While the face validity (i.e., how well does a model mimic clinical features of IBS patients) of some rodent models of visceral pain has been good, the predictive validity (i.e., how well do drug studies performed in the model predict effectiveness in humans) has been disappointing. For example, adult rats having been exposed to the maternal-separation paradigm as pups show evidence for stress-induced fecal pellet output, stress-induced visceral hyperalgesia and anxiety-like behavior, all findings homologous to those reported in IBS patients [83
]. However, several drugs that showed effectiveness in this model (e.g., antagonists for the CRF1 or NK1 receptors) have failed to show effectiveness in human models or in clinical trials. The problem of bench-to-bedside translation, however, does not simply originate in a failure of the animal models. Improved definition and classification of clinical states based on biological abnormalities are needed. Such improvement is dependent on the identification of robust endophenotypes in humans with adequate effect sizes for cardinal symptoms or global end points, which can be modeled in a transverse translational approach in rodents.
We propose a novel reverse translational approach, which begins with the identification and in-depth characterization of neurobiological endophenotypes in IBS patients or subsets of such patients. This approach does not aim to identify a rodent model of a complex, unique human disorder, but aims to use the rodent model to pharmacologically characterize the homologue of an endophenotype that has previously been characterized in patients. In contrast to biomarkers, which are thought to be specific for a particular disorder, endophenotypes are dimensional constructs that play a role across categorical disease definitions [84
]. For example, in the case of IBS, the endophenotype of enhanced responsiveness of a stress and emotional arousal circuit is likely to be found in anxiety disorders, and in other stress-sensitive disorders. Similarly, the endophenotype of ineffective cerebral cortico-limbic inhibition is likely to be found in many, often overlapping disorders characterized by physical or emotional discomfort. It has been suggested that clusters of endophenotypes may be more similar among subsets of patients with different disorders, rather than being seen in all patients of a given disorder [86
]. Rodent models of such endophenotypes are important to identify potential molecular targets, dose ranging and possible side-effect profiles of candidate compounds. As implied by the endophenotype concept, successful drug development based on this approach would be expected to be useful for subsets of patients with different syndromes, but not necessarily for all patients of a given syndrome.
Given the high incidence of mood and anxiety symptoms in IBS patients, as well as the growing acceptance of the importance of central pain amplification in the pathophysiology of IBS [87
], the development of rodent homologues of such brain endophenotypes should be important. However, the question remains, what level of inquiry of the involved brain endophenotypes in humans is most promising for the preclinical assessment of candidate IBS drugs? A behavioral change, while capturing a broad spectrum of dysfunction, may lack specificity, while a neuromolecular change may not generalize across a disease that is likely multifactorial in origin and that is characterized by subtypes with overlapping symptoms. Pseudoaffective responses (e.g., electromyography, pain behavior) themselves do not allow for the elucidation of the underlying systems-level processes by which molecular, cellular and genetic profiles bias behavior and nociception. Neuroimaging can complement such association studies by identifying the biological effects of a compound at the brain endophenotype level, such as the level of integrated neural systems and circuits.
The utility of neuroimaging in CNS drug development
Our understanding of the functional and structural reorganization of the brain in response to chronic pain, and how the brain responds to pharmacological treatment, has been significantly changed as a result of developments in neuroimaging of the CNS. The key findings of these studies can be summarized as follows:
- In several chronic pain conditions, regional changes in gray matter density have been demonstrated with anatomical imaging. Even though the underlying neuroanatomical changes remain to be determined, these findings have great potential to function as endophenotypes for persistent pain conditions, or even as biomarkers for individual syndromes;
- Alterations in brain state and response of the brain circuits to drugs have been demonstrated with PET and functional MRI (fMRI);
- Changes in neurotransmitter levels (glutamate, aspartate, glycine and γ-amino butyric acid [GABA]) have been shown with magnetic resonance spectroscopy (MRS) .
Borsook et al.
] have insightfully described how the use of fMRI, in particular, may help speed drug development for CNS indications at a number of levels that include:
- Evaluation of differential efficacy of drugs within and across pharmacological subtypes;
- Identification of potential for CNS side effects;
- Opportunities to define drug dosing and benefits of drug combinations;
- Potential for surrogate models using healthy subjects for drug evaluation;
- Setting up a potential method for re-evaluating failed drug candidates;
- An objective method to select and stratify patient populations to enable pro of-of-concept clinical investigations .
The potential applications of neuroimaging provide many opportunities for bidirectional translation between humans and rodents, which may help speed drug development for chronic visceral pain states, including IBS. An underlying assumption in this proposition is that pharmacologic subtypes and/or side effect profiles show specific brain mapping ‘signatures’ that are similar in humans and animals.
One of the potential strengths of integrating neuroimaging during the drug development process is its potential to translate findings of alterations in neural circuits across species, enabling a more focused use of animal models in research. Neuroimaging may also serve as a useful proxy measure of pain responses that, in the animal, cannot be elicited verbally. While pain imaging of the CNS has been extensively explored in human subjects, neuroimaging technologies applied to rodents to study endophenotypes of persistent pain in animals is still in its infancy. A significant gap remains in ‘bedside-to-bench’ translation of well-studied brain-mapping abnormalities in IBS patients (reviewed in [91
The choice of imaging modalities in animal models
While structural imaging has the potential to greatly increase our understanding of the functional neuroanatomy of chronic pain conditions, the current lack of understanding of the mechanisms underlying such structural changes, and the temporal characteristics of these changes, makes it currently impractical to use such end points for drug development in rodents. In this regard, functional brain mapping and chemical imaging may represent more suitable approaches. Ideally, such imaging in animals would be performed under conditions that approximate those used in human subjects – that is, nonsedated animals with minimal interference with the subject’s natural behavior. At the same time, the ideal imaging modality would optimize spatial and temporal resolution, allow for serial measurements across time, while providing 3D views of brain function. No method simultaneously meets all these criteria.
Past research on brain responses to noxious visceral stimulation in animals has relied predominantly on the measurement of early response genes, in particular c-Fos expression [92
], with a broad variability reported between laboratories [92
]. Unlike human imaging studies evaluating brain responses during acute CRD, c-Fos studies typically use prolonged exposure (>30 min) to high-intensity visceral stimuli, which may lead to the integration of a variety of nonspecific behaviors over the duration of pain exposure, including acute sensitization of the visceral afferent system. Furthermore, analysis is often limited to a few selected brain regions, lacking the whole-brain level analysis achieved in human studies. Thus, it is not surprising that translation of findings between human and animal brain mapping has been diffcult. It is noteworthy that studies examining increases in c-Fos expression in response to CRD in the lumbosacral region of the spinal cord have reported more consistent results within animals [92
], but parallels to human imaging have not been explored extensively. Other region-specific analyses of neuronal responses to CRD have been carried out using in vivo
electrophysiological recording [101
], and such brain electrical recordings may prove useful once specific brain regions of vulnerability have been determined. Spatial resolution, with microPET and advanced image reconstruction software, remains at best approximately 1.2 mm at the center of the field of view. This represents approximately 7% and 13% of the width of the rat and mouse brain, respectively, and is poorly suited for the detection for all but the broadest changes in regional cerebral blood flow (rCBF) or metabolism in rodent models. In specific instances, however, such broad changes may suffice for the testing of specific compounds, as has been demonstrated for opiates [103
]. Functional MRI and single photon emission computed tomography (SPECT, nanoSPECT), though they provide whole brain analysis and adequate temporal and/or spatial resolution, require sedation of the animal, limiting the types of brain responses that can be examined [97
]. We have advocated in the past the use of autoradiographic methods of perfusion mapping, as this method can be applied in awake, nonrestrained animals and yield information at the circuit level across the entire brain, with a spatial resolution (~100 µm) appropriate for the rat or mouse models, and a temporal resolution (seconds–minutes) sufficient for capturing acute brain changes. Nevertheless, autoradiographic methods, although they provide 3D spatial information, contain no information about dynamic cerebral changes. Therefore, studies of disease progression or response to treatment using intra-animal comparisons cannot be performed. In addition, because of the need for extensive cryosectioning of the brain, the method lends itself less well to high-throughput screening than perhaps MRI, nanoSPECT or microPET.