Functional MRI and Diffusion Tensor Imaging of Brain Reorganization After Experimental Stroke

Functional MRI methods are traditionally sensitized to changes in cerebral hemodynamics in response to neuronal activity. The most frequently applied fMRI method is blood oxygenation level-dependent (BOLD) gradient-, or spin-echo MRI. Alternative techniques include arterial spin labeling (i.e., cerebral blood flow-weighted MRI) and steady-state susceptibility contrast-enhanced MRI in combination with a (super)paramagnetic blood pool agent [i.e., cerebral blood volume (CBV)-weighted MRI]. The principles of these different fMRI methods have been extensively described by Mandeville and Rosen [14] and Harel et al. [15].

Functional MRI enables assessment of human brain activation in relation to sensory, motor, or cognitive tasks or stimuli. Besides, fMRI can also be applied in animal experiments. Despite the fact that anesthesia (type, depth, and duration) may significantly influence neurovascular coupling [16], fMRI in animals has been successfully executed with anesthesia protocols that preserve hemodynamic responsiveness to neuronal activity [17]. Experimental fMRI studies have provided important insights into the physiological basis of fMRI responses, fundamentals of hemodynamic and metabolic aspects of neuronal activation, and consequences of lesions on cerebral activity [17, 18], which will be discussed below in relation to stroke injury and recovery.

We and others have applied fMRI in rodent stroke models to assess changes in sensorimotor system activation in response to limb stimulation [19‐27]. To maximize the sensitivity and specificity of the detection of activation responses, most of these fMRI studies have taken advantage of high magnetic field strength (≥7 T) [22, 23, 26] and/or a (super)paramagnetic blood pool agent [19‐25]. This not only increases the signal-to-noise and/or contrast-to-noise ratio of the fMRI signal, but also reduces the contribution of unspecific signal from large blood vessels [14, 15]. Consistent with fMRI studies in stroke patients, these fMRI experiments in rodent stroke models have revealed diminished activation in the ipsilesional sensorimotor cortex [19‐21, 24, 26, 27], enhanced contralesional activation [19‐21], and chronic reinstatement of perilesional activity [20, 21, 26, 27].

Functional MRI-detected activation response to forelimb stimulation may be transiently absent in the representational somatosensory cortex after experimental stroke, despite normal structural appearance on diffusion- and T₂-weighted MR images and intact vasoreactivity [20, 21, 26, 27]. Functional denervation, depression, or deterioration of surviving neurons, as well as derangement of neurovascular coupling, could explain the temporary lack or responsiveness. The preservation or reinstatement of activation within the ipsilesional sensorimotor cortex has been found to be associated with functional recovery after transient middle cerebral artery occlusion in rats [21, 26], which emphasizes the significance of reorganization of affected cortical representational fields for post-stroke functional recovery.

Some studies have reported a rise in contralesional fMRI activity after experimental unilateral cerebral ischemia [19‐21], which correlated with the extent of tissue injury in the ipsilesional sensorimotor cortex [21]. The relevance of enhanced activation in the unaffected hemisphere remains controversial; it may hypothetically contribute to functional recovery, but could also be a direct pathophysiological consequence (e.g., due to broad disinhibition) following stroke.

To assess neuronal signal transmission between the contralesional and ipsilesional sensorimotor cortex after unilateral stroke, we have measured perilesional BOLD signals measured in response to direct intracortical stimulation in the contralesional hemisphere of rat brain [28]. Despite disturbed sensorimotor function in the first days after stroke, BOLD responses were preserved in the perilesional sensorimotor cortex, although the time-to-peak was significantly delayed, possibly due to deficient neurovascular coupling (Fig. 1). These results indicated that interhemispheric functional connectivity is sustained and that—in spite of dysfunction—perilesional cortical tissue can be activated subacutely after unilateral stroke.

Besides assessment of neural correlates of functional deficits and recovery, fMRI may also provide a means to evaluate treatment effects on brain function after stroke. For example, restoration of activation responses to electrical forelimb stimulation has been demonstrated with fMRI in rats treated with a Ca²⁺ antagonist after permanent MCA occlusion [25]. In studies on the therapeutic efficacy of delayed albumin or lithium administration in a rat stroke model, enhanced fMRI activity was detected in the perilesional somatosensory cortex in albumin- or lithium-treated animals, while lesion size was not different from saline-treated animals [22, 23].

