Fluorine-19 Nuclear Magnetic Resonance Imaging and Spectroscopy of Sevoflurane Uptake, Distribution, and Elimination in Rat Brain

All animal protocols were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh. Adult, male Sprague-Dawley rats weighing 250-300 g were anesthetized with 55 mg/kg sodium pentobarbital subcutaneously, after which the trachea was nontraumatically intubated. The lungs were ventilated with 100% Oxygen²using a rodent ventilator (Model 683, Harvard Apparatus, Boston, MA) at 60 strokes/min. When the first sign of awakening from pentobarbital anesthesia was observed, sevoflurane administration was started at a predetermined dose (4-10% in oxygen), and the time was arbitrarily designated as zero. A calibrated sevoflurane vaporizer (Ohmeda Sevotec 3, BOC Health Care, England) was used in conjunction with a Vetroson small animal anesthesia machine (Model 6610, Summit Hill Laboratories, Navesink, NJ). Pancuronium was given intraperitoneally for neuromuscular blockade, at a dose of 1 mg initially and at 0.25 mg every 2 h thereafter. Uptake of sevoflurane was terminated at a designated time (2.5-85 min), either by discontinuing anesthetic administration or by a 1-ml intracardiac bolus injection of 4 M KCl to induce cardiac arrest. The former method was used for continuous observation of sevoflurane washout, whereas the latter was used to image the anesthetic distribution at specific time. Animals were placed prone in all experiments. A heating pad maintained a body temperature of 37 degrees Celsius.

All experiments were carried out at the Pittsburgh NMR Center for Biomedical Research, using a 4.7-Tesla, 40-cm horizontal bore Bruker BIOSPEC NMR spectrometer. The resonance frequency for¹⁹Fluorine was 188.316 MHz. Definitions and details of NMR methodology are contained in the Appendix. The NMR detection coil, a home-built, 12-element birdcage resonator tunable to¹Hydrogen and¹⁹Fluorine, was used for most MRI and MRS experiments. A single-turn, 2 cm-diameter, surface-coil probe was used in selected experiments to confine the imaging volume to the brain. A three-dimensional (3D) Fourier transform (FT) imaging sequence and a two-dimensional (2D) FT imaging sequence, both with an echo time, TE, of 1.7 ms, were developed and used to minimize brain signal loss due to short T². Continuous T²relaxation measurements were made during washin and washout. Each complete T²measurement, consisting of 16 or 18 TE values ranging from 200 micro second to 200 ms, was best fit using Equation 2to determine the washin and washout kinetics that were T²-specific.

(Figure 1(left)) shows three representative¹⁹Fluorine head images, acquired at 2.5, 7.5, and 62 min, respectively, after the onset of sevoflurane administration. Corresponding¹Hydrogen images (right) are also shown for anatomic reference. The¹⁹Fluorine image at 2.5 min was the sum of five coronal slices of a three-dimensional image, and the other two¹⁹Fluorine images were acquired using the two-dimensional sequence shown in Figure 1(A) in the Appendix. The B¹inhomogeneity seen in the uppermost¹Hydrogen image arose because the probe was not returned to the¹Hydrogen frequency.

To visualize the intracranial, regional distribution of sevoflurane, a localizing surface-coil probe was placed closely above the head. The radiofrequency pulse length was preadjusted, using a spherical phantom, to have a maximum excitation profile about 7 mm from the center of the coil. Figure 2(a) shows a representative coronal slice from a three-dimensional image after a 10-min administration of 7% sevoflurane; an anatomic reference appears in Figure 2(b).

The¹⁹Fluorine signals of sevoflurane in the rat head contained two T²components. Nonlinear regression of T²measurements using multiple-exponential decay functions of TE produced two different time constants, with T^2fastin the range of 3-5 ms and T sub 2slow in the range of 105-130 ms. Typical examples of such regressions using Equation 2are given in Figure 3(A). The percentages of the two T²components in the total intensity depended on the receiver coil used and on the uptake time. The insert to Figure 3(A) compares the T sub 2 relaxation measured with the birdcage coil and with the surface coil on the same animal, which was killed 6.5 min after the sevoflurane administration. Clearly, the surface coil detected a higher percentage of the T^2fastcomponent than did the birdcage coil. This remained true during the course of uptake.

