Available imaging systems for use in preclinical toxicology studies increasingly show utility as important tools in the toxicologic pathologist’s armamentarium, permit longitudinal evaluation of functional and morphological changes in tissues, and provide important information such as organ and lesion volume not obtained by conventional toxicology study parameters. Representative examples of practical applications in toxicology research and preclinical studies are presented for ultrasound, PET/SPECT, optical, MRI and MALDI-MSI imaging. Some of the challenges for making imaging systems GLP-compliant for regulatory submission are presented. Use of imaging data on case-by-case basis as part of safety evaluation in regulatory submissions is encouraged.
The use of non-invasive in vivo imaging and the option of using ex vivo imaging of fresh and fixed specimens provide a unique adjunct to conventional histopathology evaluation in preclinical animal studies. The number of available imaging methods has increased in recent years, and include ultrasound, optical imaging, positron emission tomography (PET), single photon emission computed tomography (SPECT), computed tomography (CT) and magnetic resonance imaging (MRI). Images provide detailed morphological and functional insights in addition to 2D, 3D and 4D quantitative data on progression and regression of lesions in preclinical disease models and in conventional toxicity and carcinogenicity studies. The aim of this opinion paper is to provide a brief overview of imaging modalities with examples applicable to toxicologic pathology and our recommendations for regulatory submission of imaging data. Representative imaging applications have been selected from an International Academy of Toxicologic Pathology educational course presented at the 13th European Congress of Toxicologic Pathology in Surrey, UK, September 2015.
Overview of selected imaging modalities useful in preclinical studies
Ultrasound as an imaging modality utilizes high frequency sound waves that travel through tissues, organs and the body to produce images. A transducer (probe) placed against a body converts electrical signals to sound waves, sends them into the body, and the sound waves that are reflected back are turned into electrical signals for processing by a computer to generate an image. The pulses are sent into the animal at recorded time intervals. The distance and direction of the reflected return echo and its time of arrival back to the transducer permits the construction of a two-dimensional image of internal structures in the animal. Resolution increases while depth of penetration decreases at higher frequencies. Use of high frequency sound waves permits real-time monitoring of blood flow. Gas-filled microbubbles can be used to increase the signal when imaging blood flow. Attaching specific proteins to microbubbles (such as antibodies directed against endothelial cell proteins) allows ultrasound to be used to target specific intravascular biochemical processes (Kiessling et al., 2012). Advantages of ultrasound include relative low cost, portability of equipment, good temporal resolution, good safety profile and excellent sensitivity when microbubbles are used. Limitations are poor imaging of bone and airand limits on depth penetration.
PET and SPECT
Positron emission tomography (PET) and single-photon emission tomography (SPECT) utilize radionuclides to allow study of biochemical changes and levels of molecular targets in living animals (Khalil et al. 2011; Yao et al. 2012; Vaquero and Kinahan 2015). For PET to work well one must identify and produce a radiolabeled imaging agent that is specific and selective for the target of interest. A common PET radionuclide is 18F with a half-life of just under two hours. Other radionuclides include 64Cu, 76Br, and short-lived radionuclides 11C, 13N and 15O. A small quantity of the specific radiolabeled agent is given intravenously to the test subject with subsequent tracing of its distribution in the body using a PET camera. The short-lived radionuclide is generated by a cyclotron and decay of the radionuclide occurs by positron emission and subsequent collision of the positron with an electron. Each collision (annihilation) generates two photons that have significantly higher energy (511 keV) than conventional X-rays. A ring-like detector surrounding the subject detects annihilation events and, after detector normalization and other corrections, converts their electrical signal to tomographic images.
Primary use of PET in the clinic is to image cancer most often using 18F-labeled glucose. The assumption is that the metabolic demands of the cancer cells exceed that of normal cells and 18F-glucose will preferentially localize in the cancer cells. Preclinical use of PET is in basic research and in rodent models to investigate mechanism of human disease (Yao et al. 2012). Miniaturized PET scanners for preclinical use have a relatively low spatial resolution of 1 to 2 mm (Moses 2011) but a high sensitivity.
