Male C57BL/6J mice were purchased from Janvier (Le Genest-St. Isle, France). HIF1a^fl/fl , HIF2a^fl/fl (provided by Prof. Dr. Ben Wielockx) were crossed with Albumin Cre transgenic mice, and the offspring was intercrossed to generate HIF1a^fl/fl Albumin Cre^Tg/+ (HIF1a^AlbKO), HIF2a^fl/fl Albumin Cre^Tg/+ (HIF2a^AlbKO), and HIF1aHIF2a^fl/fl Albumin Cre^Tg/+ (HIF1aHIF2a^AlbKO) mice, all in a C57BL/6J background. All offspring was genotyped by PCR on genomic DNA isolated from toe biopsies. Mice were housed in a temperature-controlled, specific pathogen free (SPF) air-conditioned animal house with 14 and 10h light/dark cycles and received food and water ad libitum. All mice were used at the age of 8 – 12 weeks, and all experiments were approved by the institutional ethics committee for animal welfare of the Faculty of Sciences, Ghent University, Belgium.

A hypoxia reporter plasmid was purchased from Addgene (plasmid #26731). The plasmid contained a cassette containing three Hypoxia Responsive Elements (HRE) derived from the mouse Pgk1 gene (sequence HRE: TGTCACGTCCTGCACGACTCTAGT), followed by a mini TK promoter (27), firefly luciferase cDNA and SV40 polyA, flanked by 2 chicken beta-globin HS4 insulator core sequences (28) on both sides. The reporter plasmid is considered to be specific for hypoxia signals (27). The insulators were flanked with 800 bp homology arms to the TIGRE locus (29) and by NotI restriction sites. The 5629 bp cassette was made synthetically (Genscript) and cloned in a pUC57 backbone vector. The cassette was removed from the vector by NotI digest, gel extracted and purified using phenol-chloroform extraction and ethanol precipitation. The fragment was dissolved in TE buffer pH 7.5.

The purified fragment (1.5 ng/µl) was injected in C57BL/6J zygotes together with Cas9 protein (60 ng/µl, VIB Protein Core) and cr/tracrRNA duplex to the TIGRE locus (5’ TAACTTTAATTCTAGCGATC 3’, 40 ng/µl). Founders were identified by PCR amplification of toe DNA with primers to the luciferase cDNA identifying integration of the cassette in the genome: 5’ GGAAGACGCCAAAAACATAA 3’ and 5’ GGAAGACGCCAAAAACATAA 3’. Correct integration in the TIGRE locus was identified with a PCR over the left homology region with a primer in the TIGRE locus 5’ GCCTGGAACTCACTATACAA 3’ and a primer in the cassette 5’ TTAATATGCGAAGTGGACCT 3’ on the one hand and a PCR over the right homology region with a primer in the cassette 5’ TAAAAAACCTCCCACACCTC 3’ and a primer in the TIGRE locus 5’ AACTAAGAAGAAACGCCTCC 3’.

Polymicrobial sepsis was induced in mice by performing a CLP procedure, as previously described by Rittirsch et al. (2009) (30). Briefly, mice were anesthetized by isoflurane inhalation and a midline incision was made in the abdomen. Then, the cecum was exposed, 75% ligated, and a single through-and-through puncture was made with a 21-Gauge needle. During the procedure, a small amount of cecal content was extruded. The abdominal musculature and skin were closed by applying simple running sutures and metallic clips, respectively. During lethality experiments, mice were injected intraperitoneally (i.p.) with broad-spectrum antibiotics (25 mg/kg ceftriaxone and 12.5 mg/kg metronidazole, Sigma) in 100 µl phosphate buffered saline (PBS) 8h and 24h after CLP onset. For organ isolation experiments, a sham procedure was also performed. Here, the cecum of mice was exposed but not ligated or punctured. Mice were euthanized via cervical dislocation at the indicated timepoints after sepsis initiation, and plasma and organs were collected.

LPS from Salmonella abortus equi was purchased from Sigma-Aldrich N.V. (L-5886). For in vivo DEX injection, Rapidexon (Medini N.V.) was used. LPS and DEX were diluted in PBS. Luciferin (XenoLight™ D-Luciferin - K+ Salt) was purchased from Caliper Life Sciences.