As the uptake time increased, the ratio of the two T²components changed. Figure 3(B) depicts, on a log scale, I^T2fast(t)/I^total(t) x 100% (see Appendix) as a function of sevoflurane uptake time (t). Twelve rats, denoted by different symbols in Figure 3(B), were measured with the birdcage or the surface coil or both. Although the total¹⁹F-NMR intensity in the rat head increases with increasing uptake time (Figure 3(C)), the percentage of the T^2fastcomponent contributing to the total intensity decreases exponentially (linear on the semi-log plot seen in Figure 3(B)). For example, the T^2fastcomponent contributes initially to almost 60% of the total¹⁹F NMR signal intensity in the head, as measured with the birdcage head coil (or 82%, as measured with the 2 cm-diameter surface coil). Ninety minutes later, this contribution decreased to less than 28% when measured with the birdcage coil (or to approximately 55% with the surface coil).

(Figure 4) depicts typical NMR washout data for three rats after 67-89 min of 4% sevoflurane general anesthesia. Total NMR signal intensities, normalized by initial values for each washout, were decomposed using Equation 2to provide the elimination kinetic constants for the T^2fastand T^2slowcomponents. Solid lines show best fits with single- or double-exponential functions. Note that the washout of the T^2fastcomponent, which constitutes 25% of the total intensity at the beginning of the washout process, is single-exponential and occurs at a faster rate than that of the T^2slowcomponent, which washes out at two different rates.

The¹⁹F NMR images in Figure 1and Figure 2are the first to show in vivo detection of anesthetic distributions in brain parenchyma. Almost immediately after beginning administration of the anesthetic, sevoflurane is seen mostly in the brain tissue (Figure 1, top left). As uptake time increases, sevoflurane can be seen to accumulate in the surrounding muscle and fat (Figure 1, middle and bottom left). Note that the decrease in relative intensity in the brain region with increasing uptake time does not represent an absolute reduction in the total amount of sevoflurane in brain. Rather it represents a relative reduction due to slower accumulation of sevoflurane in the surrounding tissues. The sequence of images is consistent with perfusion-limited models of anesthetic uptake, [5]wherein saturation of the vessel-rich group (e.g., CNS) occurs significantly earlier than that in the vessel-poor group (fat). The surface coil images (Figure 2), with better image quality and partial localization to the brain, show nonuniform distribution of sevoflurane in brain tissue 10 min after the onset of sevoflurane administration. Corpus callosum and dentate gyrus occur in this image as relatively hypointense regions (compare Figure 2(a and b)). Detailed anatomic identification and quantitation of concentrations, however, will require improved image resolution.

The ability to conduct¹⁹F MRI of the brain has significant implications. For example, if fluorinated anesthetics were administered by microdialysis, one might use these new¹⁹F MRI techniques to visualize affected regions, so as to correlate anesthetic intensity with neurophysiologic effects. Another potential application of the new technique is to evaluate anesthetic distribution in the presence of cerebral blood flow abnormalities, e.g., slowed uptake after stroke or head injury.

Similar to results with halothane and isoflurane, [9-13]our data indicate that sevoflurane resides in at least two distinct microenvironments in the rat head, as characterized by the two T²time constants. The T^2fastcomponent results from relatively immobile sevoflurane molecules, possibly bound to macromolecules. [29]In contrast, the T^2slowcomponent results from relatively mobile sevoflurane molecules. Taken together, the early abundance of imaging intensity in the brain, the later change of the intensity to adipose tissue, and the shift in total signal distribution from the T^2fastcomponent to the T^2slowcomponent during uptake indicate that sevoflurane in brain tissue contributes primarily to the T^2fastcomponent, whereas that in adipose tissue contributes primarily to the T sub 2slow component. This conclusion is consistent with the observations that partial localization with surface coil to the brain region results in a much higher percentage of T^2fastthan that from the entire head (Figure 3(A)) and that the relative decrease in T^2fastcomponent as a function of uptake time is less steep with the surface coil than with the head coil (compare the two slopes in Figure 3(B)).

Elimination of sevoflurane can be described phenomenologically using apparent first-order kinetics. [10,11]Thus, the overall rate of elimination can be characterized by the half time, t¹/2: Equation 1where k is the apparent first-order elimination rate constant. By decomposing the total washout intensities based on the T²values, distinct washout kinetics were revealed for sevoflurane in different microenvironments in the rat head. From regions where sevoflurane is relatively immobile, as characterized by T^2fast, the washout is single-exponential and has a t¹/2 of 13 min. In contrast, in the relatively mobile environments, sevoflurane exhibits two washout rates, with t¹/2 being 15 and 100 min, respectively. Thus, after 1 h of washout, sevoflurane is almost completely eliminated from the immobile environments (Figure 4); and after 2 h of washout, 99.8% of the total sup 19 Fluorine signal remaining in the rat head originates from the T^2slowcomponent.

