In Vivo Imaging of Hypoxia-Inducible Factor Regulation in a Subcutaneous and Orthotopic GL261 Glioma Tumor Model Using a Reporter Gene Assay

Abstract

Intratumoral hypoxia changes the metabolism of gliomas, leading to a more aggressive phenotype with increased resistance to radio- and chemotherapy. Hypoxia triggers a signaling cascade with hypoxia-inducible factor (HIF) as a key regulator. We monitored activation of the HIF pathway longitudinally in murine glioma tumors. GL261 cells, stably transfected with a luciferase reporter driven under the control of a promoter comprising the HIF target gene motive hypoxia response element, were implanted either subcutaneously or orthotopically. In vivo experiments were carried out using bioluminescence imaging. Tumors were subsequently analyzed using immunofluorescence staining for hypoxia, endothelial cells, tumor perfusion, and glucose transporter expression. Transient upregulation of the HIF signaling was observed in both subcutaneous and orthotopic gliomas. Immunofluorescence staining confirmed hypoxic regions in subcutaneous and, to a lesser extent, intracranial tumors. Subcutaneous tumors showed substantial necrosis, which might contribute to the decreased bioluminescence output observed toward the end of the experiment. Orthotopic tumors were less hypoxic than subcutaneous ones and did not develop extensive necrotic areas. Although this may be the result of the overall smaller size of orthotopic tumors, it might also reflect differences in the local environment, such as the better intrinsic vascularization of brain tissue compared to the subcutaneous tissue compartment.

MALIGNANT GLIOMA is the most aggressive and not yet curable primary brain tumor. On average, glioma patients die within 12 to 24 months after diagnosis, even with multimodal treatments, including tumor resection and radio- and chemotherapy.^1,2 Hypoxia appears to be linked to tumor aggressiveness and poor outcome as it enhances resistance to radiation and chemotherapy.^3–5 Gliomas form hypoxic areas in particular in domains characterized by high proliferative activity.^6–8 Tumor cells exposed to hypoxia need to adapt to this condition to maintain the energy homeostasis essential for survival. This is achieved by activation of a signaling cascade in which hypoxia-inducible factor (HIF) is the key regulator. ⁹

Therefore, HIF and mediators controlling HIF activity have been suggested as attractive therapeutic targets. ¹⁰

HIF is a heterodimeric protein complex consisting of one of three oxygen-dependent α units (HIF-1α, −2α, −3α) and an HIF-β unit. Both subunits are constitutively expressed, yet under normoxic conditions, levels of HIF-α are modulated through hydroxylation of prolyl residues within the oxygen-dependent degradation domain by prolyl hydroylases (PHDs), which promotes binding to the von Hippel-Lindau protein, initiating proteosomal degradation. Under hypoxic conditions, the PHD enzyme activity is inhibited, leading to stabilization of the HIF-α subunit, which then enters the cell nucleus, where it dimerizes with the HIF-β subunit. The fused HIF protein binds to promoter regions on the DNA comprising the hypoxia response element (HRE) motif. The HIF–HRE interaction activates the expression of target genes, which regulate angiogenesis, metabolism, pH homeostasis, immune response, cell survival, and cell death, as well as metastasis formation.^9,11,12

Bioluminescence assays are widely used in preclinical approaches to investigate genetic alterations in relation to tumor growth or treatment therapies.^13,14 Biolimescence arises from the activity of an enzyme of the luciferase family (e.g., firefly luciferase), which is expressed under the control of a gene of interest or its promoter. Firefly luciferase oxidizes its substrate D-luciferin, leading to photon emission. ¹⁴ Although oxygen is a necessary substrate for this assay, the dependence on tissue oxygen levels appears to be relatively weak. Even under severe hypoxic conditions (oxygen tension 0.2%), bioluminescent signal output of approximately 50% of its normoxia value could be detected. ¹⁵ The bioluminescence reporter gene assay has evolved as a powerful tool, in particular for studying molecular processes in tumors.^13,16,17

Given the central role of hypoxia and HIF signaling, the objective of this study was to develop and characterize a reporter gene assaybased on the bioluminescent reporter gene firefly luciferase for monitoring HIF activity longitudinally in GL261 glioma tumor allografts in mice. In particular, we wanted to evaluate (1) to what extent the implantation site determines the degree of tumor hypoxia, HIF signaling, and the induction of HIF downstream products and (2) whether, similar to observations reported for colon cancer xenograft, ¹⁸ tumor HIF activity diminished over time, indicative of a temporospatial disconnection between hypoxia and hypoxia signaling. For this purpose, murine GL261 glioma cells were stably transfected with an HRE–firefly luciferase construct, which were shown to express the reporter on PHD inhibition, leading to stabilization of HIF-1α and activation by HIF signaling. GL261 cells stably transfected with a SV40-firefly luciferase construct, leading to constituent expression of the reporter, were used as the control. Bioluminescence imaging has been used to monitor activation of the HIF signaling pathway during tumor growth. We compared both a subcutaneous and an orthotopic implantation site to evaluate the influence of the tumor's microenvironment on development and in particular hypoxia signaling.

