Activation of Hypoxia Inducible Factor 1 Is a General Phenomenon in Infections with Human Pathogens

Background Hypoxia inducible factor (HIF)-1 is the key transcriptional factor involved in the adaptation process of cells and organisms to hypoxia. Recent findings suggest that HIF-1 plays also a crucial role in inflammatory and infectious diseases. Methodology/Principal Findings Using patient skin biopsies, cell culture and murine infection models, HIF-1 activation was determined by immunohistochemistry, immunoblotting and reporter gene assays and was linked to cellular oxygen consumption. The course of a S. aureus peritonitis was determined upon pharmacological HIF-1 inhibition. Activation of HIF-1 was detectable (i) in all ex vivo in biopsies of patients suffering from skin infections, (ii) in vitro using cell culture infection models and (iii) in vivo using murine intravenous and peritoneal S. aureus infection models. HIF-1 activation by human pathogens was induced by oxygen-dependent mechanisms. Small colony variants (SCVs) of S. aureus known to cause chronic infections did not result in cellular hypoxia nor in HIF-1 activation. Pharmaceutical inhibition of HIF-1 activation resulted in increased survival rates of mice suffering from a S. aureus peritonitis. Conclusions/Significance Activation of HIF-1 is a general phenomenon in infections with human pathogenic bacteria, viruses, fungi and protozoa. HIF-1-regulated pathways might be an attractive target to modulate the course of life-threatening infections.


Introduction
Mammalian cells adapt to oxygen deprivation by the activation of hypoxia inducible factor (HIF)-1, the key transcription factor during hypoxia. Subsequently, the expression of hypoxia-inducible genes involved in angiogenesis [e.g., vascular endothelial growth factor (VEGF)], glycolysis [e.g., hexokinase (HK)], proliferation and survival [e.g., adrenomedullin (ADM)] and erythropoiesis (e.g., erythropoietin) is transcriptionally regulated [1,2]. The heterodimeric transcription factor HIF-1 is composed of the two subunits HIF-1a and HIF-1b. While HIF-1b is constantly present in the nucleus, HIF-1a levels are affected by changes in the cellular oxygen partial pressure (pO 2 ). The key mechanism involved in HIF-1 activation has been identified to require inhibition of the enzymatic activity of ''prolyl hydroxylase domain''-containing proteins (PHDs) during hypoxia. PHDs mediate hydroxylation of the prolyl residues Pro402 and Pro564 of the HIF-1a subunit which results in the binding to the von-Hippel-Lindau protein and subsequent proteasomal degradation under normoxic conditions.
In contrast, hypoxia results in PHD-inhibition and subsequent stabilization of HIF-1a, binding of the HIF-1 heterodimer to promoter regions of hypoxia-inducible genes and corresponding gene induction [2]. Iron deprivation [induced by iron chelating compounds, e.g., desferrioxamine (DFO)] has been found an alternative strategy of HIF-1 activation [3]. The molecular explanation for this phenomenon is that PHDs contain iron as an essential cofactor for their enzymatic activity; therefore, iron chelation inhibits PHD activity resulting in the subsequent activation of HIF-1 [4].
A constantly growing body of evidence suggests that HIF-1 plays a novel and important role in infectious and inflammatory diseases [5]. HIF-1 is essential for the bactericidal capacity of phagocytes and for controlling systemic spread of bacteria in mice [5,6]. Interestingly, HIF-1 activation occurs during bacterial infections with the angiogenic bacterium Bartonella henselae (causing the vasculoproliferative disorder bacillary angiomatosis) [7] and this phenomenon is linked with the expression of the Bartonella adhesin A [8][9][10]. In infections with humanpathogenic Enterobac-teriaceae, HIF-1 activation is the result of iron-competition between bacteria and host cells caused by secreted bacterial siderophores; here, HIF-1 plays an important role in the defense of Yersinia enterocolitica infections [11]. Using keratinocytes, it was recently demonstrated that HIF-1 activation results in the expression of cathelicidin, an antimicrobial peptide mediating protection against gram positive group A streptococci [6]. LPS from gram negative bacteria can trigger HIF-1 activation in macrophages [12,13] and this HIF-1 activation is crucial for the development of a LPS triggered sepsis [14,15]. Interestingly, LPS-dependent HIF-1 activation does not occur using epithelial or endothelial cell-based infection models [9][10][11].