In vivo NMR measurements of halothane [10]and isoflurane [11]washout using surface coils showed that t¹/2 depended on the size of coil used. This dependence was attributed to unequal sampling and weighting of various tissues by different coils. By reducing the coil size to focus on brain tissue, Litt et al. [10]and Mills et al. [11]estimated elimination rates of 34 min for halothane from rat brain and 36 min for isoflurane from rabbit brain, respectively. As pointed out by the same authors, these times are considerably longer than those (t¹/2 [nearly equal] 18 min) estimated from meticulous but invasive studies using gas chromatography, [20]autoradiography, and radiochromatography. [30-32]In our washout studies, however, a homogeneous birdcage coil was used to sample the entire rat head, with equal weighting of all tissue types. Although the retention time of sevoflurane in the rat head may differ slightly from that of halothane or isoflurane, the general agreement between the elimination rate of the T sub 2fast component and that found invasively from brain tissue is striking. This agreement further confirms our earlier assessment that sevoflurane in brain tissue contributes mostly to the T^2fastcomponent.

Based on the Meyer-Overton rule and minimum alveolar concentration additivity, [5]at equal minimum alveolar concentration doses, similar molar concentrations of anesthetic molecules must be present at crucial sites in the CNS where clinical anesthesia is produced. By a similar argument, recovery from general anesthesia would require elimination of general anesthetics from these sites. Our ability to separate "short-life" (T^2fast) from "long-life" (T^2slow) sevoflurane signals during uptake and elimination points to a new, noninvasive way to study the pharmacokinetics of anesthetic action. The parallel relationship between rapid awakening from sevoflurane anesthesia (data not shown) and the rapid elimination of the short-life (T^2fast) component suggest that sevoflurane general anesthesia results from sevoflurane binding to certain macromolecules in the brain (although the T²measurements in this study cannot distinguish specific from nonspecific binding). Inter-molecular dynamic polarization transfer* between nuclei in anesthetic molecules and in brain tissue has the potential to reveal the specificity of binding.

In conclusion, this¹⁹F NMR study used newer techniques at a higher magnetic field to demonstrate the intracerebral spatial distribution of sevoflurane. Sevoflurane distribution in the brain is heterogeneous under the experimental conditions described. There are two anesthetic microenvironments, characterized respectively by T^2fastand T^2slow, which exhibit different elimination kinetics. Both MRI and MRS data suggest that the T^2fastcomponent of sevoflurane is related to general anesthesia.

The authors thank Dr. Etsuro K. Motoyama, for providing a sevoflurane vaporizer for this study; Ken Lee Foon, for machining the parts of the tunable birdcage probe; Maryann Butowicz, for assisting in animal preparation; and Dr. Chien Ho and Dr. Alan Koretsky, for suggestions and discussion.

MRI and MRS exploit an intrinsic nuclear property called nuclear spin. In an external magnetic field, nuclei with non-zero spins are polarized to build up spin energy. This energy is usually too weak to detect individually. Signals measured in MRI and MRS result from the so-called resonance phenomenon, which requires that an ensemble of spins act coherently to absorb excitation energy established by a radiofrequency coil. In modern NMR technology, the coherence occurs in response to radiofrequency impulses and is often a complicated function of resonance frequency and the "lifetime" of the coherence. For a given nucleus, the resonance frequency (often expressed as "chemical shifts" relative to a standard frequency) is directly proportional to the magnetic field strength, B⁰. In an MRI experiment, B⁰can be systematically varied by magnetic field gradients (i.e., changes in field strength per unit distance) to register the spatial information. The lifetime of the coherence is characterized by two relaxation times: a longitudinal relaxation time (T¹) measuring how fast the nuclear spin energy is converted into non-spin energies, collectively referred to as lattice energy, and a transverse relaxation time (T²) measuring how fast the excitation energy is interchanged among nuclear spins. The resonance frequencies and relaxation times characterize the molecular environments of the nuclei. In general, a long T²indicates a relatively free (or more mobile) microenvironment, whereas a short T²indicates a relatively bound (or less mobile) microenvironment.