Materials and Methods

Cell Culture

The murine glioma cell line GL261 was provided by M. Weller, University Hospital Zurich. Cells were maintained in high-glucose Dulbecco's Modified Eagle's Medium, (DMEM) containing 4.5 mg/mL glucose (PAA, Pasching, Austria) with 10% heat-inactivated fetal calf serum.

Cell Lines and Production of Stable Transfected Glioma Cell Lines

GL261 glioma cells were transfected using polyethylenimine (Polysciences Inc., Warrington, PA), as previously described, ¹⁸ with a pGL3 promoter vector (SV40-firefly luciferase) construct from Promega (Dübendorf, Switzerland), and the electroporation was done using an Eppendorf multiporator with the combination in a ratio of 1:10 of linearized neomycin resistance gene containing pcDNA3.1 (Invitrogen, Basel, Switzerland) and a linearized pGL(P2P)95bp (HRE-firefly luciferase) construct (kindly provided by D. Stiehl, Institute of Physiology, University of Zurich), as described by Lehmann and colleagues. ¹⁸ Single-cell clone detection was performed as previously described. ¹⁸

Luciferase Activity Assay

Stably transfected GL261 cells expressing firefly luciferase were grown in six-well plates and treated with dimethyloxallyl glycine (DMOG) (Cayman Chemical, Brunschwig, Switzerland) at 1 mM in dimethyl sulfoxide (DMSO) (Sigma-Aldrich, Steinheim, Germany) or only with DMSO for 24 hours in triplicate. Probes were collected, and the luciferase signal was measured using a luminometer (Berthold Technologies, Regensdorf, Switzerland) and the Promega reporter assay kit according to the manufacturer's instructions.

Western Blotting

Treated cells were washed twice with phosphate-buffered saline (PBS) before adding high-salt lysis buffer containing 0.4 M NaCl, 0.1% Nonidet P-40, 10 mM Tris-HCl pH 8.0, 1 mM ethylenediaminetetraacetic acid, 1 mM phenylmethylsulfonlyl fluoride, and protease inhibitor cocktail (Roche, Reinach, Switzerland), scratched, and placed immediately on ice. Protein extraction concentration was estimated using a Bradford (Bio-Rad, Cressier, Switzerland) assay and separated by loading 20 μg/well on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gel. For the performance of the immunodetection, antibodies luciferase (ab16366; Abcam, Cambridge, UK) 1/1,000 and β-actin (MAB1501; Millipore AG, Zug, Switzerland) 1/50,000 were used. As secondary antibodies, horseradish peroxidase–labeled antibodies were used. Immune complexes were detected by SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific, Reinach, Switzerland) and exposed to x-ray films (Thermo Scientific).

Mouse Model and Glioma Cell Implantation

The study was performed in compliance with the Swiss Law of Animal Protection and approved by the Cantonal Veterinary Office in Zurich (License ZH 168/2010). For the subcutaneous tumor model, 2.5 × 10⁶ stably transfected GL261 cells with 30% Matrigel (BD, Allschwil, Switzerland) were injected subcutaneously into the left flank of 9- to 11-week-old CD-1 nude mice (Charles River, Sulzfeld, Germany), with nine animals in the HRE-luciferase and in the SV40-luciferase group. For the intracranial tumor model, 9- to 11-week-old CD-1 nude mice were anesthetized with isoflurane (Abbott Laboratories, Cham, Switzerland) and fixed into a stereotactic frame. A skin incision was made along the midline, the scalp was retracted, and a hole into the skull was made byhand with a 25-gauge needle 1 mm anterior to the bregma and 1.5 mm right to the midline. A 26-gauge syringe (Hamilton, BGB Analytik AG, Bockten, Switzerland) was inserted 4 mm deep into the brain and retracted 1 mm backward. Two microliters of stably transfected GL261 cells (10⁴ cells/μL) were injected over a period of 2 minutes. The syringe needle was left in position for another 2 minutes and then slowly retracted with a velocity of 1 mm/min. The hole in the skull was closed with bone wax (Ethicon, Johnson & Johnson, St. Stevens Woluwe, Belgium), and the skin wound was closed with two stitches using a polypropylene twine (Premilene, B. Braun, Tuttlingen, Germany). For one week following the surgery, the mice were treated with an antibiotic in their drinking water (1 mL Borgal [Intervet, Veterinaria AG, Pfaffikon, Switzerland] per 500 mL drinking water). Five mice were used per group (HRE-luciferase and SV40-luciferase).