Although several aspects have been analyzed in the last years, the exact role of HIF-1 in the course of infection with human pathogens remains widely unclear. Therefore, we investigated the activation of HIF-1 in infections with human pathogens more generally using human skin biopsies, cell culture techniques and by employing Staphylococcus aureus infection models. Our data reveal that HIF-1 might play an important and previously underestimated role in many infectious diseases.

Activation of HIF-1 by human pathogens is a general phenomenon in infections
First, we analyzed ex vivo whether pathogen-triggered HIF-1 activation is detectable in biopsy samples of patients suffering from various bacterial, viral, fungal or parasitic skin infections. For this purpose, paraffin-embedded patient skin biopsies were collected according to the following strict inclusion criteria: (a) a clinically suspected infectious process, (b) the histopathological diagnosis of an infection (no malignancy, no autoimmunological disease) and (c) a microbiologically positive laboratory result confirming the presence of specific pathogens in the lesions. In total, four samples of bacterial skin infections, two samples of viral skin infections, three samples of fungal skin infections and two samples of protozoic skin infections were investigated by HIF-1a-specific immunohistochemistry (Fig. 1). In control samples (healthy patients, bioptic sample taken for other reasons, Fig. 1A), a rare and faint nuclear HIF-1a staining occurred in the epidermis whereas the dermis was negative for HIF-1a. In contrast, in all cutaneous infections [here: infections with S. aureus (Fig. 1B), coinfection with S. agalactiae and S. aureus (Fig. 1C), coinfection with Acinetobacter baumanii and E. coli (Fig. 1D), infections with Borellia burgdorferi (Fig. 1E), Varicella zoster virus (Fig. 1F), Human Herpes Virus-8 (Fig. 1G), Tinea rubrum (Fig. 1H,I), C. albicans (Fig. 1J) and Leishmania donovani (Fig. 1K,L)] a strong nuclear HIF-1a signal was detected in keratinocytes (predominantly of the spinal cell layer), dermal capillaries, neutrophils, dermal lymphocytes and macrophages and sub-corneal neutrophils (for details please see legend of Fig. 1).
Activation of HIF-1 in host cells upon infections with B. henselae and several members of the family of Enterobacteriaceae is triggered by oxygen-dependent and -independent mechanisms [7,11]. Now, we were interested whether and by which mechanisms various human pathogens lead to the activation of HIF-1 in host cells. For this purpose, we infected HeLa-229 cells and NHEKs (Normal Human Epidermal Keratinocytes) with a collection of bacterial pathogen reference strains (S. aureus ATCC 33592, S. epidermidis ATCC 12228, E. coli ATCC 25922, P. aeruginosa ATCC 27853, S. agalactiae SK 43) and with the fungal pathogen C. albicans ATCC 90028. In all these in vitro infection models, a robust activation of HIF-1 was detectable via HIF-1a immunoblotting (Fig. 2, 3) and congruent results were obtained via immunohistochemistry using Hela-229 cells grown on coverslips (Fig. S1, not all data shown). HIF-1 activation was correlated with increased oxygen consumption of the respective infected host cells strongly arguing for oxygen-dependent HIF-1 activation mechanisms. This suggestion is confirmed by the observation that HIF-1 activation by S. aureus, P. aeruginosa and E. coli was overcome by culturing infected cells in gas-permeable dishes (data not shown). Accordingly, induction of the HIF-1-regulated VEGF was detected in bacterial infections with S. aureus, P. aeruginosa and E. coli by VEGF mRNA induction whereas HIF-1a mRNA transcript levels itself again appeared unaffected (Fig. S2). Infections with S. pyogenes did not lead to oxygen consumption nor HIF-1 activation (data not shown).
Infection of host cells with S. aureus results in HIF-1regulated gene programming Infection of host cells with B. henselae or Enterobacteriaceae (Y. enterocolitica, Enterobacter aerogenes, Salmonella enterica) results in the activation of HIF-1 and a subsequent HIF-1-dependent angiogenic gene programming in vitro and in vivo [7,[9][10][11]. As HIF-1 activation was also detectable in infections with S. aureus (see above), a clinically most important pathogen, we focussed on this bacterium to elucidate which mechanisms and which biological consequences underlie this process.