All injections were given i.p., except for the hydrodynamic intravenous (i.v.) tail injection of the DNA plasmid. Injection volumes were always adapted to the bodyweight of the mice. In lethality experiments, mice were monitored by measuring rectal body temperature. Mice with body temperature below 28°C were euthanized using cervical dislocation. Blood was taken via cardiac puncture after sedation of the mice with a ketamine/xylazine solution (Sigma-Aldrich N.V.) or via retro-orbital eye bleeding after sedation with isoflurane. To obtain mouse plasma, blood samples were collected in EDTA-coated tubes, and samples were centrifuged at 3.000 rpm for 15 minutes at 4°C. Plasma samples were stored at -20°C for biochemical analysis. For sampling of liver, mice were killed by cervical dislocation at indicated time points.

Mice were randomly assigned to the normoxia group and hypoxia group. The normoxia group was exposed to room air (21% O₂), whereas the hypoxia group was placed in a ventilated hypoxic chamber with 7% O₂ and 93% N₂ for the indicated time points. The oxygen levels were monitored with a Greisinger GOX 100 oxygen sensor (Conrad).

Mice were injected in the tail vein over five seconds with a HRE-luciferase reporter plasmid solution (Addgene, #26731; 10 µg/ml in sterile, endotoxin-free PBS) or PBS (control) in a volume equivalent to 10% of the body weight, as described by Van Bogaert et al. (2011) (31). The HRE-luciferase plasmid contains three hypoxia response elements (24-mers, TGTCACGTCCTGCACGACTCTAGT) from the mouse Pgk1 gene upstream of firefly luciferase. Five hours after transfection, mice were subjected to a sham or CLP procedure, or injected with PBS or LPS, and visualized at indicated time points. Briefly, mice were injected with 200 μl of a 15 mg/ml potassium salt luciferin solution. 10 minutes after injection, livers were isolated and visualized via the imaging chamber of the IVIS Spectrum In Vivo Imaging System (Caliper Life Sciences). Photon emission was integrated over a period of 2 minutes and recorded as pseudo-color images. Living Image (Caliper Life Sciences) was used for image analysis. The regions of interest were selected based on the luciferase signal (purple) detected over all images. To confirm the specificity of the technique used for the injection of the HRE-luciferase reporter plasmid, liver was also visualized. Data were acquired as photons/cm²/s and log(Y) transformed before statistical analysis. Results are normalized to the PBS control group.

Liver was isolated, put in RNA later (Life Technologies Europe), and stored at -20°C before RNA was isolated. Total RNA was isolated with the Aurum total RNA mini kit (Biorad) according to manufaturer’s instructions. RNA concentration was measured with the Nanodrop 8000 (Thermo Fisher Scientific), and 1000 ng RNA was used to prepare cDNA with Sensifast cDNA Synthesis Kit (Bioline). cDNA was diluted 20 times in ultrapure water for use in RT-qPCR reactions. RT-qPCR primers for used targets are listed in Table 1 . RT-qPCR reaction was performed with sensiFast Sybr no-ROX mix (Bioline) and was performed in duplicate in a Roche LightCycler480 system (Applied Biosystems). The stability of the housekeeping genes (HKGs) were determined by Genorm. Results are given as relative expression values normalized to the geometric mean of the HKGs, calculated in the qBase+ software (Biogazelle).

Blood glucose and ketone body levels were measured in tail blood with the use of OneTouch Verio glucose meter (LifeScan) and Freestyle Precision Neo meter (Abbott), respectively. Free fatty acids (Abnova) were measured in mouse plasma with the use of colorimetric assays according to manufacturer’s instructions.

Data were expressed as means ± standard errors of the means (SEM). Statistical significance was evaluated with a two-way ANOVA in GraphPad Prism 9.0 software (GraphPad Software, San Diego, CA). If applicable, two-way ANOVA analysis were followed by post-hoc analysis to correct for multiple testing during the pairwise multiple comparisons using the Šídák’s multiple comparisons test. Fold changes or ratios were log(Y) transformed before statistical analysis. Survival curves were subjected to the Log-Rank (Mantel-Cox) test to investigate whether statistical significance could be observed during different groups. Determining if two sets showed a significant overlap as approached as a (gene) set enrichment analysis, the hypergeometric test was used to obtain a p-value of the overlap. As the population size for the test, we used the total number of genes (13.000 genes) for which we can reliably obtain (normalized counts > 1 in 50% of the samples or all samples of one condition) gene level counts in the liver.

