|Home | About | Journals | Submit | Contact Us | Français|
Hexamethyldisiloxane (HMDSO) has been identified as a sensitive proton NMR indicator of tissue oxygenation (pO2) based on spectroscopic spin-lattice relaxometry. A rapid MRI approach has now been designed, implemented, and tested. The technique, proton imaging of siloxanes to map tissue oxygenation levels (PISTOL), utilizes frequency-selective excitation of the HMDSO resonance and chemical-shift selective suppression of residual water signal to effectively eliminate water and fat signals and pulse-burst saturation recovery 1H echo planar imaging to map T1 of HMDSO and hence pO2. PISTOL was used here to obtain maps of pO2 in rat thigh muscle and Dunning prostate R3327 MAT-Lu tumor-implanted rats. Measurements were repeated to assess baseline stability and response to breathing of hyperoxic gas. Each pO2 map was obtained in 3½ min, facilitating dynamic measurements of response to oxygen intervention. Altering the inhaled gas to oxygen produced a significant increase in mean pO2 from 55 Torr to 238 Torr in thigh muscle and a smaller, but significant, increase in mean pO2 from 17 Torr to 78 Torr in MAT-Lu tumors. Thus, PISTOL enabled mapping of tissue pO2 at multiple locations and dynamic changes in pO2 in response to intervention. This new method offers a potentially valuable new tool to image pO2 in vivo for any healthy or diseased state by 1H MRI.
There is increasing evidence that hypoxia stimulates angiogenesis and metastasis and that hypoxic tumors are more aggressive (1). Furthermore, extensive hypoxia has been associated with poor clinical prognosis for several tumor types, notably cervical and head and neck, based on electrode measurements (2–4). Extensive hypoxia has also been identified in prostate, breast, and brain tumors (5–7). It is expected from definitive observations in cell culture (8) and preclinical investigations in rats and mice (1,9–11) that hypoxic tumors resist radiotherapy. Measurement of tumor hypoxia is becoming increasingly pertinent, as therapy can now be tailored to the characteristics of individual tumors, i.e. personalized medicine. An adjuvant intervention may be applied to patients with hypoxic tumors, e.g. hyperoxic gas breathing to reduce hypoxic fraction. Alternatively, for tumors that resist modulation, a radiation boost may be applied using intensity modulated radiation therapy, or a hypoxic-cell-selective cytotoxin, such as tirapazamine, may be administered.
The ability to measure tissue oxygen tension (pO2) non-invasively may be important in understanding the physiology, pathophysiology, and, potentially, clinical prognosis of diseases such as cancer and stroke. To date, many assays have examined hypoxia, rather than pO2 itself. Radionuclide approaches using fluoromisonidazole and copper diacetyl-bis(N4-methylthiosemicarbazone) can identify hypoxia and have shown predictive value in clinical studies (1). Likewise, immunohistochemistry of biopsy samples after pimonidazole or EF5 (a fluorinated derivative of etanidazole) administration and trapping in tumor tissues have been correlated with outcome (7,12). MRI is particularly suitable for multiple repeat measurements for observing dynamic changes in tissue oxygenation in response to intervention, and blood-oxygen-level-dependent (BOLD) contrast gives an indication of vascular oxygenation, albeit usually qualitative (13–15). 19F NMR can provide quantitative oximetry based on spin-lattice relaxation of perfluorocarbons (16), although it is currently limited to preclinical studies, as reviewed in detail (17,18). The technique has been used to evaluate the ability to manipulate tumor pO2 based on hyperoxic gas breathing. Most significantly, correlations have been shown between pO2 at the time of irradiation and growth delay in Dunning prostate R3327-HI and R3327-AT1 rat tumors (10,11) using hexafluorobenzene (HFB) as a reporter molecule.
Although 19F-MR oximetry is well established and continues to make important contributions to basic research, its clinical translation is hampered by the continuing lack of 19F capability on most clinical MRI scanners. Recently, hexamethyldisiloxane (HMDSO) was identified as a 1H-NMR probe of pO2, and the feasibility of tissue oximetry was presented using 1H-NMR spectroscopic relaxometry of HMDSO, after direct intra-tissue injection (19). With the use of the spectroscopic approach, localization was achieved by virtue of a discrete injection site. The present study demonstrates the implementation of an imaging-based method: proton imaging of siloxanes to map tissue oxygenation levels (PISTOL). As proof of principle, phantom studies are presented and this pO2 reporter molecule is used to investigate dynamic changes in pO2 in rat thigh muscle and syngeneic Dunning prostate R3327-MAT-Lu tumors in response to respiratory challenge with oxygen. A comparative 19F-MR oximetry study was also carried out in rat thigh muscle using HFB.
