Animals and tumors
Female C3H mice, produced by the National Cancer Institute Animal Production Area (Frederick, MD), were used for this study. The mice were 7–9 weeks of age at the time of experimentation and weighed between 20–30 grams. All experiments were carried out under a protocol approved by the National Cancer Institute Animal Care and Use Committee, and were in compliance with the Guide for the Care and Use Of Laboratory Animal Resource, (1996) National Research Council. Tumors were grown in the mice by a subcutaneous (s.c.) injection of a single-cell suspension of 2 × 105 squamous cell carcinoma (SCCVII) cells in the right hind leg. Tumors grew to a size of ~1.0 cm diameter 7–10 days after injection. Tumor volumes were determined prior to PET scanning, by measuring three orthogonal diameters using calipers. For all imaging studies ~1 cm diameter tumors were used.
PET scanner
PET images were obtained with the ATLAS (Advanced Technology Laboratory Animal Scanner), a dedicated small animal PET scanner developed at NIH with an axial field-of-view (FOV) of 2 cm, a transverse FOV of 6.8 cm and an aperture of 8 cm [28–30].
The image data acquired with the ATLAS PET scanner were reconstructed using the three-dimensional ordered subset expectation maximization (OSEM) reconstruction algorithm, including a model of the system resolution [31]. All reconstructions were performed on the NIH Biowulf computer cluster, utilizing sixteen subsets and taking 10 iterations. Under these conditions, a spatial resolution of 1.6 mm full-width-at-half-maximum (FWHM) was achieved.
PET scan experimental set-up
The mice were fasted at least 3 hours prior to injection of 18F-FDG. Each mouse was anesthetized with isoflurane (induction; 2.0%, maintenance; 1.5%) carried by air at 750 ml/min and delivered via a nosepiece. An anesthetic delivery/scavenging device was continuously used throughout the experiment. Once anesthesia was established, further isoflurane was carried by either air (20.9% oxygen) or carbogen (95% oxygen/5% CO2). In order to inject18F-FDG intravenously (IV), a cannula was inserted into a tail vein. The IV lines consisted of a 30 gauge 1/2" needle inserted into one end of a 60 cm length of PE10 tubing (IntraMedic Polyethylene) (internal diameter = 0.28 mm; external diameter = 0.61 mm) and the needle of another removed from the hub inserted into the opposite end of the tubing. IV lines were filled with heparin (100 USP units/mL) prior to cannulation. To verify the integrity of the line, a bolus of heparin was given prior to 18F-FDG injection. A bladder catheter (PE10 tubing) was used to collect urine output and a rectal temperature probe (FISO Technologies Fiber Optic Temperature Gauge Model FOT C-PEEK) was used to monitor and maintain core body temperature. Once these procedures were completed the mouse was immobilized in a jig that secured the feet. This jig was inserted into a cylindrical Lucite chamber, fixed with a port for the attachment of thermostat-controlled, warm air circulator (Nikon Model ITC-32, Nikon Inc. Japan) in order to maintain the mouse's core temperature. The entire assembly was fixed to a computer-controlled motorized gantry capable of precisely moving the animal into the scanner. After core temperature of 37°C was established, PET image collection was started, and a 100 μl bolus of 18F-FDG (~400 μCi in 0.9% saline) was injected. The injected doses were recorded for each animal imaged. Initial pilot studies were performed on 4 mice by centering the tumor in the field-of-view prior to injection. Two of these first mice were studied only with carbogen and two only with air. For each mouse, a dynamic acquisition of 5 min/frame was acquired for 90 min total. In subsequent experiments each mouse was studied sequentially on both air and carbogen. These sequential studies, performed on an additional 5 mice were more complex, each consisting of 5 phases. Initially, each animal was anesthetized using either air or carbogen as the carrier gas for the isoflurane (3 with air and 2 with carbogen). While this gas was continuously administered, the scanner was positioned at the level of the heart, and dynamic imaging begun immediately prior to injection of 18F-FDG (12 frames/10 sec for 2 min, 6 frames/30 sec, followed by 45 frames/1 min), continuing for 50 min, in order to capture an image-derived input function. The scanner was then moved to the level of the tumor (the lower extremities), and dynamic imaging (5 min/frame) resumed for an additional 40 min, while continuing to administer the same mixture of isoflurane and either air or carbogen. At the end of this acquisition, the carrier gas for isoflurane was switched from air to carbogen, or vice versa, and the animal was left to equilibrate on the new gas mixture for 10 min. When the 10-minute equilibration period had ended, the scanner was again moved to the level of the heart and dynamic image collection commenced as described above, beginning approximately 30 sec prior to injection of a second bolus of 18F-FDG, first over the heart and then over the tumor. All the studies described above were performed while the mouse's body temperature was kept at 37°C.
An additional 2 mice were studied using this same two injection methodology but while their body temperature was kept at 30°C. We have previously shown that C3H SCCVII tumor-bearing animals placed on isoflurane experience an approximate 7°C decrease in core body temperature 15 min after initiation of isoflurane [27]. To conduct low temperature studies animals were prepared as described above and were administered isoflurane without temperature control until the core body temperature dropped to 30°C at which point the warm air circulator was adjusted to maintain a core body temperature of 30°C. This temperature was maintained throughout the two injection study.
