Abstract
This session at the 2010 joint symposium of the Society of Toxicologic Pathology (STP) and the International Federation of Societies of Toxicologic Pathologists (IFSTP) explored modern neuropathology methods for assessing the neurotoxicologic potential of xenobiotics. Conventional techniques to optimally prepare and evaluate the central and peripheral neural tissues while minimizing artifact were reviewed, and optimal schemes were set forth for evaluation of the nervous system during both routine (i.e., general toxicity) studies and enhanced (i.e., specialized neurotoxicity) studies. Stereology was introduced as the most appropriate means of examining the possible impact of toxicants on neural cell numbers. A focused discussion on brain sampling took place among a panel of expert neuroscientists (anatomists and pathologists) and the audience regarding the proper balance between sufficient sampling and cost- and time-effectiveness of the analysis. No consensus was reached on section orientation (coronal sections of both sides vs. a parasagittal longitudinal section with several unilateral hemisections from the contralateral side), but most panelists favored sampling at least 8 sections (or approximately double to triple the current complement) in routine toxicity studies.
This half-day session at the 2010 Joint Symposium of the Society of Toxicologic Pathology (STP) and the International Federation of Societies of Toxicologic Pathologists (IFSTP) explored many deceptively simple questions related to toxicologic neuropathology. What is the best method to prepare and adequately sample the nervous system for neuropathology evaluation? What structures or neuroanatomical levels should be sampled? Which special stains and techniques strike the best balance between sensitivity of lesion detection and appropriate allocation of time and money? Perhaps more than any other organ system, accuracy and precision in neuropathological evaluation require careful advance planning, experienced sample handling, appropriate sectioning, and careful selection of diagnostic techniques. This session focused on processing, sectioning, and staining of both the central (CNS) and peripheral (PNS) nervous systems in both routine (general toxicity) and enhanced (specialized neurotoxicity) studies and explored modern image analysis techniques and evolving regulatory trends as applied to these processes. These techniques are covered in more detail in individual manuscripts from each speaker (Boyce, R. W. et al. in press; Boyce, J. T. et al. in press; Jordan et al. in press; Jortner in press; Switzer, in press).
Preparation and Analysis of the Central Nervous System
The first lecture, given by Dr. William Jordan (Veterinary Pathology Services, Inc., Greenfield, IN), provided a concise, practical overview of key steps in preparing the CNS for microscopic examination. If brains and spinal cords are not collected and fixed correctly, the subsequent processing and analytical steps will be compromised and perhaps completely invalidated. High-quality histological evaluation hinges on proper collection, which requires rapid exsanguination, minimal handling of CNS tissue, and rapid and thorough immersion or perfusion fixation (Dorman et al. 2002; Fix and Garman 2000). Dr. Jordan described a rapid method of removing rodent spinal cords using hydraulic pressure that eliminates the potential for artifacts induced from prior bone removal (de Sousa and Horrocks 1979; Levine 1965; Meikle and Martin 1981).This method requires tight insertion of a blunt needle into one end of the vertebral column followed by rapid expulsion of physiological saline in the attached syringe to blow the organ out of the vertebral canal. This technique produces intact spinal cords with remarkably few artifacts, although dorsal root ganglia and spinal nerve roots are not available for analysis.
The second critical part of CNS handling is consistent histological sectioning to survey multiple anatomic sites (Greaves 2007). Histology laboratory personnel need a basic familiarity with neuroanatomy to ensure that major brain features are sampled at the same specified levels in every study animal. Hematoxylin and eosin (H&E) is the cornerstone stain for many toxicologic neuropathologists, whereas other special stains (techniques widely available in print or via the Internet) to detect specific disease processes (e.g., Fluoro-Jade B or C or silver stains for neurodegeneration), cell types (e.g., cresyl violet for neurons, Luxol fast blue for myelin), or cell type-specific markers (e.g., immunohistochemical detection of glial fibrillary acidic protein [GFAP, an astrocyte marker]) are frequently used to answer discrete questions (Haschek et al. 2010; Schmued and Hopkins 2000; Switzer 2000). Fluorescent examination of H&E-stained brain sections with a 450- to 480-nm FITC filter (wide band) is an underused technique to detect neuronal degeneration, as dying neurons are often autofluorescent; Fluoro-Jade staining enhances this fluorescence.