Next, we introduced a genetically defined small colony variant of S. aureus (S. aureus DhemB) and the S. aureus DhemB hemBcomplemented strain (S. aureus pCX19VhemB) in our experiments. S. aureus DhemB shows a small-colony phenotype and is auxotrophic for the synthesis of hemin [16]. Therefore, as hemin is essential for the biosynthesis of cytochromes, S. aureus DhemB is deficient in aerobic metabolism [17]. Interestingly, S. aureus DhemB was neither capable in activating HIF-1 nor in VEGF induction whereas the complemented strain was fully restored in its biological effects (oxygen consumption, HIF-1 activation and VEGF induction) suggesting oxygen-dependent mechanisms in the activation of HIF-1 by S. aureus. Furthermore, S. aureus strains defective in certain cell wall components known to be important for pathogenicity [e.g., lipoteichoic acid [18], wall teichoic acid [19], extracellular adherence protein [20]] were not impaired in their ability of inducing HIF-1 activation (Fig. 5).
To exclude that the above described phenomenon might be restricted to this particular S. aureus laboratory strain, we employed several other S. aureus small colony variants (clinical SCVs) originating from patients suffering from chronic infections and their respective parental wild-type strain (Fig. 6, not all data shown). Data clearly revealed that only wild-type but not clinical SCV bacteria lead to the activation of HIF-1 underlining the general aspect of our observations. Further analysis revealed that the clinical SCV S. aureus OM 1b is auxotrophic for menadione [21] and S. aureus A22223 II is unable to synthesize hemin [22]. Both auxotrophies are linked to the activity of the electron transport chain and affect thereby the aerobic metabolism.
To prove whether HIF-1 activation by S. aureus is an oxygendependent process, we determined cellular hypoxia upon a bacterial infection using a hypoxia-sensitive dye and by quantification of the dissolved O 2 concentration in cell culture supernatants. In fact, cellular hypoxia occured in infections with S. aureus 8325-4 and the S. aureus DhemB hemB-complemented strain but not when cells were infected with S. aureus DhemB. Here, a significantly increased oxygen consumption in cells infected with S. aureus 8325-4 and the S. aureus DhemB hemB-complemented strain was detectable (Fig. 7A, B). To analyze whether such HIF-1 activation by S. aureus might be overcome by oxygen, an infection model using conventional versus gas-permeable cell culture dishes was introduced allowing to investigate oxygen-dependent andindependent mechanisms of HIF-1 activation. As expected, activation of HIF-1 by S. aureus 8325-4 and the S. aureus DhemB   hemB-complemented strain did not occur in gas-permeable culture dishes and this correlated clearly with the oxygen partial pressure in the respective cell culture supernatants. For S. aureus DhemB no HIF-1 activation nor oxygen consumption was detectable in infected host cells (Fig. 7C, D). Taken together, these results show that (i) S. aureus wildt-ype but not S. aureus SCV induces a HIF-1 regulated gene programming in host cells and that (ii) HIF-1 activation is in fact an oxygen dependent process.
Finally, we employed a murine intravenous S. aureus infection model in which the S. aureus bacteremia results in kidney abscess formation [23]. After seven days upon infection, macroscopically visible kidney abscesses occurred. Kidneys were removed, fixed and embedded in paraffin. Serial sections were processed by hematoxylin-eosin (H&E)-and HIF-1 staining. Massive abscess formation was also detected in H&E staining (Fig. 8B) and HIF-1a positive nuclei were most abundant in the abscess-surrounding areas (Fig. 8A). HIF-1 activation was also detectable in the peritoneum of mice suffering from a S. aureus-peritonitis (see below).
Taken together, we conclude from the above described in vitro, ex vivo and in vivo data that HIF-1 activation is a general phenomenon in infections with human pathogenic bacteria, viruses, fungi and protozoae and that this phenomenon is most likely due to hypoxic signaling.