SPECT uses radionuclides that decay with emission of single gamma rays. Typical radionuclides include 99mTc, 123I and 111In and a gamma camera that rotates around the subject is used to capture data from differing positions to allow tomographic reconstruction. Technical implementation (use of a collimator) to cancel out photons that are spatially uninformative can markedly decrease the sensitivity of SPECT in contrast to PET but sensitivity can be increased using micro-pinhole apertures to maintain sensitivity with reasonable spatial resolution. Miniaturized SPECT instruments are available for preclinical studies and have been used for imaging cancer xenografts in mice, rat hearts following ischemia-reperfusion, and for imaging dopamine transporters in rat brain (Scherfler et al., 2002). New versions of SPECT instrumentation for preclinical use have been developed with spatial resolutions of less than a millimeter (van der Have et al., 2009).
The advantage of PET and SPECT is that they detect biochemical changes that typically precede anatomical changes (Bernsen et al. 2014; Franc et al. 2008). Another advantage is the possibility of continuous acquisition of scan data from a radionuclide or radionuclide-labeled experimental substance. Temporal data may be acquired in the form of a dynamic scan. For example, an uptake of a specific radiolabeled antibody-drug conjugate by the tumor can be recorded at multiple time points after the injection, including minutes after the drug delivery, followed by hours and days post-injection in the same animal – providing data which are impossible to obtain using other methods (ter Weele et al., 2015).
A major limitation of PET and SPECT is poor anatomical localization. The lack of an anatomical reference frame is easily overcome by combining PET or SPECT with CT or MRI (Chatziioannou 2005; Shao et al. 1997). The radiation exposure safety concern of using radionuclide imaging for PET and SPECT limits clinical use but is of less concern for preclinical studies provided care is taken regarding exposure of preclinical technical staff.
Optical fluorescence and bioluminescence
Optical imaging, a technique for non-invasively looking inside the body, measures light produced by optical reporters that are visualized through the tissues (typically the skin) of a live animal (Dufort et al. 2010). This imaging platform allows monitoring of function in its natural structural context. There are two broad categories of optical imaging: fluorescence and bioluminescence. Fluorescent technology uses non-ionizing visible, ultraviolet, or infrared light and the special properties of photons to obtain detailed images of organs, tissues, cells, and even molecules. Optical imaging is useful for visualizing soft tissues and takes advantages of different colors of light to simultaneously measure multiple features in an organ or tissue. Bioluminescence relies on introduction of one of several luciferase enzymes into the subject that then interact with its substrate to produce photons of visible light (Sadikot and Blackwell 2005). Both are cost-effective, non-invasive, sensitive, physiological, and high throughput with spatial and temporal resolutions up to 0.25 mm, and can be designed to target a specific protein or pathway of interest. The primary limitation of optical imaging is its shallow depth of penetration limiting its use to visualize internal organs in animals larger than rodents. Novel optical imaging techniques are constantly evolving providing capability to monitor dynamic functional processes at the cellular and molecular level in biological systems and the whole animal. Different varieties of optical, in vivo imaging modalities include optical coherence tomography (Vakoc et al., 2009, Wenzel et al., 2015), photo-acoustic microscopy (Stein et al., 2009), photo-acoustic tomography (Burton et al., 2013, Hudson et al., 2014), multi-photon microscopy (Breckwoldt et al., 2014, Harb et al., 2013), hybrid methods such as laser speckle and fluorescence (Kalchenko et al., 2011, Kalchenko et al., 2010, Towle et al., 2012, Zhongchan et al., 2013), and bioluminescence (Corson et al., 2014, Hirano, 2016) imaging.
CT or CAT- Computed Tomography (Computer-Aided Tomography or Computed Axial Tomography)
Computed tomography uses X-ray imaging of sections or slices (tomography) through the body or a specific tissue to produce a 3-D image of the subject being imaged. CT is basically an anatomic imaging process where the X-ray source is coupled to a detector array and both rotate around the animal. Tissues that strongly absorb X-rays such as bone show up as white, air shows up as black, with soft tissues showing up in shades of gray. The degree of X-ray attenuation by different tissue components is assigned Hounsfield units (HUs) based on their differences in density and composition. Software automatically assigns HUs to every voxel on the CT scan to assist in image interpretation. An iodinated contrast agent can be used during CT imaging to improve spatial resolution and soft tissue contrast. Scaled-down versions of hospital CT scanners are used in preclinical research for small animals, including invertebrates.
