Quantitative evaluation of oxygen metabolism in the intratumoral hypoxia: 18F-fluoromisonidazole and 15O-labelled gases inhalation PET

Background Intratumoral hypoxia is one of the resistant factors in radiotherapy and chemotherapy for cancer. Although it is detected by 18F-fluoromisonidazole (FMISO) PET, the relationship between intratumoral hypoxia and oxygen metabolism has not been studied. The purpose of this study was to evaluate the intratumoral perfusion and oxygen metabolism in hypoxic regions using the rat xenograft model. Ten male Fischer rats with C6 glioma (body weight = 220 ± 15 g) were investigated with 18F-FMISO PET and steady-state inhalation method of 15O-labelled gases PET. The tumoral blood flow (TBF), tumoral metabolic rate of oxygen (TMRO2), oxygen extraction fraction (OEF), and tumoral blood volume (TBV) were measured under artificial ventilation with 15O–CO2, 15O–O2, and 15O–CO gases. Multiple volumes of interest (1-mm diameter sphere) were placed on the co-registered 18F-FMISO (3 h post injection) and functional 15O-labelled gases PET images. The TBF, TMRO2, OEF, and TBV values were compared among the three groups classified by the 18F-FMISO uptake as follows: group Low (L), less than 1.0; group Medium (M), between 1.0 and 2.0; and group High (H), more than 2.0 in the 18F-FMISO standardized uptake value (SUV). Results There were moderate negative correlations between 18F-FMISO SUV and TBF (r = −0.56 and p < 0.01), and weak negative correlations between 18F-FMISO SUV and TMRO2 (r = −0.38 and p < 0.01) and 18F-FMISO SUV and TBV (r = −0.38 and p < 0.01). Quantitative values were as follows: TBF, (L) 55 ± 30, (M) 32 ± 17, and (H) 30 ± 15 mL/100 mL/min; OEF, (L) 33 ± 14, (M) 36 ± 17, and (H) 41 ± 16%; TMRO2, (L) 2.8 ± 1.3, (M) 1.9 ± 1.0, and (H) 2.1 ± 1.1 mL/100 mL/min; and TBV, (L) 5.7 ± 2.1, (M) 4.3 ± 1.9, and (H) 3.9 ± 1.2 mL/100 mL, respectively. Intratumoral hypoxic regions (M and H) showed significantly lower TBF, TMRO2, and TBV values than non-hypoxic regions (L). OEF showed significant increase in the severe hypoxic region compared to non-hypoxic and mild hypoxic regions. Conclusions This study demonstrated that intratumoral hypoxic regions showed decreased blood flow with increased oxygen extraction, suggesting the need for a treatment strategy to normalize the blood flow for oxygen-avid active tumor cells in hypoxic regions. Electronic supplementary material The online version of this article (doi:10.1186/s13550-017-0263-6) contains supplementary material, which is available to authorized users.


Background
It is well known that the hypoxic regions in tumors can be resistant to radiotherapy as well as chemotherapy [1]. PET imaging of a hypoxic tracer, such as 18 F-fluoromisonidazole (FMISO), can be used to detect hypoxia in patients with head and neck cancer and predict an adverse prognosis against hypoxia-targeting agents [2]. 18 F-FMISO is a 2-nitroimidazole, which accumulates in the tissue environment at low oxygen concentrations by binding to macromolecules after reductive activation [3]. The rate of its accumulation increases as partial pressure of oxygen (pO 2 ) falls below 10 mmHg [4]. Clinical hypoxia thresholds based on pO 2 measurements (2.5 and 5.0 mmHg) corresponded to the 18 F-FMISO standardized uptake value (SUV) (2.0 ± 0.6 and 1.4 ± 0.5, respectively) [5]. However, 18 F-FMISO PET cannot reflect regional oxygen metabolism, and in vivo quantitative imaging of oxygen consumption in the hypoxic tumor has not been investigated.
Tumor hypoxia is divided into two types, acute hypoxia and chronic hypoxia. Acute hypoxia is the result of pronounced heterogeneity of the tumor blood flow distribution, whereas chronic hypoxia is observed at the limits of oxygen diffusion (100-200 μm) away from blood vessels [6]. It has been suggested that intratumoral oxygen metabolism is heterogeneous depending on the tumor microenvironment and quantitative measurement of oxygen consumption is essential to elucidate the mechanism of tumor hypoxia as it is the shortage of oxygen supply by blood flow against oxygen demand. In the hypoxic regions, two conditions were assumed, one is the shortage of blood supply with reduced blood flow and maintained oxygen consumption; the other is increased oxygen demand with a maintained blood flow and increased oxygen consumption. In this study, we aimed to evaluate the relationship between intratumoral oxygen demand and blood supply in hypoxic regions by in vivo 15 O-labelled gases and 18 F-FMISO PET using the rat xenograft model of C6 glioma, focusing on the oxygen extraction.

