Introduction

Recent advances in imaging technologies are leading to noninvasive assays for monitoring the progression of biological processes in living, intact small animals. Over the past 10 years, transgenic manipulation of mice has created phenotypes with the potential of linking specific genes to molecular, cellular, and organ functions during both good health and disease. Application of the transgenic or gene “knock-out” mouse mutant technology to studies of the biochemistry and physiology of the brain, heart, musculosketetal, and metabolic systems has historically been through necropsy studies, which require the sacrificing of many animals and the performance of tedious assays. Contemporary noninvasive imaging methods of radionuclide emission tomography [positron emission tomography (PET), single photon emission tomography (SPECT)], magnetic resonance imaging (MRI), magnetic resonance spectroscopy (MRS), and optical imaging (near-infrared imaging and tomography, fluorescence imaging, optical tomography, and bioluminescence imaging) have great potential for the study of transgenic animals as small as the 30-g mouse.

Perhaps the area where the combination of molecular biology and advanced imaging is progressing most rapidly is the use of reporter strategies to image gene expression. The use of reporter genes has contributed greatly to our understanding of how genes are expressed in cells in culture and in vivo. Regulation of gene transcription is mediated by complex interactions among positive and negative regulatory elements, which are ultimately directed by the environment of the cell. But studies of gene regulation in mammalian cells have generally been performed in the absence of the contextual influences of the entire organism, because reporting systems used in animals have required ex vivo assays and thus have been limited to biopsy or postmortem tissue samples. Further, such studies are arduous, and results may be delayed until a lengthy analysis has been completed. Fortunately, three modes of noninvasive imaging can now be used to study gene expression: radiolabeled reporter gene imaging, MRI and MRS of enzyme kinetics or contrast agent accumulation and photon imaging, and detection of fluorescence and bioluminescence reporter genes (luciferase and green fluorescence protein).

This review first presents a technological description of the character of the image data that can be acquired from current or feasible instruments and, second, gives examples of how noninvasive imaging methods have been applied to studying physiology and gene expression in small animals, particularly transgenic animals.

Transgenic Animal Preparation

There are two ways in which the genome can be manipulated to create genetically altered mice (47C). The simplest is to inject DNA directly into the nucleus of single-celled embryos. This DNA occasionally integrates into the genome giving rise to a transgenic mouse. By proper choice of DNA coding region and DNA control region virtually any protein or RNA can be expressed in any cell type.

By using gene targeting via homologous recombination in murine embryonic stem cells, a specific gene can be either ablated or modified, and the mouse carrying the specifically modified genome can be generated for subsequent studies. Thus, a gene suspected of playing an important role in development, organ function, or pathogenesis can be either knocked out or modified in a particular domain or even in a single amino acid residue, and the resultant phenotype can be studied over the animal’s lifetime in the absence of any acute pharmacological or surgical interventions. While these interventions are most common in the mouse, they have been performed in other animals (e.g. rat, rabbit, pig).

Because these animal models are the results of a stable genetic modification, the primary genetic etiology of any changes that are observed is known. Furthermore, these models are stable as the mutations are propagated through the germ line. Genetic modifications lead to the creation of new animals whose numbers can be expanded and subsequently disseminated to the research community for a variety of assays and studies. A large variety of biochemical, tissue and physiological assays have been performed on transgenic or knockout mice. However, recently there is rapid growing interest in application of noninvasive imaging to study genetically modified animals. The emphasis in this review is on application of imaging to transgenic models as well as on the characteristics and potentials of functional imaging instrumentation for in vivo studies of small animals.

Emission Imaging of Radiopharmaceuticals

The use of radiotracers as blood pool and metabolic probes for imaging is the basis of clinical nuclear medicine studies, which are known in the tomographic mode as SPECT and PET. The conventional projection or planar imaging with standard collimators for human studies and the contemporary PET and SPECT scanners used for human patient studies do not have sufficient resolution for transgenic animal studies; thus there has been, over the past 10 years, an effort to develop imaging devices with both the resolution and sensitivity to allow in vivo animal studies.

Single Photon Emission Cameras for Planar Imaging

The simplest adaptation of the current technology of standard nuclear medicine instruments is to use a pinhole of 1-mm diameter in a tungsten plate between the animal and the camera (Figure 1A). High resolutions of 1–2 mm in projection images will be obtained, but at the cost of very poor sensitivity. For example, a 1-mm pinhole located 50 mm from the organ will have a sensitivity of <0.003% that of an ideal detector completely surrounding the animal and ∼200-fold less than a cylinder of detectors for PET. Nevertheless, it is possible to achieve high counts in dynamic studies of blood pool changes, with a simple pinhole camera and rapid injection of a blood pool tracer used in the anesthetized animal.

Despite the <2-mm resolution obtainable from pinhole collimation with single or multiple cameras, the sensitivity is too low to allow imaging studies of high-affinity receptor systems such as the neuroreceptors of the brain. To remedy this problem, a new technology that uses arrays of small (3- × 3-mm) detectors such as CsI(Tl) crystals as scintillators and silicon photodiodes as light detectors has evolved with improved intrinsic resolution over conventional gamma cameras (Figure 1B) (35, 50, 97). These devices with promise for projection imaging are in the prototype stage. They have an improved spatial resolution, are inexpensive, and are easily adapted to a bench-top device. Solid-state cameras with the expensive cadmium zinc telluride system are also candidates for laboratory animal studies, but they are very expensive (65).

Devices such as those shown in Figure 1 have much greater resolution than the contemporary gamma cameras used for clinical studies. Because the contemporary radionuclides used have energies of 80 keV and above, the inherent diffusion of photons in detector systems leads to intrinsic resolutions of 4 mm. However it is possible to acquire projection images with 2- to 3-mm resolution by using radionuclides with energies in the range of 35 keV (e.g. iodine-125). Iodine-125 can be attached to many different molecules involved in biologic and even therapeutic activity. The low energy has the advantage of requiring less collimation and less dense or thinner detector systems (i.e. wire chambers).

Cardiac Blood Pool Dynamics in Small Animals by Planar Imaging

To evaluate cardiac performance in small animals, the transit of blood through the ventricles is measured from moment to moment by gated radionuclide images, using a pinhole collimator (2-mm aperture) fitted to a conventional gamma camera. The feasibility of this approach to measure left-ventricular ejection fraction and ejection rate was determined with and without drug interventions by using 5 mCi of Tc-99m pertechnetate in 400-g laboratory animals (53). With Tc-99m blood pool agents or the very short half-life (9.3 min) tantalum 178 obtained from the tungsten/tantalum generator (62), high-quality images of ventricular function have been obtained by using 1-mm pinhole collimation and gamma cameras (67) or a compact wire chamber system (Proportional Technologies, Inc.; M Dae, personal communication).

Planar Imaging in C-neu-transgenic Mice and Mouse Xenografts

The c-neu oncomouse is a transgenic mouse that spontaneously develops breast tumors. By using Tc-99m sestamibi, a common radioisotope probe for breast tumor detection in human subjects, and a gamma camera with a single 1-mm-pinhole collimator, semiquantitative studies of tumor retention of the tracer were obtained with 60-min imaging time and 5–20 mCi of injectate (19). The reason imaging can be accomplished in this low-sensitivity system is that the doses injected are 2000-fold the mass-normalized doses used in human subjects.

Use of pinhole collimation and conventional gamma cameras has been found effective in evaluation of monoclonal antibody B72.3, which reacts with a tumor-associated antigen found on human breast and colon carcinomas. The tracer can be iodine-125, which allows evaluation of the kinetics and behavior under therapeutic maneuvers in mouse cancer xenografts.