In Vivo Bioluminescence Imaging

GL261 tumor–bearing mice were anesthetized with isoflurane (Abbott Laboratories, Chicago, IL) at a concentration of 1.5% in an oxygen/air mixture. Ten minutes after intraperitoneal injection of 100 μL firefly D-luciferin (15 mg/mL; Caliper Life Sciences, Oftringen, Switzerland) in PBS (DMEM), bioluminescence signals were detected for different exposure periods between 300 and 10 seconds with small binning by an IVIS 100 imaging system (Caliper Life Sciences, Hopkinton, MA) for the subcutaneous tumor–bearing mice, and an IVIS Spectrum imaging system (Caliper Life Sciences) for the intracranial tumor-bearing mice. Images of 30 seconds' exposure time for subcutaneous tumors and 3 minutes' exposure time for intracranial tumors were analyzed with Living Image software 2.6 (Caliper Life Sciences). A region of interest (ROI) was drawn around the area comprising the bioluminescence intensity (threshold 5% of maximum intensity), and the total number of counts within the ROI and its area were determined.

Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) experiments were performed on a 4.7 T animal magnetic resonance system (Pharmascan 47/16 MRI system, Bruker BioSpin GmbH, Ettlingen, Germany) using a cryogenic quadrature transmit-receive coil. Anesthetized mice (isoflurane 3%; Abbott Laboratories, Cham, Switzerland) were placed on a cradle specially designed for mouse imaging. To maintain the anesthesia for the duration of the experiment, 1.5% of isoflurane in an air/oxygen mixture (80%/20%) was provided to the animal via a face mask. Stereotactic fixation was used to avoid head movement during imaging acquisition, and the body temperature was maintained at 37°C using warm-water pipes. Spin echo images using rapid acquisition with relaxation enhancement (RARE) were acquired with the following parameters: repetition time/echo time = 1,000/8 ms, pixel dimension = 110 μm × 110 μm, and slice thickness = 600 mm in coronal, sagittal, and horizontal orientation 1 minute following the intravenous administration of contrast agent (Dotarem, 0.5 mmol gadolinium/mL, Guerbet S.A., France) via the tail vain.

Tumor Volume Determination

The volume of the subcutaneous tumors was calculated according to tumor volume = (length × width ² )/2 (assuming a prolate shape) using caliper measurements. Growth rates were estimated by carrying out a regression analysis assuming exponential tumor growth (y = aċ(kċx)). The intracranial tumor volume was estimated from the MRIs. The tumor area within individual slices was determined using ImageJ software version 1.47n (National Institutes of Health, Bethesda, MD). The procedure was repeated for the data sets recorded in transverse, sagittal, and horizontal orientation, and the respective values were averaged. As MRI measurements were carried out only at three time points, volume values for intermediate points were estimated using regression curve, assuming exponential tumor growth. For the analysis, only data sets from mice measured at all time points were included (subcutaneous, n = 9; intracranial, n = 5).

Immunofluorescence on Tissue Sections

For histologic analysis of hypoxia on the tumors, tumor-bearing mice received an intravenous injection of the hypoxia marker pimonidazole (1.2 mg of pimonidazole in 100 μL of PBS; Hypoxy Probe, Natural Pharmacia International, Burlington, MA). After 1 hour, the mice were then intravenously perfused with Hoechst 33342 stain (Sigma-Aldrich GmbH, Buchs, Switzerland; 0.5 mg in 100 μL PBS per animal) for 2 minutes before being euthanized by cervical dislocations. Tumors were excised, dry ice frozen, and sectioned into 12 μm thick cryosections.

Cryosections were stained according to standard immunofluorescence procedures. The following primary antibodies were used: CD31 (NB100-1642; Novus Europe, Cambridge, UK) 1/1,000; pimonidazole (fluorescein isothiocyanate-MAb1, Hypoxyprobe-1 kit, Natural Pharmacia International) 1/1,000; HIF-1α (NB100-479; Novus) 1/1,000; HIF-2α (NB100-122; Novus) 1/1,000; glucose transporter 1 (GLUT1) (ab14683.; Abcam) 1/1,000; and carbonic anhydrase 9 (CA9) (NB100-417; Novus) 1/4,000. As secondary antibodies, Alexa Fluor 594–coupled IgGs (Molecular Probes/Invitrogen, LuBioScience, Lucerne, Switzerland) were used at a dilution of 1/1,000. Tumor sections were imaged using a Mirax Midi Slide Scanner (Carl Zeiss AG, Feldbach, Switzerland). For the immunofluorescence analysis, only tumors in which both preterminal injections (pimonidazole and Hoechst stain) could be successfully performed were included (subcutaneous tumors, n = 13; intracranial tumors, n = 10).