NMR experiments were performed using a Varian Inova® 4.7 T horizontal-bore system equipped with actively shielded gradients. A chemical-shift selective (CHESS) spin-echo sequence was used to identify the location of HMDSO. A spin-echo echo planar imaging (EPI)-based pulse sequence (Fig. 1) was then used for measuring T1 values for this slice location. The sequence consisted of an initial pulse-burst saturation recovery (PBSR) preparation sequence with 20 non-selective saturation pulses (inter-pulse delay = 50 ms) followed by a variable delay, t, for magnetization recovery. Three CHESS (20) pulses can be included at the end of τ for optional frequency-selective saturation of water and fat. A spin-echo EPI acquisition, consisting of a frequency-selective π/2 pulse (on-resonance for HMDSO), a slice-selective π pulse, and an EPI readout, follows τ. A long echo time (~50 ms) was used, which aided suppression of the fat resonance. This combination of PBSR with frequency-selective excitation EPI (HMDSO) and suppression (water, fat) allowed T1 mapping of HMDSO in 3½ min. For 19F-MR oximetry experiments, FREDOM (fluorocarbon relaxometry using echo planar imaging for dynamic oxygen mapping) was applied using a standard EPI sequence with PBSR (17). The location of HFB was easily determined by using a standard spin-echo sequence, because of the lack of 19F background signal. In both cases (1H and 19F), T1 values were obtained using the corresponding sequence with the ARDVARC (alternating relaxation delays with variable acquisitions for reduction of clearance effects) protocol (21). Varying τ in the range 0.1–55 s, gave a total acquisition time of 3½ min per T1 measurement for PISTOL. T1, R1 (=1/T1) and pO2 maps were computed on a voxel-by-voxel basis using a home-built program written in Matlab (Mathworks Inc., Nattick, MA, USA). For a given voxel, the T1 value was obtained by a three-parameter least-squares curve fit of the signal intensities corresponding to 16 τ values using the Levenberg–Marquardt algorithm. The R1 maps were converted into pO2 maps using previously published calibration curves (19).
A phantom consisting of tubes containing water, mineral oil (to simulate fat), and HMDSO was used to optimize the pulse sequence and test water and fat suppression. To measure the pO2 vs R1 calibration curve by imaging, a second phantom consisting of four gas-tight John Young NMR tubes (Wilmad Labglass, Buena, NJ, USA) containing 1 mL HMDSO each bubbled with different concentrations of O2 (0%, 5%, 10%, and 21% calibrated gases; Airgas Southwest, Dallas, TX, USA) was used. Temperature was kept constant with aD2O-filled circulating water pad and monitored using a fiber-optic temperature probe (FISO Technologies Inc., Quebec City, Quebec, Canada). Mean intensities of each region of interest corresponding to each tube were obtained from the T1 maps and converted into R1 values. Measurements were repeated six times to provide mean R1 values to obtain a calibration curve.