An alternative fMRI method, known as resting-state fMRI (rs-fMRI), assesses spatial functional correlations within neural networks without the need of a stimulation paradigm [29]. During the last decade, rs-fMRI has been increasingly applied as a tool to study alterations in the brain’s intrinsic functional architecture as potential physiological correlates of neurological and psychiatric disorders [30‐32]. Spontaneous fluctuations in baseline (“resting-state”) neuronal signaling are reflected in low-frequency fluctuations (<0.1 Hz) of the BOLD signal and show temporal coherence between anatomically connected brain regions within a particular neuronal network, such as the sensorimotor system [29, 33, 34]. Throughout the gray matter, the extent of synchronization between these low-frequency BOLD fluctuations is related to functional connectivity. Correlation of these signals with electroencephalographic brain activity has indicated that these slow hemodynamic fluctuations are associated with neuronal function [35‐37].

Several groups have shown that functional connectivity measurements with rs-fMRI are feasible in anesthetized rats [38‐42], which has formed the basis of our recent rs-fMRI studies in rats to depict spatiotemporal characteristics of reorganization in functional neuronal networks after ischemic brain injury [43, 44]. We found that strong correlations between rs-fMRI signals in ipsilateral and contralateral cortical sensorimotor regions disappear in the first days after unilateral stroke in rats, indicative of loss in interhemispheric functional connectivity [44], which is in correspondence with findings in human patients subacutely after stroke [45]. In subsequent weeks and coinciding with recovery of sensorimotor function, interhemispheric functional connectivity was partially retrieved [44] (see Fig. 2a). On the other hand, intrahemispheric functional connectivity between primary somatosensory and motor cortices was preserved in the lesion borderzone and moderately enhanced contralesionally, which may reflect intact and/or new formations of local neuronal connections. This was supported by additional manganese-enhanced MRI (MEMRI) experiments [43]. MEMRI is an MRI-based in vivo tract tracing method that is based on the detection of neuronal distribution of paramagnetic Mn²⁺, a Ca²⁺ analog that can enter active neurons through Ca²⁺ channels [46]. Injection of Mn²⁺ in the brain has been shown to result in neuronal uptake and transsynaptic transport along connective pathways [47]. In our combined rs-fMRI and MEMRI study, we found that injection of MnCl₂ in the contralesional sensorimotor cortex resulted in locally increased uptake, indicative of locally enhanced structural connectivity between neurons, as well as decreased transcallosal transfer, corroborative of the loss of interhemispheric functional connectivity [43]. These findings underline the link between functional and structural reorganization of neuronal networks after stroke.

Functional connectivity has been most often computed by calculation of correlations between the low-frequency BOLD fluctuations in an a priori-selected brain region and low-frequency BOLD fluctuations from other voxels in the brain. Beside region-based functional connectivity analyses, functional connectivity may also be assessed at a whole network level by means of graph analysis of rs-fMRI data [32, 48, 49]. Networks can be represented as graphs containing nodes and edges. In the case of rs-fMRI, the image voxels represent the nodes and the intervoxel correlations between low-frequency BOLD fluctuations represent the edges. The network’s structure can be assessed by measuring its clustering coefficient, a measure of segregation that reflects the degree to which nodes are clustered, and the shortest path length, a measure of integration that reflects the minimum number of edges between any pair of nodes. A high clustering coefficient and low average shortest path length indicate a small-world network topology, which is proposed to be an optimal network configuration for global information transfer and local processing [48‐50]. Graph-based analysis of rs-fMRI data from chronic human stroke subjects showed a decrease in clustering coefficient and a shift towards a random network configuration in the motor system [51]. Graph analysis of our rs-fMRI data from rats recovering from unilateral stroke points towards increases in clustering coefficient, average shortest path length, and small-world index in the intact bilateral sensorimotor cortices at early stages, which normalize chronically (Fig. 2b–d) (unpublished observations). This emphasizes that large-scale networks outside the ischemic area rearrange over time. Speculatively, higher values of the calculated network parameters subacutely after stroke may be a sign of diffuse overconnectivity that precedes maturation towards a normally functioning circuitry within the reorganized network.

Abovementioned fMRI studies in experimental stroke models are based on measurement of cerebral activity under baseline conditions or in responses to somatosensory stimuli. Alternatively, neuronal activation may be evoked by pharmacological stimulation. Pharmacological MRI (phMRI) is a variant of fMRI that assesses hemodynamic changes associated with cerebral activity in response to centrally acting pharmacological agents [52]. A pioneering phMRI study in rats after stroke has shown a strongly diminished cerebral activation response to injection of the GABA_A antagonist bicuculline, indicative of functionally compromised brain tissue [53]. To investigate the long-term effect of focal cerebral infarction on the cholinergic system, which has been implicated in post-stroke cognitive impairment, we have carried out phMRI in combination with administration of pilocarpine—a muscarinic cholinergic agonist—to measure the activation of the cholinergic system after photothrombotic infarction in rat cerebral cortex [54]. Hemodynamic responses to cholinergic muscarinic receptor stimulation were significantly reduced inside the ischemic area at 1 and 3 weeks and in areas distant from the lesion at 3 weeks after right-sided frontal cortical infarction, while cholinergic cell number was unchanged. Furthermore, we have observed reduced cholinergic muscarinic receptor responsiveness in structurally preserved somatosensory projection cortex at 2 weeks after subcortical ischemic damage induced by transient occlusion of the right middle cerebral artery in rats (Fig. 3) (unpublished observation). These phMRI data demonstrate significantly altered cholinergic receptor responsiveness distant from focal cerebral ischemic infarction.

The previous sections deal with functional MRI methods that can be employed to identify altered patterns of brain activity after stroke. Changes in functional brain organization, however, are often closely associated with structural modification of neuronal elements in the brain. DTI offers an MRI-based means for the assessment of neuroanatomical changes associated with brain injury and repair. DTI informs on the three-dimensional displacement of tissue water, mathematically characterized by an effective diffusion tensor consisting of nine matrix elements, which can be exploited to assess the microstructure of gray and white matter tissue (see Basser and Jones [55] and Mori and Zhang [56] for reviews). As the diffusion of tissue water is restricted by the presence and orientation of biological barriers such as cell membranes and myelinated fibers, stroke-induced structural modifications therein can significantly alter the characteristics of tissue water diffusion, such as the DTI-derived axial, radial, and mean diffusivity (MD), as well as fractional anisotropy (FA) [57]. This may involve different processes at different post-stroke stages, ranging from acute cell swelling (causing MD decrease) and subacute cell lysis and demyelination (causing MD increase and FA decrease), to chronic axonal regeneration or remyelination and gliosis (causing FA increase) [13, 57, 58].

Since white matter tracts are composed of highly oriented fibers, which cause relatively high anisotropy of diffusing tissue water, DTI is very suitable to measure effects on white matter integrity. Studies in patients and animals have reported loss of FA in ipsilesional white matter subacutely after stroke, which has been linked to demyelination or axonal loss [6, 12, 59]. A few serial DTI studies in experimental stroke models have demonstrated that this initial decrease of FA may be chronically followed by normalization or elevation in the ischemic lesion borderzone [58], which could be further enhanced by treatment with neural progenitor cells [60], sildenafil [61], or erythropoietin [62]. This area revealed a high density of axons and myelin on postmortem histological sections [60, 61]. Furthermore, part of the ipsilesional internal capsule with elevated FA in post-stroke rat brains also showed significant manganese enhancement on T₁-weighted MRI after injection of the paramagnetic neuronal tracer in the perilesional sensorimotor cortex [58]. This suggests that rearrangement of white matter in the ischemic boundary is accompanied by preservation or restoration of neuronal connectivity. Figure 4 shows postmortem high-resolution FA maps that we collected at 11 weeks post-stroke. These data clearly illustrate the FA rise in perilesional white matter at chronic stages after stroke. In accordance with these findings in animal stroke models, elevated FA in ipsilesional corticospinal tracts has also been found in chronic stroke patients, which was associated with improved motor function [63]. Evidently, structural integrity of the corticospinal pathway appears critical for a favorable outcome in sensorimotor performance after stroke.

The DTI-derived diffusion anisotropy and principal diffusion direction in white matter tissue can be used to model the architecture of neuronal fibers, visualized by orientation-based color-coded FA maps (see Fig. 4) or three-dimensional fiber tract maps computed with tractography algorithms [56]. However, the angular resolution of DTI, which determines the extent to which fiber orientation can be estimated, is limited [64]. This can significantly affect the accuracy of tractography. DTI provides a single principal diffusion direction per voxel, which complicates the resolution of complex white matter architecture with crossing, kissing, bending, or fanning fibers within a voxel. In addition, DTI assumes a Gaussian model of diffusion, which may not be appropriate for restricted diffusion in biological tissues [65]. Alternative diffusion-imaging schemes, such as q-ball imaging [66], spherical deconvolution [67], and diffusion spectrum imaging [68] have been developed to address these issues of multiple fiber orientations in a single voxel and to improve the quality of fiber tracking. A recently introduced method, diffusion kurtosis imaging, allows quantification of the degree to which tissue water diffusion is non-Gaussian [65]. This enables further characterization of the complexity of tissue microstructure in gray and white matter by accounting for the diffusion kurtosis as a result of the present cellular compartments and membranes [69, 70]. These advanced diffusion imaging techniques, however, require high hardware standards and long scan durations, as well as special analysis software. Currently, this impedes widespread application in preclinical and clinical settings.