Statistical Analysis

For statistical analysis of the results of the luciferase assay and the bioluminescence induction study, an unpaired two-tailed Student t-test was performed using GraphPad Prism software version 5.02 (GraphPad Software, La Jolla, CA). Also, a linear mixed model analysis was performed for the bioluminescence induction study using the statistical software R version 2.13.0 (The R Foundation for Statistical Computing, Vienna, Austria) using the lme4 package. A model was designed with time, group, and their twofold interactions as fixed effects and the animal intercepts as random effects. Square root transformation was applied on the signal to improve the normality of the residuals. The effect of the group × time interaction and the two factors was tested with likelihood ratio tests. Residual analysis of the mixed models was performed with QQ-plots to inspect normal distribution, Tukey-Anscombe plots for the homogeneity of the variance and skewness, and scale location plots for homoscedasticity (i.e., the homogeneity of residual variance). The assumption of normally distributed residuals was considered plausible in the subcutaneous group and, to a lesser extent, the intracranial group. A statistically significant difference was indicated by a p value of less than .05.

Results

HIF Pathway Regulation of Stably Transfected Murine Glioma Cells In Vitro

Following stable transfection of GL261 glioma cells with either the HRE-Luc or the SV40-Luc construct, the reporter expression was analyzed using a bioluminescence and Western blot assay. Both cell lines yielded a signal under baseline condition (Figure 1). The cells were then exposed to DMOG, which inactivates the PHD-induced HIF degradation, leading to elevated HIF-α levels and thereby increased HIF activity. On administration of the substrate D-luciferin, the bioluminescence signal exhibited by GL261/HRE-Luc cells was significantly increased by a factor of 3, 24 hours after DMOG. In contrast, DMOG administration did not affect the signal intensity in the GL261/SV40-Luc control cells (see Figure 1).

Figure 1.

Bioluminescence imaging and Western blot of stably transfected GL261/HRE-Luc and GL261/SV40-Luc cells. Comparison of bioluminescence intensity (A) and Western blot (B) assay under control conditions (treatment with DMSO) and under hypoxia-mimicking conditions following administration of the prolyl hydroxylase inhibitor DMOG for GL261/HRE-Luc- and GL261/SV40-Luc-transfected cells for hours. Comparing the bioluminescent signal intensities, an increase in luciferase induction by a factor of 3 was observed for the GL261/HRE-Luc cells 24 hours after administration of DMOG compared to the control condition. In contrast, GL261/SV40-Luc cells show no difference in light output and corresponding luciferase levels for the two treatments. All values are shown as mean ± SD.

HIF Pathway Regulation In Vivo in Subcutaneous and Intracranial Tumors

Stably transfected GL261 cells were injected subcutaneously into the left flank or intracranially into the right striatum of nude mice to monitor the HIF pathway regulation during tumor growth. Bioluminescence signals originating from subcutaneously and orthotopically implanted tumors could be detected 13 and 11 days following implantation, respectively. The bioluminescent signal intensity of both groups (HRE-Luc and SV40-Luc) increased with time (Figure 2, A and B). Intracranial tumor volumes were determined using MRI at days 10, 17, and 24 following implantation (Figure 2C). Intermediate volumes, required for the estimation of the normalization bioluminescence signal, were obtained by interpolation assuming an exponential growth curve. The relative growth rate of both subcutaneous and intracranial tumors was similar, with growth rates corresponding to a volume-doubling time of 4.9 days and 3.6 days for the GL261/SV40-Luc line, respectively, and 6.1 and 4.2 days for the GL261/HRE-Luc tumors, respectively. It appears that transfection with the HRE-Luc construct reduced the growth rate of both subcutaneously and orthotopically implanted tumors, although the difference between the two cell lines did not reach statistical significance due to the large interindividual variability (Figure 3, A and B). We compared the light output per unit volume for the GL261/HRE-Luc and GL261/SV40-Luc tumors for both implantation sites to estimate the activation of the HIF pathway during tumor growth. To account for the specific anatomic condition of each tumor, all measurements were normalized to the first measurement point for each animal. In subcutaneous GL261/HRE-Luc tumors, we observed a sevenfold increase in light output at day 22 followed by a decrease until the end of the experiment (day 27). In contrast, the normalized photon count per volume remained constant for the control GL261/SV40-Luc group apart from increased fluctuations toward the end of the experiment. A similar pattern was observed for the intracranial tumors. The bioluminescent signal of the GL261/HRE-Luc group initially increased approximately 3.2 times between days 11 and 18, whereas the values for the control GL261/SV40-Luc group remained at the basal level. For advanced disease stages (> day 18 following implantation), there was increased variability in the bioluminescent signal in both groups (Figure 3D). Statistical differences were observed at the peak levels of the subcutaneous GL261/HRE-Luc group compared to the corresponding GL261/SV40-Luc group analyzed by t-test (Figure 3C). Analysis of normalized bioluminescent signal for the GL261/HRE-Luc tumors compared to GL261/SV40-Luc over time revealed a significant signal difference indicative of HIF induction for both implantation sites with p values of .0004 for subcutaneous (see Figure 3C) and .0411 for intracranial (see Figure 3D) tumors.