Methods
Preparation of 18 F-FMISO and 15 O-labelled gases 18 F-FMISO was produced by the method as described previously with minor modifications [7,8]. In brief, 18 F-fluoride was produced with a cyclotron by the 18 O(p, n) 18 F nuclear reaction using 18 O-H 2 O. 18 F-fluoride was trapped with an ion-exchange resin QMA sep-pak light cartridge and eluted from the resin by 210 mmol/L K 2 CO 3 solution (0.2 mL) and 31 mg/mL Kryptofix 2.2.2 in acetonitrile solution (K222, 0.7 mL). Water was removed by azeotropic distillation. To the residue, repeated addition of acetonitrile solution (0.2 mL) and azeotropic distillation was performed. The dried 18 F-KF/K222 complex mentioned above was added to the precursor (3-(2-nitroimidazol-1-yl)-2-O-tetrahydropyranyl-1-O-toluenesulfonylpropanediol 5 mg) in acetonitrile solution (0.7 mL) and reacted at 110°C for 10 min. After cooling, the reaction mixture was added to 0.5 M HCl (0.7 mL). The mixture was hydrolyzed by heating for 5 min at 100°C. After cooling, the mixture was neutralized with 1.5 M sodium acetate and transferred to the high-pressure liquid chromatography (HPLC) injector. Crude product was purified with semi-preparative HPLC (column: YMC-ODS-AM 10 × 250 mm, mobile phase: 3% EtOH/H 2 O, flow rate: 5.0 mL/min). The purified fraction was evaporated to dryness, and the residue was dissolved in saline containing benzyl alcohol (24 μL). The specific activity ranged from 30 to 300 GBq/μmol at the end of synthesis. Radiochemical purity was higher than 98%. 15 15 O-CO 2 ) added N 2 gas target at a 12-MeV proton 7 μA current accelerated by the CYPRIS R HM 12S in-house cyclotron (Sumitomo Heavy Industry).

Animal preparation and implantation of C6 glioma cells
Male Fischer rats from Charles River Japan, Inc. were housed under a 12-h light/12-h dark cycle and had free access to food and water. A rat glioma C6 cell line, which was derived from gliomas induced by N-nitrosomethylurea, was provided by the RIKEN BRC. The cells were cultured in MEM medium (Sigma-Aldrich) with 10% fetal bovine serum (Sigma-Aldrich) at 37°C in a humidified incubator containing 5% CO 2 . Cultured cells were collected after washing in PBS and harvested with trypsin. Tumor xenograft models were established by the subcutaneous injection of 1 × 10^6 tumor cells suspended in 0.2 mL of culture medium and Matrigel (1:1; BD Biosciences) into the bilateral shoulder region of F344 rats (a total of two sites in the left and right shoulder per one rat). PET experiments were performed 2 to 3 weeks after the implantation of C6 glioma cells, when the tumor diameter had reached 1-2 cm. 18