To investigate the presence of HIF signaling on a genome-wide level in the liver of septic mice, bulk RNA-SEQ analysis was performed on the livers of mice 6h and 24h after a CLP or sham procedure, and 6h and 24h after hypoxia (7% oxygen) or normoxia ( Figure 1 ). After 6h, 896 and 2556 genes were significantly upregulated (adjusted P-value (P) < 0.05, LFC > 0), while 910 and 2244 genes were significantly downregulated (P < 0.05, LFC < 0) after hypoxia or CLP, respectively ( Figure 1A ). We identified the upregulation of 2490 and 4183 genes (P < 0.05, LFC > 0), and the downregulation of 2070 and 4169 genes (P < 0.05, LFC < 0) 24h after hypoxia or CLP ( Figure 1B ). The overlap between the hypoxia and CLP dataset demonstrates that there is a significant enrichment of hypoxia signaling in the up- (6h: 253/2556, P = 3.68e^-11 and 24h: 942/4183, P = 4.7e^-12) and downregulated (6h: 223/2244, P = 2.28e^-9 and 24h: 1081/4169, P = 2.01e^-96) genes in CLP-induced polymicrobial sepsis at both timepoints ( Table 2 ). As expected, Enrichr analysis of the shared upregulated genes shows a clear enrichment in pro-inflammatory responses as well as hypoxia at both timepoints ( Table 3 ). The log fold changes (LFCs) of genes significantly upregulated by CLP and hypoxia reported by the Enrichr analysis, are shown in the heatmap of Figure 1C . When analyzing the pathways induced by these genes, Enrichr revealed HIF signaling pathway, as expected, but also metabolic pathways such as glycolysis. Furthermore, the mRNA expression levels of Hif1a ( Figure 1D ) and Epas1 ( Figure 1E ), the genes encoding HIF1α and HIF2α respectively, are significantly higher after CLP. In contrast, the dominant-negative regulator of the HIF pathway HIF3α (34), encoded by Hif3a, is hardly expressed in the liver of mice isolated after CLP or sham ( Figure 1F ). In contrast, the impact of deep hypoxia on the transcriptional levels of Hif1a, Epas1 and Hif3a is quite minimal. After 6h and 24h of hypoxia, we detected a small (but non-significant) increase in Hif1a mRNA expression levels of 6% and 15%, respectively. Epas1 mRNA levels did not increase at both time points. Also in the presence of hypoxia, Hif3a is hardly expressed in the liver of these mice (data not shown).

Based on the RNA-SEQ analysis, HIF signaling is present in the liver of septic mice. We have generated an HRE-luciferase reporter mouse. A hypoxia reporter plasmid was purchased from Addgene (plasmid #26731). This cassette containing 3 HREs, followed by a mini TK promotor (27), firefly luciferase cDNA, and SV40 polyA, flanked by chicken insulator sequences (28) was injected in C57BL/6J zygotes (1.5 ng/µl) and inserted via random integration (29). Several founder lines were obtained in which the construct had integrated. The function of the HRE-luciferase activity was measured in the germline transgenic reporter mice via luciferin injection and optical imaging using the IVIS SpectrumCT system as a proof-of-concept ( Figure 2A ). Heterozygous HRE-luciferase transgenic reporter (HRE-Luc^Tg/+) mice were put in hypoxia (7% oxygen) or normoxia, and visualized after 2h, 6h and 24h. We were able to detect a clear luciferase signal under hypoxic conditions, while only a limited amount of luciferase activity was observed in normoxia ( Figure 2B ). Furthermore, the luciferase signal could be detected in several organs such as the brain, the heart and lungs, isolated from the mice after 24h of hypoxia ( Figure 2C ).

Next, using the HRE-Luc^Tg/+ reporter mice, luciferase signals were investigated after CLP-induced polymicrobial sepsis. Therefore, a CLP or sham procedure was performed on HRE-Luc^Tg/+ mice and wild-type littermates (HRE-Luc^+/+). In the latter mice, no signal was observed ( Figure 2D ). A significant increase in luciferase reporter activity was detected after 6h and 24h of CLP when imaging the entire animals and their livers compared to sham-operated mice ( Figures 2D–F ), strongly suggesting that hypoxia signaling is present in the livers of septic mice and that HIF proteins are transcriptionally active in sepsis.