For the final step, microscopic evaluation, the most important ingredient is a knowledgeable and experienced pathologist who is well versed in regional brain anatomy and function, the staining processes used to demonstrate various tissues and disease processes, and the response patterns of the CNS to injury. This basic anatomic knowledge is necessary to reliably recognize the major brain regions and thus efficiently and accurately employ one of the many species-specific brain atlases (available in print and online) (Franklin and Paxinos 2007; Paxinos and Watson 2005; Paxinos et al. 2000). Precise anatomical localization of a lesion within the brain is imperative in toxicity studies because of the heterogeneous nature of the brain and the differing chemistry and responses of its many site-specific cell populations. Well-defined localization of lesions also facilitates communication between pathologists, neuroscientists, and regulatory officials and is an important component in determining pathogenesis and evaluating pharmacological activity.
Preparation and Analysis of the Peripheral Nervous System
The second lecture was an overview by Dr. Bernard Jortner (Virginia Polytechic Institute and State University, Blacksburg, VA), who described a number of unique features that must be considered for proper sampling and evaluation of the PNS, as well as many common toxicant-induced findings. The PNS is composed of cells similar in structure and function to the CNS, including neurons and myelinating cells, and is intimately connected to the CNS. However, PNS tissue sampling, processing, and evaluation often require techniques not commonly employed for any other body system, including the CNS. Even embedding of the PNS can be specialized, with epoxy resins necessary to provide the sensitivity for identifying certain subtle lesions. Specifically, Dr. Jortner covered fixation, differential anatomic sampling, tissue processing, and microscopic evaluation. Special techniques including nerve fiber teasing, immunohistochemistry, and electron microscopy were also reviewed.
Ideally, peripheral nerve sampling should include not only proximal samples, such as the sciatic nerve section that is evaluated in most routine toxicity studies, but also a more distal nerve section to increase the likelihood of detecting early ascending axonopathies and neuropathies. Avoidable artifact can be minimized by handling peripheral nerves only from the cut ends (to minimize compression and distortion of the delicate fibers) and by fixing peripheral nerve segments while they are adhered to stiff material such as cardstock (to minimize contraction artifact) (Jortner 2000). High-quality peripheral nerve sections for light microscopic evaluation are best achieved by post-fixation with osmium, embedding in epoxy resin, sectioning at 1 µm, and staining with toluidine blue (Aparicio and Marsden 1969, with many other references in print and online). Morphometric methods can be easily employed on resulting cross-sections to determine nerve fiber sizes and myelin thickness (Broxup et al. 1990). A promising technique adapted from clinical medicine that is used to evaluate neuropathies of nonmyelinated nociceptor fibers is to count fiber numbers in specially stained skin biopsies. This evaluation can be done serially over time, causes minimal distress to the patient or animal, and has been used to detect toxicant (e.g. cisplatin)-induced sensory neuropathy as well as diabetic neuropathy (Griffin et al. 2001; Lauria et al. 2005).
Toxicant-induced PNS lesions can involve neuronal cell bodies, axons, myelin, or a combination of all three. Pyridoxine induces a primary neuronopathy affecting sensory ganglia that results in secondary degeneration of axons of the affected cells (Chung et al. 1998). Axonopathies include the well-studied organophosphate ester-induced delayed neurotoxicity (OPIDN), which produces lesions similar to traumatic severing of axons and so is sometimes termed “chemical transection” (US EPA 1998a, 1998b). Tellurium was used to illustrate myelinopathy, as administration to developing rats impairs Schwann cell cholesterol synthesis, which transiently halts ongoing myelination and destabilizes already formed myelin sheaths (Lampert et al. 1970).