Figure 2.

Tumor signal and size imaging. Bioluminescence images of subcutaneous (A) and intracranially (B) implanted GL261/HRE-Luc and GL261/SV40-Luc tumors in nude mice. Images were acquired repeatedly with the time points indicated in the figure referring to days postimplantation. C, Contrast-enhanced MRIs of nude mice with intracranially implanted GL261/HRE-Luc and GL261/SV40-Luc tumors. Images were recorded following administration of the gadolinium-based contrast agent Dotarem, which significantly enhanced the contrast to noise ratio between tumor and brain tissue as it does not cross the intact blood-brain barrier. The time points indicated refer to days postimplantation.

Figure 3.

Tumor growth curves and intensity of bioluminescence signal. Tumor volumes are displayed as a function of time for GL261/HRE-Luc and GL261/SV40-Luc tumors implanted subcutaneously (A) or intracranially (B). Subcutaneous tumors were measured using a caliper at the time points indicated, whereas the volume of the intracranial tumors was derived from the MRI data. The solid lines represent the results of the regression analysis assuming an exponential growth curve. The results of the regression parameters are indicated in the figure. Induction factor for subcutaneously (C) and orthotopically implanted GL261 tumors (D). Induction factors were computed by dividing the volume-normalized bioluminescent signal intensity at a time t by the corresponding value at baseline. Subcutaneous GL261/HRE-Luc tumors display a transient signal increase, reaching a maximum induction by a factor of 7.6 at day 22, whereas no significant changes with regard to the baseline value were observed in GL261/SV40-Luc tumors (C). Similarly, a transient signal increase was observed for intracranial GL261/HRE-Luc tumors, which reach a maximum induction by a factor of 3.2 at day 18. For GL261/SV40-Luc tumors, the induction factors remained at the baseline value irrespective of the implantation site. Following day 20, the induction factors started to increase for both tumor lines (D). Open squares represent GL261/HRE-Luc tumors; filled squares represent GL261/SV40-Luc tumors. Group–time (g:t) interaction effect analysis revealed a significant difference in normalized luciferase signal between GL261/HRE-Luc and GL261/SV40-Luc for both implantation sites, with the respective p values given in the figures. All values are shown as mean ± SD. BLI = bioluminescence imaging.

Immunofluorescence Analysis of Tumor Tissue

Tissue sections from harvested tumors were labeled for immunofluorescence analysis. Mice received preterminal injections with pimonidazole, an agent that is reduced in a hypoxic environment and covalently bound to thiol-containing proteins in hypoxic cells, ¹⁹ thus indicating hypoxic tumor regions, and a perfusion marker (Hoechst stain). Figure 4 shows stainings of a representative GL261/ HRE-Luc tumor for each implantation site, subcutaneous and intracranial. In subcutaneous tumors, pimonidazole-positive areas (hypoxic areas) were clearly separated from regions containing blood vessels, as revealed by the presence of CD31-positive endothelial cells (see Figure 4Aa). Hypoxic tumor areas could be detected in only half of the brain tumors, independent of size. Orthotopic tumors were consistently less hypoxic. In tumors displaying hypoxic regions, there was a clear segregation between blood vessels and hypoxic tumor tissue, similar to subcutaneous tumors (see Figure 4Ba).

Figure 4.

Immunofluorescence analysis. Overview images of immunofluorescently stained cryosections of a subcutaneous (A) and an intracranial (B) GL261/HRE-Luc tumor. The insert for the high-resolution images is indicated by a white frame. All tumor sections were stained with the perfusion marker Hoechst stain (blue) and the hypoxia marker pimonidazole (pimo; green) as a standard reference. In addition, tumor sections were stained for CD31 as an endothelial marker (a), HIF-1α (b), HIF-2α (c), and the HIF targets GLUT1 (d) and CA9 (e) (all shown in red).