Potentials for Brain Neuroreceptor Planar Imaging

An example of in vivo rodent studies that are enabled by high-resolution animal imaging systems adjusted to image iodine-125 is the use of ligands, which have a high affinity for the dopamine transporter (DAT) system. The distribution of binding sites of labeled (¹²⁵I or ¹²³I) RTI-55 [3 beta-(4-iodophenyl)tropan-2 beta-carboxylic acid methyl ester], a phenyl tropane analog of cocaine, and the selective labeling of the DAT have been studied by in vitro and ex vivo autoradiography in the rat whole brain. RTI-55 binds not only to DAT but also to the serotonin transporter (30). The behavior of this tracer in cocaine abuse as well as in human subjects with compulsive violent behavior is a subject of intensive contemporary interest (83). This tracer is also a probe for investigation of the DAT in the spontaneous hypertensive rat, which has been found to increase DAT before the onset of hypertension (94). Imaging studies of RTI-55 in the rat brain are currently underway, using a planar camera with an array of detectors and a high-resolution collimator as shown in Figure 1B.

Single Photon Emission Tomography in Animal Studies

The tomographic modes of PET and SPECT differ not only in the types of radionuclides used but also in the obtainable sensitivity. SPECT requires a lead collimator between the object and the subject and in most designs requires revolution of the gantry containing the imaging system around the animal. Even with multiple cameras surrounding the animal, the sensitivity of SPECT will be >70-fold less sensitive than that of PET (11); thus, for tomographic applications, PET systems have received the most attention as instruments for studying small animals. Nevertheless, using large doses and relatively long imaging times, successful tomographic studies with SPECT have been reported.

Despite the very low sensitivity of SPECT pinhole imaging, successful biomedical studies of the myocardium, of skeletal bone, and of tumors have been conducted with pinhole SPECT (45, 78, 95). The sensitivity is about 3 counts per Ci (Tc-99m) for a 2-mm aperture with 40- to 50-mm rotation radius. The expected resolution is 2.2 mm for a 2-mm aperture (78). Whereas 30 min of data collection was required to evaluate myocardial infarction in rats with Tl-201 (111 MBq) and a single pinhole on a rotating gamma camera at 40-mm radius (103), <8 min would be required for multiple pinhole systems recently described (Figure 1C) (45). In the system with four pinholes, a resolution of 1.65 mm has been obtained with a 1-mm pinhole aperture and a 180-mm radius of rotation.

It has been shown that imaging of herpes simplex virus type-1 (HSV-1) thymidine kinase (tk) gene expression is possible by using a clinical gamma camera in planar imaging mode and by SPECT (84). Iodine-131–labeled 2′-fluoro-2′-deoxy-1-beta-D-arabinofuranosyl-5-iodo-uracil (FIAU) becomes trapped in the cell if thymidine kinase is present during the phosphorylation reaction. A tumor that has been transfected in vitro with the HSV-1 tk gene and injected into an animal will show accumulation of iodine-131–FIAU-derived activity, whereas tumors not transfected will not show activity. The magnitude of the accumulation corresponds to the in vitro measured sensitivity of cell lines to the therapeutic agent ganciclovir. Studies of HSV-1 tk gene expression with PET (see below) give improved image quality with long-lived (4-d) iodine-124 or even the shorter-half-life (2-h) fluorine-18.

Positron Emission Tomography

Most of the emphasis on instrumentation for animal studies has been to build specialized PET systems, owing to the abundance of metabolically interesting probes of mammalian metabolism and neuroreceptor activity (81A) and to the good sensitivity of PET (12). Below we discuss the main idea of PET and then review the systems being proposed.

Positron tomography involves detection of the two 511-keV photons emitted when a positron annihilates with a nearby tissue electron (Figure 2; see color figure). The positron emission is one of the natural modes of decay of man-made isotopes such as carbon-11, nitrogen-13, oxygen-15, or fluorine-18. Note that each of these isotopes has an atomic number that is one less than the corresponding stable elements carbon-12, nitrogen-14, oxygen-16, and fluorine-19, respectively. These new isotopes are unstable because they are missing a neutron and, to reach stability, they decay by converting a proton to a neutron, which involves losing the positive charge of the proton in the form of a positive electron known as the positron. When the positron encounters a free electron within a few tenths of a millimeter from its origin in the nucleus, an annihilation occurs, with the release of two photons or gamma rays whose energy is equal to that of the mass of the positron or electron, in agreement with E = mc ². These 511-keV photons fly off at 180° from each other and are detected by dense crystals known as scintillators. The 511-keV photons are converted to photons of wavelengths near that of visible light, and these photons are detected electronically by photoelectron multiplier devices or other methods of converting light to an electric signal. The fact that the two photons are detected in time coincidence means that the activity must have arisen along a line between the two detectors (Figure 2). About 500,000 of these lines of response (or coincidence lines) are used along with reconstruction tomography mathematics similar to those used in X-ray computed tomography (CT) to create an image of the spatial distribution of the radionuclide. Theoretically the resolution of a PET system can be between 1 and 2 mm and, unlike SPECT, the sensitivity is not dependent on the resolution but on the surface area of detectors that can be placed around the animal.

Over the past 10 years there has been a concerted effort to develop a small, compact, inexpensive, and high-resolution PET scanner for animals (6, 21, 27, 40, 46, 49, 93, 96). Six designs are shown in Figure 3. All of these designs endeavor to place small scintillation detectors in coincidence across the animal. The more detectors, the higher the sensitivity. The conventional PET system (Figure 3A) will have a resolution of 4–5 mm and a sensitivity acceptable for studying tumors in the flank but not practical for studies of brain and heart metabolism because the resolution is comparable to the size of the organs of a mouse. An exception is the Lawrence Berkeley National Laboratory (LBNL) PET600 built in 1986 for human head and animal studies (22). The resolution at the axis is 2.6 mm with a decrease to 4 mm at 10 cm from the axis. That system can accommodate only one slice owing to the problem of packing small detectors in more than one ring.

To improve resolution, smaller detectors are used, but these detectors require very small, high-efficiency photoelectron multiplier tubes that are not practical; thus position-sensitive larger photoelectron multiplier tubes are used. Systems modeled after the conventional PET but smaller in diameter use various combinations of scintillator crystals, consisting of bismuth germanate (BGO) or lutetium oxyorthosilicate (LSO) coupling arrangements to single or multiple photoelectron multiplier tubes, or to position-sensitive photoelectron multiplier tubes (43, 94). The use of position-sensitive photoelectron multiplier tubes with very small detectors requires light pipes, as shown in Figure 3E (13, 14), to place the detector ring close to the animal and at the same time accommodate the bulky position-sensitive photoelectron multiplier tubes. The most available modern animal PET system is known as MICROPET (Figure 3E).

The MICROPET design is being commercialized (Concorde Microsystems, Knoxville, Tenn.) for use by the scientific communities. The detector systems have a field of view of 100 mm and a diameter of either 40 or 80 mm. The image resolution is 2 × 2 × 2 mm, and the sensitivity is 450 events per second per mCi for the 40-mm rodent system.

The current PET animal systems do not have the sensitivity to allow the entire spectrum of PET studies, because the specific activity of many radiopharmaceuticals is not high enough to allow imaging without some saturation of the biological receptor sites, particularly in studies of the neurochemical signaling systems of the brain. Thus a very high-sensitivity and high-resolution system has been proposed, wherein the animal is completely enclosed in a rectangular volume formed by 6 planar banks of PET detector modules (40). Each detector module uses an array of 0.3 × 0.3 × 3 cm–LSO crystals coupled on one end to a photoelectron multiplier tube and on the opposite end to an array of 3-mm–square silicon photodiodes, as shown in Figure 3F. The sensitivity of MICROPET is given in terms of noise-equivalent count rate, as shown in Figure 4 as a function of the activity density. The sensitivity of the improved system currently under construction is 700-fold that of MICROPET. The commercialization of the high-sensitivity system will require a few years.

