In Vivo Small Animal Imaging: A Comparison to Gross and Histopathologic Observations in Animal Models

2.1

Basic Principles of MRI/MRM

Both MRI and MRM are based on the principles of nuclear magnetic resonance (NMR), a spectroscopic technique used by scientists to obtain structural information for molecules. NMR is a phenomenon of absorption and emission of energy in the radio frequency range of the electromagnetic spectrum by certain atomic nuclei when placed within a magnetic field. Clinical MRI uses magnetic resonance of hydrogen atoms to obtain detailed anatomical information in living tissue. High-field preclinical MRI/MRS can also use different nuclei such as ¹³ C, ²⁹ Na, ³¹ P, ¹⁹ F, etc., to obtain unique biological information in animal models of human diseases.

An MRI scanner consists of the several key components. The first is a permanent magnet capable of producing a strong static magnetic field. Typically, superconductive magnets produce magnetic fields in the range from 1.5 to 3T (Tesla) for clinical imaging and up to 17.6T for preclinical MRI. The second constituent is a radiofrequency (RF) system consisting of transmitter(s) and receiver(s) that excites the nuclei in the sample and receives the emitted signal; an RF coil or arrangement of multiple RF coils is used to measure signals from the whole body or a sample region to be scanned. Finally, the scanner features a gradient system that includes three orthogonal controllable magnetic field gradients to encode the spatial information required to form an image. The strength of the magnetic field is one of the most crucial factors dictating image quality. Higher magnetic fields enhance signal-to-noise ratio (SNR), allowing higher spatial and temporal resolution with shorter scanning times for rodent models.

Many institutions have imaging core facilities with MRI equipment dedicated for small animal imaging, with higher magnetic field strengths ranging from 3 to 11 T. In comparison, clinical MRI facilities for human patients have significantly larger available diameter and provide a lesser resolution within a large field of view. For animal imaging, just as in human imaging, specially designed RF coils are placed around the whole body or a specific region of the body that is to be scanned. The inherent isotropic three-dimensional (3D) nature of MRI allows detailed analysis of organs and retrospective studies through any arbitrary plane within the object, whereas individual two-dimensional (2D) planar images can also be recombined into 3D data sets for volumetric rendering. Moreover, MRI has the ability to obtain exquisite contrast between various types of tissue (e.g., gray vs. white matter in the brain) by exploiting the inherent differences in their tissue-specific properties, such as relaxation time, water/fat ratio, diffusion, spectral signatures, etc., by using specific MRI image acquisition modes (“pulse sequences”). The most widely used properties are the longitudinal relaxation time (T ₁ ) and the transverse relaxation time (T ₂ ). Depending on the pulse sequence, either T ₁ or T ₂ contrast can be emphasized in the image, producing MRI images with differential enhancement of contrasting soft tissues. When applied to excised tissue samples, these contrast enhancement methods have been coined “proton staining,” as the unique proton contrasts provided by MRM enable direct examination of the state of water protons in tissues ( ).

One of the most important determinants of the SNR is the choice of the RF coil. High-quality coils are made up of multiple smaller coil elements (i.e., array coils). Superconductive RF coils can also improve SNR for micro-MR imaging. However, in spite of these instrument innovations, low signal sensitivity issues continue to be a hurdle for some rodent MRI studies.

The signal sensitivity and contrast can sometimes be improved by the use of contrast agents in MR imaging. The use of contrast agents based on gadolinium (Gd) or iron oxide particles can influence the relaxation times of water molecules in their immediate vicinity. Iron oxide nanoparticles have been used to track cell migration, angiogenesis, apoptosis, and gene expression in various organs. Targeted (“smart”) contrast agents have been developed to focus particles to specific cell-surface receptors using conjugated antibodies or smaller high-affinity molecules, such as “affibodies.” For example, molecular imaging of HER2 can be undertaken successfully using a novel class of multiple modality imaging affibodies that include a Gd ³⁺ -binding protein or iron oxide nanoparticles and near-infrared (NIR) fluorescent probe Cy5.5 ( ; ). The protein-based MRI contrast moiety (ProCA1) agent greatly enhanced the MR imaging, was specific for HER2, and required less Gd ³⁺ , thus reducing the chances for nephrotoxicity ( ).