The animal investigations were approved by the Institutional Animal Care and Use Committee. Ten healthy Copenhagen-2331 rats (Harlan, Indianapolis, IN, USA) were used to obtain pO2 data in the thigh muscle (six rats for HMDSO studies and four separate rats for HFB studies). A further six male Copenhagen rats were implanted with Dunning prostate R3327 MAT-Lu tumors subcutaneously on the thigh, to obtain pO2 data in tumors. Tumors were allowed to grow to a range of sizes from 1.2 to 10.6 cm3 (five were > 3cm3). For MRI, rats were maintained under general gaseous anesthesia (air and 1.5% isoflurane; Baxter International Inc, Deerfield, IL, USA). For pO2 measurements in vivo, 50 μL HMDSO (99.7%; Alfa Aesar, Ward Hill, MA, USA) was administered along two or three tracks in the thigh muscle (n = 6) or MAT-Lu tumors (n = 6) in a single plane using a Hamilton syringe with a 32G needle, as described in detail previously for the analogous 19F-NMR approach (17). For comparative 19F-MR pO2 measurements, 50 μL HFB (99.9%; Lancaster Co., Pelham, NH, USA) was administered in the thigh muscle, in a separate cohort of animals (n = 4), as above. The rats were placed in the magnet in the prone position, and body temperature was maintained using a warm water blanket. The thigh or tumor was placed inside a size-matched single-turn 1H/19F tunable volume coil. A cross-section through the thigh or tumor was imaged after HMDSO or HFB was located, as described above. In order to modulate tissue oxygenation, the rats were subjected to respiratory challenge in the sequence, air (20 min) – oxygen (30 min) – air (30 min), and T1 datasets were acquired every 5 min. pO2 values were obtained from the R1 values. Typically, 16 pO2 maps were obtained over a period of 80 min. The statistical significance of changes in pO2 was assessed by using analysis of variance on the basis of Fisher's protected least-squares difference test at 95% confidence level (Statview, SAS Institute, Carey, NC, USA).
Suppression of water and mineral oil signals using the spectrally selective spin-echo EPI sequence (Fig. 1) was successful in a water-filled phantom containing smaller tubes of HMDSO and mineral oil (Fig. 2a,b). T1 measurements from the phantom comprising sealed HMDSO tubes with different oxygen concentrations (at 36.5°C) yielded T1 values essentially identical with those reported previously by spectroscopy (19) (Fig. 2c,d,e). A linear fit to the data yielded a calibration curve R1 = (0.108 ± 0.001) + (0.00130 ± 0.00001) × pO2 at 36.5°C.
HMDSO was readily observed in thigh muscle and tumors by PISTOL with complete suppression of fat and water signals. Discrete distribution of HMDSO was seen in thigh muscle (Fig. 3a,b) and a MAT-Lutumor (Fig. 3f,g) using the CHESS spin-echo sequence. These appear spatially similar to the images of HMDSO and the corresponding pO2 maps acquired using PISTOL (Fig. 3c–e and 3h–j). The imaging data revealed the pO2 distribution, showing the effect of breathing oxygen. Baseline pO2 values were obtained by averaging four baseline pO2 measurements while the rats breathed air. In rat thigh muscle (n = 6), mean baseline pO2 ranged from 27 to 71 Torr (mean = 55 ± 17 Torr), but was stable in any given muscle (mean variation = ±4 Torr over 20 min). On alteration of inhaled gas to oxygen, mean pO2 increased significantly (Fig. 4a) and continued to increase over 20 min. For the group of thigh muscles, mean pO2 was significantly elevated (P < 0.05) compared with baseline by the first measurement (5 min) after the switch of inhaled gas to oxygen and reached values of 163–290 Torr (mean pO2 = 238 ± 59 Torr) after 30 min of breathing oxygen. On return to air breathing, mean pO2 had decreased significantly by the first measurement (5 min) and had returned to a value not significantly different from baseline by the second measurement (10 min). Measurements of pO2 in thigh muscle using HFB yielded similar results (Fig. 4b). For this group (n = 4), mean baseline pO2 ranged from 23 to 51 Torr (mean = 35 ± 11 Torr), and was stable in any given muscle (mean variation = 3 Torr over 20 min). In response to oxygen breathing, mean pO2 was significantly elevated (P < 0.05) compared with baseline by the first measurement (5 min) after the switch of inhaled gas to oxygen and had reached values of 152–305 Torr (mean pO2 = 211 ± 79 Torr) after 30 min of breathing oxygen. On return to air breathing, mean pO2 had decreased significantly by the third measurement (15 min) and had returned to a value not significantly different from baseline by the fourth measurement (20 min).