F-FMISO PET
Ten rats with tumor xenografts (8 weeks age, body weight = 220 ± 15 g) were anesthetized with 2% isoflurane with room air, and a Terumo 24-G indwelling cannula was inserted into the tail vein. PET/CT data were acquired with a small-animal PET system (Inveon PET/ CT system, Siemens Medical Solutions) [9]. The animals were placed in a feet-first supine position in the PET scanner. 18 F-FMISO (69.9 ± 10.1 MBq) was administered intravenously via the catheter cannula, and dynamic PET scans (60 min) were started at the same time (n = 4). Delayed static scans (10 min) were performed at 3 h post injection in all rats (n = 10). CT acquisition was performed before or after the PET acquisition. 15

O-labelled gases inhalation PET
The next day after FMISO PET, the tumoral blood flow (TBF), tumoral metabolic rate of oxygen (TMRO 2 ), oxygen extraction fraction (OEF), and tumoral blood volume (TBV) were measured with a steady-state inhalation method under artificial ventilation of 15 O-CO 2 , 15 O-O 2 , and 15 O-CO gases mixed with room air and oxygen [10,11]. Rats were anesthetized with 2% isoflurane with room air during the arterial cannulation in the femoral artery, later switched to intramuscular injection of midazolam (1.2 mg/kg body weight), xylazine (4.8 mg/kg body weight), and butorphanol (1.6 mg/kg body weight). The stability of the anesthesia was already established in the previous study [10]. Tracheotomy was performed, and artificial ventilation was started (tidal volume 2-3 mL, frequency 60/min).
Inhalation of the 15 O-CO 2 (200 MBq/min) or 15 O-O 2 (400 MBq/min) gas was continued during the PET measurements for 16 min. The inhalation time of the 15 O-CO gas (400 MBq/min) was 3 min, and the PET measurements were continued for a total of 13 min. Arterial blood sampling (0.1 mL) was performed during the steady-state PET acquisition in the 15 O-CO 2 and 15 O-O 2 gases studies (13 and 16 min after the start of the scanning) and 10 min after the start of the scanning in the 15 O-CO gas study (A total of five blood samplings per rat). The radioactivity and weight of the whole blood and plasma after centrifugation (3000 round/min (800g), 3 min) were measured with a well scintillation counter (BeWell, Molecular Imaging Labo). CT acquisition was performed after the PET acquisition for the attenuation and scatter correction.
Systemic blood pressure (BP) and heart rate (HR) were indirectly measured with a tail-cuff apparatus (BP-98A-L, Softron). Arterial blood gas (ABG) analysis was performed from the blood samples (i-STAT system, Abbott Point of Care Inc). The body temperature was maintained with a heating sheet system.

