Stroke and Neurodegeneration Induce Different Connectivity Aberrations in the Insula
Background and Purpose—Stroke and neurodegeneration cause significant brain damage and cognitive impairment, especially if the insular cortex is compromised. This study explores for the first time whether these 2 causes differentially alter connectivity patterns in the insular cortex.
Methods—Resting state–functional magnetic resonance imaging data were collected from patients with insular stroke, patients with behavioral variant frontotemporal dementia, and healthy controls. Data from the 3 groups were assessed through a correlation function analysis. Specifically, we compared decreases in connectivity as a function of voxel Euclidean distance within the insular cortex.
Results—Relative to controls, patients with stroke showed faster connectivity decays as a function of distance (hypoconnectivity). In contrast, the behavioral variant frontotemporal dementia group exhibited significant hyperconnectivity between neighboring voxels. Both patient groups evinced global hypoconnectivity. No between-group differences were observed in a volumetrically and functionally comparable region without ischemia or neurodegeneration.
Conclusions—Functional insular cortex connectivity is affected differently by cerebral ischemia and neurodegeneration, possibly because of differences in the cause-specific pathophysiological mechanisms of each disease. These findings have important clinical and theoretical implications.
The insular cortex (InsC) is a main structural–functional hub of the brain.1 It features widespread connections with cortical and subcortical regions, and it is implicated in varied domains, such as body sensation, language, emotion, and social cognition.2 Notably, the InsC abounds in von Economo neurons, which facilitate rapid integration of information. Together with its topological centrality,1 these features render the InsC as a core region for overall brain dynamics and cognition.
InsC damage after stroke or neurodegeneration can considerably impair cognitive function. Stroke can disturb behavioral, emotional, and sensory domains,3 whereas early neurodegeneration (including the InsC) in behavioral variant frontotemporal dementia (bvFTD)4 usually leads to social and emotional dysfunction.5 Despite such general differences, little is known about the distinct insular connectivity alterations caused by each of these causes. A direct comparison of InsC connectivity patterns in patients affected by stroke and neurodegeneration6 may reveal theoretically and clinically relevant differences between each pathophysiological process.
In fact, at the molecular level, these conditions differ in timing and physiopathology. Stroke involves tissue loss, neuronal reorganization, and plasticity.7 Instead, neurodegeneration in bvFTD results from protein aggregation, inducing physiopathological events (eg, axonal degeneration, synapse loss, and dendritic retraction) and propagation of misfolded proteins.8 These differences at the molecular level suggest that InsC connectivity may be differentially affected by stroke and bvFTD.
Previous studies have reported aberrant long-range connectivity (mostly hypoconnectivity) in both stroke and bvFTD.4,7 Nevertheless, local connectivity is a key feature of brain dynamics.9 Unlike global connectivity (which may be similarly altered across pathologies), this dimension may better discriminate between diseases. Crucially, it modulates global connectivity10 and is fundamental for the emergence of small-world properties. Yet, local connectivity research has been rather sparse in neurological populations, especially in patients with stroke. In neurodegeneration, local connectivity disturbances may constitute a key disease marker9 and a useful dimension to characterize regions of early atrophy. