Epilepsies are a heterogeneous group of neurological disorders that can be divided into symptomatic, presumed symptomatic, and idiopathic epilepsies based on the underlying etiology (Engel, 2002). They are characterized by a transient and recurrent failure of normal brain function and simultaneous activation of a large population of neurons resulting in electrographic and/or behavioral seizure activity (Engel, 1989). Epileptic process in symptomatic epilepsies often includes an initial brain damaging insult such as head trauma, stroke, brain infection, or status epilepticus. This is followed by a latency period that can last from a few weeks to several years, during which a large number of various neurobiological changes can take place, including neuronal death, gliosis, neurogenesis, axonal sprouting and injury, reorganization of extracellular matrix, and vascular changes. These alterations eventually lead to the appearance of spontaneous seizures, that is, development of epilepsy. Experimental evidence suggests that some of these cellular changes can continue even after epilepsy diagnosis (Pitkanen and Sutula, 2002). Magnetic resonance (MR) techniques have shown great potential for noninvasive detection of structural, metabolic, and functional abnormalities during the epileptic process. In the present review, we will mainly focus on structural magnetic resonance imaging (MRI) in animal models of symptomatic epilepsies, including quantitative relaxation and diffusion MRI studies, while other publications in this supplement will more widely cover topics such as functional and metabolic imaging.
Summary: Small animal magnetic resonance imaging (MRI) has opened a window through which brain abnormalities can be observed over time in rodents noninvasively. We review MRI studies done during epileptogenesis triggered by status epilepticus in rat. Most of these studies have used quantitative T2, diffusion, and/or volumetric MRI. The goal has been to identify the distribution and severity of structural lesions during the epileptogenic process, that is, soon after status epilepticus, during epileptogenesis, and after the appearance of spontaneous seizures. Data obtained demonstrate that MRI can be used to associate the development of brain pathology with the evolution of clinical phenotype. MRI can also be used to select animals to preclinical studies based on the severity and/or distribution of brain damage, thus making the study population more homogeneous, for example, for assessment of novel antiepileptogenic or neuroprotective treatments. Importantly, follow-up data collected emphasize interindividual differences in the dynamics of development of abnormalities that could have remained undetected in a typical histologic analysis providing a snapshot to brain pathology. A great future challenge is to take advantage of interanimal variability in MRI in the development of surrogate markers for epilepsy or its comorbidities such as memory impairment. Understanding of molecular and cellular mechanisms underlying changes in various MRI techniques will help to better understand complex progressive pathological processes associated with epileptogenesis and epilepsy.
BASIC PRINCIPLES OF SIGNAL FORMATION, LOCALIZATION, AND IMAGE CONTRAST FORMATION
While the complete description of the physical principles of signal formation, localization, and MRI contrast formation is beyond the scope of this review article, it is useful to understand some basic principles in order to evaluate the limitations and benefits of MRI. MRI is based on a physical phenomenon at the level of atomic nuclei called nuclear magnetic resonance (NMR). When an object is placed into a high static magnetic field, signal from hydrogen nuclei (or other nuclei with nonzero spin quantum number) can be detected after disturbing populations of nuclei in the different energy states in the sample by applying a radio frequency electromagnetic pulse. When the system returns toward equilibrium it produces detectable NMR signal. Localization of the signal source can be performed using magnetic field gradients. The contrast in the image is determined by multiple factors, related to the amount and mobility of water molecules, which can be tailored during measurement.
Some of the physical principles presented above favor the suitability of MRI for experimental studies of epilepsy. MRI is operating in the radio frequency range of the electromagnetic spectrum, and thus has very little interaction with tissue, resulting in excellent penetration and no adverse effects. This allows noninvasive repeated longitudinal studies in a manner that is not possible for imaging methods that exploit ionizing radiation. The signal in MRI originates most often from hydrogen nuclei, which are the most abundant nuclei in the body because of the high water content of soft tissue. This makes NMR signal relatively strong despite the fact that in the currently used magnetic field strengths only 1 out of 105 nuclei is active in the signal formation. When this relatively high signal is localized using magnetic field gradients, high resolution images with good signal-to-noise ratio can be obtained. The typical resolution in small animal MRI imaging is in the order of 100 μm in plane and 1 mm in slice thickness, but can be squeezed down to 50 μm isotropic resolution with optimal combination of hardware and measurements techniques.
The contrast in MRI is determined both by the amount of hydrogen nuclei (water content) in the sample as well as many other features associated with dynamic behavior of water within the time scale of MRI measurement (millisecond to second). The water dynamics influence so-called relaxation times (T2, transversal; T1 longitudinal relaxation time) that determine how hydrogen nuclei returns toward equilibrium after excitation by a radio frequency pulse. By modifying the excitation and acquisition technique (so-called pulse-sequence) the contrast can be sensitized to relaxation and diffusion of water, blood flow, magnetic coupling between free and bound water, and many other endogenous contrast mechanisms. The contrast can be further modified by using exogenous contrast agents that alter the relaxation properties of water in the tissue and/or blood. Contrast agents are most often composed of paramagnetic substances such as Gd3+ chelates and Mn2+ that shortens predominantly T1 relaxation time producing hyperintensity in T1-weighted imaging, or superparamagnetic iron oxide particles that produce negative contrast in T2-weighted imaging. Considering of all imaging contrasts available in the tissue, it is reasonable to conclude that MRI is the most versatile of the current biomedical imaging methods and provides an armamentarium of techniques that can be potentially used for in vivo characterization of different phases of the epileptogenic process.
MRI STUDIES IN ANIMAL MODELS OF ACQUIRED EPILEPSY—FOCUS ON STATUS EPILEPTICUS-INDUCED EPILEPTOGENESIS
In rodents, epileptogenesis can be triggered by various clinically relevant etiologies such as inducing abnormalities in genes encoding the development of cortex (Engel, 1989; Lee et al., 1997), status epilepticus, stroke, or traumatic brain injury (for a review, see Pitkanen et al., 2007). So far, most of the MRI data come from studies in which the epileptogenic process has been initiated by chemically or electrically induced status epilepticus (Table 1), and data were collected during a relatively short period following the induction of prolonged seizure activity.
|Major finding||Animal model||Time after induction||Reference|
|Increased T2-weighted signal as a consequence of vasogenic oedema||Kainic acid injection into rat striatum||1 day||King et al., 1991|
|Cytotoxic oedema caused by seizures can be detected by diffusion-weighted imaging||Intraperitoneal bicuculline injection in rat||During status epilepticus||Zhong et al., 1993|
|T2 increase without neurodegeneration||Prolonged febrile seizure in postnatal day 11 rat pups||24 h – 8 days||Dube et al., 2004|
|Delayed diffusion increase associated with delayed tissue damage||Kainic acid injection into rat hippocampus||28 days||Tokumitsu et al., 1997|
|Initial T2 increase is followed by normalization and secondary increase during epileptogenesis||Intraperitoneal lithium-pilocarpine injection in rat||6 h – 9 weeks||Roch et al., 2002|
|In vivo detection of mossy fiber plasticity by manganese enhanced MRI||Intraperitoneal kainic acid injection in rat||1 month||Nairismagi et al., 2006b|
The first NMR studies performed in seizure models dates back to the 1980s (Petroff et al., 1984). These studies used 31P NMR-spectroscopy to monitor energy metabolism during bicuculline induced status epilepticus, and thus did not really fill the criterion of “imaging” studies. However, these studies demonstrated decreased phosphocreatine levels without changes in the levels of ATP and pioneered the idea that NMR techniques can be used for in vivo detection of seizure consequences. The first MRI studies in epilepsy models were conducted in the early 1990s (Karlik et al., 1991; King et al., 1991). Status epilepticus was induced by chemical convulsants such as kainate or pilocarpine, and animals were followed with MRI for a few days. First papers described increased signal intensity in T2-weighted imaging, consistent with increased T2 relaxation time of water (King et al., 1991; Zhong et al., 1993). These changes were typically seen for a few days after induction of status epilepticus and were attributed to edema formation (King et al., 1991; Bouilleret et al., 2000) or tissue damage.