In rat prostate MAT-Lu tumors (n = 6), mean baseline pO2 ranged from −0.5 to 41 Torr (mean = 17 ± 16 Torr), but was stable in each tumor (mean variation = ±2 Torr over 20 min). Tumors 1–5 showed a mean baseline pO2 = 12 ± 12 Torr with a mean variation of ±2 Torr. For these five tumors, mean pO2 was significantly elevated (P < 0.05) compared with baseline by the fifth measurement (25 min) after the switch of inhaled gas to oxygen (Fig. 5). When oxygen was breathed for 30 min, mean pO2 reached 30 ± 27 Torr (range 7–69). On switching back to air from oxygen, mean pO2 had returned to a value not significantly different from baseline by the second measurement (10 min). Tumor 6 exhibited particularly high pO2 response to breathing oxygen (mean pO2 = 323 ± 44 Torr after 30 min of oxygen breathing), which was very different from the other tumors. Comparing the 1H anatomical (H2O) and selective HMDSO images showed that the HMDSO was deposited in the tumor periphery and was probably not representative of the tumor bulk. Hence, it has been excluded from Fig. 5. Figure 6 shows histograms of pO2 distributions for pooled voxels from rat thigh muscle (Fig. 6a) and MAT-Lu tumors (Fig. 6b) and changes in the distribution after 30 min of oxygen breathing. Figure 7 shows the changes in representative voxels (five each) from a representative muscle (Fig. 7a) and tumor (Fig. 7b, same tumor as shown in Fig. 3). Data for individual tumors are summarized in Table 1.
HMDSO has previously been shown to be a promising 1H-MR-based pO2 reporter molecule for spectroscopic in vivo studies based on the linear dependence of its spin-lattice relaxation rate R1 on pO2 at a given temperature in the range 26–46°C (19). The present study demonstrates the feasibility of imaging dynamic changes in tissue oxygenation using frequency-selective EPI of HMDSO, after direct injection into tissue.
In seeking a proton NMR pO2 reporter molecule, HMDSO was selected for its similarity to traditional 19F-NMR perfluorocarbon agents. HMDSO is hydrophobic and is essentially immiscible with aqueous solutions. Thus, gas exchange with the surrounding tissue occurs without exchange of ions, which might influence the spin lattice relaxation; thus, the validity of in vitro calibrations is maintained for in vivo determinations. HMDSO is readily available, inexpensive, and easy to store. Although HMDSO was used here, other symmetric, hydrophobic siloxanes may also be effective pO2 reporter molecules, and this new concept is open to development and optimization. In the perfluorocarbon field, many different molecules have been exploited for in vivo oximetry over the years (22,23). Initially, perfluorocarbon blood substitute emulsions were favored for their biocompatibility, but multi-resonance molecules, such as perfluorotributylamine and perflubron (perfluoro-octylbromide) were suboptimal for oximetry, because of relatively low R1 dependence on pO2 and high dependence on temperature (22). Moreover, the multi-resonance spectra caused considerable signal loss for imaging (24). Ultimately, HFB and perfluoro-15-crown-5 ether were identified as superior because of their single resonances (18,25,26), and HFB is particularly attractive because of its low temperature dependence and ready commercial availability. Just as the perfluorocarbon oximetry field has evolved, other siloxanes may be identified or developed that may be superior to HMDSO.
Mapping pO2 was demonstrated using frequency-selective excitation of the HMDSO resonance with efficient frequency-selective fat and water suppression in vitro and in vivo. The calibration curve obtained here with imaging compares well with the calibration curve obtained previously by spectroscopy (19). Clearance of HMDSO from tissues (half-life ~35 h) is relatively slow compared with clearance of HFB used in the analogous 19F-MR oximetry (FREDOM) (17,27), so that there is minimal clearance during a typical investigation of oxygen dynamics in response to acute interventions (19). Thus, although FREDOM provides sensitive assessment of acute changes in response to respiratory challenge or vascular targeting agents, the slower clearance of HMDSO may facilitate studies of chronic changes in tumor oxygenation accompanying tumor growth or long-term chemotherapy. Application of HMDSO to breast studies could be difficult in the presence of silicone implants because of the similar chemical shifts.