Blood sampling
After anesthetizing the mouse with isoflurane carried by medical air (1.5%, 700 mL/min), the skin and integument over the right or left ventral surface of the neck was cut with surgical scissors and the cleidomastoideus and sternomastoideus muscles were dissected to expose the left external jugular vein, removing fat and/or cauterizing peripheral vessels to minimize blood loss. A cannula, like that described previously for tail vein cannulation, was then inserted into the external jugular vein and fixed in place with Vetbond tissue adhesive. A junction was made by soldering a 30 G 1/2" needle into a 23 G 1" needle. The 30 G 1/2" needle end could be inserted into the end of the central line, while the 23 G 1" needle end could then be inserted into a small piece of PE50 tubing, into which a 10 μL Hamilton syringe could be inserted for a more precise collection of blood. After administering 18F-FDG (~400 μCi, 100 μl) bolus, 10 μL of blood were drawn from the central line every ten minutes for one hour. Each blood sample was then diluted in 1 mL of saline, and its activity was read in a Cobra II AutoGamma scintillation counter.
Image-derived and blood-draw derived input functions
It is possible that the input function (the concentration of arterial 18F-FDG as a function of time) would change depending on whether the mouse was breathing air or carbogen. The area under the input function is a measure of the amount of 18F-FDG available for the tumor to metabolize. In order to compare air and carbogen 18F-FDG uptakes, the ratio of mean air to carbogen uptake in the tumor volume of interest (VOI) was computed, normalizing by the integral of the input function from time zero up to the time of the measurement of uptake. The input function was estimated and used for normalization as described below.
PET images at early time points revealed a large vessel dorsal and/or caudal to the heart. Comparison with CT images indicated that the vessel was the inferior vena cava, but this could not be determined with absolute certainty due to possible small mis-registrations between PET and CT. A strict determination of whether the vessel was arterial or venous was thought unnecessary, apart from any first transit uptake by the lung (known to be low), especially since the bolus was slow compared to the expected rapid cardiac transit times [32].
To determine a measured blood time activity curve, Am(t), using the major blood vessel, regions of interest were drawn around this vessel at several transaxial levels to determine the relative mean activity concentration in the vessel over time. Only relative concentrations could be determined because the diameter of this vessel was thought to be small compared to the spatial resolution of the scanner, making partial volume effects important. Furthermore, tissues adjacent to the vessel could have potentially contaminated the activity measured in the region of interest [33]. Corrections for these effects were made in the following way. A square region of interest (ROI), 2.8 mm on a side and encompassing the aorta was drawn and used to generate an aortic time activity curve. Two concentric rectangles were then drawn around this first ROI (7.3 mm and 9.6 mm on a side) and a background region was obtained by subtracting the smallest ROI from the largest ROI. Another time activity curve was created from this background region in order to estimate the background activity, B(t). It was assumed that the measured input function, Am(t) was equal to some combination of the true input function, At(t), and the background activity, B(t):
Am(t) = a* At(t) + b*B(t). equation 1
The constant "a" is less than unity because some of the counts blur out of the vessel region of interest due to the partial volume effect. Similarly "b" represents blurring of background counts into the vessel's region of interest. B(t) was only a negligibly small fraction of A(t) until late times (> 4 min) so the correction was only important at late times.
Since we were only interested in estimating the ratio of the blood input functions as a result of two different physiological challenges, air and carbogen breathing, we did not need to correct the input functions for "a" because the size and placement of the ROI was identical in each case:
Thus, we only needed to correct the input function for background contamination. Four venous blood samples were drawn from 30–60 min post injection. These samples were thought to be at late enough times so that venous activity concentrations would be nearly identical to arterial concentrations, so At(t) = Blood(t) at late times. This was confirmed by the very slow variation of Am(t) versus t at these times. Therefore, for each dataset, air or carbogen, the scaling factor b was obtained considering the true blood activity derived from the blood draws at late time, the actual measured input function and the background activity:
To make the correction more robust, 4 time points were considered in order to estimate b.
For logistical reasons, blood samples could not be drawn from the same mouse. Therefore a cohort of 6 mice of the same age, size, and tumor development as the mice imaged were injected with the same volume of FDG, and each breathed either air or carbogen under the same conditions as the imaged mice. The time activity curves for all mice (n = 3, air; n = 3, carbogen) were normalized by the activities of injected 18F-FDG and the curves were fit temporally with a bi-exponential function. These fitted normalized blood samples were then used in the equation 3 to compute b. Once b was determined for each imaged mouse, the true arterial ratio between air and carbogen could be computed from equation 2.
Volumes of interest
Volumes of interest around the whole tumor were defined using a semi-automatic, three-dimensional, threshold based, region-growing algorithm (MedX, Sensor Systems Inc). The threshold was based on a percentage of the average maximum intensity in the tumor. Average maximum intensity was defined as the maximum pixel in the tumor averaged with neighboring pixels in a 0.5625 mm radius sphere, to reduce statistical noise. A 30% threshold generated a VOI that visually included the entire tumor, but VOIs using other thresholds were also studied.
The very small in-plane mis-registrations of certain mice that occurred between the air and the carbogen measurements were corrected by maximizing the 3D correlation between the two image sets (FLIRT linear registration tool; Image Analysis group, Oxford University). The final series of images collected during carbogen breathing were aligned to the respective final series of images collected during air breathing. Image data outside the tumor were not used for registration. The 3D-to-3D registration was performed using six degrees of freedom (rigid body), with 3 translations and 3 rotations. After proper normalization for input function, the air data set was then subtracted from the carbogen data set, and the alignment of tumor signals was judged by eye. Although not used for alignment, the femoral vessels also appeared to be present in each data set, and the signal in those regions was zero after the subtraction, lending support to accurate registration. Registration resulted in at most a 2 mm total displacement in the region of the tumor.