At the time of the first experimental MRI studies in epilepsy, a new MRI technique called diffusion-weighted imaging (DWI) became available. It is based on sensitizing the MRI measurement to water diffusion with pulsed magnetic field gradients. DWI showed great potential in detecting acute cerebral ischemia earlier than T2-weighted imaging (Moseley et al., 1990). Not surprisingly, the technique was rapidly also applied to epilepsy research. Zhong et al. were among the first to demonstrate that diffusion MRI can detect brain alterations in rats with status epilepticus earlier than methods measuring T1 and T2 relaxation changes (Zhong et al., 1993). Diffusion was decreased by 14–18% from normal only minutes after bicuculline injection. Interestingly, this is within the timeframe when the changes are observed in acute cerebral ischaemia after occlusion of the feeding vessel. Several follow-up studies confirmed these early findings in different animal models of induced seizures (Righini et al., 1994; Nakasu et al., 1995; Zhong et al., 1995; Wang et al., 1996; Wall et al., 2000). These studies also showed that initial diffusion decrease typically normalizes within 1–7 days. The speed of recovery varied between the brain areas. For example, Wall et al. (2000) reported that pilocarpine induced seizures resulted in acute diffusion decrease in the amygdala and the piriform cortex, which are the primary affected areas, and that the diffusion decrease normalized within 24 h. In the hippocampus that was not acutely damaged in the model, diffusion remained normal during the first hours after bicuculline injections and increased only at 24 h. Importantly, these data show that delayed processes launched by status epilepticus may lead to delayed diffusion increase in select brain areas during the progression of epileptic process (Tokumitsu et al., 1997; Nairismagi et al., 2004).
Recent quantitative relaxation and diffusion studies have confirmed the early findings showing changes in diffusion and T2 relaxation after status epilepticus. More detailed analyses have also shown that the MRI alterations depend, for example, on the animal model, time point selected for data collection, and the brain region analyzed. Interestingly, a deviation from the general view of increased T2 following status epilepticus has been recently shown by several research groups. Bhagat et al. found that T2 was decreased at 3 h to 12 h after soman induced status epilepticus in rat (Bhagat et al., 2001). Similar T2 decrease was also detected in the lithium-pilocarpine model at 3–5 h after induction of prolonged seizure activity (van Eijsden et al., 2004). This was attributed to so-called negative BOLD effect, which is shown to occur in the rat brain when the ratio of oxygen consumption and oxygen delivery increases from that at normal physiological state leading to increased concentration of paramagnetic deoxyhemoglobin in blood (Payen et al., 1996; Grohn et al., 1998). The finding was confirmed by measuring increased lactate concentration by MRS as a marker of anaerobic metabolism (van Eijsden et al., 2004). The interpretation of T2 increase a few days after status epilepticus as a marker of irreversible cell death has also been recently challenged (Dube et al., 2004; Nairismagi et al., 2004). In a rat model of prolonged febrile seizures, Dube et al. showed that increased T2 values in the limbic regions at 24 h after induction of seizure at P10 were not associated with neurodegeneration, as assessed using the Fluoro-Jade B method (Dube et al., 2004). Dissociation of T2 changes from cell death was also found by Nairismägi et al. (2004) who reported increased T2 2 days after induction of status epilepticus with electrical stimulation of the amygdala. However, T2 normalized during the following days when there was still an active ongoing neurodegeneration as shown by Fluoro-Jade B staining (Pitkanen et al., 2002). Furthermore, T2 increase in the hippocampus was recently detected in the rat amygdala kindling model, which is associated with minimal cell loss and tissue destruction (Jupp et al., 2006). These studies show that interpretation of the MRI findings have to be done with great caution, and conclusions about the cellular substrates of MRI alterations cannot be adopted directly from, for example, stroke studies where significantly increased T2 at 24 h after insult is generally accepted as a marker of irreversible cell or tissue damage (Welch et al., 1995).
LONG-TERM MRI FOLLOW-UP STUDIES DURING EPILEPTOGENESIS AND EPILEPSY
The epileptogenesis phase is an attractive target for MRI studies considering the large number of cellular alterations that take place during this phase of the epileptic process. In addition to neurodegeneration, there are alterations in axons that can sprout, degenerate and change their orientation. Blood vessels can undergo morphologic changes and angiogenesis. Finally, the extracellular matrix is constantly being reshaped to allow the sprouted axons and newly-born neurons to move to their target sites. As our literature review shows, there are only a small number of studies that have tried to address these alterations with MRI during epileptogenesis. Rather, the focus has been on assessment of the occurrence and extent of neurodegeneration.