Quantitative data calculation
All PET data were reconstructed with 3-dimensional ordered-subset expectation maximization followed by maximum a posteriori (OSEM3D-MAP) (16 subsets, 2 OSEM3D, and 18 MAP iterations) with attenuation and scatter correction. The image matrix was 128 × 128 × 159, which yielded a voxel size of 0.776 × 0.776 × 0.796 mm. Quantitative values of TBF, TMRO 2 , OEF, and TBV were calculated by the steady-state inhalation method by Frackowiak et al. [10,11]. Detailed calculation formulas were written in our previous study [10]. In brief, TBF was calculated from the 15 O-CO 2 PET using the radioactivity counts of the whole blood during the steady state. TMRO 2 and OEF were calculated from the 15 O-CO 2 and 15 O-O 2 PET using the radioactivity counts of the whole blood and plasma during the steady state. TBV was calculated from the 15 O-CO PET using the radioactivity counts of the whole blood. Functional images of the TBF, TMRO 2 , OEF, and TBV were reconstructed using in-house software from the last 6-min frame of each 15 O-labelled gases study. A tissue-blood partition coefficient for water was fixed at 0.91 mL/g [12]. The correction value of the hematocrit between the great vessels and the tissue was fixed at 0.70 [13]. The TBV data were used to correct for intravascular hemoglobin-bound 15 O 2 [14].
Quantitative PET data analysis 18 F-FMISO PET images were co-registered to 15 O-labelled gases PET with reference to the CT image using AMIDE software (Ver. 1.0.1) and PMOD software (Ver. 3.404). To compare the distribution of the two PET images, multiple small volumes of interest (VOIs, 1-mm sphere) were manually placed over the tumor of 18 F-FMISO PET images (static images 3 h post injection) and functional images of 15 O-labelled gases PET according to a previous study [15]. In setting VOIs, co-registered coronal images of 18 F-FMISO (window level 0-3) and TBF (window level 0-80) were displayed using view tool of PMOD and 1-mm sphere VOIs were paved on PET images over the entire tumor where there is an uptake of 18 F-FMISO or TBF (Additional file 1: Figure S1). In addition, 8-mm sphere VOIs were placed on the forelimb muscle and used for the evaluation of background tissue activity to determine the threshold level of significant FMISO uptake. The quantitative values of TBF, TMRO 2 , OEF, and TBV were plotted against the 18 F-FMISO SUV for correlation analysis. Next, VOIs were divided into three groups according to the 18 F-FMISO SUV: group Low, low 18 F-FMISO uptake region (less than 1.0 in SUV); group Medium, moderate 18 F-FMISO uptake region (between 1.0 and 2.0 in SUV); and group High, high 18 F-FMISO uptake region (more than 2.0 in SUV). The threshold value between group Low and Medium was determined by the background tissue activity, using the forelimb muscle uptake of 18 F-FMISO SUV (the value obtained by adding the average and twice the standard deviation). The threshold value between group Medium and High was determined according to the previous study which reported the 18 F-FMISO SUV of 2.0 corresponded to the pO 2 measurement of 2.5 mmHg [5].

Immunohistochemistry
After 15 O-labelled gases PET/CT, pimonidazole (hypoxia marker, HypoxyprobeTM-1, 60 mg/kg) was injected 60 min before euthanasia. The tumor tissue samples were resected and fixed overnight with 4% paraformaldehyde and later cryoprotected in 30% sucrose in PBS before storage at −80°C. Frozen sections (10-14 μm thick) were obtained with a cryostat (CryoStar NX70, Thermo Scientific) and mounted on glass slides. After blocking with blocking agent (Dako REAL Peroxidase-Blocking Solution) for 5 min at room temperature, incubations with the primary antibody against pimonidazole (FITC-MAb1, 1/100 dilution) for 30 min and the second antibody (HRP linked to rabbit anti-FITC, 1/100 dilution with Dako REAL Antibody Diluent) for 30 min were performed in order. Mounted sections were coverslipped after dehydration in ethanol and clearance with xylene, followed by 3,3′-Diaminobenzidine (DAB) (Dako Envision) and hematoxylin staining.

Statistical analysis
Correlation analyses of TBF, TMRO 2 , OEF, and TBV against the 18 F-FMISO SUV were performed with Spearman's correlation coefficient. The quantitative values of TBF, TMRO 2 , OEF, and TBV were compared among the three groups with multiple comparison tests (Games-Howell test). Probability values of less than 0.05 were considered to denote statistical significance using SPSS version 19.0 (SPSS).