Inhibition of HIF-1 by 17-DMAG increases survival of mice in a S. aureus peritonitis model Mice are highly susceptible for a S. aureus peritonitis [24]. For this purpose, we infected mice intraperitoneally with S. aureus wild-  type resulting in a 100% lethality of mice (n = 8). Infection with the related S. aureus SCV (S. aureus ermBVhemB) strain (not activating HIF-1, see above) did not cause such a massive infection as all mice (n = 8) survived this infection (Fig. 9A).
As only S. aureus wild-type and not S. aureus SCV infections led to the activation of HIF-1 in vitro (see above), we finally analyzed the biological role of HIF-1 in this infection model. 17-DMAG is known to be a potent HIF-1 inhibitor [25]. Therefore we first established experiments elucidating the inhibition of HIF-1 activation by 17-DMAG in our experimental in vitro setting. In fact, addition of 17-DMAG prevented the activation of HIF-1 by DFO in a dose-dependent manner. A minimum concentration of 10 mmol/L 17-DMAG was effectual for inhibition of DFOinduced HIF-1 activation (Fig. 9B, not all data shown).
In vivo, the administration of 17-DMAG in fact prevented the peritoneal activation of HIF-1 during infection and this phenomenon was accompanied by a significant better survival of mice suffering from a S. aureus peritonitis. In detail, mice treated twice with 17-DMAG survived to a significantly higher percentage (10/ 38; 26%) compared to mice which did not receive 17-DMAG (1/ 38; 3%, Fig. 9C).

Discussion
In this work, we demonstrate that HIF-1 activation in infections with human pathogenic microorganisms is a general phenomenon not restricted to certain pathogens. A robust and reliably detectable HIF-1 activation was shown (i) ex vivo in biopsies of patients suffering from skin infections, (ii) in vitro using cell culture infection models and (iii) in vivo using murine S. aureus infection models. Moreover, inhibition of bacterial induced HIF-1 activation resulted in higher survival rates in a S. aureus peritonitis model suggesting that the modulation of HIF-1-regulated pathways might influence the course of infections greatly.
In the past, hypoxia and iron-deprivation have been identified as the major conditions leading to HIF-1 activation [2,3]. PHDs (mainly PHD-2) were shown to represent the molecular regulators of HIF-1 activity regulating ubiquitination and subsequent proteasomal degradation of the HIF-1a subunit [26][27][28]. According to these observations, we demonstrated earlier that HIF-1 activation by bacteria is either the result of cellular hypoxia following bacterial infections with e.g., B. henselae [7] or of ironcompetition between bacteria and host cells in infections with certain Enterobacteriaceae [11]. From the herein described experiments it became obvious that HIF-1 activation by pathogens is a general phenomenon occurring in infections with bacteria, viruses, fungi and protozoae (see Fig. 1-4). This hypothesis is supported by various reports showing that HIF-1 activation occurs also in infections with, e.g., Chlamydia pneumoniae [29], Helicobacter pylori  [30] and respiratory syncytial virus (RSV) [31]. Using bacterial and fungal infection models (HeLa-229 cells, NHEKs), we were able to link this HIF-1 activation with increased oxygen consumption [via (i) pimonidazole staining, (ii) direct quantification of oxygen partial pressure in cell culture supernatants and (iii) by overcoming cellular hypoxia using gas-permeable cell culture dishes, see Fig. 2, 3, 7] and strongly arguing for a role of PHD-2 and against transcriptional HIF-1 induction (excluded by quantification of HIF-1a mRNA, see Fig. 4, Fig. S2). Interestingly, in our experiments no increased HIF-2 activation was detectable when cells were infected with P. aeruginosa, E. coli and S. aureus (data not shown) and, until now, no reports are published indicating a role of HIF-2 in such settings.
