In the present work, a novel method for detecting hypoxia in tumors, phosphorescence quenching, was used to evaluate tissue and tumor oxygenation. This technique is based on the concept that phosphorescence lifetime and intensity are inversely proportional to the oxygen concentration in the tissue sample. We used the phosphor Oxyphor G2 to evaluate the oxygen profiles in three murine tumor models: K1735 malignant melanoma, RENCA renal cell carcinoma, and Lewis lung carcinoma. Oxygen measurements were obtained both as histograms of oxygen distribution within the sample and as an average oxygen pressure within the tissue sampled; the latter allowing real-time oxygen monitoring. Each of the tumor types examined had a characteristic and consistent oxygen profile. K1735 tumors were all well oxygenated, with a peak oxygen pressure of 37.8 ± 5.1 Torr; RENCA tumors had intermediate oxygen pressures, with a peak oxygen pressure of 24.8 ± 17.9 Torr; and LLC tumors were all severely hypoxic, with a peak oxygen pressure of 1.8 ± 1.1 Torr. These results correlated well with measurements of tumor cell oxygenation measured by nitroimidazole (EF5) binding and were consistent with assessments of tumor blood flow by contrast enhanced ultrasound and tumor histology. The results show that phosphorescence quenching is a reliable, reproducible, and noninvasive method capable of providing real-time determination of oxygen concentrations within tumors.
- oxygen measurement
- Oxyphor G2
- tissue oxygenation
the ability of a solid tumor to grow beyond ∼1 mm in diameter is dependent on the formation of a new blood supply, due to limits of diffusion of oxygen, nutrients, and growth factors (13). Tumor vessels have abnormal morphology and function. Often, distinct arteries and veins are not identifiable within tumors, and the vessels may be tortuous, disorganized, and dilated and may exhibit arteriovenous shunting (5). Tumor vessel walls often lack smooth muscle or pericyte coverage and are often highly permeable and leaky, leading to an increase in tumor interstitial fluid pressure (8, 18). The abnormal vessel morphology leads to sluggish blood flow and inadequate perfusion to the tumor tissue, and many tumors, including carcinomas of the uterine cervix (20, 32), squamous cell carcinomas of the head and neck (3), prostatic carcinomas (30), and glioblastoma multiforme (11), have been shown to develop areas of severe hypoxia.
The presence of hypoxia within tumors is correlated with a poor prognosis. Studies that used polarographic needle electrodes have demonstrated that hypoxic carcinomas of the head and neck (3, 33) and of the uterine cervix (15, 19, 20, 26) are more likely to have poor therapeutic outcomes than are well-oxygenated tumors after treatment with surgery or radiation therapy. Studies of soft tissue sarcomas have shown an increased likelihood for distant metastasis in hypoxic tumors (2, 20), suggesting that hypoxia confers a more aggressive phenotype.
The relationship between poor prognosis and tumor hypoxia has led to many attempts to develop assays for determining the oxygenation status of tumors, but few clinically useful methods for accurately measuring oxygenation exist. Two such methods are the Eppendorf needle electrode and measurement of nitroimidazole binding. The Eppendorf needle electrode has been used in numerous studies linking oxygen pressure and prognosis in tumors (19, 20). Electrodes for tissue oxygen measurements are invasive and measure only over a small area at the tip of the needle electrode at one point at a time, making it a challenge to evaluate tissue heterogeneity. A second method for measuring hypoxia in tumors is the use of nitroimidazole drugs [for review, see Bussink et al. (4)]. In the presence of hypoxia, nitroimidazoles become reduced by nitroreductases and the products trapped intracellularly by covalent binding to intracellular proteins (9, 10, 24, 41). Drug binding can then be detected by use of fluorescently labeled specific monoclonal antibodies (29, 41) or, in some cases, can be directly radiolabeled for noninvasive detection with positron emission tomography or single photon-emission computed tomography imaging (27, 39, 47). One limitation of nitroimidazole measurements is that data are obtained only after several hours of metabolism of the drug. As a result, the measurements obtained provide a time-averaged measurement of tissue oxygenation over the period of metabolism.