Computed tomography allows for rapid acquisition time and high spatial resolution. Micro-CT imaging systems used for preclinical studies offer resolutions as low as 3.25-9 µm (Lombardi et al., 2014; Rueckel et al., 2014). There is essentially limitless depth of penetration. Technical advances in X-ray detection and use of helical scanning and multi-detector systems offer increasingly lower levels of radiation to obtain suitable images. However, exposure to radiation remains a limitation in clinical applications. While soft tissue contrast can be enhanced by use of iodinated contrast agents, soft tissue detail is less than that from MRI. CT remains an important and highly efficient preclinical imaging modality and when combined with PET, SPECT, optical or MRI modalities provides an anatomical reference frame for these other imaging modalities. Numerous examples of CT imaging applications in toxicologic pathology are available in the published literature (Badea et al., 2004; Badea et al., 2005; Vasquez et al., 2013; van Deel et al., 2016; Solomon et al., 2016; Ashton et al., 2015; Brown et al., 2008; Martiniova et al., 2010; Wang et al., 2016; Cavanaugh et al., 2004; Wise et al., 2010).
MRI – Magnetic Resonance Imaging
MRI uses a powerful magnet and radio frequency (RF) energy to image atomic nuclei within the body. It is based on the constant angular momentum of protons and neutrons as they spin about their axes. In addition to their angular momentum, these nuclear particles have a small magnetic field called a magnetic moment. Both the angular momentum and the magnetic moment are vector quantities and, thus, have directionality and spin or wobble features like the wobble of a spinning toy top. When placed in a powerful magnetic field such as in an MRI magnet, these nuclear particles align either parallel or anti-parallel to the MRI magnetic field, absorb energy when aligned by a pulse from one of the MRI magnet coils, and return the absorbed energy as RF pulses, called relaxation, while returning to their normal alignment where an additional gradient coil locates the X, Y and Z orientation of the tissue MR signal after each magnetic pulse from the MR instrument.
The most abundant atomic nucleus in the body is the hydrogen proton, 1H. Thus, it is the most commonly used nucleus for MRI. 1H imaging is dependent upon the concentration of 1H, its degree of polarization, and its gyromagnetic properties. 1H anatomic images are based on the hydrogen in tissue water. MR imaging can be adapted to use other less abundant atomic nuclei, including 31P, 13C, 23Na, 19F and 17O2, as well as hyperpolarized nuclei such as He. A variety of pulse sequences, timing variations, and image acquisition parameters allow capture of two major types of relaxation as the atomic nuclei return to their normal equilibrium. The two types of relaxation are called spin-lattice (longitudinal) and spin-spin (transverse) and the images are generated due to tissue differences in the longitudinal (referred to as T1) and transverse (referred to as T2) relaxation times. By varying the timing parameters of RF pulses (pulse sequences), an acquired image can be either T1 or T2 weighted to enhance contrast of specific tissue pathologies.
A variety of MRI techniques with and without use of MRI contrast agents or super-paramagnetic nanoparticles permit assessment of dynamics of blood flow, oxygen status in areas of perfusion, degree of tissue cellularity, areas of necrosis, and inflammation. Functional MRI (fMRI) can identify functional versus dysfunctional areas in the brain based on an increase in oxygenated blood in areas of increased neural activity. MRI soft tissue anatomy is excellent with resolutions on the order of 25 to 100 µm in preclinical animal studies. MRI is a powerful non-invasive method for assessment of phenotypes and therapeutic efficacy in animal models of disease used in preclinical research and drug development. In vivo MRI provides an opportunity for longitudinal evaluation of tissue changes and phenotypic expression in experimental animal models to monitor progression, regression, and therapeutic responses non-invasively. Ex vivo MRI of formalin-fixed specimens permits a thorough examination of multiple digital slides while leaving the specimen intact for subsequent conventional H&E histology of regions of interest (Nyska et al., 2014; Tempel-Brami et al., 2015; Ramot et al. 2017a; Ramot et al., 2017b).
For conventional MRI systems, equipment acquisition, operation and maintenance are relatively expensive and are limited to institutional facilities where these systems can be isolated for safety reasons. These limitations have recently been overcome with the development of more compact MRI systems that are designed for operation in conventional laboratories without the cost, complexity, and infrastructure of conventional MRI systems (Tempel-Brami et al., 2015; Ramot et al. 2017b).