There are some disadvantages of PET for animal studies. These studies require a supply of short–half-life radionuclides, which are generally available if the tracer uses fluorine-18; however, for carbon-11 and nitrogen-13, an accelerator source is required. A problem with the use of animals for the evaluation of a PET radiopharmaceutical or for use of a radiopharmaceutical for the evaluation of neuroreceptors is that the sensitivity of the system might be too low for the statistics needed, considering the resolution and saturability of the receptor site being studied. This problem is ≥200-fold greater for SPECT. When using experimental animals in PET, the amount of radiotracer injected can be infinitely increased to compensate for the need for increased statistics or for the low half-life of the tracer (e.g. 20 min for C-11). The amount of activity is limited by the amount of stable compound that accompanies the injection or the specific activity of the radiopharmaceutical, because the stable compound can saturate the receptors especially when high-affinity ligands are used.

The resolution obtainable with F-18 ligands and contemporary designs is 2 mm, whereas the resolution of the in vitro autoradiographic approach with F-18 and the PhosphorImager is 0.1 mm. Although the in vitro PET method requires animal sacrifice and dynamic studies require three to five animals per time point owing to physiological variability, the requirements for anesthesia and low resolution of in vivo PET might give the experimental advantage to in vitro methods for some experimental goals.

Animals must be anesthetized to minimize motion artifacts, and the effect of the anesthetic itself must be evaluated when interpreting the data. Problems of coregistration from one study to another must also be solved. Many studies can be done by using one MRI volume set as a registration standard for each animal during a longitudinal study.

Examples of Animal PET Studies

Although PET has not been used extensively in transgenic animals, its potential for characterization of the phenotypes relative to specific mutations involved in, for example, Huntington’s disease, Parkinson’s disease, multiple sclerosis, congestive heart failure, or creatine kinase (CK) gene knock-outs can be appreciated by the quality of contemporary images. Because pinhole planar imaging with single photons could achieve resolutions that allow studies on animals as small as mice, using conventional technologies, most previous radiopharmaceutical studies have used pinhole tomography. Recently the major trend in small-animal studies with radiopharmaceuticals has been toward PET with specialized systems. Figure 5 (see color figure) shows a whole-body distribution of a glucose analog uptake with 2 mCi of F-18 fluorodeoxyglucose in the rat. Figure 6 shows the bone metabolism by the accumulation of ionic fluorine-18, an approach that has potential for studying models of osteoporosis.

PET has been used to study changes in post-synaptic D2-receptor affinity and monitor effects of growth factor on presynaptic dopaminergic function in a rat model of Parkinson’s disease (44, 64). PET has been used to monitor tissue grafting in a rat model of Huntington’s disease (85). A number of published reports used PET with rodents to screen or evaluate new or proposed radiopharmaceuticals. For example the evaluation of cocaine analogs such as dopamine-transported ligands (100); high-affinity antagonists for the 5-HT2A and 5-HT 1a receptors (3, 42); a ligand for the 5-HT uptake site (79) and antagonists for the alpha-2 adrenoreceptor (24, 41).

Noninvasive imaging of the expression of reporter genes in living animals using PET and SPECT has been underway for the past few years. Because radionuclide-labeled ganciclovir is metabolized by thymidine kinase and thereby trapped in the cell, the expression of the HSV-1 tk can be demonstrated by PET imaging with I-124–labeled FIAU (84) or 8-[¹⁸F]fluoroganciclovir (31) as reporter probes of adenovirus-directed expression of the HSV-1 tk reporter gene in living mice. The experience with FIAU has been in cancer cell lines transfected with HSV-1 tk and grown in rats in which gene expression is revealed as accumulated activity relative to the sensitivity to ganciclovir. There was also shown to be a correlation between the percent injected dose of 8-[¹⁸F]fluoroganciclovir retained per gram of liver and the levels of the hepatic HSV-1 tk reporter gene expression. Repeated, noninvasive, and quantitative imaging of PET reporter gene expression appears to be a valuable tool for studies of gene therapy, at least in animals, and studies of both environmental and behavioral modification of gene expression in transgenic mice. The use of I-124 is attractive because of the much longer half-life than F-18; however, the intrinsic behavior of I-124 (high-energy positron and relative low positron yield) will lead to less resolution potential than F-18 (84).

Magnetic Resonance

A second general category of imaging and noninvasive evaluation useful in the study of transgenetic animals is the use of magnetic resonance. The excellent soft tissue contrast, the ability to readily obtain resolutions less than 0.2 mm, the flexibility for gating or fast image acquisition by MRI, and the growing list of physiological factors that can be quantitated make this modality especially attractive for imaging cardiac and brain anatomy and function. In addition, the ability to monitor a large number of small molecules with MRS allows a number of metabolic problems to be studied. We divide MR techniques into five sections. The first section discusses the attributes of magnetic resonance imaging relative to other modalities. It is difficult in the space provided to give a full description of MR techniques and the reader is referred to the excellent book by Gadian (30A) for further details.

A second section describes some applications of MRI to the study of transgenic animals. The obtainable resolution of <0.2 mm and gating capabilities of MRI make this modality attractive for cardiac function and anatomy. The third section discusses the unique attributes of MRS. Then, in the fourth section, we show some contemporary applications of MRS to the characterization of transgenic animals. Finally, there is a section emphasizing newer techniques for animal MRI. An earlier review on applications of MRI and MRS to transgenic mice considered three models in detail (47B). This review cites additional examples but is not meant to completely cover all applications.

Magnetic Resonance Imaging

The physical principle that forms the basis of MRI is the interaction of nuclei, which have a magnetic moment and angular momentum with a magnetic field. The nucleus of an atom such as the hydrogen atom of water has a positive charge and is considered as a charge spinning about an axis. This motion of the nucleus is equivalent to an electric current flowing in a loop of wire, and this generates a small magnetic field. The small fields line up with an externally applied magnetic field like a bar magnet would line up; however, the interaction of the angular momentum associated with nuclei that have spin (¹H, ¹³C, and ³¹P) with the magnetic field results in a precession in the same way a rotating top precesses about the Earth’s gravitational field. The frequency of this precession is specific for each nuclear species and proportional to the external magnetic field. Thus, for a field of 1.5 T (15000 G), the frequency is 63 MHz. The basic relation is ν = γ B , where ν is the resonant frequency of the precession oscillation, γ is the usual gyromagnetic rate constant specific to each nuclear species, and B is the field.

The quantum theory requires that, for hydrogen, two magnetic-moment values occur resulting in an alignment of the nuclear spins with and against the external field. The hydrogen nuclei are said to populate corresponding lower and upper energy states with a very small population difference (about 1 in 1 million). If energy is applied in the form of an external, rotating magnetic field associated with the radiofrequency (RF) pulse whose frequency exactly matches the precession frequency, then resonant absorption of energy takes place, and this boosts nuclei from the lower state to the higher state. After this applied field is turned off, the excited “spins” return to their equilibrium position after losing energy by relaxation processes (the T₁ and T₂ relaxation). The signal received from the relaxation is the magnetic resonance signal. The more nuclei or the higher the water concentration, the greater the signal. Because the difference in the excited and lower states is only 1 in 1 million, a strong signal requires a high concentration of nuclei. This is the case for hydrogen nuclei in tissue because the concentration is ∼90 M or about 10¹⁸ per mm³. The substantial signal from hydrogen is 10,000-fold that from phosphorous and sodium, which means that high-resolution imaging of nuclear species other than ¹H is problematic and requires an extremely high magnetic field because the signal does increase in proportion to field.