Volumes of the hippocampus, cingulate cortex, retrosplenial granular cortex, and lateral ventricles were recently determined with high intrarater reliability at 10 days after kainic acid-induced status epilepticus, that is, during the time when animals are undergoing epileptogenesis or some of the rats can already have spontaneous seizures (Wolf et al., 2002). Authors reported ∼12% decrease in hippocampal volume and increased ventricle volume, which is consistent with hippocampal sclerosis.
A few longer-term studies analyzing the volumetric changes during the epileptic process have been conducted. Status epilepticus induced using Li-pilocarpine approach at P12 resulted in volume reduction in the temporal lobe regions in a subpopulation of rat when assessed 3 months later (Nairismagi et al., 2006a). In the amygdala stimulation model, progressive decrease in the thickness in the piriform cortex/amygdala region was detected starting at day 9 after stimulation and continuing throughout the 6 month observation period (Nairismagi et al., 2004). Furthermore, gradually increasing volume of the cortical lesion and ventricles accompanied by decreasing hippocampal volume was recently detected in a 6 month longitudinal study following traumatic brain injury (Koskinen et al., 2005), which is known to initiate epileptogenesis in a subpopulation of animals (Kharatishvili et al., 2006). Decreased hippocampal volume accompanied by decreased volume of the thalamus, putamen, cortex, and perihinal/entorhinal/piriform cortices was also detected 60 days after pilocarpine-induced status epilepticus (Niessen et al., 2005). Interestingly, the severity of hippocampal volume loss correlated with the severity of impairment of spatial learning in the Morris water-maze test. These studies clearly indicate that progressive atrophy, which is most likely associated with neuronal death, coincides with epileptogenesis for several months after initial insults.
There are several studies reporting the progressive increase of lesion size during epileptogenesis although the exact anatomical extent of the lesion has not usually been analyzed. This is obviously because a more complex analysis is required. Furthermore, it is difficult to produce statistically meaningful information from animal models that typically display relatively large variation in lesion size and location. However, it is possible that the anatomical location is a more important factor for epileptogenesis than the lesion size. The naturally occurring variability in the lesion extent may, however, provide important information about the time-dependent involvement of different brain regions to epileptogenesis.
Quantitative relaxation and diffusion MRI is, in general, a more sensitive indicator of tissue changes than volumetric MRI. Long-term studies (>2 weeks) indicate that initial changes caused by status epilepticus or other initial insults are often reversible (Roch et al., 2002; Dube et al., 2004; Nairismagi et al., 2004; Koskinen et al., 2005). However, secondary changes in these parameters occurring weeks after status epilepticus often take place either before or at the time of the appearance of spontaneous recurrent seizures. Tokumitsu et al. (1997) were among the first groups describing the delayed response. Their study demonstrated increased water diffusion at 28 days after intra-hippocampal kainate injection. In the pilocarpine model, reversible acute T2 increases were detected during the first 24 h, followed by normalization over a 5-day period and secondary increase in chronic phase several weeks later (Roch et al., 2002). Similar results were obtained in our recent study in the amygdala stimulation model (Nairismagi et al., 2004). Both the primary focal area (amygdala) as well as the monosynaptically and polysynaptically connected areas (piriform cortex, midline thalamus and hippocampus) showed increased T2, T1ρ, and diffusion values 2 days after status epilepticus induction. Most of these abnormalities were reversed by day 9 and abnormalities reappeared after day 20. Interestingly, a novel relaxation parameter called T1ρ (Sepponen et al., 1985), which has been shown to be a sensitive indicator of cell death in stroke (Grohn et al., 1999) and glioma gene therapy (Hakumaki et al., 2002) was analyzed in this study and found to be slightly more sensitive to tissue changes than conventional relaxation parameters. All animals in the study expressed spontaneous seizures 3 months after amygdala stimulation. The deviation of MRI parameters greatly increased with the onset of spontaneous seizures and was attributed to transient changes in relaxation and diffusion caused by recurrent seizures that could occur in clusters of 20 per day. This is probably one of the reasons why it was difficult to correlate the MRI parameters with histological findings.
