Focal brain damage, such as after ischemic stroke or in epilepsy, not only affects local neuronal networks, but can also have significant effects on the function of remote brain regions. For instance, constrained behavioral output (Carter et al., 2010) and cognitive comorbidities (Tatemichi et al., 1994; Motamedi and Meador, 2003) after stroke and epilepsy have partly been attributed to functional disturbances of distant brain regions that were not primary affected. In addition, spontaneous partial recovery of sensorimotor and behavior output in the weeks and months following stroke may be explained by functional plasticity of remote and unaffected brain regions (Langhorne et al., 2009; Stinear, 2010). As brain functions are believed to emerge from interacting regions within large-scale neural networks, the brain's intrinsic capacity to compensate for focal injury is linked to reorganization of surviving functional networks at a global level (Grefkes and Fink, 2011).

A sophisticated set of measures, using the concept of graph theory, has been used to describe alterations in functional network organization following focal brain damage (Rubinov and Sporns, 2010), with promising opportunities for applications in clinical settings (Fornito et al., 2013). Obviously, changes in functional networks are strongly dependent on the structural status of neuronal connections that link brain regions. Recent studies have reported a high similarity between functional network connectivity and underlying structural neuronal connections (Achard et al., 2008; Honey et al., 2009). However, it is largely unknown how functional networks reorganize in response to transient focal disturbances without structural damage. Characterizing the impact of transient localized functional brain disruptions on large-scale neural networks can provide important insights into a network's resilience, which might lead to improved understanding of underlying causes of comorbidities in diseases with localized brain damage, such as focal epilepsy and ischemic stroke.

Global neural networks are known to have a hierarchical structure with the existence of a number of well-connected and highly central hub regions (Hagmann et al., 2008; Bullmore and Sporns, 2012) that form a so-called “rich-club,” which reflects a network's capacity to segregate and integrate multisensory information in a united manner (Van Den Heuvel and Sporns, 2011; Harriger et al., 2012). These well-connected regions play a key role in global signal processing between different parts of the network, and functional disturbance of hub regions can have critical impact on overall network functioning. In this study we aimed to map the impact of transient functional brain disruption on global neural networks in the absence of structural damage in a newly developed reproducible rat model. To that aim, we characterized the brain's functional network organization with resting-state fMRI (rs-fMRI) after unilateral pentobarbital-induced brain “silencing.” We specifically focused on effects on well-connected hub regions using a relatively new framework that provides quantitative measures of the organization of highly connected rich-clubs (Van Den Heuvel and Sporns, 2011; Harriger et al., 2012; Collin et al., 2013). So far it has remained unknown whether rodent brain networks contain rich-club connections and if so, how these connections should be interpreted. Our hypothesis was that (i) rat brains have functional rich-club connections, and (ii) that transient hemispheric functional disruption affects the rich-club organization in both the disrupted and unaffected network regions.

Perfusion MRI data demonstrated that the catheterized carotid artery predominantly supplied the ipsilateral hemisphere allowing for selective hemispheric brain silencing with pentobarbital (Figure 1). Perfusion was strongest in the cortical areas, whereas levels in subcortical regions were lower. We also detected some contrast agent arrival in the anterior contralateral hemisphere, which can be explained by delivery via the circle of Willis.

Our pilot study shows that transient functional global network disruptions in response to unilateral and reversible functional brain disturbance can be effectively studied using a straightforward rodent model in combination with rs-fMRI. Our rat data support the concept that densely connected rich-club regions play a central role in global brain communication, as has been previously shown in humans. In addition, we found that global network hub configurations can be significantly affected by temporary functional disruption of a hemisphere.

A rodent model that allows selective and reversible functional hemispheric silencing has not been described before. Our results show that this model can be useful to assess effects of (transient) brain impairment on functional networks. The concept of this model was based on recent FC and network data presented by Douw et al. (2009, 2010). They showed that FC within and between hemispheres changed in persons with epilepsy during an intra-arterial amobarbital injection—the WADA test—as measured with EEG (Douw et al., 2009). They concluded that the effects of unilateral anesthesia on FC in the contralateral hemisphere should not be ignored. Direct comparison of low-frequency rs-fMRI connectivities with the different EEG frequency bands is difficult, and more data are required to elucidate differences and similarities between rs-fMRI and EEG patterns. Nevertheless, the observed decrease in interhemispheric EEG FC in the delta and theta bands by Douw and colleagues corresponds with our interhemispheric EEG results (Figure 5). Over the ipsilateral hemisphere the delta and theta band synchronization increased after amobarbital injection in humans, which also agrees with our findings of increased ipsilateral intrahemispheric FC in rat brain after pentobarbital injection. Apparently, unilateral functional brain disruption by amo- or pentobarbital injections is confined to subverted functional outcome (that is, speech and language comprehension in humans) rather than absolute inhibition of low-frequency FC. This idea is further supported by recent FC measurements at an anesthetic depth characterized by unresponsiveness, where reductions in connectivity were shown to be partial and not complete (Hudetz, 2012).