In rat thigh muscle, the range of baseline pO2 values measured by PISTOL and pO2 response to oxygen challenge were similar to those measured here by the well-established 19F-MR oximetry technique and those reported previously using 1H spectroscopy of HMDSO (19) or 19F MRI of HFB (28), needle electrodes, fiber optical probes (29,30), or electron paramagnetic resonance (31,32). The response of individual voxels and muscles in separate rats to oxygen breathing depends on voxel location (Figs. 3, ,4,4, ,6,6, and and7),7), but the baseline values and response are generally higher than seen in tumors. As also reported by Yeh et al. (33), based on measurements using oxygen electrodes, the pO2 of Dunning prostate tumors tends to be lower than in skeletal muscle. The baseline pO2 values and pO2 response to oxygen challenge are quite similar to those reported previously for large MAT-Lu tumors using 19F-MR oximetry (34). Different rat prostate and breast tumor types are reported to exhibit a range of baseline oxygenations, and the response to interventions is highly variable. In some cases, hypoxic fractions resist modulation with hyperoxic gas breathing [e.g. Dunning prostate R3327-AT1 tumors (10,35)]. In other tumors, notably with a well-developed and highly perfused vasculature, hyperoxic gas essentially eliminated hypoxia [large Dunning prostate R3327-HI tumors (11)].
Direct intratumoral injection of the reporter molecule has benefits and drawbacks. It allows immediate measurement of any region of interest after minimally invasive administration. By contrast, reporter molecules administered systemically initially report vascular oxygenation (36) and even after clearance tend to sequester in well-perfused regions biasing measurements (37). Of course, direct injection does limit measurements to accessible tissues. It would be preferable to exploit endogenous molecules as used in BOLD contrast, but this reveals vascular oxygenation and is subject to variations in vascular volume, hematocrit, and flow (13,14). Direct measurements of tissue water T1 are attractive (38,39), but they may be influenced by many factors in addition to pO2. Like perfluorocarbons, HMDSO is lipophilic and is essentially immiscible in aqueous solutions. It could be emulsified for systemic delivery to help circumvent the need for direct intra-tissue injections, and such an attempt is currently underway. Some silanes are highly reactive, whereas HMDSO is quite inert, and is reported to exhibit minimal toxicity (40,41). In a 13-week subchronic inhalation toxicity study in Fischer 344 rats exposed to 5000 ppm of HMDSO, no treatment-related signs of toxicity or mortality, or other significant histological changes were found (40). After oral (300 mg/kg) or intravenous administration (80 mg/kg, as emulsion) of HMDSO, various polar metabolites were found in the urine as a result of the oxidation of the Si–CH3 bond (42). Another study reported no irritation in Draize tests of skin or eye irritancy and no acute toxicity in rats (LD50 > 3.8 g/kg) (43). In our studies, we saw no overt signs of inflammation or discomfort, although no microscopic analyses were performed. For future routine use as an intra-tissue-injected pO2 reporter molecule, further investigation of possible local inflammatory response after direct injection of siloxanes is warranted.
In summary, PISTOL is a sensitive, quantitative 1H-MR method for imaging oxygen tension and dynamic changes in response to interventions. This new technique opens up further opportunities to evaluate pO2 in vivo. Rapid translation of this method to the clinical setting is feasible with current state-of-the-art MR hardware, as clinical instruments can routinely generate effective water and fat suppression as used in detection of metabolites such as choline, lactate, and citrate. A further advantage of PISTOL is that it will now be possible to add quantitative oximetry to a protocol consisting of other 1H-MR-based functional techniques routinely used for research as well as clinical diagnosis, such as dynamic contrast enhancement, diffusion measurements, and MRS.
We are grateful to Jennifer McAnally for assistance in tumor implantation and to Drs Dawen Zhao, Weina Cui, and Mark Conradi for constructive suggestions. We appreciate the help of Glenn Katz in preparing the figures. We also thank Dr Maj Hedehus (Varian Inc., Palo Alto, CA, USA) for helpful tips in pulse sequence programming. J.P.T. gratefully acknowledges a fellowship from the Communidad de Madrid. This work was supported in part by the DOD Prostate Cancer Initiative (W81XWH-06-1-0149). The MR investigations were performed at the Advanced Imaging Research Center, an NIH BRTP facility (P41RR02584) in conjunction with the Small Animal Imaging Research Program (NCI U24 CA126608).
Contract/grant sponsor: DAMD; contract/grant number: W81XWH-06-1-0149.
Contract/grant sponsor: NIH; contract/grant number: P41RR02584.
Contract/grant sponsor: NIH; contract/grant number: U24 CA126608.
†Presented in part at the 14th Annual Meeting of the International Society of Magnetic Resonance in Medicine, Seattle, 2006