Currently there are no experimental studies that had systematically taken into account the influence of spontaneous seizures on quantitative MRI parameters, which clearly compromises the interpretation of long-term studies in spontaneously seizing animals. Without verification of the occurrence of the last seizure relative to MRI by using continuous (video-)EEG monitoring it is impossible to differentiate to what extent the MRI changes are due to slowly progressing tissue alterations, and to what extent due to transient changes caused by seizures. Ideally, EEG should be recorded also during MRI measurement. More data are needed about the effects of various anesthetics used during MRI scanning on spontaneous seizures. In fact, repeated anesthesia can be a confounding factor in long-term studies because of its possible effects on epileptogenesis.
TIME COURSE AND HISTOPATHOLOGICAL CORRELATES OF RELAXATION AND DIFFUSION CHANGES
By combining data from several different models, we can picture a general pattern of the behavior of quantitative MRI parameters after status epilepticus (Fig. 2). Interestingly, the pattern of changes resembles that presented for cerebral ischemia (Welch et al., 1995). However, there are some exceptions. First, unlike in stroke, an increased T2 relaxation measured a few days after status epilepticus is not necessarily a marker of irreversible cell death. Second, the time course of the secondary increase of diffusion can be delayed for up to several months.
The interpretation of histopathological correlates of relaxation and diffusion changes is a demanding task. Both relaxation and diffusion probe the mobility and interaction of water molecules with different cellular and molecular structures in a complex environment like epileptogenic tissue undergoing circuitry reorganization. Many changes during epileptogenesis and epilepsy, including cytotoxic and vasogenic edema, cell death, gliosis, neurogenesis, axonal and dendritic plasticity, reorganization of extracellular matrix, and angiogenesis are likely to influence relaxation and/or diffusion of water molecules. In most cases, it is probably not possible to explicitly link any quantitative MRI changes to any single process that takes place during progressive pathology. Still it may be possible to find correlations in certain specific cases and use these as a surrogate marker, for example, for cell death or gliosis.
In the primary affected area, diffusion drop takes place almost immediately. The exact mechanism underlying the diffusion drop is still debatable. However, there is reasonable consensus in the MRI community that initial diffusion drop in cerebral ischemia, and most likely also in status epilepticus, is due to depolarization of the cells leading to the cytotoxic edema, in which extracellular water enters the intracellular space (van der Toorn et al., 1996; Eidt et al., 2004). As water diffusion in the intracellular space is more restricted than in the extracellular space, and simultaneously the tortuosity of the extracellular space increases, the net result is decreased diffusion. There are evidences that this may be accompanied also with reduced diffusivity in the intracellular space and changes in intracellular streaming (Duong et al., 1998). Diffusion increase in tissue is most often associated with decreased amount of cellular membranes and other normal cellular barriers that restrict water displacement due to Brownian motion. In stroke, diffusion increase is associated with cell-death (Welch et al., 1995). In tumor treatment studies, almost a linear correlation has been found between the diffusion coefficient and cell density (Kauppinen, 2002). Also in epilepsy studies, increased diffusion seems to occur simultaneously with several destructive processes leading to cell-death, and is found to be elevated in the anatomical regions where the most severe tissue damage has been found in histology (Tokumitsu et al., 1997; Nairismagi et al., 2004). Proliferating glial populations may also contribute to the time course of diffusion changes.
Increased T2, which generally peaks at 1–2 days after status epilepticus, is most often associated with vasogenic oedema formation. This is a plausible explanation as free water content in tissue is strongly correlated with relaxation times, and profound oedema in tissue is likely to overwhelm more subtle cellular changes that may also contribute to relaxation. Resolution of the oedema leads to normalization of the T2 relaxation. A few recent studies compared delayed T2 increase with histopathology (Roch et al., 2002; Dube et al., 2004; Nairismagi et al., 2004). T2 increase coincided with gliosis and neurodegeneration. Any change in tissue leading to increased rotational mobility of water and/or water content will lengthen T2 relaxation time, and thus it is understandable that it was impossible to unequivocally link T2 increase to any specific cellular level change.
BEYOND CONVENTIONAL STRUCTURAL MRI
It is evident that conventional structural MRI can provide information about spatiotemporal involvement of different anatomical regions immediately after status epilepticus, during epileptogenesis, and epilepsy. However, beyond detecting overall tissue abnormalities, specificity of the relaxation and diffusion-based MRI is not very good, and as we have discussed, several different kinds of alterations in the tissue can cause similar changes in these quantitative MRI parameters.