The hypoxia-dependent levels of HIF-1α and HIF-2α colocalized poorly with the hypoxic areas in subcutaneous GL261 tumors (see Figure 4, Ab to 4bc). A similar observation was reported for HIF-1α in the context of a colon tumor model. ¹⁸

As hypoxia triggers HIF signaling and the expression of HIF target genes such as GLUT1 and CA9, we evaluated the colocalization of pimonidazole-positive areas and the areas stained positively for GLUT1 or CA9, as published previously.^20–22 In both the subcutaneous and the intracranial tumors, we observed high expression of GLUT1 and CA9 (see Figure 4, Ad to 4Ae and 4Bd to 4Be). For subcutaneous tumors, GLUT1-, CA9-, and pimonidazole-positive areas were found predominantly at the border of necrotic tumor regions, in agreement with earlier reports.^23–25 GLUT1 and CA9 expression was largely confined to hypoxic tumor regions. The colocalization of pimonidazole-stained areas was better with CA9-than with GLUT1-positive regions (see Figure 4, Ad and 4Ae). In contrast to subcutaneous tumors and in line with the smaller bioluminescent intensity observed in vivo, pimonidazole staining was significantly less pronounced for intracranial tumors and rather confined to small, patchy areas. In general, GLUT1- and CA9-positive regions corresponded to hypoxic areas, although we also observed significant GLUT1 expression in regions that were largely devoid of pimonidazole staining (see Figure 4, Bd and 4Be).

Discussion

Given the central role of hypoxia signaling in tumor development and its effect on the outcome of tumor therapy, it is not surprising that a number of imaging assays targeting full-length HIF-1α ¹⁸ or its oxygen-dependent degradation domain (ODDD),^26,27 as well as assays assessing HIF activity by targeting the DNA binding motif HRE, ²⁸ have been developed. The feasibility of these approaches was demonstrated by assessing the HIF-1 activity in solid tumors grown predominantly subcutaneously in mice. These studies revealed that such reporter gene assays enable studying HIF signaling by monitoring either the stabilization of the key regulating factor HIF-1α or by evaluating the expression of HIF downstream products in vivo. Such assays are relevant for investigating the role of tumor hypoxia and hypoxia signaling in a longitudinal manner and for evaluating the mechanistic aspects of tumor therapy response.

Bioluminescence imaging in combination with a luciferase reporter gene assay has evolved as an attractive tool to explore elements of the HIF signaling pathway in different tumor types,^18,26,29 including gliomas.^27,30 A major concern when using luciferase-based assays is the dependence of the enzymatic reaction on the availability of oxygen as a cosubstrate. Yet this dependence has been reported as rather weak: even at very low oxygen levels (0.2%), approximately 50% of the normoxia photon count may still be recovered. ¹⁵ Only under these harsh conditions would the luciferase activity be underestimated and, correspondingly, the biological event the assay is reporting on. The minimal impact of these effects is also illustrated by the fact that tumors expressing luciferase under the control of the SV40 promoter yield constant volume-normalized light output throughout most of the experimental period for both subcutaneously and orthotopically implanted tumors. As these tumors display a similar degree of hypoxia to the ones expressing the HRE-luciferase construct, we conclude that hypoxia had a negligible effect on the luciferase readout in both cases.

Although subcutaneous tumors are attractive from an experimental point of view, they are of limited value as the tumor develops in a nonnatural environment. It is therefore important to compare the results obtained with analogous studies using orthotopic implantation, in which tumor cells encounter the appropriate microenvironment. Two stably transfected GL261 cell lines were used to investigate the HIF pathway activity during the tumor growth in the subcutaneous and intracranial tumor mouse model. The GL261/HRE-Luc cells should lead to reporter expression only on HIF-mediated signaling, which was confirmed in vitro. In vitro experiments demonstrated the feasibility of monitoring HIF regulation: inhibition of PHD by DMOG led to a significant increase in light output in this cell line, whereas it had no effect on light output when luciferase was expressed under the control of the SV40 promoter (see Figure 1A).