The time required for the nuclei to regain equilibrium after receiving a resonant RF pulse is dependent on the interaction of the resonant nuclei with the magnetic fields of the surrounding nuclei in the molecular matrix of the tissue. This is the T₁ effect; T₁ is about 600–800 ms in the brain at 1.5 T and 2000 ms in cerebrospinal fluid. A second relaxation time known as T₂ is related to the interactions of the nuclear spin with adjacent nuclear spins, which results in a loss of coherence or in-step rotation of each nuclear spin in the ensemble. This effect is noted in the rate of decay of the emitted signal after the RF pulse. Because these two relaxation times vary with the type of tissue and even the state of tissue oxygenation, it is possible to glean image contrast from analysis of the received signals after sending in RF pulses of prescribed durations and intervals.

The spatial information from MRI comes from the fact that the frequency of the signal varies with the local field. If the field through the patient is modified in a known direction and a known magnitude, then the received frequency will be proportional to the spatial position of the proton concentration in the magnetic field (frequency encoding). The method of varying the external field is through field gradients applied successively and rapidly in different directions. It is also convenient to vary the phase of the resonant frequency in a known manner by applying short pulses of magnetic field gradients that alter the resonance frequency by a small amount momentarily. This causes the phase of the signal to be altered in a spatially dependent manner (phase encoding). Finally, it is possible to selectively excite spins in a defined region of the sample by application of frequency selective pulses in magnetic field gradients (selective excitation). Varying the type of RF pulses and pulsed gradient fields in time and direction allows the acquisition of three-dimensional data using a combination of these methods. The ability to manipulate nuclear spins in so many diverse ways leads to the great richness inherent in MRI.

Thus the anatomical position of different tissues, mainly gray matter, white matter, and cerebrospinal fluid, are determined at resolutions of <1 mm by using routine MRI. In addition, it is possible to differentiate flow or motion in blood vessels and microscopic diffusion through observation of how the MRI signal changes after prescribed RF and gradient pulses. In vivo microscopy with MRI has sufficient resolution to evaluate the anatomy of small animals as well as that of embryos and cells to the 10-μm level.

Applications of MRI to Animal Studies

There has been growing interest in developing instrumentation specific for small animal studies. In principle, any MR instrument can be used for small animal imaging as long as it has the capabilities to manipulate magnetic field gradients and RF in the appropriate manners. Indeed, some of the earliest small animal in-vivo longitudinal studies used conventional chemistry NMR equipment wherein the shim coils provided the gradients for imaging (70). Vertical 5–9 cm bore magnets of 9.4 T–14 T used for chemical studies are also available for mouse studies. Animals have been imaged using instrumentation commonly available for human imaging such as 1.5 T–3 T, 100 cm bore MRI units. Animal imaging in a clinical environment will be further enhanced by the recent development of 7 T–8 T magnets for human use. The development of 19 T–22 T small bore magnets now in progress will allow small animal imaging at extremely high fields.

Perhaps the most useful MR systems have been instruments developed specifically for animal imaging which include the horizontal bore 4.7 T 40 cm bore magnets and 7 T magnets with up to 21 cm diameter bores. The new development of 9.4–11.7 T magnets with horizontal bores of 21–40 cm will substantially expand the applications of MRI and MRS to transgenic animal studies. Currently, the most common use of MRI in transgenic mice has been for cardiac imaging. Mice with 0.2 g hearts beating at rates over 350 bpm have been studied successfully with very accurate anatomical detail using very high field (7 T–9.4 T) with resolutions < 100 μm (35A, 73, 73A, 76A) (Figure 7). Conventional magnets at 1.5 T have been used also with capabilities for 200 μ resolution. In the horizontal bore 1.5 T systems the animals were positioned supine over a 4 × 8 cm endorectal coil and electrocardiograph (ECG) leads are placed on the fore paws and a hind paw. Multislice, multiphase cine MRI was used to collect a series of 1.6 mm thick short-axis slices which had an in-plane resolution of 190 μm (field of view is 50 mm and images are 256 × 256 pixels). The ECG gating is set for a rate of 250/min and 5 phases are collected by introducing a trigger based on the heart rate. The temporal resolution was 39 ms. Retrospective evaluation of the data sets was used to select the frame for the largest ventricle chamber and the smallest ventricle as diastole and systole respectively. In the experiments at 7T (76A) a similar approach was taken, however, faster gradients allowed 11 ms temporal resolution and data covering the whole heart were used to analyze cardiac mass. Different acquisition schemes have allowed even faster temporal resolution (35A). Indeed, there do not seem to be any obstacles to cardiac imaging at physiological heart rates for the mouse even under dobutamine stress (35A, 73A, 97A). It should be straightforward to implement the full range of functional cardiac imaging techniques to assay wall motion and regional blood flow in the mouse heart.

One of the first demonstrations of the capability of MRI to quantify the size of the heart chambers and wall thickness in the living mouse was the study of induced cardiac hypertrophy by aortic constrictions in the CD-1 mouse 1–3 months before imaging (76). Four 1-mm-thick anatomic levels were acquired on a 9.4-T, wide-bore GE/Omega magnet, using an EKG-gated spin warp acquisition sequence with an echo time of 18 ms and a field of view of 40 mm with a 256 × 256 matrix (resolution of 0.16 mm). Four minutes were required for each level. This work demonstrated the feasibility of MRI as an alternative to transthoracic echocardiography. MRI has a greater accuracy for three-dimensional quantitation than conventional emission imaging techniques. In addition to aortic constriction induction of hypertrophy, MRI has followed hypertrophy induced by adrenergic agents (76A).

Transgenic models of hypertension are proposed to mimic human hypertension better than previous models, which use intravascular volume expansion by salt, renal ablation, or constriction of the aorta. MRI or ultrasound evaluation of the left ventricular mass and function of 30-g animals with 0.2-g hearts allows serial evaluations without the necessity for necropsy and a large number of animals. The inherent three-dimensional capabilities of MRI make it an accurate technique for heart mass quantification.

The results of a cardiac function MRI study of the knock-out mouse with a dysfunctional guanylyl cyclase A gene compared with the wild-type siblings are shown in Figure 8 (26). The knock-out mice develop hypertension and cardiac hypertrophy associated with their inability to transduce the signals from atrial naturetic peptide. At 8 months the left ventricular mass of the knock-out mice was on average 50% greater (226 mg) than the left ventricular mass of the normal animals (156 mg), but the ejection fractions of 75% and chamber volumes were similar. This study, while successful in demonstrating feasibility, can be improved in terms of spatial and temporal resolution if a higher-field magnet with improved gradient magnitudes and rates is used.

A model of familial hypertrophic cardiomyopathy is transgenic mice with an arginine to glutamine missense mutation of β-cardiac myosin chain (76B). Perfused isovolemic heart function and MRS studies of bioenergetics have been performed. These studies could also be performed by MRI functional studies similar to those cited above.