Diffusion-weighted MR imaging (DWI) is an MRI method that is capable of detecting the extent of apoptosis and potential edema without using a contrast agent. DWI may also be utilized to monitor changes in tissue volume and/or fluid content caused by cell swelling, tumor shrinkage, necrosis, hydrocephalus, and other processes that can occur in the course of therapy. In addition, diffusion tensor imaging can be used to image microstructure of white matter and other organs based on spatially anisotropic diffusion of water molecules along tissue fibers. Alternatively, apoptosis has been detected by MRI using contrast agents based on annexin V conjugated to nanocrystals of iron oxides, Gd-containing liposomes, or quantum dots (QDs) with a paramagnetic lipid coating. Cell death induced by domoic acid, carbonyl sulfide (COS), or acetaminophen can be easily detected by MRI ( ; ).

The images obtained with MRM are often correlated with gross or histopathologic findings. MRM shares the same physical principles as MRI. The delineation of high-resolution MRI and MRM is somewhat unclear, but does relate to the higher resolution provided by MRM. To obtain the higher resolution, MRM requires much stronger magnetic gradients, typically about 50- to 100-fold higher than those of most clinical imaging systems. MRM can be utilized to image both live and fixed tissues. Usually, the resolution of MRM images is less than 100 μm depending on the length of the scanning time. For example, image acquisition from a fixed adult mouse with a uniform resolution of 43 μm would take roughly 30 min. Further increase in the spatial resolution would require significant increase in the acquisition time to maintain the same SNR in images (theoretically proportional to the power 6 of the resolution increase ratio). MRM is a superb imaging modality for toxicity evaluation in the nervous system, as lesions revealed by imaging correlate well with those observed using light microscopy ( ; ).

MRI/MRM has broad applications in studying rodent models of various neurologic injuries and disorders, including hydrocephalus, stroke (especially ischemic), head injury, brain tumors, and spinal cord injuries. MRI can be applied to optimize interventional treatment schedules, validate functional endpoints, or to select the most appropriate sites for tissue harvest and pathology sampling. Longitudinal studies using MRI to define the most appropriate time for tissue harvesting have the capacity to greatly reduce the number of animals required in research studies. In toxicological studies, routine evaluation by conventional histopathology is problematic due to the anatomical heterogeneity of the central nervous system (CNS) and PNS as well as the relatively limited number of planar (2D) sections that can be analyzed from any single subject (see Nervous System, Vol 3, Chap 9). In contrast, a comprehensive MRI or MRM study of the brain or spinal cord offers a complementary approach to histopathology by providing individual planar images of a lower resolution that can be combined into a 3D composite image that permits assessment of the entire nervous system as well as digital “dissection” of 2D planes at any angle. The MRI modality is especially well suited for neuroimaging because it can clearly distinguish gray and white matter, and thus can be used to quantitatively map brain regions—especially in the course of developmental neurobiology and aging experiments in rodents conducted in a longitudinal fashion ( ).

2.2

Advantages

Compared to CT, MRI provides much greater contrast and more detailed resolution between the soft tissues of the body, making it especially useful for imaging neurological, musculoskeletal, cardiovascular, and oncology models. Additionally, MRI, unlike CT, uses no ionizing radiation. MRM is nondestructive and takes advantage of a unique “proton staining” as define earlier. Both MRM and MRI data are inherently 3D and digital. Thus a 2D section can be sliced along any plane. Since no distortion from dehydration or sectioning occurs in MRM, morphometric measurements are more accurate than those from traditional histologic preparations.

Movement of the subject during image acquisition can cause artifacts and confound the contrast needed for signal interpretation in MRI/MRM. Consequently, there has been a significant effort to reduce movement artifacts by developing improved animal handing and monitoring methods, as well as novel gated or navigator imaging pulse sequences. Gated sequences synchronize MRI acquisition with respiratory motions and cardiac cycle measured by a physiology monitoring system. Alternatively, navigator techniques monitor motion in real time using an additional pulse to selectively produce echo signals from the diaphragm region and to trigger image acquisition at the end of expiration or to retrospectively select out imaging data that were acquired at end of expiration ( ). Increasingly sophisticated anesthetic delivery systems are able to provide physiological monitoring and supportive care for rodent species during imaging studies. These monitoring devices typically employ fiber optic technology, making them MRI compatible while monitoring respiratory rate, heart rate, body temperature, tidal volume, and electrocardiographic readings during image acquisition. Furthermore, spatial resolution can be significantly enhanced during MRI of rodents by cardiac gating and respiratory synchronization, thus decreasing the motion artifact while images are being acquired. Complete control of respiration requires intubation or tracheotomy and mechanical ventilation of the experimental animals; these procedures are not typically employed for rodents. Finally, cardiac (i.e., electrocardiographic) and thoracic motion information is processed by a scan management system to control timing of the MR sequence recording.

2.3

Disadvantages

As applied to small animals, MRM has two major weaknesses. The first is related to the “newness” of the method. The interpretation of MR images must be built upon experience from comparison of tissue sections from diseased organs to the images produced with MRM. For example, using MRM, some lesions cause a hypointense signal, while others produce hyperintense signals. Consequently, from our experience, there is a fairly steep learning curve to acquire the skills needed to interpret MR signals and correlate them with tissue damage that is observed with histopathology.

The second weakness is the accessibility and affordability of the modality. The equipment required is expensive, and the personnel required for development and operation of the facility are in short supply. It is our experience that to establish a high-field MRI/MRM facility, a minimum of $3 million in capital equipment with a one- to two-year lead time is needed to establish a functional center. Furthermore, the higher field strengths in MRI/MRM come with greater maintenance costs and raise increased safety concerns around the use of accessory equipment, as all devices housed in the vicinity of the magnet need to be devoid of magnetic ferrous components.

2.4

Correlation of MRI or MRM to Gross or Histopathological Lesions

Several studies have shown the potential for comparing MRM imaging to histology ( ; ; ; ; ; ; ). One signature publication evaluates the utility of MRM in detecting and characterizing COS neurotoxicity in rats ( ). Male F344 rats were exposed to 400 ppm of COS for 4 weeks (6 h/day), anesthetized, and then imaged with MRM utilizing a Gd-based contrast dye. As shown in Figure 13.2 , exposure to COS was associated with a hyperintense (bright) signal unilaterally in the parietal cortex, where the Gd was able to diffuse; this hyperintense focus corresponds to the region of malacia seen in the corresponding H&E-stained tissue section. Without contrast, hypointense MRM signals in the brain have been shown to correspond to foci of microgliosis with or without neuronal loss, while hyperintense signals correspond to hemorrhage. Lesions with a hyperintense rim surrounding a hypointense center correspond to peripheral hemorrhage and hemosiderin-laden microgliosis surrounding necrotic cores. Complete evaluation of the entire brain using 3D digital MRM images acquired in vivo was critical for defining a strategy for trimming the brain for subsequent histopathological analysis. This capability is likely to be especially useful in future studies where the heterogeneous parenchyma of the brain will show site-specific toxic responses that might be missed using conventional brain sampling schemes for histopathology. Currently, standard histopathological sampling methods include seven coronal sections for evaluation with placement using brain gross external anatomical landmarks. In the rat neurotoxicity study, only three coronal sections were originally being collected for microscopic examination (i.e., the conventional method at that time) and MRM was able to uncover a previously unidentified region of the brain with lesions which likely would not have been discovered during the previous routine trimming and microscopic assessment.