Determination of Neuron Number in Neurotoxicology Studies
Dr. Rogely Boyce (WIL Research Laboratories, Ashland, OH, currently with Amgen, Inc., Thousand Oaks, CA) delivered the first part of a talk describing the essentials of stereology. Small changes in cell number within the nervous system can have profound effects on function, and most neural tissues have significantly limited capacity for regeneration. However, conventional qualitative examination is often an inadequate means of assessing structure volume and population number. In the absence of secondary reactive change, most trained neuropathologists cannot detect a loss until ~40% or more of the neurons in a given area have been lost. When changes in cell number are large, two-dimensional (2D) morphometric measurements acquired from routine slides or digitized images have the advantage of providing concrete data that can be compared between animals but are prone to the same errors in estimating tissue volume and cell number as routine qualitative methods. Stereology allows for reliable, repeatable, and simple quantification of cell numbers in the total three-dimensional (3D) volume of a neural structure.
Estimating cell numbers from routine slides requires a number of assumptions that are often not correct in biological systems. Major suppositions are that any cell loss occurs uniformly, that normal cell density is equal throughout the structure, and that the total volume of the structure in question is predictable and not subject to artifactual swelling or shrinkage. Stereological analysis transcends these often erroneous assumptions. To accurately estimate cell numbers through the use of multiple sections, systematic uniform random sampling is undertaken by which the first section is positioned randomly within the tissue, but subsequent sections are taken at regular intervals through the region of interest. Other concepts covered included the physical disector, in which two parallel thin sections separated by a known distance are evaluated to count the number of objects that are confined to one section, and the optical disector, in which a stack of optical sections are evaluated to enumerate the unique objects. This latter method suffers from its requirement for difficult-to-process thick (40–60 µm) sections. The physical disector can be used on standard paraffin-embedded thin sections, which are more practical for processing, evaluation, and archiving purposes. Another important stereology concept is the unbiased counting frame, which prevents a single cell from being counted more than once, and also eliminates bias toward larger cells.
Although stereological methods have not been widely applied in toxicologic pathology, recent advances have made it possible to implement these methods in a regulatory toxicology setting, particularly methods for estimation of total cell number. It is now feasible to make estimates of cell number in a time-effective way using thin paraffin sections by integrating precise and efficient stereological methods using automated whole slide imaging, automated section sampling, computer-assisted measurements, and automated capture and registration of physical disectors (Boyce, R. W. et al., in press).
The second part of the talk was given by Dr. John Boyce (WIL Research Laboratories, Ashland, OH), who discussed regulatory concerns with respect to data generated from neurostereological analysis (Boyce, J. T. et al., in press). He emphasized that validation and reporting of these data need not be complicated, because proper validation can be accomplished through the mathematical proof of the analysis methods and performance of a validation study. Stereology reports should document the tissue processing methods, counting technique, and the statistical methods used. Dr. Boyce warned that simply counting more sections does not necessarily increase accuracy; it might make the results more precisely inaccurate! A further caution is that although stereological methods are unbiased, bias may be inadvertently introduced during technical implementation. Stereology methods using thin sections have the advantage that the slides can be easily and routinely archived and so are available as raw data.
Enhanced Neuroanatomic Sampling
The final lecture, presented by Dr. Robert Switzer III (NeuroScience Associates, Knoxville, TN), provided a rational foundation for selecting the number of brain sections to be evaluated in toxicity studies so that a balance is struck between the adequacy of sampling of this heterogeneous organ and the cost (in money and time) to screen a sufficient number of sensitive sites. He presented a method for comprehensively sampling the over 600 distinct subanatomic areas in the brain, so that reasoned risk-assessment decisions can be made in as cost-effective a manner as possible (http://www.neuroscienceassociates.com). Properly employing these techniques requires consistent, evenly spaced brain cross (coronal)-sections that are taken with knowledge of their subanatomic variability. It is widely believed that adequate toxicological sampling of body tissues reduces the risk that a toxic test article, environmental agent, or condition might go undetected.