Douw et al. also assessed EEG data acquired during WADA-test using whole-brain functional network analysis (Douw et al., 2010). Their results revealed an increased random network organization after unilateral functional disruption with amobarbital in people with epilepsy (Douw et al., 2010). However, the network topologies within the injected and non-injected hemispheres were not studied separately in this group of patients. In our study we found no difference in rich-club characteristics between pre- and postbarbital whole-brain networks. Differences in weighted rich-club effects were only found in sub-network analyses of individual hemispheres (Figure 3). The asymmetric presentation of prebarbital hub regions (that is, mostly contralateral sensorimotor areas) is in line with studies in healthy rats that reported right-sided population bias and functional lateralization (Glick and Ross, 1981; Alonso et al., 1991).

A positive weighted rich-club effect characterizes the tendency of network hubs to engage in stronger functional connectivities among them than expected by chance. We found significantly decreased ρ^w(s) values in the ipsi- and contralateral sub-networks after unilateral pentobarbital injection. Similar positive rich-club organizations have been described at a neuronal level in the nematode worm Caenorhabditis elegans (Towlson et al., 2013) and at a larger scale in human neocortical networks (Van Den Heuvel and Sporns, 2011). Our findings in rodent brain networks, consolidate the hypothesis that connections according a rich-club layout may be a general and scale-invariant principle of brain network organization (Towlson et al., 2013). Nonetheless, further validation of this hypothesis is essential, given the small sample size used in this study.

The rich-club connections in the structural human brain network are considered to be the backbone for global brain communication (Van Den Heuvel et al., 2012). The majority of these connections are long-distance neural pathways, such that the rich-club phenomenon is costly in terms of consumed energy and space. However, the trade-off for higher cost is higher performance in terms of network integration; up to 70% of all network's shortest paths travel through rich-club connections.

Whether structural network principles also hold for functional rich-club networks remains to be determined, but we expect the functional rich-club organization to follow the underlying structural organization to a large extent. On the other hand, functional rich-club connections may be much more flexible in acutely responding to environmental changes, such as induced by anesthesia. This is, for instance, underscored by the differences in our whole-brain and sub-network data (Figures 3B,C). Unilateral brain silencing did not affect the rich-club effect at a whole-brain level; only the club members changed dramatically (Figure 2). On the other hand, the hemispheric sub-networks did reveal change in rich-club connectivity after hemispheric silencing. Apparently the level of network definition is critical for detecting changes in functional rich-club organization after functional brain lesioning. A better understanding of this phenomenon may require the acquisition of structural diffusion tensor imaging data alongside rs-fMRI in future studies. This will allow comparison of functional networks and their rich-club organization with data from the intact structural networks.

Limitations in this proof-of-principle methods study should be acknowledged. First, experiments were conducted at 1.0% isoflurane anesthesia. We do not know how this baseline anesthesia may have interacted with the pentobarbital-induced brain silencing. We expect this effect to be relatively small as cerebral glucose utilization, cerebral blood flow coupling (Maekawa et al., 1986; Lenz et al., 1998) and FC, including the long distance connections (Liang et al., 2012), are largely preserved at 1.0% isoflurane anesthesia (Wang et al., 2011). We also know that the neurovascular coupling in rat brain is similar across distinct types of anesthetics including pentobarbital and isoflurane (Franceschini et al., 2010), however further validation studies, including robust dose-response experiments, are needed. Rs-fMRI acquisition may be ideally executed in awake rats, however, this is complicated due to motion artifacts related to post-surgical pain, stress and discomfort.

Second, anesthesia-based functional brain disruption by unilateral injection through the carotid artery may not be exclusively unilateral due to presence of the circle of Willis and recirculation effects. Nonetheless, the small injection volume used in our study limited this effect, which was confirmed by the strongly lateralized results for hemispheric FCs.

Third, the likelihood that a statistically significant result reflects a true effect is reduced in studies with a small sample size (Button et al., 2013). The large changes in functional network connectivity and organization that we observed in this rat model alleviate this problem. Nonetheless, firm conclusions on directionality of network changes and size of effect require additional studies with larger groups of animals. This is emphasized by the variation that we found between FC correlation coefficients obtained from the individual networks (Figure 4). This variation is however comparable to the variation found between individual interhemispheric FC in rats recovering from a unilateral stroke, as studied with rs-fMRI with a similar experimental setup and design (Van Meer et al., 2012).

In conclusion, our pilot study suggests that changes in rich-club organization following functional network disruption might affect the capacity to efficiently segregate and integrate multisensory information. These changes are unlikely to depend on the presence of structural damage. We speculate that functional rich-club alterations may reflect an underlying mechanism of neurological impairments in patients with no or limited structural brain damage, such as after transient ischemic attack.