Here, we describe the use of a noninvasive, real-time optical technique for measuring oxygen pressures within the microvasculature of murine tumors based on oxygen-dependent quenching (40) of the phosphorescence of Oxyphor G2 (7, 36). The rate of decay of the phosphorescence signal following a short pulse of excitation light, the phosphorescence lifetime, depends on the concentration of the quenching molecules in the solution. In biological samples, oxygen is the primary quencher, and the measured phosphorescence lifetime may be converted to oxygen pressure by the Stern-Volmer relationship (7). This approach has several advantages over oxygen measurement by oxygen electrodes or by nitroimidazole binding in that it allows real-time measurement of oxygen levels, including measurements of the distribution of oxygen pressures within a given tissue volume.
Oxyphor G2 is a 2,400-Da hydrophilic molecule, with 16 negative charges on its surface; for this reason, the phosphor remains intravascular after administration (7, 36). Because Oxyphor G2 is an intravascular molecule, oxygen measurements obtained from phosphorescence quenching are of the oxygen pressure within the mixed venous and capillary blood supplies. This oxygen measurement differs from those of oxygen electrodes, which measure a mixed vascular-extracellular space-intracellular oxygen pressure, and from those by nitroimidazole binding, which measures the intracellular oxygen. In the present paper, we describe the oxygen distributions in three different murine tumors, comparing the values measured by phosphorescence quenching with those by EF5 binding and with vascular morphology and blood-flow patterns.
MATERIALS AND METHODS
BALB/c, C3H/HeN, and C57Bl/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME) or from Harlan Sprague Dawley (Indianapolis, IN). All animal experiments were approved by the University of Pennsylvania Animal Care and Use Committee. All mice were females, 6–8 wk of age, and were maintained in microisolator cages under sterile conditions. The K1735 (28) murine melanoma, RENCA (31) renal cell adenocarcinoma, and Lewis lung carcinoma (LLC) cell lines (syngeneic with C3H/HeN, BALB/c, and C57/Bl6 mice, respectively) were cultured in DMEM supplemented with 10% FCS and 1% penicillin-streptomycin at 37°C and 5% CO2. For the creation of tumors, 2 × 106 (K1735, LLC) or 4 × 106 (RENCA) tumor cells were injected subcutaneously into the right flank of anesthetized (140 mg/kg ketamine and 1.3 mg/kg xylazine administered intraperitoneally) syngeneic mice. When established tumors reached a diameter of 10–12 mm, generally 2–3 wk after cell inoculation, oxygen measurements were made.
Oxyphor G2, a two-layer glutamate dendrimer of Pd-tetra-(4-carboxyphenyl) benzoporphyrin (7, 36), was used for phosphorescence quenching measurements. The chemical structure is shown in Fig. 1. Oxyphor G2 exhibits two absorption maxima in the visible and near-infrared region, 450 nm and 636 nm (blue and red), with an extinction coefficient of ∼50 mM−1·cm−1 for the maximum at 636 nm. The phosphorescence emission is maximal at ∼800 nm, and the quantum efficiency of phosphorescence in the absence of oxygen is ∼15%. The phosphor has a lifetime at zero oxygen of 255 μs and a quenching constant of 280 Torr oxygen−1·s−1 when dissolved in blood at 37°C.
Phosphorometer and oxygen measurements.
Phosphorescence lifetime measurements were performed with the PMOD 5000 phosphorometer (Oxygen Enterprises, Philadelphia, PA) as described by Vinogradov et al. (43) and proprietary software. The PMOD 5000 is a frequency domain instrument that measures phosphorescence lifetimes rather than intensity because lifetime measurements are insensitive to the presence of tissue chromophores and fluorophores (42, 46). The excitation light is carried to the animal through one 3-mm-diameter glass fiber bundle, and the emission is collected through another 3-mm-diameter glass fiber bundle. The emitted light is filtered through a 695-nm long-pass glass filter (Schott Glass, Elmsford, NY) and detected by a 3-mm-diameter avalanche photodiode (Hamamatsu, Bridgewater, NJ). The resulting photocurrent is converted into a voltage, amplified, digitized, and directly transferred to the computer memory (PCI bus).