Infections with S. aureus strains deficient in their aerobic metabolism [S. aureus DhemB [16], clinical S. aureus SCVs [21,22]] did not influence the oxygen partial pressure in cell cultures and neither led to HIF-1 activation and oxygen consumption. HIF-1 activation was fully restored when infections with the respective parental wild-type or genetically complement-  ed S. aureus strains were performed (Fig 4, 6, 7). Obviously, cell wall components of S. aureus (LTA, WTA, EAP) are not involved in HIF-1 activation (Fig. 5) arguing again for a crucial role of infection-triggered hypoxia. Interestingly, when using a murine peritonitis model [24], SCV-strains (deficient in their capacity of HIF-1 activation) turned out to be non-pathogenic (see Fig. 9A). Clinical SCVs are usually isolated from patients suffering from chronic infections (e.g., from lung infections due to cystic fibrosis or from osteomyelitis) [32]. It might be speculated that acute infections (e.g., with S. aureus wild-type) are linked to a rapid and strong HIF-1 activation whereas such activation is missing in chronic infections (e.g., with SCV bacteria) suggesting that the activation of HIF-1 in infections correlates directly with pathogenicity and the course of infection.
Although there is solid evidence for the impact of HIF-1 on the host response, the exact role of HIF-1 in infectious diseases is still largely unknown. The activation of HIF-1 by bacteria was demonstrated in vitro and ex vivo for the angiogenic bacterium B. henselae causing the vasculoproliferative disorders bacillary angiomatosis and peliosis hepatis [7,10]. Here, B. henselae induces an angiogenic reprogramming of host cells via HIF-1. Infected cells secrete angiogenic compounds (e.g. VEGF) leading to the proliferation of endothelial cells [9], the assumed habitate of B. henselae [33,34]. In contrast, for some members of the family of Enterobacteriaceae (Y. enterocolitica, E. aerogenes and S. enterica spp. enterica) an oxygen-independent HIF-1 activation was demonstrated. Here, bacterial siderophores (secreted by these Enterobacteriaceae) compete with host cells for iron affecting PHD2-activity and resulting in the activation of HIF-1 [11]. As mice with deletion of HIF-1a in the intestinal epithelium showed a significant higher susceptibility to orogastric Y. enterocolitica infections, bacterial HIF-1 activation appears to represent a host defense mechanism [11]. In fact, HIF-1 plays an important role in cellular host defense and innate immunity as HIF-1 is crucial for activation and stimulation of T cells by dendritic cells [35], regulates the bactericidal capacity of phagocytes [6] and the production of antimicrobial peptides (e.g., cathelicidin) [14]. Such defensins of the intestinal mucosa have been demonstrated to provide an antibacterial barrier to prevent infections with, e.g., Enterobacteriaceae [36]. Therefore, a lacking intestinal HIF-1 activity should finally result in a higher susceptibility for infections with, e.g., Y. enterocolitica. Similar mechanisms might also be operating in inflammatory bowel diseases (IBDs) where HIF-1 regulates the barrier function of the intestinal mucosa [37].
In contrast to the protective effect in intestinal infections, HIF-1 activation or induction of HIF-1 regulated genes seems to be detrimental in severe systemic infections. Strongly elevated VEGF serum levels were detected in septicaemic patients [e.g., suffering from a methicillin-resistant S. aureus (MRSA) infection] and in patients with severe meningitis [38][39][40]. HIF-1 activation was also observed in a lipopolysaccharide (LPS)-induced murine sepsis model [14]. Interestingly, functional blocking of serum VEGF by soluble VEGF-receptors decreased the mortality in a LPSsepticaemia model dramatically [15]. Concluding these observations and our results gained from the murine S. aureus peritonitis model (Fig. 9A, C) we suggest that an overwhelming HIF-1 activation (and subsequent VEGF secretion) is detrimental in severe infections. A mechanistic explanation for this hypothesis might be hidden in the vasculopermeability-increasing effects of VEGF which was originally described as vasculopermeability factor (VPF) [41]. In fact, VEGF-dependent vascular leakage was already demonstrated in patients suffering from septic shock [42].
Taken together, HIF-1 activation is obviously a general phenomenon in severe infectious diseases caused largely by hypoxia-dependent mechanisms. Our ex vivo, in vitro and in vivo data from infections with humanpathogenic bacteria, fungi, viruses and protozoae point towards a most important role of HIF-1 in the host defense against various pathogens. Extensions of these findings will determine the exact mode of HIF-1 activation by pathogens and the related biological effects in critically ill patients. The application of HIF-1 inhibiting compounds for the treatment of severe infections (such as peritonitis) remains to be elucidated in further studies.