MRI was used to show the evolution of lethal myocarditis in mice overexpressing tumor necrosis growth factor alpha (TNF-α) (48). This is an extremely relevant model for studying congestive heart failure mechanisms. TNF-α is a proinflammatory cytokine that has been found in end-stage congestive heart failure, and it is believed that the cardiomyocytes are a major source of this cytokine. TNF-α exposure causes a substantial decrease in cardiac contractility. The transgene construct was made, containing the murine α-myosin heavy-chain promoter with complementary DNA to murine TNF-α. The animals were studied at ages of 9–24 weeks by using a Bruker 7-tesla spectrometer modified to acquire images. Eight oblique 1-mm-thick slices were obtained by spin echo imaging and R wave gating, such that images of end diastole to mid-systole were obtained. Because the field of view was 15 mm and the matrix size was 128 × 128, the in-plane resolution would be 0.12 mm. The MRI image data allowed quantification of a progressive increase in end-diastolic and end-systolic volumes with age in these transgenic mice. The corresponding decrease in ejection fractions could be calculated, and the response to isoproterenol shown to decline. This model and the in vivo MRI monitoring are excellent examples of MRI functional imaging of small animals by using contemporary NMR resources.

A transgenic mouse (beta-S-alpha-knockout) has altered red-cell destruction, hematological changes, and progressive organ damage. This model has great potential for evaluation of sickle cell disease therapy, including gene therapy (25), and can be compared with other transgenic mouse models of sickle cell disease by oxygen-induced MRI (25). MRI spin echo imaging at room air and 100% oxygen demonstrated the presence of blood hypoxia through the detection of paramagnetic deoxygenated hemoglobin in images of liver and kidneys.

A strain of mice that express a myelin-basic, protein-specific T-cell receptor spontaneously acquires experimental allergic encephalomyelitis, which is considered an animal model for the human demyelinating autoimmune disease and multiple sclerosis. The pathology of experimental allergic encephalomyelitis shows inflammatory infiltrates of T cells, B cells, and macrophages in white matter meninges and perivascular regions. The transgenic mouse spontaneously acquires the disease in some environments (34), otherwise the disease is induced with injection of myelin-basic protein. Because the white matter of the central nervous system including the spinal cord is so severely affected, the proton magnetic resonance signal is changed from that associated with normal water environments and water diffusion in subhuman primate models (71) and rodents (59), and these features are similar to those seen in various stages of human multiple sclerosis. The general features have been related to the T2 relaxation time effects in most human and animal models; however, diffusion imaging has also been shown to give information independent of that revealed by T2 changes (89).

In the first paper that reported the appearance of white matter lesions in the transgenic mouse (1), diffusion tensor imaging, in addition to other forms of MRI, was used to evaluate the spinal cord at very high resolution (20 × 20 × 300-μm-thick slices), using an 11.7-T, 9-cm-bore Bruker magnet equipped with gradients of 290 G/cm and ex vivo samples of the spinal vertebrae.

In vivo MRI of this animal model would not achieve the 20- to 30-μm resolution of MRI of the ex vivo samples owing to magnetic susceptibility, RF inhomogeneity, and animal motion associated with breathing and cardiac motion. But the disease progression and possible response to proposed therapies could be monitored in this transgenic animal by similar MRI methods applied to the brain. This work established a transgenic animal that can be studied noninvasively to evaluate the progression of disease by new MRI methods of visualizing the direction, diffusion anisotropy, and organization of neuronal fibers.

The potential of diffusion tensor imaging for visualizing experimental allergic encephalomyelitis pathology originates from the substantial diffusion anisotropy that exists in white matter fibers of the central nervous system (36, 56, 60, 61, 68, 82). These studies set the foundation for future longitudinal in vivo studies in this transgenic mouse and demonstrate that diffusion tensor imaging may offer intriguing new possibilities for visualization of white matter organizational changes in growth and aging in other transgenic and eugenic models.

Diffusion weighted MRI has also found widespread application to visualizing defects in brain due to focal ischemia (98A). Nitric oxide has been implicated to increase tissue damage when produced at high levels. Wild-type mice and mice deficient in neuronal nitric oxide synthase were imaged at 2 T after permanent middle cerebral artery occlusion (104). An apparent diffusion coefficient (ADC) was quantitated from MRI data. Histological staining of ischemic regions correlated best with ADC changes of greater than 25%. The area of the brain exhibiting ADC changes greater than 25% was the same in both wild-type and mice lacking neuronal nitric oxide synthase. Interestingly the volume of tissue showing ADC changes between 10 and 25%, defined as the peri-infarct zone, was approximately 25% smaller in the knockout mice than in the wild-type mice. These results are consistent with the notion that increased production of nitric oxide associated with focal ischemia causes metabolic disturbances that lead to more tissue damage.

Magnetic Resonance Spectroscopy

MRS gives information on the chemical composition of a sample or a specific region in the body. It provides a method for evaluation of neurochemical composition in localized regions (e.g. choline, choline-containing compounds, creatine, N-acetylaspartate, N-acetyl–containing compounds, glutamate, glutamine, gamma aminobutyric acid, lactate, etc). The principles of magnetic resonance discussed above state that the frequency of a particular element will depend on the magnetic field. But the field at a specific nuclear site in a molecule is usually slightly different from the applied external field owing to the local magnetic field from the orbiting electrons (Figure 9). The net effect is to shield the nucleus somewhat from the external field, and this shielding will depend on the position of the nucleus in the molecule. For example, the resonant frequency for phosphorus at 1.5 T is 25.9 MHz, but the actual frequency for the three phosphorus nuclei in ATP and the one phosphorus of creatine phosphate will be slightly different from inorganic phosphorus owing to the different electronic- or magnetic-field environments associated with the molecular positions, as shown in Figure 10, where these differences in frequency are measured as parts per million relative to a standard.

As discussed above, the population difference between the nuclear spins in the upper and lower energy states is only about 1 in 10⁶. Because the concentration of most metabolites in tissue is in the millimolar range as opposed to hydrogen, which is in the 100 M range, the signal from the different phosphorous nuclei is about 10⁴-fold less than the proton signal used in imaging. To compensate for this low sensitivity and to achieve a wide dispersion of the individual spectral components, it is necessary to use higher fields available on magnets produced for biomedical and chemical studies. In human subjects at 1.5 T, meaningful studies of the brain composition of choline-containing compounds, creatinine, and N-acetyl-aspartate can be accomplished in 0.2-cc volumes (in 6- × 6- × 6-mm regions) because the proton (hydrogen nucleus) signal is much greater than that from carbon and phosphorous compounds at the same magnetic field. The signal-to-noise ratio is proportional to the volume, the field, and the square root of the acquisition time.

The three most popular nuclei for MRS in vivo have been ³¹P, ¹H and ¹³C (47A). ³¹P MRS gives a wealth of bioenergetic information including measurement of ATP, phosphocreatine, Pi, pH and Mg⁺² and has been applied extensively to study muscle and heart. Thus, it is no surprise that ³¹P MRS has been applied to a wide variety of transgenic mice. ¹H MRS has found widespread application to brain metabolism studies because of the ability to detect a large number of metabolites important to brain function. ¹³C MRS offers the advantages of being able to study metabolic fluxes by infusing specific ¹³C labeled compounds. Recently this has been used to assess glucose metabolism in the brains of mice overexpressing phosphofructokinase (66).

Applications of MRS to Studies of Transgenic Animals

Creatine kinase (CK) is normally found at high levels in muscle and brain, and catalyzes the reaction: phosphocreatine (PCr) + MgADP + H⁺ ⇔ creatine (Cr) + MgATP. High levels of CK ensure that PCr is rapidly converted to ATP during periods of increased ADP levels, thus enabling PCr to serve as a high energy phosphate reservoir.

CK is not normally found at high levels in liver. The CK isoenzyme family has a multifaceted role in cellular energy metabolism and is characterized by a complex pattern of tissue-specific expression and subcellular distribution (88).