To obtain mean oxygen measurements in tissue, measurements were made with the PMOD 5000 in the two frequency modes in which the excitation light is modulated at two frequencies at the same time. Details of this technique have been previously described (42). The phosphorescence lifetimes are calculated with the assumption of a single exponential decay (a single phosphorescence lifetime) and converted to the oxygen pressure by the Stern-Volmer equation. The method for obtaining the distribution of phosphorescence lifetimes, and therefore the distribution of oxygen, has been described by Vinogradov et al. (42). The distribution of phosphorescence lifetimes is used to calculate the oxygen pressure histograms. The result is a table of oxygen pressure values and the amount of phosphorescence (in arbitrary units) at each oxygen pressure. These data are used to construct histograms of the fraction of the phosphorescence arising from phosphor in an environment with each oxygen pressure from −5 to 100 Torr. The values for the amount of phosphor at each oxygen pressure are normalized such that the sum of all values (integral) for each of the histograms had the same value. This normalizing procedure removes the dependence on excitation light intensity and phosphorescence collection efficiency, providing a value that is proportional to the fraction of the blood volume in the sampled tissue (volume fraction) at that oxygen pressure.
Oxygen measurements in vivo.
Tumor-bearing mice were anesthetized with 1.25% isoflurane in air and injected with 0.1 ml of Oxyphor G2 (2.5 mg/ml) via a tail vein. The day before phosphorescence measurements, the hair overlying the tumor and the hair over the flank muscle on the opposite leg were removed with a depilatory salve. Two light guides, separated by a center-to-center distance of 8 mm, were placed on both tumor tissue and on normal muscle. Mean oxygen measurements were made in normal muscle tissue every 10 s until the oxygen pressure reached a normal level, ∼10 min. The oxygen measurements made at 10-s intervals provide an average oxygen pressure over the sampled tissue volume. Oxygen histograms were then obtained from each of four quadrants of the tumor and then at two different locations in normal muscle of the nontumor-bearing hind leg. These sets of measurements were repeated three times. Data are presented as the distribution of oxygen (oxygen histogram) in the sampled tissue volume.
Tumor histology and immunohistochemistry.
EF5 (provided by Dr. Cameron Koch, University of Pennsylvania, Philadelphia, PA) staining was performed to determine tumor cell hypoxia. Immunofluorescent analysis of EF5 in tumor sections were performed as previously described (9), using a cyanine 3-conjugated anti-EF5 monoclonal antibody (provided by Cameron Koch). To determine the relationship between blood vessel morphology and tumor oxygenation, frozen thin sections (10 μm) were stained for endothelial cells by a rat anti-mouse CD31 monoclonal antibody (PharMingen, San Diego, CA) as described previously (17). Portions of the tumors were fixed in 4% paraformaldehyde overnight and subsequently paraffin embedded and sectioned for hematoxylin and eosin staining. All histological specimens were viewed under a Nikon E600 Eclipse (Nikon, Melville, NY) equipped with a krypton-argon laser and optical filters for visualization of FITC, Texas red, and cyanine 3 fluorescence. Images were acquired by a Photometrics Coolsnap CF charge-coupled device camera (Roper Scientific, Trenton, NJ) and image acquisition software (IP Lab, Scanalytics, Fairfax, VA).
Contrast-enhanced ultrasound imaging.
To determine the relationship between oxygen measurements determined by phosphorescence quenching and tumor blood flow, representative tumors in each group were subjected to contrast-enhanced ultrasound imaging. Mice were anesthetized with 1.25% isoflurane in air and the hair overlying the tumor was removed with a depilatory salve. The imaging transducer (7–15 MHz transducer, ATL 5000 from Philips Ultrasound, Bothell, WA) was aligned with the long axis of the tumor. A 5-mm acoustic standoff between the transducer face and tumor was achieved by generous application of acoustic gel. Initial scanning of each tumor was performed in B-mode (grayscale ultrasound) to define the boundary of the tumor mass. A rectangular area was then placed around the tumor and surrounding tissue, denoting the region in which power Doppler data would be acquired. One hundred microliters of microbubble ultrasound contrast agent (Optison, Amersham, Princeton, NJ) were then injected via tail vein catheter, and Power Doppler images were obtained with the scanner gated externally at 20-s intervals in a sequence of 0.5, 2, and 4 Hz. The area of contrast enhancement denotes perfused regions in the tumor. Images were recorded on videotape (S-VHS format) and digitized frame by frame at 24-bit resolution by a Macintosh AV-7600 frame grabber.