For performing infection experiments, cells were detached with 0.05% Trypsin-EDTA (Gibco). After trypsinization of NHEKs, Trypsin-EDTA was neutralized by adding Trypsin Neutralizing Solution (PromoCell). Cells were seeded the day before infection in cell culture media without antibiotics (to allow bacterial growth). Infection experiments were performed in cell culture media without antibiotics and without FCS to avoid unspecific HIF-1 activation. The following pathogens were used: S. aureus, S. epidermidis, S. pyogenes, C. albicans
To detect HIF-1a protein in infected cells, high-amplification immunohistochemistry was used as described previously [44]. HeLa-229 cells or NHEKs were seeded on coated glass slides (Superfrost Plus, Menzel, Germany) the day before infection. Four hours upon infection, cells were fixed with freshly prepared 3.75% paraformaldehyde (PFA; pH 7.4; dissolved in PBS) for 30 minutes and rinsed in PBS. Antigen retrieval was performed for 2 minutes in preheated target retrieval solution (Dako) using a pressure cooker. For detection of HIF-1a, monoclonal mouse anti-human HIF-1a (a67; Novus) and biotinylated secondary anti-mouse antibodies (Dako) were used. For signal amplification and visualisation, a catalyzed signal amplification system (CSA-Kit, Dako) based on a streptavidin-biotin-peroxidase reaction was used according to the manufacturers instructions. Between incubations, specimens were washed two to three times (50 mmol/L Tris-HCl, 300 mmol/L NaCl, 0.1% Tween-20, pH 7.6). DAB was used as chromogen for peroxidase-reaction. Untreated cells were used as negative controls.
Reporter assays were carried out using the dual-luciferase reporter assay (Promega, Mannheim, Germany). A pGL3-based hypoxia responsive element (HRE) plasmid containing four tandem HIF-1 enhancer sequences from the 39-region of the erythropoietin gene upstream of the firefly luciferase gene [11] was cotransfected with a plasmid encoding a Renilla luciferase reporter construct for normalization [45]. Cells were transiently transfected using the ExGen 500 in vitro Transfection Reagent (Fermentas, St. Leon-Rot, Germany) and incubated for 24 hours under cell culture conditions. Cells were lysed four to six hours after infection with Passive-Lysis-Buffer (Promega). HIF-1-dependent luciferase activities normalized to Renilla activity were determined using a Packard TopCount NXT (BioScience, Groningen, Netherlands).

RNA isolation, Reverse Transcription of Messenger RNA and Polymerase Chain Reaction
Total RNA of control and infected cells was isolated with the RNeasy Mini Kit (Qiagen, Hilden, Germany). To remove contaminating DNA from RNA preparations, the Turbo DNAfree TM Kit (Applied Biosystems, Darmstadt, Germany) was used. Reverse transcription was carried out using SuperScript TM III Reverse Transcriptase (Invitrogen, Karlsruhe, Germany), RNa-seOUT TM Recombinant Ribonuclease Inhibitor (Invitrogen), oligo-d (pT) 18 mRNA primers (New England Biolabs, Frankfurt am Main, Germany) and nucleotides (10 mmol/L dNTP Mix; Fermentas, St. Leon-Rot, Germany). HIF-1a, VEGF and actin gene expression was quantified using a Light Cycler 480 System (Roche Diagnostics, Mannheim, Germany). Primers and standard probes were obtained from LC Search (Heidelberg, Germany) as described earlier [11].

Detection of Cellular Hypoxia and Oxygen Consumption
NHEKs and HeLa-229 cells were cultivated using conventional polystyrene dishes or special gas-permeable dishes (Lumox; Greiner Bio-One, Frickenhausen, Germany) with a hydrophilic tissue culture treated bottom membrane as described [46]. The pO 2 (mmHg) was quantified in the medium of infected or uninfected control cells using the blood gas analyzer ABL-77, sensor cassettes and calibration packs (Radiometer, Willich, Germany). For internal control, cells were harvested in parallel four to six hours after infection, and whole cell extracts were prepared for the detection of HIF-1 activation by immunoblotting as described above.