The driving force for the energy-requiring process of mammalian metabolism is generated through ATP hydrolysis by the appropriate ATPase (enzyme that releases energy from hydrolysis of ATP), but the local ATP/ADP ratio is kept high by ATP regeneration via the nearby bound CK enzyme. The maintenance of high ATP/ADP ratios, even during periods of high cellular activity, is of thermodynamic and kinetic importance because it aids in keeping both the free energy of ATP hydrolysis and the ATPase activity high.

³¹P magnetic resonance is also able to measure the in vivo flux through the CK reaction, by using magnetization transfer techniques. Such methods use selective perturbation of either the PCr or the γ-ATP peak in the ³¹P MRS spectrum with a frequency-selective radio frequency field that leads to a modification of the intensity of the resonance concerned. Via the chemical exchange reaction catalyzed by CK, the frequency-selective perturbation is partially transferred to the exchanging partner: PCr when the irradiating field is at the γ-ATP frequency and γ-ATP when irradiating at the PCr frequency. This magnetization transfer experiment enables the quantitation of the forward and reverse flux through the CK reaction, as explained in Figure 11 (see color figure). The CK system of the mitochondria greatly facilitates the diffusion of free-energy equivalents through the cell and thereby also accelerates and smoothes transitions between different work states and dampens oscillations in the concentrations of ATP and ADP (101). Applications of MRS to evaluation of mutants with modified CK enzyme systems are discussed next.

Recently developed transgenic-mouse models of the CK isoenzyme system open novel opportunities to assess the functional importance of the individual isoenzymes and their relative contribution to the total in situ flux through the CK reaction (63).

Five different CK isoenzymes (1 mM concentration) have been identified in mammalian tissues. In adult striated muscle, the most abundant CK species is in the cytosol (MM-CK). In the brain and many other tissues, the cytosolic CK isoform is BB-CK. The B-type isoform abundant in infant muscle largely disappears in adulthood. Mitochondrial CK (Mi-CK) is located in the intermembrane space. In muscle, part of the MM-CK is freely dispersed in the cytosol whereas another part is associated with various subcellular structures (Figure 11; see color figure).

In nonmutant wild-type tissue, all CK isoenzymes may be expected to contribute to the measured CK flux. A differentiation of the relative contributions of the mitochondrial and the cytosolic CK species has not been possible until recently. Transgenic models of CK enable separate measurement of the cytoplasmic and mitochondrial components of the CK system in the in vivo situation.

To understand the physiological role of the CK-PCr system in muscle bioenergetics, a null mutation of the muscle CK (M-CK) gene was introduced into the germline of mice. Mutant mice show no alterations in absolute muscle force, but lack the ability to perform burst activity. Their fast-twitch fibers have an increased intermyofibrillar mitochondrial volume and an increased glycogenolytic/glycolytic potential. PCr and ATP levels are normal in resting M-CK–deficient muscles, but rates of high-energy phosphate exchange between PCr and ATP are at least 20-fold reduced. Strikingly, PCr levels decline normally during muscle exercise, suggesting that M-CK–mediated conversion is not the only route for PCr utilization in active muscle (87). Mice that are deficient in Mi-CK and lack both MM-CK and Mi-CK isoforms have been generated and the skeletal muscle studied (77). In MM/Mi-CK-/-, the skeletal muscle defects found in MM-CK-/- were more severe.

³¹P-NMR measurements of the flux through the CK reaction were carried out on intact skeletal and heart muscle from wild-type mice and from transgenic mice, homozygous for a complete deficiency of the muscle-type cytoplasmic CK isoform (MM-CK-/-). Skeletal muscle and heart muscle were compared because they differ strongly in the relative abundance of the Mi- and M-CK isoforms. The kinetic properties of cytoplasmic and Mi-CK are substantially different (88).

³¹P-NMR spectra were obtained from control and transgenic muscles overexpressing B-CK during mechanical activation, and there were no NMR-measurable differences detected. These results indicate that a 50% increase in CK activity owing to expression of the B subunit does not have large effects on skeletal muscle metabolism or contractile function. Therefore, control muscle has sufficient CK activity to keep up with changes in cellular high-energy phosphates except during the early phase of intense contractile activity because the rise time as an isometric contraction was 28% faster in transgenic muscle (72).

Heart studies in knockout transgenic mice (e.g. MM/MiCK-/-; and MiCK-/-) were studied for contractile function and energetics by MRS (74C). Despite a similar contractile response, MM/MiCK hearts increased [ADP] by 95%, whereas wild-type and MMCK-/- hearts maintained [ADP] at baseline levels. In contrast to what has been reported in skeletal muscle, MM/MiCK-/- hearts were able to hydrolyze and resynthesize phosphocreatine.

Both the BB-dimer of CK (CKB) (47, 54) and the Mi-CK (55) have been expressed in the livers of transgenic mice. It has also been demonstrated in transgenic mice expressing CK in the liver that the production of PCr in animals fed creatine protected ATP from falling during fructose infusion and during hypoxia (10, 54). ³¹P-NMR spectroscopy demonstrated that the expression of Mi-CK in the liver resulted in animals that can use phosphocreatinine

³¹P-NMR spectroscopy was also used to study the energy metabolism of dystrophin-deficient skeletal muscle of mdx mice, an animal model of Duchenne muscular dystrophy, in which expression of a truncated form of utrophin (trc-utrophin) has been obtained through transgenesis technology (33). The mdx muscles showed a more severe intracellular acidosis during exercise and a slower and incomplete postexercise recovery of normal intracellular pH (pHi). In contrast, in trc-utrophin muscles, the kinetics and amplitude of pHi changes were remarkably close to normal behavior. Those studies showed that the impaired proton washout that is present in mdx muscles is corrected to a great extent by the expression of trc-utrophin.

One of the most recent examples of noninvasive functional imaging with MRS is the evaluation of cerebral metabolism in a transgenic animal having some attributes of Down’s syndrome. The gene for the liver-type subunit of phosphofructokinase (PFKL) resides on chromosome 21 and is overexpressed in Down’s syndrome patients. Transgenic-PFKL mice with elevated levels of PFKL had an abnormal glucose metabolism with reduced clearance rates from blood and enhanced metabolic rates in brain. By infusion of [1-¹³C]glucose followed by in vivo NMR measurements of brain glucose metabolism, the initial rate of glucose utilization in the brain of the transgenic mice was 58% faster than that in control non-transgenic mice. The faster utilization of glucose in transgenic-PFKL brain is similar to the increased rate of cerebral glucose metabolism found in the brain of young-adult Down’s syndrome patients, which may play a role in the etiology of their cognitive disabilities (66).

MRS of the brain in transgenic mice was used to study the possible role of putrecine in development of neuronal damage in forebrain ischemia. Cerebral ischemia causes activation of ornithine decarboxylase followed by accumulation of putrecine, and it is not clear whether putrecine is protective or neurotoxic. To address this issue, transgenic mice overexpressing ornithine decarboxylase were created (36a). These mice had 70-fold higher levels of ornithine decarboxylase and greatly increased levels of putrecine. No significant changes in a number of metabolites measured by ¹H or ³¹P MRS were detected except for a 40% lower level of Mg⁺² as measured by chemical shift of the ATP peaks (46). These mice demonstrated an elevated threshold to both chemically and electrically induced seizures indicating that increased putrecine levels were not damaging to the brain and in some circumstances were protective (36B). The effects of elevated putrecine on ischemia was studied in these transgenic mice.