Using MRI to image tissues in an anesthetized mouse can be useful to evaluate toxicity or to monitor tumor development and progression. For example, male mice have been evaluated for bladder tumors using dynamic MRI with a Magnevistcontrast agent. Transgenic mice expressing the Harvey Ras oncogene driven by the urothelial-specific uroplakin II promoter develop papillary tumors in the urinary bladder. The mice were scanned on a Bruker 4.7T MRI machine for 20–30 min to acquire T ₁ -weighted images. Figure 13.3 is a composite of images providing comparison of the MRI, gross, and histopathologic features of these lesions. Although MRI can demonstrate a tumor in the bladder cavity, it is difficult to appreciate that the tumor is a papilloma due to the lower resolution with MRI versus histopathological examination. Additionally, the MR image has a band artifact related to the contrast agent on the urine surface. Nonetheless, MRI can be used to screen live rodents for carcinogenic effects of toxicants to provide guidance for when and where to focus the microscopic evaluation.

3.1

Basic Principles of CT

X-ray CT can produce high-quality 3D anatomical images. CT is a mature imaging modality with the first human clinical system having been introduced in 1972 ( ). More recently, numerous improvements and advancements in the field of CT imaging have allowed development of high-resolution preclinical CT systems for rodents and small specimens; these instruments have been dubbed “micro-CT” due to their spatial resolution in the micrometer scale (typically in the 50–100 μm range for in vivo scans or down to ~100 nm for specimens). The detailed anatomical images produced by CT and micro-CT systems allow this technology to be applied to toxicology studies to provide unique anatomical and/or functional information.

A complete discussion of CT physics and image production is beyond the scope of this chapter, and the reader is directed to several references for additional information ( ; ; ). In general, the production of CT images includes two major steps: acquisition and reconstruction. For image acquisition, an X-ray source is mounted opposite an X-ray detector, and the pair is rotated around a subject positioned in the center. In some micro-CT systems, the X-ray source and detector are stationary and the specimen is rotated. As the X-rays pass through the subject, the radiation is differentially absorbed or scattered depending on the material properties of the different tissues in the path. The attenuated X-rays then leave the subject and strike the detector, after which the incident radiation is detected at a single angular position. This radiation detection at a single position is termed a projection and can be thought of as being similar to a planar radiograph. The X-ray source and detector are then rotated to collect additional projections around the subject (usually several hundreds of projections collected in an arc ranging from 180 to 360 degrees). The CT projections are then converted into tomographic 3D image slices taken through the subject in a process termed reconstruction. This mathematical process takes 2D image elements (pixels) from the projections and then “back-projects” them to form reconstructed 3D image elements (voxels) in the CT image space. Reconstruction is computationally expensive, so methods have been developed to perform these calculations on CPU (central processing unit)-based computer clusters, or more recently, on GPU (graphics processing unit)-based systems to substantially reduce reconstruction times. During the reconstruction process, various image filters can be applied to the input projection data and/or the output image slice data to improve image quality. The output image slices may also be scaled to CT units, termed Hounsfield units (HU), that are typical of clinical CT systems. The HU scale is linear and is defined with air having a density value equal to −1000 HU and water being equal to 0 HU. After reconstruction, the CT images can then be viewed and analyzed by any number of image analysis software programs.

3.2

Image Information

A powerful aspect of the CT platform is that the resultant image data are quantitative both in regard to voxel intensity scaling and for defined voxel dimension. Combined with relatively high spatial resolution compared to other imaging modalities, these CT image attributes can be exploited to obtain unique data. For example, the 3D image can be used to obtain quantitative spatial information, such as measurements of length, width, and volume. The very high spatial resolutions permit “virtual histology” evaluations to be performed by micro-CT for such tiny objects as biopsy specimens and mouse embryos ( ; , ; ). Given the linear nature of CT image voxel intensity scaling, the image voxel values can be calibrated using density phantoms within the field of view of the scan as scales for quantifying chemical composition of certain structures. For example, rods of known hydroxyapatite concentration can be used to derive a calibration equation to convert voxel values from HU to units of bone mineral density (e.g., mg hydroxyapatite/cm ³ ). The exquisite detail offered by micro-CT permits the structural analysis of complex features, such as bone microarchitectural parameters in safety toxicology studies ( ). In addition, micro-CT measurements of anatomical parameters at the organ level can be employed to obtain functional data, such as ejection fraction, stroke volume, and cardiac output for the heart and lung ventilation ( , , , ). Dynamic CT has been used to measure renal parameters relevant to parenchymal (e.g., glomerular filtration rate, fractional tubular volumes, and urine formation rates) and vascular (renal blood flow) functions ( ). One group has even correlated CT imaging findings to gene expression in liver cancer where a set of distinctive CT features or “traits” specific for individual hepatocellular carcinomas were identified ( ). The traits were further filtered based on their frequency, prominence, interobserver agreement between two radiologists, and independence. A module network algorithm was employed to identify associations between filtered trait combinations and expression levels of over 6000 genes as determined by microarray analysis.