Oxygen measurements in normal muscle.
Immediately after the induction of anesthesia, real-time mean oxygen measurements were begun on quadriceps muscle of the nontumor-bearing hindlimb. These measurements were obtained to ensure that fluctuations in tissue oxygenation resulting from the induction of anesthesia had resolved before we obtained histogram measurements. Measurements at 10-s intervals were continued until the oxygen measurements were within a narrow (within ∼5 Torr) range for at least 2–3 min. Many of the mice had an initial decrease in tissue oxygen levels immediately after induction of anesthesia, but these increased again over several minutes, and tissue measurements typically achieved a steady state within 10–20 min. In the majority of mice, oxygen levels in the microvasculature of normal tissue stabilized between 35 and 45 Torr. The average final oxygen pressure was 36.5 (SD 10) Torr. The variability seen in the final stabilized values among mice is most likely due to differences in body temperature and respiratory rate. Representative tracings acquired during the initial 20 min of anesthesia in four mice are shown in Fig. 2A.
Histogram measurements of normal muscle tissue were obtained in triplicate in all mice (n = 19) from both the distal and the proximal parts of the quadriceps muscle of the nontumor-bearing hind leg. Figure 2B shows a representative histogram for muscle, which provides the distribution of oxygen within the volume of tissue examined. Each curve on the histogram represents one measurement from the same mouse. The nearly identical readings obtained for each of the measurements demonstrate the reproducibility of the technique. In the majority of mice, the peak of the histogram (or the oxygen pressure most frequently represented in the sampled area) occurred between 30 and 45 Torr. The average value among all mice was 42 (SD 12) Torr.
K1735 tumors were grown in C3H/HeN mice for 2–3 wk until the tumors reached a size of 10–12 mm diameter, at which time phosphorescence quenching measurements were performed (n = 6 mice). Representative histogram measurements for K1735 tumors are shown in Fig. 3A. Each curve on the histogram represents one measurement within the representative tumor. The data show that K1735 is very homogenous regarding its intravascular oxygenation, as there is almost no variability among different areas within a tumor. Because the tumors were homogenous in their oxygenation, individual histograms from each tumor were averaged to obtain one representative histogram for each mouse to determine average peak tumor oxygenation (peak of the histogram) and the percentage of each tumor that had oxygen pressures below certain thresholds.
All K1735 tumors evaluated had histograms that were nearly identical to those obtained from normal muscle, with an oxygen peak between 32 and 46 Torr. The average peak oxygen pressure of all K1735 tumors was 37.8 (SD 5.1) Torr (Table 1). The small standard deviation relative to the mean oxygenation is indicative of the consistent measurements obtained in K1735 tumors. In the K1735 tumors, only 6.1 (SD 2.3)% of the tumor volume had oxygen pressures <15 Torr, whereas 24 (SD 6.3)% had oxygen pressures <30 Torr (Table 1). To determine the relationship between blood perfusion and tumor oxygenation, contrast-enhanced Doppler ultrasound imaging was performed on K1735 tumors (Fig. 3B). All of the tumors examined were uniformly perfused with few or no areas that were devoid of blood flow.
To determine the relationship between intravascular oxygenation as measured by phosphorescence quenching and intracellular oxygen status, EF5 assessments of oxygen were performed on two of the tumors (one representative section shown in Fig. 4A). Previous studies in our laboratory demonstrated that K1735 malignant melanoma tumors grown in mice do not exhibit hypoxia or necrosis (16). EF5 staining on the tumors in these studies was consistent with the previous results. Only small portions of the K1735 tumors had any detectable EF5 binding (shown in red), and these regions occurred at some distance from the blood vessels. Staining with CD31 [platelet endothelial cell adhesion molecule (PECAM), shown in green] demonstrates that the K1735 vasculature is relatively well organized, with very few regions that are devoid of vessels. Hematoxylin and eosin staining showed no evidence of necrosis (Fig. 4B). These data demonstrate that, in this tumor model, there is good correlation between intravascular oxygen status and intracellular oxygenation. The lack of necrosis suggests the supply of nutrients and oxygen is adequate to meet the metabolic requirements of the tumor cells.
Lewis lung carcinoma.