Results
The number of animals and tumor regions which were included in the analysis are summarized in Table 1. The time activity curves of 18 F-FMISO PET showed increased accumulation in the tumor and wash out from background tissue over time (Fig. 1a). Hypoxic regions in the tumor were clearly visualized at 3 h post injection of 18 F-FMISO (Fig. 1b).
The mean systolic and diastolic BP, HR, PaO 2 , PaCO 2 , hemoglobin concentration (Hb), hematocrit (Ht), and oxygen saturation (SaO 2 ) during 15 (Fig. 2). Figure 3a shows mild but almost the same distribution pattern in TMRO 2 as compared to TBF, partial elevation of OEF in the area close to the center, and a high TBV spot in the marginal zone of the tumor. Co-registered images of 18 F-FMISO and TBF showed that intratumoral hypoxic regions with FMISO uptakes were located as lining the inside of the high TBF area and there was no accumulation in the central area inside the FMISO uptake, suggesting a necrotic region (Fig. 3b-d).
Relationships between 18 F-FMISO uptake and quantitative values of 15 O-labelled gases PET are shown in Fig. 4a. There were moderate negative correlations between 18  Intratumoral hypoxic regions (group Medium and High) showed significantly lower TBF, TMRO 2 , and TBV values than non-hypoxic regions (group Low). OEF showed significant increase in the severe hypoxic region compared to non-hypoxic and mild hypoxic regions. These findings suggested that oxygen demand was relatively maintained by increased oxygen extraction in the hypoxic region.
The pimonidazole-staining immunohistochemistry revealed a heterogeneous intratumoral distribution of the hypoxic regions, and high intensity staining was observed mainly at the edge of the viable region of the tumor section (Fig. 5a-c). The central area of the tumor showed necrotic regions partially accompanied by hypoxic cells (Fig. 5d). Pimonidazole staining corresponded to 18 F-FMISO uptake regions, surrounded by blood flow preserved regions (Fig. 6).