Cerebral high-energy metabolites, intracellular pH, and lactate were monitored by means of ³¹P- and ¹H-NMR spectroscopy, respectively. Changes after 15 min of carotid occlusion included an increase in the inorganic phosphate/PCr ratio, intracellular acidification from a pH of ∼7.1 to ∼6.5, and an increase in lactate concentration from <1 to ∼10 mmol/kg in both control and transgenic mice. After deocclusion, recovery of energy metabolites, normalization of intracellular pH and lactate were similar in both animal groups. Ornithine decarboxylase activity rose nine- and threefold in nonmutant and transgenic mice, respectively, 6 h after ischemia. These results indicate that elevated levels of putrescine might be playing a protective rather than a neurotoxic role. (52).

New MR Techniques for Small Animals

The increase in interest in imaging small animal models in MR is leading to new developments specifically to image small animals. In particular, there is growing interest in imaging gene expression and cell migration.

Perhaps the first work that explicitly demonstrated that MR could detect the expression of a gene was the detection of the PCr resonance in livers of transgenic mice expressing CK (10). The expression of creatine kinase in liver has had little adverse health effects suggesting that CK might be a suitable reporter protein for MR imaging of gene expression in humans. The ability to accurately quantitate the amount of CK in a tissue using magnetization transfer techniques means that the amount of reporter expressed can be quantitated and that increased levels of CK can be detected even in tissues which normally express high levels of CK. The major drawback in using CK as an MR gene expression reporter is that ³¹P MRS detection of PCr is not very sensitive so only images with resolution of 0.5 cm can be obtained.

An alternative approach is to affect contrast in normal ¹H MRI by expression of reporter genes. Preliminary results indicate that overexpression of the transferrin receptor to increase cell iron levels may be a useful approach (47D). Excess iron is deposited in ferritin in cells, and ferritin is known to be a good MR contrast agent. Furthermore, contrast agents specific for detecting the transferrin receptor have been developed and shown to be useful for detecting expression of the transferrin receptor in cultured cells (56A). The transferrin receptor may be a useful MR reporter gene to change endogenous contrast in tissues or to use exogenous agents to detect expression.

Another new approach is to design contrast agents that are locally activated by expression of a specific enzyme. A gadolinium chelate has been designed so that a carbohydrate moiety shields the gadolinium from surrounding water rendering it an ineffective MRI contrast agent. Upon exposure to the enzyme β-galactosidase the carbohydrate moiety is cleaved, enabling water proton access to the gadolinium, thus turning it into an effective contrast agent. Thus contrast is changed locally wherever β-galactosidase is expressed leading to a potentially effective strategy for imaging the expression of β-galactosidase. This is a very general approach to contrast agent design and it may be possible to design agents that are activated by other gene expression reporters (55A), a new class of agents capable of both transfecting genes into cells and enhancing the contrast of the targeted cells for MRI. DNA is used to encode a marker gene and as a molecular scaffold, which electrostatically binds polylysine conjugated to transferrin, an iron uptake protein, and polylysine modified with gadolinium chelated to diethylenetriaminepetaacetic acid (46A).

A number of important physiological and pathophysiological processes depend on cell migration. Being able to image the migration of cells in a living animal would have widespread application to monitoring developmental processes and immune cell function. The general approach by MRI to cell migration has been to load specific cells with MRI contrast material. This has been done by direct injection of contrast agents into specific cells in developing animals (45A), or by isolating cells and loading them with MR contrast agents (97B, 101A). By isolating T-cells and loading them with a contrast agent and re-introducing these cells into animals, migration of T-cells to sites of inflammation has been detected by MRI (101B). Using T-cells loaded with iron oxide particles in phagocytic vesicles, it was possible to detect the presence of single cells in MRI with 50–100 micron resolution (22A). These results clearly indicate that MRI is poised to have a major impact in problems that require visualization of cell migration in live animals.

Optical Transmission and Emission Imaging

Light can be used to probe the function and structure of tissues by a variety of approaches with similarities to X-ray CT, ultrasound, and emission imaging. The methods that have relevance to imaging small animals are listed here in five categories.

This section reviews the attributes of optical imaging, including transmission tomography, fluorescence imaging, coherent optical tomography, and bioluminescence.

Near-Infrared Spectroscopy

In laser optical tomography or near-infrared spectroscopic tomography, data are collected by applying a laser source at one or different locations around the object of interest and then detecting the light emitted in one or several locations to determine certain characteristics of the medium transversed by the beam. The method is analogous to X-ray CT. Light in the near-infrared range (wavelength from 700 to 1200 nm) penetrates tissue and interacts in many complicated ways, with the predominant effects being absorption and scattering, but coherence and polarization are also changed. Light transmitted through a turbid medium comprises ballistic, snake, and diffusive components (91, 102), which are displayed schematically in Figure 12 (see color figure). Ballistic photons scatter coherently, and snake photons scatter only slightly in the forward direction. Both travel in the incident direction and arrive sooner than the multiple-scattered diffuse photons. These early-arriving, forward-propagating photons carry image information and do not obey the diffusion equation. The difference in transit time and changes in the characteristics mentioned above are exploited for selecting the image-bearing light. Because the image-bearing photons change the least, the idea is to devise a gate that will let the photons with a specific initial property through but block others. Realization of this idea has led to the development of space, time, polarization, nonlinear optical, coherence, and absorption gates. A space gate exploits the fact that the ballistic and snake photons come out of the tissue in the incident direction, whereas the multiple-scattered light emerges in all directions (74A, 92). A small aperture centered on the line of incidence and placed after the sample can effectively discriminate against most of the off-axis scattered light. A particular embodiment of the space gate is the Fourier gate (74A, 92), wherein scattered photons are masked from the image by an aperture in the lens back focal plane.

At resolutions in the range of 10–40 mm, there are many applications to human clinical situations, but application of near-infrared spectroscopy to small animals is problematic. Differential absorption of light or infrared spectroscopy has allowed imaging of intracranial hemorrhage in newborns and of brain functions in adults. Optical detection of hematoma has been shown to benefit adults after surgery for head trauma (32), and transmission optical imaging has been used successfully for identification of cerebral hemorrhages in neonates at increased risk for such bleeding. The method is known as near infrared spectroscopy, wherein two wavelengths are used to quantitate differences in the amount of hemoglobin in blood or tissues. A practical embodiment of the use of spectral imprints for identification of tissue composition is the motionless tomographic optical system, consisting of an imaging headband constructed from a web of fibers sewn into a soft band (5). This band can be wrapped around the head for extended periods without discomfort or difficulty. Other applications in which optical tomography in its present state has started to yield useful results, and where high spatial resolution is not required, include the measurement of tissue oxygenation for the study of muscular dystrophy, assessment of tissue perfusion in the extremities of diabetics, and the study of brain activity during specific tasks.

Historically, optical-concentration estimates have been rendered nonquantitative by the long, irregular photon paths caused by scattering. Unfortunately, in-vivo optical images generated by simple transmission of light have poor resolution despite the many innovations to improve data collection and mathematically reconstruct spatial distributions of attenuation and diffusion (2).

Fluorescent Optical Imaging

Many substances, such as hemoglobin and cytochromes, exhibit characteristic absorption spectra that depend on whether the molecule is in its oxidized or reduced state. Indocyanine green has been used to detect vascular permeability in animal tumors (35A). NAD/NADH exhibits fluorescence properties that allow for their detection after excitation by light. As these substances play crucial roles in metabolic processes at the cell level, the ability to discern them through indirect measurements is a biomedical tool of great potential.

Contrast agents such as indocyanine green, which absorbs at 800 nm and fluoresces at a slightly shorter wavelength, can be detected in very small amounts. The fluorescent properties of certain dyes depend on the intensity of the excitation and biological parameters such as oxygenation state and local electric potential. For example, when a cell in contact with other cells is electrically stimulated, the evoked excitation neural cell network can be visualized by using a voltage-sensitive fluorescent dye (e.g. RH795). The optical techniques have been limited to ex vivo tissue preparations (37), but it is feasible to consider minimally invasive endoscopic approaches to physiological measurements in small animals, using fiber optic systems. Fluorescent-imaging approaches to small animals will find applications with endoscopic probes particularly in evaluation of atherosclerosis.