Phosphorescence quenching measurements were performed on LLC tumors (n = 7 mice) of similar size. A representative histogram is shown in Fig. 3C. The LLC tumor histograms are nearly identical in all tumor regions examined; therefore, individual histograms were averaged to produce one representative histogram for each tumor for calculation of oxygen pressures. The oxygen profile obtained with LLC tumors is very different from those obtained for K1735. The average peak oxygen pressure was 1.8 (SD 1.1) Torr. The range was from 1 to 4 Torr, indicating that all of the LLC tumors evaluated had substantial volumes of very hypoxic tissue. An average of 44.7 (SD 5)% of the volume of LLC tumors had oxygen measurements of <30 Torr, and 29.6 (SD 4.8)% had oxygen measurements of <15 Torr (Table 1). The small standard deviation is indicative of consistently low oxygen pressures within LLC tumors. The tumors were sufficiently small so that phosphorescence measurements probably included some of the surrounding normal tissue. Thus the fractions of the tumors with low oxygen pressures may be underestimated. Contrast-enhanced ultrasound imaging was performed in several LLC tumors, which showed large gaps in tumor perfusion (Fig. 3D).
Oxygen measurements using EF5 were performed on LLC tumors and demonstrated that there is a large amount of tumor tissue hypoxia present compared with the K1735 tumors (Fig. 4C). PECAM staining shows a large number of small, tortuous vessels, with hypoxia both at some distance from the vessels and immediately adjacent to them. Hypoxia present immediately adjacent to vessels suggests defects in vessel perfusion. Hematoxylin and eosin staining of LLC tumors demonstrates areas of tumor tissue necrosis, consistent with regions of severe chronic hypoxia (Fig. 4D).
RENCA tumors (n = 6) were subjected to phosphorescence quenching measurements at a similar size as the K1735 and LLC tumors. Figure 3E shows a representative histogram of a RENCA tumor. Measurements obtained in four different regions of a tumor varied widely, unlike measurements in K1735 or LLC tumors. Some regions of RENCA tumors were severely hypoxic, whereas other regions appeared well oxygenated. The oxygen measurements in different regions of these tumors were averaged to establish an “average” oxygen concentration distribution across the entire tumor. The average peak oxygen pressure was 24.8 (SD 17.9) Torr. The large standard deviation relative to the peak oxygen measurements is consistent with the large amount of variability within RENCA tumors. An average of 40.0 (SD 12.1)% of the volume of LLC tumors had oxygen measurements of <30 Torr, and 21.8 (SD 10.2)% had oxygen measurements of <15 Torr (Table 1). Ultrasound examination demonstrated a similar pattern of heterogeneity in blood flow (Fig. 3F).
Oxygen measurements with EF5 in the RENCA tumors were consistent with the oxygen histograms. There are areas of the RENCA tumor that are well vascularized, with no evidence of tissue hypoxia (Fig. 4E, left), whereas other portions of the tumor have large regions of hypoxia at some distance from the blood vessels (Fig. 4E, right). Hematoxylin and eosin staining also demonstrate regions of necrosis adjacent to regions of healthy tumor cells (Fig. 4F).
Comparison between tumor types and between tumors and normal muscle.
There was no significant difference between the peak oxygen pressure of K1735 tumors and that of normal muscle (P = 0.196), whereas the peak oxygen pressures of RENCA and LLC tumors were significantly lower (P = 0.006 and P < 0.0001, respectively) (Table 1). This confirms that K1735 tumors are well oxygenated, whereas LLC and RENCA tumors are much more hypoxic than normal tissue. K1735 tumors have significantly higher peak oxygen pressure (P < 0.0001) and a significantly lower percent of the tumor volume at <30 Torr (P < 0.0001) and <15 Torr (P < 0.0001) compared with LLC tumors. There was no significant difference between the peak oxygen concentration of K1735 tumors compared with that of RENCA tumors (P = 0.0672). However, K1735 tumors had a significantly lower percentage of the tumor volume at <30 Torr oxygen pressure (P = 0.010) and <15 Torr (P = 0.0031), indicating that they were significantly less hypoxic than RENCA tumors.