One way to increase specificity of MRI is to use contrast agents. We recently used manganese-enhanced MRI (MEMRI) (Lin and Koretsky, 1997; for more details of this technique see review by Obenaus and Jacobs in this supplement) for the detection of one of the hallmark features of temporal lobe epilepsy, namely mossy fiber sprouting (Nairismagi et al., 2006b). Even though its role in the development of epilepsy is still under dispute, there is vast evidence that mossy fiber sprouting is present both in tissue resected from patients with drug-refractory temporal lobe epilepsy as well as in the most commonly used animals models of acquired epilepsy (Mathern et al., 2002; Pitkanen et al., 2007). MnCl2 was injected into the enthorhinal cortex, from where paramagnetic Mn2+ was actively transported via the perforant pathway to the mossy fiber pathway (Fig. 3). This resulted in specific labeling of mossy fibers making cellular level targets visible in T1-weighted MRI. Two weeks after kainic acid injection, increase in enhanced area in the dentate gyrus and in the CA3 subfield was correlated with histopathological assessment of the severity of mossy fiber sprouting. This example clearly demonstrates that if contrast agents can be targeted to a specific cell type (or even to molecular targets) (Bulte and Kraitchman, 2004) then the specificity of MRI can be greatly improved.
Diffusion tensor imaging (DTI) is a method that allows visualization of different anatomical features as compared to conventional MRI contrast (Basser et al., 1994; Mori and Zhang, 2006). For more details of this technique, see review by Obenaus and Jacobs in this supplement. So far, DTI has not been widely used in experimental MRI studies in the epilepsy field. However, recent demonstrations of unique contrast obtainable, for example, in the rat hippocampus ex vivo (Shepherd et al., 2006) make it a very attractive method to study cellular level reorganization in epilepsy.
We have reviewed results obtained using MRI methods that are sensitized to structural tissue changes. One of the greatest advantages of MRI is that it can combine several different measurements within the same imaging session. Not only structural but also functional and metabolic information can be obtained by MRI. Indeed, drawing a strict line between structural and functional techniques is in some cases artificial. BOLD fMRI and dynamic contrast-enhanced (DCE) perfusion techniques can detect direct hemodynamic consequences of seizures (Tenney et al., 2003; Engelhorn et al., 2005) and also potentially depict slowly progressing changes in vascular system and metabolism. In humans, also changes related to resting state activity of the brain (Waites et al., 2006) have been detected. Magnetic resonance spectroscopy (MRS) is capable of detecting cell specific metabolites that can help to differentiate, for example, between neuronal loss and gliosis. The resolution of the spectroscopic imaging techniques (MRSI) has improved and provides spatial localization in the order of 5 μl (Liimatainen et al., 2006). Novel molecular imaging approaches may allow imaging of specific molecular targets in future. While this may become available also by MRI, currently the sensitivity of positron emission tomography (PET) and single photon emission computed tomography (SPECT) provide better means for molecular imaging. Multimodal imaging taking full advantage of complementary information available from different in vivo imaging modalities is becoming available in biomedical imaging centres dedicated to small animal imaging, and is likely to increase our understanding of complicated progressive disease processes associated with the development of epilepsy.
Overall, the number of experimental MRI papers related to epilepsy is relatively small if compared, for example, to MRI studies in experimental stroke. Undoubtedly, there are a large number of relevant questions that can be addressed by combining modern MRI techniques with carefully validated animal models. The use of paramagnetic tracers or DTI in detection of changes in axons, blood vessels and the extracellular matrix, as well as taking advantage of multimodal imaging, remain to be explored in epilepsy research. The variability in the lesion size between the animals undergoing epileptogenesis provides an exiting opportunity to use MRI techniques for identification of surrogate markers for epileptogenesis or development of comorbidities in individual animals. It remains to be seen if such techniques can also be used to follow-up the effects of new antiepileptogenic and neuroprotective treatments. Finally, an important feature of MRI techniques is that they are relatively easily transferable from experimental to clinical settings—a critical issue when translating data from laboratory to clinic.
Acknowledgment: Research by authors is supported by Academy of Finland (AP, OG), Emil Aaltonen Foundation (OG), and the Sigrid Juselius Foundation (AP). We thank Riikka Immonen and Jaak Nairismagi for obtaining MRI data for Figs. 1 and 3.