Detection of hypoxia was also displayed by visualization of the hypoxic cell state using the hypoxia-sensitive marker pimonidazole hydrochloride (Natural Pharmacia International, Burlington, USA). Pimonidazole hydrochloride (which is a 2-nitroimidazole) forms adducts with thiol groups in proteins, peptides and amino acids in hypoxic cells (pO 2 ,10 mmHg) [47]. Here, 200 mmol/L pimonidazole hydrochloride (Hypoxyprobe-1) was added to HeLa-229 cells immediately before infection as described earlier [7]. After six hours of infection, cellular hypoxia was visualized using a primary monoclonal antibody IgG1 that detects protein adducts of pimonidazole in hypoxic cells (Hypoxyprobe-1Mab1; NPI, Burlington, USA) and secondary Cy3-conjugated goat antimouse IgG antibodies (Dianova, Hamburg, Germany). Visualization was performed using a Leica DM IRE2 confocal laser scanning microscope (CLSM; Leica Microsystems, Wetzlar, Germany).

Intravenous infection of mice with S. aureus
Mouse infection experiments were done in accordance with the animal experiment proposal (H2/06, Tuebingen, Germany, approved by the Regierungsprä sidium Tübingen, Germany). The Institutional Animal Care and Use Committee approved this protocol. Female NMRI mice (33-35 g) were infected intravenously with S. aureus 8325-4 (0,3610 6 bacteria/g body weight) resuspended in 200 ml conditioned medium as described previously [23,24]. Seven days after infection, mice were euthanized and kidneys were taken out when a macroscopically visible abscess formation had occurred. Organ fixation was performed using 3.75% PBS-buffered (pH 7.4) paraformaldehyde (PFA).

Selection of patient samples
Formalin-fixed, paraffin-embbed diagnostic skin biposies were collected from the files of the Department of Dermatology, University of Tuebingen. Only samples for which a microbiological diagnosis of the underlying infection was made in parallel were included in the study. Microbiological pathogen identification was performed using standard microbiological techniques [48].
Histology and HIF-1 immunohistochemistry of patient and murine samples For murine kidney samples, serial sections (thickness of the sections: 2 mm) were done and processed by hematoxylin-eosin and HIF-1a staining. Human samples were processed directly from paraffin-embedded blocks.
Paraffin sections (2 mm) were dewaxed in xylene, rehydrated in a series of ethanol washes, and placed in distilled water before staining procedures. Slides were coated with 3-aminopropyl-triethoxysylane. For detection of HIF-1a isoforms, a monoclonal mouse anti-human HIF-1a antibody was used for patient samples (a67; Novus Biologicals) and polyclonal rabbit anti-mouse HIF-1a antibodies were used for murine samples (PM16, obtained from a rabbit immunized with a peptide containing amino acids 553 to 669 of mouse HIF-1a; friendly gift from Patrick Maxwell, Imperial Hospital, London, UK). Detection of bound antibodies was performed by using biotinylated secondary anti-mouse (patient samples) or anti-rabbit (murine samples) antibodies and a catalyzed signal amplification system (Dako) based on the streptavidin-biotin-peroxidase reaction, according to the instructions provided by the manufacturer. Antigen retrieval was performed for 90 s in preheated target retrieval solution (Dako) using a pressure cooker. All incubations were performed in a humidified chamber. Between incubations, specimens were washed two to four times in buffer (50 mmol/L Tris-HCl, 300 mmol/L NaCl, 0.1% Tween-20, pH 7.6). For counterstaining, hematoxylin-eosin staining of the corresponding section was done using standard laboratory procedures.

Statistical analysis
For statistical analysis, the unpaired, two-tailed Student t test was used. Analysis of the survival rates of mice was performed by Kaplan-Meier analysis using the GraphPad Prism Software (GraphPad Software Inc., La Jolla, USA). For all assays, a value of P,0.05 was considered statistically significant. Figure S1 Induction of HIF-1 activation in HeLa-229 cells by bacterial pathogens. Hela-229 cells were seeded on glass slides infected with P. aeruginosa ATCC 27853 or C. albicans ATCC 90028. HIF-1 activation was detected by nuclear accumulation of HIF-1a via immunohistochemistry six hours upon infection.