In order to confirm whether HIF1α and/or HIF2α is/are responsible for the luciferase signal detected in the liver after CLP, mice with a conditional knock-out of HIF1α (HIF1a^AlbKO) or HIF2α (HIF2a^AlbKO), or both (HIF1aHIF2a^AlbKO) in hepatocytes were generated. To validate the hepatocyte specific knock-out mice used in these experiments, Hif1a and Epas1 mRNA levels were measured via RT-qPCR in the liver of HIF1a^AlbKO, HIF2a^AlbKO, HIF1aHIF2a^AlbKO mice and wild-type littermates ( Supplementary Figure 1 ). As expected, Hif1a and Epas1 mRNA levels were significantly downregulated in the respective knock-out mice ( Supplementary Figures 1A, B ). Both genes were significantly downregulated in the liver of HIF1aHIF2a^AlbKO mice ( Supplementary Figure 1C ). We also detected a downregulation of Epas1 mRNA in the liver of HIF1a^AlbKO mice ( Supplementary Figure 1A ), suggesting that HIF1α depletion also causes some HIF2α reduction under normoxic conditions.

First, we confirmed the presence of hepatic HIF activity in septic mice by injecting the HRE-luciferase reporter plasmid (which was used to generate the transgenic reporter mice) or PBS (control) via the tail vein under high pressure, leading to hepatocyte-specific transfection. These mice were randomly assigned to a sham or CLP procedure. Luciferase activity was measured at the indicated timepoints ( Figure 3A ). In sham mice, low luciferase signals were detected in mice injected with the reporter plasmid. As soon as 6h after CLP, the luciferase activity strongly increased and remained high until 24h post-surgery ( Figures 3B, C ), suggesting strong HIF transcriptional activity. Furthermore, we compared HRE-luciferase activity between CLP and LPS-induced endotoxemia. Mice were injected with the reporter plasmid via the tail vein followed by an LPS injection or a CLP procedure. A sham operation or PBS injection was performed as a control. The luciferase activity tended to increase to the same extent in both mouse models of systemic inflammatory response syndrome (SIRS) and sepsis ( Supplementary Figure 2 ). Finally, we studied if the HRE-luciferase activity induced by sepsis is comparable with mice in hypoxia (7% oxygen). Therefore, mice were put in normoxic or hypoxic conditions, or were subjected to a sham or CLP procedure after high-pressure injection of the HRE-luciferase reporter plasmid. 6h after CLP, the HRE-luciferase signal was significantly increased compared to sham, remained high until 24h, and was comparable to the signal induced by hypoxia ( Figures 3D, E ). In contrast to their mRNA expression levels, HIF proteins do accumulate in hypoxic conditions.

To investigate which HIF protein is involved in the HRE-luciferase activity observed during sepsis, we measured the reporter activity in the liver of HIF1a^AlbKO, HIF2a^AlbKO and HIF1aHIF2a^AlbKO 24h after CLP-induced polymicrobial sepsis using the reporter plasmid ( Figure 4 ). The HRE-luciferase activity increased in HIF1a^AlbKO ( Figures 4A, D ) and HIF2a^AlbKO ( Figures 4B, E ) 24h after CLP, to the same extent as in wild-type mice. However, when both HIF proteins are absent in hepatocytes, we were no longer able to detect a significant increase in the HRE-luciferase activity in sepsis ( Figures 4C, F ), suggesting that both HIF1α and HIF2α are responsible for the HIF activity in sepsis and that perhaps both proteins can functionally compensate for the loss of the other in hepatocytes.

Several studies have shown that a conditional HIF1α or HIF2α knock-out in myeloid cells protects against LPS-induced endotoxemia (35–37). However, the role of hepatic HIF proteins in polymicrobial sepsis has been poorly studied. Therefore, we first investigated whether HIF1α and/or HIF2α expression in the liver contribute to sepsis mortality. Mice with a conditional knock-out of HIF1α (HIF1a^AlbKO) or HIF2α (HIF2a^AlbKO), or HIF1α and HIF2α (HIF1aHIF2a^AlbKO) in hepatocytes and wild-type littermates were subjected to a CLP procedure. However, no survival benefit was observed in the absence of HIF1α and/or HIF2α in hepatocytes after CLP ( Figure 5A ). Furthermore, when these mice were injected with a lethal dose of LPS (11.25 mg/kg), no significant impact on survival between hepatocyte-specific HIFa^AlbKO mutant and wild-type mice was observed ( Figure 5B ).