LLC tumors had a significantly lower peak oxygen pressure compared with RENCA tumors (P = 0.0058). There was no significant difference in the percent of the tumor volume below 30 and 15 Torr (P = 0.37 and 0.1, respectively), suggesting that, although the RENCA tumors have significantly higher peak oxygen pressures compared with LLC tumors, both tumor types have large regions that are significantly hypoxic.
The ideal oxygen measurement technique would be 1) comprehensive, in that it would sample essentially all of the tissue volume; 2) have good temporal resolution; 3) be capable of continuous measurement or repetitive measurements at frequent intervals; and 4) measure the distribution of oxygen within the tissue. Oxygen-dependent quenching of phosphorescence, the method we have described here, appears to meet many of these requirements. The signal increases dramatically as the oxygen pressure decreases. This results in increasing sensitivity with decreasing oxygen pressure and makes it a very sensitive to regions of relative hypoxia. Because the phosphor both absorbs and emits near-infrared light, measurements can be made through several centimeters of tissue, permitting evaluation of oxygenation in large volumes of tissue. Phosphorescence quenching is capable of real-time measurements and thus can measure transients in tissue oxygenation and response to treatment. With appropriate selection of the sites for introducing the excitation light and collecting the phosphorescence, comprehensive measurements can be made of the fraction of the tissue volume at each oxygen pressure. Oxygen distributions (histograms) can be obtained with each measurement, accurately indicating the presence and relative volume of any regions of hypoxia, even when these are only a small fraction of the total volume. Calibration is absolute, eliminating measurement drift and/or the necessity for repetitive calibrations. One limitation of phosphorescence quenching, as with any optical technique, is that light is strongly absorbed by melanin. We have attempted phosphorescence quenching experiments with highly melanotic B16 malignant melanoma tumors and found that phosphorescence emission was undetectable (data not shown).
The data presented here represent the first use of phosphorescence quenching for comparing the oxygenation of a variety of murine tumors. Previous work has demonstrated the feasibility of this method for evaluating intravascular oxygenation in one rodent tumor model (46). The data presented here show that each murine tumor type studied has a characteristic oxygen profile. The K1735 tumor has a well-oxygenated phenotype, with an oxygen histogram that is nearly identical to that of normal tissue, whereas the LLC tumor had severely hypoxic readings throughout the tumor volume. Finally, the RENCA tumors tended to be heterogeneous in their oxygenation, with some severely hypoxic, some moderately hypoxic, and some well-oxygenated regions. Therefore, these tumor types provide models for the behavior of tumors with very different patterns and levels of oxygenation.
In the present study, we have used a new phosphor synthesized by Vinogradov and coworkers (7, 36), Oxyphor G2. The large size of this molecule (2,400 Da) and its strong negative charge (−16) restrict it to the vascular space. Therefore, oxygen measurements obtained are indicative of the oxygen pressure within capillaries and small venules, i.e., intravascular oxygenation. Nitroimidazole binding (such as that obtained with EF5, pimonidazole, or misonidazole) and phosphorescence quenching are complementary techniques, as the former measures intracellular oxygenation. In the present work, we examined tumor oxygenation with both phosphorescence quenching and a nitroimidazole (EF5) to determine whether this correlation is observed. Data for the three tumor types studied show a good correlation between oxygenation measured by nitroimidazole binding and by phosphorescence quenching. It is possible for these two techniques to provide discordant results, as intracellular hypoxia could exist without intravascular hypoxia. This could occur if the tumor vessels were formed and functioning well, but their numbers were not adequate to supply the entire tumor volume. In rapidly growing tumors, inadequate numbers of vessels could leave large areas of tumor undersupplied with oxygen, leading to diffusion-limited hypoxia and tissue hypoxia in the absence of intravascular hypoxia. A similar situation could develop in tumors treated with antiangiogenic agents, where new vessels are prevented from developing. Angiogenesis inhibition is expected to result in the development of tumor cell hypoxia (16). However, it may also result in vascular “normalization” (22) due to residual tumor blood vessels that are more organized and less permeable than those found in untreated tumors, resulting in improved blood flow. In these situations, nitroimidazole binding would show tumor cellular hypoxia, whereas phosphorescence quenching would not show evidence of intravascular hypoxia. The converse situation, of poor oxygenation within tumor blood vessels without resulting cellular hypoxia, is unlikely to occur. If abnormalities in tissue oxygenation are caused by abnormalities in blood flow or vascular anomalies, as appears to be the case for many tumors, there should be a strong correlation between oxygen levels in tumor vessels and tumor cells.
