleaner (Sigma #Z244910) for 20min, followed by 0.45μm filtration. The mice were anesthetized with isoflurane (Piramal Critical Care) and retro-orbitally i.v. injected with the liposome-DNA complex at a dose of 50μg DNA/100μg liposome per mouse. Size (hydrodynamic diameter) was 234.3±8.3nm for liposome only and 240.1±3.4 for liposome-DNA complex (as shown in Supplementary Fig. ¹⁰). Polydispersity Index was 0.400±0.027 for liposome-Vector complex and 0.414±0.009 for liposome-Sox17 complex (as shown in Supplementary Fig. ¹²). The surface charge (Zeta Potential) was 39.8±0.2mV for liposome only and 34.8±0.2mV for liposome-DNA complex (as shown in Supplementary Fig. ¹¹). n=6. Supplementary Fig. ¹³ shows the organ biodistribution of the transgene expression delivered by liposomes, where lung is the organ with highest plasmid gene expression. n=4. All these parameters were measured by dynamic light scattering (Zetasizer Nano ZS, Malvern). Serum was first ultracentrifuged for 2h at 140,000 × g (Optima TLX Ultracentrifuge, Beckman Coulter) to remove extracellular vesicles that might interfere with size measurements. We used liposomes (DDAB/cholesterol) with a high charge ratio of+4.5 (+fatty acid/-DNA). The DNA complex was cloned with Flag-tagged Sox17 coding region following the EC-specific mouse Cdh5 promoter to ensure EC transgene expression.

Statistics

Immunoblot bands were analyzed for optical density using ImageJ (NIH) software. Quantification of replicate experiments is presented as the mean±SEM. The student t-test, one-way and two-way ANOVA with Bonferroni post-tests were used to determine statistical significance, with a P value threshold of less than 0.05. Significance levels are indicated in the figures as *P<0.05, **P<0.01, and ***P<0.001. Based on our experience, we expect changes in the gene/protein expression and function measurements to be detected with 3 mice per group, so the effect size was determined as n=3 or n>3. The variance between the groups that are being statistically compared was similar.

The studies were supported by NIH grants R01-HL45638 (to A.B.M.), P01-HL60678 (to A.B.M. and J.R.), T32-HL007829 (to A.B.M.), R01-HL118068 (to J.R. and A.B.M.), R01-CA211095 (to E.V.B.), R01-HL90152 (to J.R. and A.B.M.) and AHA CDA grant 18CDA34110068 (to L.Z.). The Cdh5-CreERT2 mice were provided by Dr. Ralf Adams and the Scl-CreERT2 mice were provided by Dr. Joachim Göthert.