Green Fluorescent Protein Tags of Gene Expression

Green fluorescent protein (GFP) and fluorescent imaging have been used successfully in a variety of experiments, but strictly noninvasive in vivo imaging with high-sensitivity charge-coupled device (CCD) cameras is still in development. Single muscle fibers of intact mice have been made transgenic by intracellular injection of DNA expression vectors for the reporter gene lacZ and GFP. In vivo imaging allowed identification of single cells during and after injections, when the muscle was re-exposed days or weeks later. The GFP was still detected 61 days after injection (86). It has also been shown that GFP reporter can be used to monitor dynamic growth behaviors of individual olfactory-neuron growth cones (23). The potential for GFP to monitor gene therapy is evidenced by the following GFP reporter experiment. Expression of a fluorescent reporter gene has been studied with two alternate promoters to transcribe the GFP from Aequorea victoria (4). The human cytomegalovirus enhancer/promoter or the human muscle-specific CK promoter (CKMCKM) were inserted along with the GFP complementary DNA into a plasmid expression vector based on a modified adenoassociated virus genome. The plasmid DNA was injected into the hamstring muscle of mdx mice, and the expression of the gene which produces GFP was determined from fluorescence patterns obtained by quantitative fluorescence measurements on frozen tissues. Those measurements indicated a near-linear increase in the accumulation of the GFP in skeletal muscle during the 42 days of the study. Those experiments were designed to show the difference in long-term expression from the CKM vs the cytomegalovirus constructs, but they also encourage development of minimally invasive methods for in vivo fluorescence measurements to monitor gene therapy by fluorescent tags whose stability and indefinite lifetimes have advantages over radionuclide tags.

Optical Coherence Tomography

A particular form of optical imaging known as optical coherence tomography (OCT) was originally developed to image the transparent tissue of the eye (39) and, more recently, using near-infrared OCT has been shown to image nontransparent tissues (7, 9). OCT is analogous to ultrasound B-mode imaging, wherein infrared light is used rather than acoustical waves. Low-coherence interferometry is used to generate cross-sectional images of biological tissues. A beam of infrared light is directed onto the tissue, and the reflected or back-scattered intensity from microstructures is recorded as a function of depth. Because the technique is limited to tissue depths of only a few millimeters, the applications to tissues other than eye and skin are through endoscopy, with good feasibility studies for noninvasive evaluation of tissues with 4-μm resolution (29, 81). The capability of OCT for transgenic animal studies relies in good part on the use of endoscopic methods for in vivo studies.

Chemiluminescence and Bioluminescence

External detection of internal sources of light with very sensitive CCDs as photon-concentrating devices (38) has now been demonstrated in studies of chemiluminescence and bioluminescence of pathogens (17, 18) and the temporal and spatial expression of mammalian genes (18, 99), as well as expression in Drosophila (8) and fish (80).

The major tool for internal emissions is the luciferase enzyme system. The firefly produces its characteristic glow by combining energy-rich chemicals with a luciferase enzyme that liberates this energy to produce light. Many single-cell organisms use such enzymes to make light, and many different forms of these enzymes exist. The chemical pathways that allow this reaction to occur exist on genes. These genes can be moved onto the cells of a nearby organism. Sets of light-emitting genes from soil bacteria and fireflies can be moved into other cells to create new bacteria and animal cells that emit light. These new cells glow, much as a firefly does, and can be detected and quantified with cooled or intensified video-type CCD cameras in dark rooms. The use of luminescent markers placed in vivo provides a means of monitoring the course of infection and gene regulation in animal models of human diseases.

Noninvasive study of a bacterial infection in vivo is enabled by light-emitting luciferase (17). Unlike standard labels, light emission is linked to bacterial metabolism, and therefore follows the infection. As the gene is incorporated into the bacterial DNA, it provides a label that persists over many bacterial generations (Figure 13; see color figure). Infection of mice with Salmonella typhimurium, producing a well-characterized animal model of a human disease, typhoid fever, is a model for development of this bioluminescent approach. Salmonella cells were labeled through the expression of luciferase and accessory proteins obtained from the soil bacterium Photorhabdus luminescens (also called Xenorhabdus luminescens). Animals infected via direct injection into the peritoneal cavity with the virulent strain of S typhimurium (SL1344) demonstrated widely distributed bioluminescence, suggestive of a disseminated systemic disease. Animals infected orally with different strains showed different patterns of intestinal infection, such that progressive courses of infection appeared distinct from those that were persistent or abortive. The distribution of bioluminescence in persistent and progressive infections suggested that the cecum might be a critical site for maintenance of Salmonella infection, which, under some conditions, can then lead to systemic disease. The effect of treatment with the antibiotic ciprofloxacin was noninvasively monitored in real time, with a rapid reduction of signal noted within minutes of antibiotic treatment (Figure 13).

Bioluminescence has been used to track gene activation in living tissues without disturbing the native tissue state (18). In a first model system, a human T-cell line (Jurkat cells) had been earlier modified with an HIV-1 promoter gene fusion to emit light in response to intracellular replication of HIV-1 (98). After an HIV-1 inoculation of these Jurkat cells, it was demonstrated that replication in cell culture could be detected and imaged.

In an animal model system of gene expression, the promoter of HIV-1 was used. This promoter was selected because of its known expression in the skin when present in cells of transgenic mice (57). Earlier this promoter had been placed in front of the firefly luciferase gene (57).

The standard bearer for use of bioluminescence in transgenic mice is the use of transgenic animals carrying the HIV-1 long terminal repeat regulating the expression of the firefly luciferase reporter gene (58). Although in vivo imaging has not been used yet, the potential for this transgenic animal as a model for activation of the long terminal repeat and evaluation of reporter genes in the skin is great. The firefly luciferase reporter gene is highly sensitive (20). Reporter gene expression was increased in the skin by topical treatment with dimethylsulfoxide, retinoic acid, phorbol ester, UV light, and hydrogen peroxide, all of which have previously been shown to cause increased HIV production in cultured human cells. Other topically applied chemicals not known to increase gene activities did not increase the skin activity. The combination of this transgenic animal model and sensitive CCD camera technology has promise for in vivo studies.

Conclusions

Imaging methods currently used for the noninvasive studies of human disease have been adapted to small-animal studies to characterize transgenic animal phenotypes, gene expression, pharmaceutical distribution, and mammalian function and structure during good health and disease. The methods of radionuclide imaging have been modified by the introduction of new instrumentation to achieve the resolution and sensitivity needed for small-animal imaging. Magnetic resonance methods of imaging and spectroscopy have found effective use for in vivo studies of transgenic mice in the study of cardiac muscle and transgenic animals with engineered changes in CK isoforms. Optical techniques of fluorescence and bioluminescence have shown promise for in vivo studies of gene expression and gene therapy. Recognition of specific limitations for each modality has led to clever approaches to overcome animal imaging goals of less than 100-μm resolution for MRI, in vivo radiopharmaceutical studies of neuroreceptors in mice, and whole-animal bioluminescence imaging at nearly the single-photon detection sensitivity.

ACKNOWLEDGMENTS

This work is supported by the National Institute of Heart, Lung, and Blood Program Project Grant and the Department of Energy Office of Biology and Environmental Research (AC03–76sf00098). The illustrations were created by K Brennan and additional illustrations were provided by S Cherry, S Rose, R Peshock, and Concorde Microsystems.