One surprising finding in this study was the highly variable decrease in the oxygen pressures in normal tissues upon induction of anesthesia and in the recovery during the first 10–20 min of maintenance of anesthesia. The reason for the immediate drop in tissue oxygenation upon induction of anesthesia is not clear but may be related to a transient decrease in blood pressure or respiration or may be because of vasoconstriction secondary to a drop in temperature. Anesthesia-induced apnea certainly can contribute to the rapid changes in oxygenation as there is an observable correlation of intermittent breathing with sharp changes in tissue oxygenation. The physiological basis for the slower recoveries (Fig. 2A), occurring over periods of 10–20 min, remain to be established. Detection of these minute-to-minute changes in oxygenation is one of the benefits of real-time monitoring of tissue oxygenation obtainable with phosphorescence quenching.
Numerous techniques have been developed to try to determine the amount of blood flow present within tumors. These techniques include dynamic contrast-enhanced MRI (1, 34), contrast-enhanced computed tomography (14), contrast-enhanced ultrasound (12), and laser-Doppler flowmetry (38). In the work reported herein, we performed contrast-enhanced ultrasound to demonstrate perfused regions in tissues and to compare tumor perfusion with oxygenation measurements. Ultrasound imaging has advantages over other imaging techniques in that it is rapid, inexpensive, and has good depth of penetration; in addition, the use of microbubble contrast agent with Power Doppler ultrasound greatly improves visualization of capillary flow in tumors (12, 23). The ultrasound data acquired showed good qualitative correlation with oxygen measurements by phosphorescence quenching. The K1735 tumors were all uniformly perfused with no evidence of avascular regions, suggesting that blood vessel growth kept up with tumor growth. This is consistent with the fact that neither phosphorescence quenching nor EF5 showed evidence of hypoxia in these tumors. LLC tumors showed large areas without significant flow, suggesting that vessel growth did not keep up with tumor growth, that tumor blood flow was too slow to be detected, or that the tumor vessels were too small to be seen. These findings are consistent with the marked hypoxia detected with phosphorescence quenching. RENCA tumors showed marked heterogeneity by both ultrasound examination and phosphorescence quenching, as well as by EF5 staining.
Although this paper represents the first extensive use of phosphorescence quenching for oxygen measurements in murine tumors, the technique has many other potential uses. Previous work has demonstrated its applicability in measurements of cerebral, intestinal, and hepatic oxygenation during cardiopulmonary bypass in piglets (6), in imaging ischemic regions in the heart after myocardial infarctions, also in a pig model (37), in evaluating cerebral cortex oxygenation after ischemic/hypoxic injury in a feline model (44), and in evaluating oxygen delivery to muscle in a rat model (35). The technique has also been shown to be very useful for in vitro experiments, in particular for evaluating the intracellular oxygen pressures and oxygen uptake in isolated single myocytes (21, 25) and for examining the oxygen dependence of mitochondrial oxidative phosphorylation (45). This previous work points to the broad utility of this method of oxygen measurement.
The work reported herein demonstrates the ability of phosphorescence quenching to identify hypoxic regions within murine tumors. Potential clinical applications for this technology in cancer treatment and diagnosis are numerous and include the diagnosis of tumor hypoxia for more appropriate treatment planning, for patient selection of agents that are designed to overcome tumor hypoxia, for monitoring response to antiangiogenic and antivascular agents, and for providing valuable prognostic information.
The work was supported by National Institutes of Health/National Cancer Institute Grant U54 CA-105008–01 (W. Lee and D. Wilson) and by National Institutes of Health/National Institute of Biomedical Imaging and Bioengineering Grant EB-001713-01 (to C. Sehgal and W. Lee). Microscopy was performed at the National Institutes of Health research support lab Center for Molecular Studies in Liver and Digestive Diseases, supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant P30 DK-50306.
The authors thank Sosina Makonnen and Susan Schultz for technical assistance. We are indebted to Dr. Cameron Koch for the EF5 antibody and many useful discussions.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 the American Physiological Society