The growth rates of both subcutaneous and intracranial implanted tumors were similar within error limits displaying an approximately 10-fold increase in a period of 2 weeks. In vivo imaging experiments in mice carrying subcutaneously implanted GL261/HRE-Luc tumors revealed a transient induction of the luciferase reporter indicative of induced hypoxia signaling, whereas the bioluminescent signal of GL261/SV40-Luc control tumors stayed at the baseline level throughout the observation period. A significant effect was detected when analyzing the signal as a function of time after implantation for both tumor locations, although the statistical significance was higher for the subcutaneous tumors (subcutaneous, p = .0004; orthotopic, p = .0411). This is corroborated by the fact that significance was also reached for subcutaneous tumors when comparing GL261/HRE-Luc and GL261/SV40-Luc signals at a single time point (days 20–25 postimplantation). Histology confirmed the presence of clearly defined hypoxic areas as revealed by pimonidazole staining in all analyzed subcutaneous tumor tissues and in half of the orthotopic tumors. In line with the in vivo detection of luciferase induction in HRE-Luc-transfected GL261 cells, we also found the induction of the HIF downstream products GLUT1 and CA9 that were distributed across the whole tumor, demonstrating the relevance of the in vivo readout.

In subcutaneous GL261/HRE-Luc tumors, maximum luciferase induction was observed around day 22 following implantation; thereafter, the signal steadily declined until the end of the experiment. This transient induction is in line with an earlier observation in a colon cancer mouse model. ¹⁸ The authors speculated that an HIF-induced negative feedback mechanism might account for this phenomenon. Although this definitely constitutes an option, we also have to consider that tumors became necrotic, with severe hypoxia at the periphery of necrotic areas, toward the end of the experiment, which might explain the reduced light output per volume and potentially also the significant fluctuation observed for large tumors. Reproducible positioning of subcutaneous tumors between subsequent measurements is problematic, and variation in the location of necrotic areas (with no light output) with respect to the surface will translate into variations in the bioluminescent signal intensity measured at the animal's surface. The observation of necrotic areas is relevant as necrosis in glioblastoma was shown to correlate with the negative clinical outcome. ³¹ The metabolic milieu was found to be changed in the perinecrotic area with increased levels of GLUT1, one of the targets of HIF signaling, indicative of increased glycolytic activity.^25,32 This is in line with our results from histologic analysis, which show a high degree of hypoxia and high levels of GLUT1 at the periphery of necrotic areas (see Figure 4, Ad).

This pattern was largely reproduced for intracranial tumors, although the induction factor for the HRE-Luc construct was significantly smaller, hinting at a lesser degree of hypoxia and HIF activity in these tumors (see below). We observed a significant difference in the normalized luciferase signal induction as a function of time between the GL261/HRE-Luc and the GL261/SV40/Luc group. Histologic analysis revealed small, confined hypoxic areas within the tumor, in line with the much weaker induction factor observed in HRE-Luc-transfected glioma cells. In fact, small hypoxic regions were found in only half of the intracranial tumor tissues and were typically located in the central tumor zone (see Figure 4B). The low degree of hypoxia is likely due to the high vascularization of intracranial GL261 tumors, in contrast to the hypoxic area in the subcutaneous tumors, and CD31-positive cells were detected throughout the intracranial tumors, in agreement with earlier studies.^33,34

It should be pointed out that HIF activity does not necessarily reflect hypoxia; that is, HIF-1α stability and HIF activity assays yield information complementary to hypoxia readouts. Earlier studies have revealed a temporospatial disconnection between tissue hypoxia and HIF signaling when comparing measurements of tumor hypoxia using positron emission tomography with the hypoxia tracer [¹⁸F]-fluoromisonidazole with bioluminescence measurements of HIF-1α levels and HIF activity in transfected colon cancer cells. These in vivo findings were confirmed by immunofluorescence analysis of tumor specimens. ¹⁸ In our study, the correlation between the HIF-α- and pimonidazole-stained regions was rather poor for subcutaneous tumors that displayed extended hypoxic domains and revealed an up to sevenfold induction of luciferase under the control of a promoter comprising HREs. Interestingly, we observed better colocalization between hypoxic areas and regions displaying the HIF downstream products CA9 and, to a lesser degree, GLUT1. This can also be based on the higher stability of the downstream products compared to that of HIF-1α. ²¹ One factor contributing to the spatial (and temporal) disconnection is that HIF signaling may be induced even under normoxic conditions.^35,36 For example, it has been found that many glioma cell lines contain a mutation of the glycolytic enzyme isocitrate dehydrogenase 1 (IDH1).

Loss of function of IDH1 leads to reduced levels of α-ketoglutarate, a cosubstrate of PHD, ³⁷ and, correspondingly, to reduced PHD activity, leading to HIF-1α stabilization and activation of HIF signaling. The fact that we observed significant GLUT1 and CA9 expression in areas devoid of pimonidazole staining in orthotopic GL261/HRE-Luc tumors might hint at such a mechanism. Alternatively, there is evidence for downregulation of HIF signaling under conditions of persistent hypoxia, which would explain the absence of GLUT1 and CA9 staining in pimonidazole-positive areas. ¹⁸ The colocalization between pimonidazole, HIF-α, and HIF downstream products appeared better in orthotopic than in subcutaneous tumors. We might speculate that due to the reduced hypoxic stress in orthotopic tumors as revealed by the in vivo measurements and by the histologic workup, HIF signaling is not yet downregulated—hence, the better spatial correlation.