Specifically, Dr. Switzer recommended that all preclinical toxicity packages for novel compounds include at least one acute study in which approximately ten-fold more coronal sections of brain are evaluated from each control and high-dose animal. In the adult rat, this comprehensive neuropathology analysis comprises 60-65 sections taken throughout the brain at regular 0.32-mm intervals. For large animals, including monkeys and dogs, the same number of sections taken over wider intervals from one hemibrain would accomplish the same goal. Enhanced neuropathology evaluation of lower doses is unnecessary if no lesions are evident in the high-dose animals. Dr. Switzer offered some examples of neurotoxicants (e.g., carbonyl sulfide, ethanol, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine [MPTP], 2′-NH2-MPTP, and p-chloroamphetamine [PCA]) in which subtle, site-specific brain lesions would likely have been overlooked on less comprehensive sectioning (e.g., the standard range of three to twelve sections in conventional general toxicity studies) (Morgan et al. 2004; Unger et al. 2002).
Panel Discussion on Brain Sampling
Following the four talks, a panel was convened to discuss the merits of various brain-sampling options. The discussion was moderated by Drs. Sarah Hale (Covance Laboratories, Inc., Vienna, VA) and Lydia Andrews-Jones (Allergan, Irvine, CA). The panel members included experienced toxicologic neuropathologists (Drs. Mark Butt, Robert Garman, William Jordan, Georg Krinke, and Peter Little), a scientist from a regulatory agency (Dr. Karl Jensen), and an accomplished neuroanatomist (Dr. Robert Switzer III). Panel members, with audience input, were challenged to propose and defend their sampling recommendations for toxicologic neuropathology analysis from routine screening studies in both rodents and large animals to achieve the best possible balance between time, cost, and minimization of risk.
Dr. Andrews-Jones defined the term “routine” sampling to indicate the minimum number of sections of the brain that should be evaluated in studies in which a CNS effect is not predicted by clinical signs, mechanism of action, or structural similarity to other neurotoxicants. This basic approach is distinct from that used in neurotoxicity studies deemed to be “dedicated,” “enhanced,” “expanded,” or “focused,” terms that are used to indicate experiments targeted to specifically assess or interpret neurotoxicity/neuropharmacology end points. The panel discussants, recognizing the wide variety of possible questions asked by highly focused neurotoxicology studies, limited their comments largely to routine toxicity studies.
Dr. Little summarized the U.S. National Toxicology Program’s (NTP) recently proposed guidelines, which call for the number of rodent brain sections in routine studies to double, from the currently accepted three coronal sections (Solleveld and Boorman 1990) to seven (NTP 2010). Dr. Little emphasized that this seven-section scheme was for “routine” NTP studies but was not considered optimal for compounds with CNS effects. The additional four sections will ensure the inclusion of such toxicologically important brain structures as olfactory lobe, rostral (anterior) and caudal (posterior) colliculi, substantia nigra, raphe nuclei, and the area postrema.
Dr. Jordan proposed that sectioning in rodents to balance between structure and function should encompass one para-sagittal section (just off the midline) of one brain half and four coronal sections through the remaining half. He noted that adding sections will substantially increase the time and cost for routine evaluation, but he suggested that this increase could perhaps be offset by eliminating from the tissue collection list for routine studies certain non-CNS tissues that historically have a low likelihood of developing toxicant-induced changes.
Dr. Butt indicated his preference for evaluation of seven to ten coronal sections of rodent brain in routine studies. He noted that all sections could be placed on a total of two standard-sized slides, and that additional processing and evaluation time would be minimal, although this point was disputed by some audience and panel members. When sectioning the brains of large animals (monkeys and dogs), he proposed increasing from the currently routine three or four levels from one side of the brain to seven or eight levels (from one side), which would lead to only a small increase in the total number of brain sections. He also indicated that some tissues (e.g., one of the two kidney sections) could be removed from the routine tissue list to accommodate additional brain sections without greatly increasing the total number of tissue sections evaluated.
Dr. Switzer, who had given his enhanced sectioning preferences (sixty to sixty-five coronal sections for rodent brain) and rationale during the talk that immediately preceded the panel discussion, added that at least as many sections should be evaluated in large animals as in rodents. He made the sobering assertion that all known neurotoxicants have been identified through human exposure, including those discovered in the decades since routine toxicity testing has been required for product registration.