In addition to T²relaxation, external factors such as magnetic field inhomogeneity and magnetic field gradients can contribute to loss of coherence. As long as these external factors do not change randomly with time, the additional loss of coherence can be regained by reversing the effects of such factors. In a spin-echo, an inversion radiofrequency pulse (also called a 180 degrees pulse) is applied to reverse the sense of spin precession about the magnetic field. The resultant regain of coherence is called a spin-echo. The time duration from the beginning of coherence loss to the maximum refocusing is called the echo time, TE. If additional loss of coherence occurs because of the application of a magnetic gradient, a second gradient, of exactly the same amplitude but opposite in direction, is applied to cancel the effects of the first gradient. The resultant regain of coherence is called a gradient-recalled echo. The total time from the beginning of coherence to the refocusing is again denoted as TE. Spin-echo and gradient-recalled echo are sometimes combined to maximize the gain.

An NMR pulse sequence is a series of magnetic field gradient pulses, radiofrequency pulses, delays, and data acquisitions (Figure 1(A)). The effect of radiofrequency pulses is to bring nuclear spins into detectable coherence. Delays between and after pulses allow spins to evolve under the influence of their microenvironments. Such influence is reflected in the spectra or images as changes in chemical shifts and relaxation times.

Magnetic field strength can strongly affect the nuclear environment. In MRI, the static field is systematically modulated by magnetic field gradients. Because the resonance frequency of a nucleus is proportional to the magnetic field sensed by that nucleus, a linear change in the field strength along a given direction will result in a proportional change in the resonance frequencies of the nuclei along the same direction. Thus, by applying field gradients, spatial information is converted (or encoded) into frequency information, which can be displayed as images by an NMR spectrometer.

The dimension of an image refers to the number of independent directions in which encoding takes place. A two-dimensional imaging sequence usually contains two independent encoding gradients: a phase-encoding and a frequency-encoding gradient. The selection of a two-dimensional slice for imaging is achieved conventionally by using a long-duration selective excitation pulse during a third gradient. This method works only when the lifetime of the coherence (T²) is significantly longer than the selective pulse duration. Otherwise, loss of coherence during the selective pulse leads to no signal for imaging. In this study, we used a new two-dimensional sequence (see below), which is insensitive to short T²losses. The new method differs from conventional ones in that the selective pulse is not for excitation but for inversion of the thermal equilibrium population. Images are constructed from spectral differences with and without the selective inversion, so that only the inverted slice is imaged. Another alternative to reducing the T²loss is to avoid the use of selective pulses entirely. A three-dimensional imaging sequence can achieve this by adding a third encoding direction into an otherwise two-dimensional sequence. Obviously, adding a dimension will increase the scan time significantly.

NMR probes are radiofrequency devices that operate analogously to the transceivers in radio communication. Probes are designed in several different configurations; the most frequently used in vivo are birdcage resonators and surface coils. A birdcage resonator is made of capacitors and wire segments (often called elements) arranged as a cylindrical cage with open ends. The radiofrequency magnetic field generated by a birdcage resonator, often denoted as B¹, is perpendicular to the cylinder axis and is highly uniform inside the cylinder. A surface coil, on the other hand, uses a loop of wire for transmission and receiving. Its B¹field is perpendicular to the plane of the loop and is confined to the region close to the coil. This partial localization increases the detection sensitivity of the surface coil. Its drawback is poor B¹homogeneity.

In this study, a three-dimensional gradient-recalled echo imaging sequence with a short TE was used to minimize signal loss due to the short T²of¹⁹Fluorine signals in the brain. In the three-dimensional sequence, a short but strong radiofrequency pulse was applied, followed by two phase-encoding gradients and a third frequency-encoding gradient. Because no long-duration shaped pulses were involved, the shortest TE achievable was governed by the switching time of the gradient units and by eddy-current effects in the volume of interest. A 15-cm diameter gradient insert for the 40-cm bore magnet was used to reduce TE to values as short as 1.7 ms. In a typical three-dimensional experiment, time-domain data were digitized into a three-dimensional matrix (of size 128 x 32 x 16 or 8), which was then zero-filled twice in the second dimension and once in the third dimension before Fourier transformation to reconstruct the three-dimensional images. The fields of view in the three dimensions were 4 x 4 x 4 cm³. For the three-dimensional imaging experiments with the birdcage probe, a variable number of data averaging (NA), having a Gaussian distribution with respect to the second gradient dimension, was used to reduce the total imaging time. Thus, a typical NA list used was 4, 6, 8, 12, 18, 26, 34, 44, 56, 70, 82, 96, 108, 116, 124, and 128, respectively, for blocks 1 through 16 of the second dimension and symmetrically for blocks 17 through 32. The repetition delay (RD) between scans was 1 s with four initial dummy scans.