Discussion
We have reported herein on the relationship between blood supply and oxygen consumption in intratumoral hypoxic regions according to the 18 F-FMISO SUV. Our results indicated that tumor cells in the hypoxic environment try to maintain the oxygen metabolism against decreased blood supply similar to the misery perfusion which has been observed in steno-occlusive disease in the brain [10,16]. As shown in Fig. 4, we observed wide range of TBF or TMRO 2 values in the tumors. This variability is considered to result from intertumoral heterogeneity. In some tumors, we observed decreased trend of TBF and TBV, increased trend of OEF, and stable TMRO 2 against the increase of 18 F-FMISO SUV (Additional file 1: Figure S2). However, weak correlation was observed between OEF and 18 F-FMISO SUV as a whole, possibly due to the intertumoral heterogeneity of  10 20 In one rat, TMRO 2 and OEF images were not generated and excluded from the evaluation due to the technical reason blood flow and oxygen utility level. Figure 4b shows significant increase of OEF along with increased 18 F-FMISO SUV, suggesting that high 18 F-FMISO uptake regions could survive the hypoxic conditions by the mechanism of elevated OEF.
Previous studies have reported perfusion and oxygen metabolism in patients with cerebral gliomas using 15 Olabelled gases PET [17][18][19][20]. The TBF and TBV values varied widely depending on the vascularity of the tumor, whereas significantly decreased TMRO 2 and OEF values were observed in the tumor region compared with normal gray matter or white matter in the brain. Compared to the normal brain tissue, glioma cells demanded less oxygen and received excessive TBF, which has been defined as the condition of luxury perfusion. The authors concluded that the mechanism of this luxury perfusion is unknown, whether it was due to decreased oxygen consumption as a result of adaptation to hypoxic conditions or was inherent in the metabolism of the gliomas. In our study, relatively increased OEF was observed in    hypoxic regions against non-hypoxic regions, but OEF was decreased compared to the cerebral OEF (65 ± 9.1%) in the normal rat brain according to our previous study [10]. This result was consistent with the previous study in human by Mineura et al., which reported significantly decreased OEF in the tumor compared to the normal brain. The new findings of our study was that intratumoral hypoxic regions showed relatively increased oxygen extraction as well as decreased blood flow compared to the non-hypoxic regions. We selected 18 F-FMISO for a hypoxic tracer as it had been used most extensively in the previous study [21]. 18 F-FMISO PET significantly correlated with the pimonidazole immunohistochemical staining and the endogenous hypoxia related marker CA IX [22]. 18 F-FMISO has no protein binding, diffuses freely, and accumulates in those regions with a low oxygen concentration, which enables a delayed scan (2-3 h post injection) to achieve the best contrast [23,24]. We compared the two PET scans in this study, and multiple VOI-based comparison was a feasible method as reported in the previous study to prevent high variability of data in the voxel based analysis [15]. VOI size of 1 mm was selected because intratumoral distribution of 18 F-FMISO was heterogeneous, and larger VOIs led to ambiguous evaluation between the hypoxic and edge regions. In 15 O-labelled gases experiments, we need at least 2 h for total anesthesia time from the preparation of an arterial sampling line to the end of 15 O-labelled gases PET. Performing 15 O-labelled gases and FMISO studies in 1 day is an extreme overload for rats due to the prolonged anesthesia. We performed less invasive FMISO PET at first and set the interval of 1 day between FMISO and 15 O-labelled gases PET in this study. The anesthesia methods were different between FMISO PET and 15 O-labelled gases PET. We monitored the systemic blood pressure, heart rate, and arterial blood gas data during 15 O-labelled gases PET and confirmed that these factors were within physiological range and stable (Table 2). In addition, pimonidazole staining after 15 O-labelled gases PET matched with the 18 F-FMISO PET. It was considered that the difference about anesthesia had a minor effect for the evaluation of tumor perfusion and oxygen metabolism.
As for the arterial measurement during 15 O-labelled gases PET, two blood samples were taken per each scan during steady state (13 and 16 min after the start of the scanning) and the average value was used for the quantitative calculation. Percent differences of the radioactivity Fig. 5 Immunohistochemistry of the intratumoral hypoxic region (pimonidazole staining). a Overall image of the tumor (summed image of ×20 magnification), b boundary zone: red color square, c hypoxic region: yellow color square, and d necrotic region: blue color square (×200 magnification) between the two samples were 4.5 ± 3.3% for the whole blood and 6.0 ± 5.6% for the plasma. The measurements of arterial blood were considered to be sufficiently accurate and stable.
There are two types of tumor hypoxia, acute hypoxia (perfusion limited hypoxia) and chronic hypoxia (diffusion limited hypoxia) [6,21]. Acute hypoxia is a transient hypoxic condition which can change from day to day, whereas chronic hypoxia is a rather stable condition. There is a possibility that hypoxic regions could be different between 18 F-FMISO PET and 15 O-labelled gases PET. Okamoto reported the high reproducibility of tumor hypoxia by comparing two 18 F-FMISO PET scans within a 48-h interval in patients with head and neck cancer [25]. They concluded that the values for the 18 F-FMISO PET uptake (SUV and uptake ratio of tumor to muscle) and the hypoxic volume between the two scans were highly reproducible. Furthermore, we performed immunohistochemical staining with pimonidazole after 15 O-labelled gases PET, which corresponded to the 18 F-FMISO PET findings seen on the previous day. These results demonstrated that the main hypoxia in the tumor was chronic hypoxia. However, acute component cannot be excluded in the low TBF regions as we evaluated one time point evaluation of 15 O-labelled gases PET. Further study is necessary to evaluate the involvement of acute ischemia by investigating the temporal change, such as repetitive 15 O-labelled gases PET studies.