Dr. Garman favored a routine sampling scheme of eleven sections for rodent brains and indicated that every major subanatomic nervous system area should be specifically listed as having been evaluated in the pathology report to clarify for readers, including regulatory officials, exactly what was examined. Dr. Garman also pointed out that animals with ethylene oxide–induced brain tumors were more readily identified when more than the routine three sections of brain were evaluated (Snellings et al, 1984).
Dr. Krinke emphasized the need for responsible scientists to maintain flexibility and freedom when choosing how to section the CNS. He indicated his preference for unilateral sectioning in large animals and urged that further consideration be given to sectioning brains in the sagittal (rather than coronal) plane in routine studies. Although doing so would require significant re-education of pathologists regarding normal neuroanatomic features and might miss certain regional structures (especially the most lateral ones), it would have the advantage of including more distinct subanatomic areas per section and a simpler assessment of the structural integrity of tracts needed for integration between functional areas. Dr. Krinke also offered the important point that the key issue in comprehensive routine brain sectioning is not the actual number of sections but that written confirmation be included in the report that all of the major neuroanatomic areas have been sampled for each animal, thus supporting Dr. Garman’s recommendation to specifically list areas examined.
Dr. Jensen, who trained as a neuranatomist, favors trimming the entire rodent brain into two cassettes for routine evaluation. He re-emphasized the concern raised by the ethylene oxide inhalation study that the brain is prone to develop small neoplasms, the incidence of which cannot be adequately examined in the conventional three to five coronal sections. He also made the pithy but powerful point that “absence of evidence is not evidence of absence.”
There was also some discussion of optimal time points for brain sampling. Dr. Butt felt that we should be looking at more sections at two to four days post-exposure, which is earlier than is usually done in toxicology studies. Dr. Switzer suggested that acute studies should include examinations at two, four, seven, and fourteen days. Dr. Krinke felt that class of compound needed to be considered when deciding on timing and pointed out that for delayed neurotoxicity, it may take three weeks for a lesion to develop.
Audience participation was lively. Dr. Robert Sills (National Institute of Environmental Health Sciences [NIEHS], Research Triangle Park, NC) reiterated that the lesions caused by carbonyl sulfide (hemorrhage and necrosis specifically in the caudal [posterior] colliculi; Morgan et al. 2004) would have been missed on the standard three brain sections, and that they were observed only because some sections were serendipitously cut at nonstandard locations. The decision by the NTP (a division of the NIEHS) to look at more sections of the brain was driven by the increasing incidence of neurologic diseases that have been linked to xenobiotic exposure, including autism and Parkinson’s disease.
Audience suggestions included performing a retrospective review five years after adoption of the expanded NTP brain sectioning protocols for “routine” studies to see how many neurotoxicities were picked up using the expanded number of sections, including additional stains (vs. only H&E), and considering doing perfusion fixation in all toxicity studies. (Dr. Brad Bolon [GEMpath Inc., Longmont, CO] responded that routine perfusion fixation for general toxicology studies is not practical because it might interfere with evaluation of other organs and some molecular pathology analyses.)
Concerns expressed by the audience about increasing the number of sections for routine studies included training of histotechnologists and pathologists, quality of sections (when going from 4-µm-thick to 2-µm-thick sections), and additional cost in materials and time. Other audience comments included the observation that the cost of doing additional brain sections is miniscule compared with having a compound fail in Phase 3 clinical trials or later, and that functional tests and other assays in addition to histopathology are necessary to discern some neurologic effects.
Although the panel arrived at no definitive answer to the question of how many sections of brain are sufficient in routine toxicity studies, a number of important considerations in eventually arriving at this number were raised and discussed. Some proffered suggestions were that the number of sections to be evaluated should be increased from the usual three to five coronal sections; that brain should be sectioned in the longitudinal rather than the coronal plane; that routine screening in large animals should be performed on only one side; and that the number of non-CNS sections be reduced to accommodate more CNS sections without raising the total number of sections to be examined per study. Ultimately, scientists from industry and regulatory agencies will have to formulate an accepted standard for microscopic evaluation of the brain in routine toxicity studies with all of these considerations in mind.