As mentioned before, sepsis is characterized by a (failing) SR with hypoglycemia and increased levels of FFAs and KBs as a consequence (7, 8). Furthermore, HIF proteins are involved in regulating the expression of genes involved in glucose and lipid metabolism (21–23) in a direct way, or by reducing the function of PPARα and/or GR in hepatocytes (hypothesis investigated in this study). Therefore, we investigated whether the absence of HIF1α and/or HIF2α in hepatocytes has an influence on the hypoglycemia and increased FFA and KB levels during sepsis. HIF1a^AlbKO, HIF2a^AlbKO and HIF1aHIF2a^AlbKO mice and wild-type littermates were subjected to a CLP or sham procedure and glucose, FFAs and KBs were measured in the blood of these mice 24h post-surgery. As expected (10), hypoglycemia was detected in wild-type mice 24h after CLP. The degree of hypoglycemia was similar in all three mutant mice ( Figure 5C ). We could also detect a significant increase in both FFA and KB levels in the blood of wild-type mice and all three HIFa^AlbKO mice, 24h after polymicrobial sepsis was induced ( Figures 5D, E ). Although some differences in the degree of hypoglycemia, and FFA and KB increases were detected in the different groups, we conclude that, by and large, no major effects of HIF absence were found on biological effects of CLP-induced hypoglycemia and increase in FFA and KB levels.

Once polymicrobial sepsis via CLP is induced, mice develop a persistent and genome-wide GCR in the liver as well as hypoglycemia and hyperlactatemia (10). Furthermore, since HIFs are thought to be involved in increased glycolysis thereby contributing to the higher blood lactate levels during sepsis (7), and a clear crosstalk exists between GR and HIF (25), we investigated if HIF1α and/or HIF2α are involved in the GCR induced in the liver during sepsis. Therefore, HIF1a^AlbKO, HIF2a^AlbKO or HIF1aHIF2a^AlbKO mice and wild-type littermates were subjected to a CLP procedure. After 6h, a timepoint at which GCR is already present in mice (10), mice were injected i.p. with PBS or DEX (10 mg/kg), and liver was isolated 2h later ( Figure 6A ). The expression of typical GR-responsive genes was measured via RT-qPCR. In sham mice, a significant increase in the mRNA expression levels of Fam107a, Fkbp5 and Tsc22d3 after DEX stimulation was detected. As previously shown (10), these genes no longer responded to DEX after CLP. Furthermore, we were unable to detect any significant difference in gene expression levels after DEX stimulation in the liver of HIF1a^AlbKO, HIF2a^AlbKO and HIF1aHIF2a^AlbKO mice 6h after CLP ( Figures 6B–D ). These results show that the absence of HIF1α and/or HIF2α in hepatocytes of septic mice is not able to prevent GCR during sepsis.

Sepsis is also associated with a rapid decline in hepatic PPARα mRNA and protein levels, and hence a reduced hepatic FFA β-oxidation catabolism. In combination with increased lipolytic activity of WAT, this reduced β-oxidation causes lipotoxicity in liver and kidney after sepsis (11). Since hypoxia is associated with increased lipolysis, increased FFA levels in the blood (26), and impaired FFA β-oxidation (22, 23), and p300 functions as a co-activator for PPARα (38), hepatic HIF1α and/or HIF2α might be involved in the declined PPARα signaling during sepsis. Therefore HIF1a^AlbKO, HIF2a^AlbKO, and HIF1aHIF2a^AlbKO mice and wild-type littermates were subjected to CLP and livers were isolated 6h later ( Figure 7A ). In wild-type mice, a rapid decline of Ppara and several PPARα responsive genes was observed 6h after CLP ( Figures 7B–D ). In the absence of HIF1α in hepatocytes, the mRNA expression levels of Ppara and its responsive genes were decreased, although less pronounced than in wild-type littermates ( Figure 7B ). This might indicate that HIF1α contributes to the decline in PPARα and its signaling in sepsis. In the absence of HIF2α in hepatocytes, no major effect was observed on the expression of Ppara and PPARα responsive genes 6h after a CLP procedure in comparison to wild-type mice ( Figure 7C ). In line with the results obtained in HIF1a^AlbKO mice, the expression levels of Ppara and PPARα responsive genes were significantly reduced in the liver of HIF1aHIF2a^AlbKO mice and its wild-type littermates 6h after sepsis. However, the downregulation of Ppara and its targets genes is also less pronounced in the liver of HIF1aHIF2a^AlbKO mice ( Figure 7D ).

Altogether, we conclude that HIF proteins are not involved in the appearance of GCR in liver during sepsis and that HIF1α in hepatocytes of septic animals might be involved in the reduced PPARα signaling ( Figure 7E ).