As shown in Figs. 3 and 5, the central regions in the tumors were deficient in 18 F-FMISO uptake and blood flow, which corresponded to the necrotic regions. These regions were excluded from the VOIs analysis in this study as multiple small sphere VOIs were placed on the tumor with 18 F-FMISO uptake or blood flow uptake. However, since some necrotic regions were mixed with viable hypoxic cells as shown in the pimonidazole staining images, it is not always possible to clearly separate the hypoxic and necrotic regions in the histopathology.
This study demonstrated that tumor cells were resistant to the reduction in blood flow through the increased OEF and it will be an effective treatment strategy to modulate the blood flow, such as therapy using bevacizumab with a monoclonal antibody targeting vascular endothelial growth factor which normalizes vascular architecture, to deliver anti-tumor drugs more effectively and enhance the radiosensitizing effect through oxygenation. Bevacizumab is effective for some cancer patients in combination with chemotherapy. Recently, treatment with bevacizumab has been performed in patients with malignant gliomas, which led to an improved blood flow to the tumor [26,27]. However, it was reported that high doses of bevacizumab might have the risk of increased hypoxia due to the decreased perfusion in the tumor [28]. Conversely, low dose could not yield survival benefit in non-small-cell lung cancer patients treated with 15 or 7.5 mg/kg bevacizumab with chemotherapy [29]. Another study also reported that decreased perfusion also resulted in the reduced drug delivery after administration of bevacizumab and scheduling of anti-angiogenic drugs should be optimized [30]. It was suggested that optimizing the individual dose could improve patient outcome by the pre-treatment evaluation with imaging techniques, such as FMISO and 15 O-labelled gases or water PET, as perfusion and hypoxic conditions are heterogeneous in each cancer patient. In addition, there is no enough evidence about temporal change of perfusion and oxygen metabolism after bevacizumab therapy. The next step of our study is to elucidate how the oxygen metabolism could be changed after the vascular normalization with bevacizumab.
There are some limitations in this study. In smallanimal PET systems, the spatial resolution (full width at half maximum, FWHM) of 15 O is approximately doubled compared to 18 F or 11 C due to the longer positron range [31]. The spatial resolution is about 1.5 mm FWHM in this study, and the partial volume effect (PVE) might be larger on 15 O-labelled gases PET images than 18 F-FMISO PET (9). PVE is a principle problem of Fig. 6 Comparison among pimonidazole staining, 18 F-FMISO PET, TBF, and fusion image. Co-registered 18 F-FMISO and TBF images were resliced to the same section as pimonidazole staining for comparison using the PMOD software PET that has been long discussed. We must pay attention to the underestimation by PVE in the evaluation of intratumoral distribution with PET. However, in our previous study, elevated OEF with decreased cerebral blood flow was detected by 15 O-labelled gases PET in the rat stroke model, suggesting the feasibility to detect the subtle changes in tumor perfusion or oxygen consumption. Next, we fixed the tissue-blood partition coefficient as 0.91 mL/g and the correction value of the hematocrit as 0.70 according to previous studies of the brain [12,13]. If the partition coefficient was substantially different in some area, it has some influence on TBF values. However, these effects were relatively small in low TBF area, such as necrotic area. TBF was increased from 20 to 22 or 25 mL/100 mL/min when the partition coefficient was changed from 0.91 to 0.8 or 0.7, respectively. In addition, previous study reported tissue heterogeneity always resulted in underestimations of mean values of oxygen extraction fraction by the simulation [32]. It may be necessary to develop a new tracer kinetic model considering tumor heterogeneity for more accurate evaluation. Third, this study was conducted on a single tumor model without using anti-tumor drugs. It is the next challenge to investigate other types of tumor model including the relationship between glycolysis and oxidative phosphorylation and to evaluate the changes in perfusion and oxygen metabolism after the anti-tumor therapy or modulation of perfusion and hypoxia.

Conclusions
This study demonstrated the in vivo relationships among 18 F-FMISO SUV, TBF, TMRO 2 , OEF, and TBV values in the rat xenograft model of C6 glioma with an immunohistochemical evaluation. Intratumoral hypoxic regions showed decreased blood flow with increased oxygen extraction compared to the non-hypoxic regions. These findings suggested that a treatment strategy to normalize the blood flow would be a reasonable approach for oxygen-avid active tumor cells in hypoxic regions to achieve more effective drug delivery in chemotherapy and an oxygen sensitizing effect in radiation therapy.

Additional file
Additional file 1: Figure S1. Multiple VOI settings on coronal fusion image of 18 F-FMISO PET (window level 0-3 in SUV) and TBF PET (window level 0-80 ml/100 mL/min). VOIs (1-mm sphere) were manually placed over the entire tumor where there is an uptake of 18 F-FMISO or TBF on the fusion PET images. Figure S2. Relationships between the 18 F-FMISO SUV and quantitative values of TBF, TMRO 2 , OEF, and TBV of each tumor. Decreased trend of TBF and TBV, increased trend of OEF, and stable TMRO 2 against the increase of 18 F-FMISO SUV were observed. (DOC 673 kb)