The major drawback of the three-dimensional method is the long imaging time. Thus, the possibility of imaging anesthetic distribution in a single slice in the rat head was investigated. Conventional two-dimensional methods based on a selective slice excitation scheme have proved inadequate [12]because the long-duration selective excitation pulses are part of the TE period and the T²relaxation occurs inevitably during these pulses, leaving only diminishing signal for imaging. To minimize such T²losses, a new two-dimensional sequence, as shown in Figure 5, was developed and used. In this new sequence, a hyperbolic secant-shaped [33]selective inversion pulse was applied in alternating scans during a slice-selective gradient (G^slice) to prepare the magnetization for imaging. A nonselective short 90 degrees pulse was applied to create the coherence, followed by a regular two-dimensional gradient-recalled echo imaging acquisition. Images were reconstructed using the spectral differences with and without the selective inversion pulses, so that signals from regions not being inverted were canceled. Because the slice selection occurred during the longitudinal rather than the transverse relaxation, and also because the shaped pulse was not part of the TE, the shortest TE achievable in this two-dimensional sequence was the same as that in the three-dimensional sequence. The efficiency for the sequence ranges from 100% in the cases of perfect inversion to 50% in the cases when the inversion pulse acts effectively as a saturation pulse. For a typical two-dimensional image, the slice thickness was 7 mm, the field of view was 6 x 6 cm²with a time-domain data matrix of 128 x 32, which was zero-filled twice in the second dimension before Fourier transformation. An NA list, having a symmetric Gaussian distribution with respect to the phase-encoding steps and ranging from 4 (for the first and the last blocks) to 2,048 (for center blocks), was used. The RD was 1 s with four initial dummy scans.

The T²of¹⁹Fluorine in rat head was measured in a series of time segments throughout the uptake and elimination of sevoflurane. In each segment, the spin-echo sequence (90^x-TE/2-180 sub y -TE/2-ACQ), with 16 or 18 TE values ranging from 200 micro second to 200 ms, was used. For every TE value, four scans were summed with an interscan delay of 4 s. Thus, a complete run of a T²measurement was about 4.3 min, the midpoint of which was used to time the uptake and elimination processes.

For each 4.3-min run, T²values were determined from unweighted nonlinear regression of 16 or 18 points of total spectral intensities as a function of TE. Fitting was performed on a Macintosh computer using KaleidaGraph, a commercial program from Synergy Software (Reading, PA). Because of the decay nature of the spin-echo train, a biexponential decay function of TE was used as the theoretical model in the nonlinear regression [34]: Equation 2where I^T2fast(t) and I^T2slow(t), to be determined from the fitting, are the relative intensities of the two exponential components, characterized by the time constants T^2fastand T^2slow, respectively, and t is a time index for the uptake or elimination process, with t = 0 being assigned arbitrarily to the onset of sevoflurane administration for the uptake and to the termination of sevoflurane for the elimination, respectively. For the uptake process, the percentage of the T^2fastcomponent is normalized against the total intensities at each individual time point (i.e., I^T2fast(t)/I^total(t) x 100% for each sampling time t), whereas for the elimination process, the relative intensities of I^T2fast(t) and I^T2slow(t) are normalized against the total intensity at the beginning of the washout, I sub total (t = 0). After normalization, the overall washout intensities are decomposed to differentiate the elimination kinetics of the two T²components. All I^T2fast(t) and I^T2slow(t) are presented as the best estimates plus/minus SE of the estimates [34]using Equation 2.

* Kubota Y: Comparative study of sevoflurane with other inhalation agents. Anesth Prog 39:118-124, 1992.

** Shimada M: Hemodynamic effects of isoflurane, sevoflurane, enflurane, and halothane in man: An echo cardiography analysis. J Jpn Soc Clin Anesth 9:204-213, 1989.

*** Kazama T, Ikeda K: The comparative cardiovascular effects of sevoflurane with halothane and isoflurane. J Anesth 2:63-68, 1988.