It is commonly accepted that the high degree of vascularization of gliomas, although preventing the development of hypoxic areas, is contributing to their malignancy. ³⁸ Vascular endothelial growth factor (VEGF), an important proangiogenic factor, is an HIF target gene geared toward restoring blood supply to the hypoxic territories. The newly formed vessels are structurally and functionally abnormal. ³⁸ Antiangiogenic therapy based on VEGF inhibition is currently under evaluation for the treatment of gliomas.^39,40 Similarly, it has been shown that HIF-1α inhibition reduces VEGF expression and the growth rate of malignant gliomas and hence constitutes an alternative target for clinical treatment of gliomas. ⁴¹ In addition to its antiangiogenic efficacy, HIF-1α inhibition would also affect other downstream targets of the HIF signaling cascade.

Metabolic reprogramming is a characteristic feature of cancer.^42–44 For example, energy production switches from the oxidative phosphorylation to glycolysis even in normoxic regions. An enabling step for this switch is the upregulation of GLUT1 and GLUT3, both target genes of HIF. Increased cellular uptake of glucose is mandatory to cope with the reduced efficiency of adenosine triphosphate production via glycolysis.^9,45 Not surprisingly, we observed high expression levels of GLUT1 in GL261/HRE-Luc tumors. GLUT1 is highly expressed in normal brain tissues, enabling the passage of glucose, the major energy substrate of the brain, across the blood-brain barrier. ⁴⁶ Although the intrinsic background by cerebral glucose use degrades the value of [¹⁸F]-2-fluoro-2-deoxyglucose positron emission tomography (FDG-PET) for tumor detection and staging, GLUT1 as an HIF downstream readout still provides valuable information on HIF activity in the isolated tumor specimen. In fact, we observed good colocalization of GLUT1 expression and hypoxic areas in orthotopic and, to a lesser extent, subcutaneous tumors due to the reasons outlined above.

Although subcutaneous tumors are attractive from an experimental point of view, their clinical relevance is debated due to the unnatural environment in which they grow. Clearly, such studies have to be complemented by studies using orthotopic tumor models, which allow elevation effects of the proper tumor microenvironment. ⁴⁷ For example, it has been demonstrated that subcutaneously and intracranially injected tumors showed different sensitivity to chemotherapeutic interventions ⁴⁸ and that subcutaneously implanted tumors showed higher vascular permeability ⁴⁹ and lower blood flow. ⁵⁰ This is in agreement with our findings. Subcutaneously implanted GL261 tumors displayed reduced vascularization (CD31 staining) and, correspondingly, a higher degree of hypoxia (including the development of necrotic areas) and higher levels of the HIF downstream products GLUT1 and CA9 compared to the orthotopically implanted tumors. High HIF levels are associated with increased angiogenic activity, characterized by newly formed vessels displaying a high degree of vascular permeability.^38,51,52 Although the typically 10-fold difference in tumor volume and potentially different tumor stages may have contributed to this discrepancy, there are obvious differences in the tumor environment (e.g., in the vascularization of the host site), which are likely to affect the outcome. It was recently shown that the development of the tumor vascular system critically depends on the implantation site. ⁵³

In summary, we have demonstrated a significant transient upregulation of HIF activity in subcutaneously and, to a lesser extent, orthotopically implanted GL261 tumors in nude mice using a bioluminescence reporter gene assay. The signal decrease during advanced tumor stages might be related to a negative feedback loop that has recently been postulated and/or to the contribution of tumor necrosis. Histologic analysis confirmed—at least in part—the in vivo findings. Differences between subcutaneously and orthotopically implanted GL261 tumors were of both a qualitative and a quantitative nature. Subcutaneously implanted tumors revealed extensive hypoxic domains with high levels of HIF downstream products, in line with the in vivo findings. In contrast, orthotopic tumors displayed only small, patchy hypoxic domains, which explains the small effects observed in vivo. In contrast to the findings in subcutaneous tumors, high GLUT1 expression was also observed in domains that did not display significant pimonidazole staining in intracranial gliomas, indicative of nonhypoxic HIF activation. Tools enabling the noninvasive monitoring of HIF activity will contribute to the understanding of the factors regulating HIF signaling in tumors and might be of use in developing improved treatment strategies for (hypoxic) tumors.