Techniques on the cutting edge of CT technology include dynamic contrast, dual energy, and phase-contrast imaging. With dynamic contrast CT, multiple short (1–3 s) scans are acquired over a longer time frame (minutes) while a contrast agent is injected slowly into the subject. As the contrast agent enters the region of interest (ROI) (e.g., a tumor), various functional parameters can be extracted from the dynamic CT data set, such as regional blood flow and volume, mean transit time, and microvessel permeability surface area product ( ; ; ). For dual energy CT, image data sets are acquired at two different energy levels( ). Since materials (e.g., bone vs. contrast agent) absorb radiation at different amounts along the X-ray energy spectrum, these material properties can be exploited to visually remove tissue from an image data set (e.g., remove the voxels corresponding to bone, but leave the voxels corresponding to contrast agent) ( ; ). Dual energy imaging can also be used to improve temporal resolution in cardiac imaging. For phase-contrast CT, the X-ray radiation passes through the subject and then forms interference patterns based on the phase discrepancy produced by various tissues ( ; , ; ). These interference patterns have been shown to highlight soft tissue structures much better than typical CT images; however this type of CT technology is not widely available at present.

3.3

Experimental Procedures

Contrast resolution refers to the ability of an imaging modality to discern between objects of similar contrast, or gray value. In CT, similar to conventional radiographs, the image is usually scaled so that air is close to black in color, bone is visualized as white, and soft tissues adopt various shades of gray. Given the biologic contrast provided by bone and lung (air), CT imaging provides excellent information on the skeleton and lung without the need for externally administered contrast agents ( ; ). Since most soft tissue structures have a high water content, the contrast between most of these tissues is minimal (e.g., kidney vs. liver). To be able to differentiate soft tissue structures, contrast agents and protocols have been developed to selectively highlight various organs and systems ( ; ; ; ; ). For example, iodine-based contrast agents are routinely given by intravascular injection to highlight the vessels and heart. Some of these agents are filtered and excreted by the kidneys, which then allows the kidneys to be easily visualized ( ). In addition, lipid nanoparticle–based contrast agents have been developed that are taken up by hepatocytes to emphasize the liver ( , ; ). Inflation with ambient air has been used as a technique to provide negative contrast for CT-based virtual colonoscopy in mouse tumor models, while barium contrast agents have been employed as positive contrast agents to monitor for the development of tumors and inflammation in the colon in mouse models ( ; ; ). Another ex vivo technique to increase soft tissue contrast is to perfuse organs of interest with contrast-permeated silicone to form vascular casts. These casts can then be used to obtain high quality micro-CT images, while subsequent digestion of the tissue around the casts permits the direct observation of the vessel networks.

An essential experimental consideration for CT is reducing image blurring due to respiration and/or heart movement ( , ; ; ; ). Gating refers to the collection of image projections at defined positions within the respiratory and/or cardiac cycle (e.g., collecting projections only during peak inspiration) so that the moving organs appear stationary. This procedure can be accomplished by the use of a small pillow under the subject to detect respiratory frequency and electrocardiographic electrodes to provide cardiac cycle information. Software can then be used to trigger the CT acquisition only when the subject is in the correct phase of the respiratory and/or cardiac cycles. Although gating increases image acquisition times, it results in images with much higher quality.

3.4