MALDI-IMS (Matrix-Assisted Laser Desorption Ionization – Imaging Mass Spectrometry)
Matrix assisted laser desorption ionization (MALDI) imaging was described for the first time by Caprioli and colleagues (Caprioli et al., 1997). This imaging technique utilizes a sample such as a frozen or formalin-fixed tissue section that is probed with a laser in two dimensions while recording a mass spectrum. Technically, a matrix is deposited on a thin tissue section. It absorbs at the laser wavelength of the MALDI ion source and ionizes the analytes. The slide with the section is moved in two dimensions in the mass spectrometer while a mass spectrum is recorded at each position (for each laser impact). The distance between each laser impact (spatial resolution) is selected by the individual before the acquisition. Molecular images are then constructed by correlating ion intensities with the exact position of the data from the sample. The molecular distribution of each detected ion in the tissue with the selected method can be generated. This permits simultaneous measurement of the spatial distribution of multiple analytes (DNA, peptides, proteins, sugars, small molecules, large organic molecules) without destroying the sample. Imaging software imports the mass spectrometer data and places them on an optical image of the sample. This allows identification and localization of chemical species and metabolites on the optical image. A classical staining (e.g. hematoxylin and eosin stain) can be applied on the tissue section at the end of the mass spectrometry imaging (MSI) acquisition or on an adjacent tissue section. The obtained picture is then overlaid with the generated molecular distributions in dedicated software.
Two main advantages of MSI are its specificity and the fact that the samples do not have to be labeled with probes or radionuclide. It is possible to detect a drug and its related metabolites on one tissue section and to generate the images of each molecule based on its m/z (mass to charge ratio) (Bonnel et al., 2011a). This detection method doesn’t require labeling of the targeted molecules (no radioactivity, no fluorescence, and no tracer). It can usually detect more than a thousand compounds from a tissue surface in a full scan mode of acquisition. The best spatial resolution that can be applied today is around 5 to 20 µm. It is thus possible to visualize the molecular distributions within histological structures.
Matrix assisted laser desorption ionization imaging can be applied on all tissues prepared for conventional microscopic investigation. Formalin-fixed and paraffin-embedded (FFPE) tissues are today the first choice for safety evaluation. Several examples of MSI applications on FFPE tissues have been published (Buck et al., 2016, Groseclose et al., 2008, Lemaire et al., 2007, Pietrowska et al., 2016, Stauber et al., 2008).
As for all analytical technologies, the MSI sensitivity must be determined for each targeted compound or class of molecules. Desorption and ionization processes depend on the chemical structure and on the physicochemical environment of the biological tissue. To maximize sensitivity with histological localization, quantitative mass spectrometry imaging (QMSI) has been developed to bring quantification aspects to the molecular distribution Hamm et al., 2012).
Applied during the first decades in proteomics on tissue in the life sciences (Bonnel et al., 2011b, Chaurand and Caprioli, 2002, Chaurand et al., 1999, Franck et al., 2009, Stoeckli et al., 2001), MALDI imaging has been quickly used in a wide range of applications in the pharmaceutical industry (Castellino et al., 2011, Hamm et al., 2012, Kreye et al., 2012, Prideaux et al., 2011, Prideaux et al., 2010, Prideaux and Stoeckli, 2012, Solon et al., 2010, Stoeckli et al., 2007 , Sun et al., 2016). Indeed, MSI experiments today allow combining the pharmacokinetics and pharmacodynamics information of the drug, making the connection between early stage drug metabolism pharmacokinetics and late stage development. This imaging technology can also be used in clinical investigations (e.g. punch biopsies in dermatology or biopsies in oncology).
Practical imaging applications
Practical ultrasound applications
Cardiac hypertrophy in genetically modified mice
Using an over-expressing ErbB2tg model (Sorensen et al., 2016, Sysa-Shah et al., 2012) and/or surgically induced hypertrophy (Takimoto et al., 2005), ultrasound identifies the extent of ventricular hypertrophy as well as allowing documentation of functional cardiac impairment. The ultrasound characteristics of the cardiac hypertrophy in the ErbB2 model are sufficiently specific to reduce the need for genotyping the mice, thus allowing conduct of subsequent studies knowing which mice are carrying the appropriate genotype (Figures 1 and 2). In support of this research, cardiac ultrasound can be used to define the functional aspects of cardiac hypertrophy in characterizing this disease model.
In clinical cardiology, echocardiography is a gold standard in the diagnostics of most cardiac diseases. In preclinical studies, echocardiography allows for characterization of anatomical changes without a need to sacrifice multiple groups of animals. Instead, the stages of the development of the pathological process can be tracked using echocardiography, at the same time obtaining information about functional changes, which in many cases, precede the anatomical changes. Echocardiography can be further complemented with electrocardiography, which provides additional information about developing cardiac hypertrophy (Sysa-Shah et al., 2012).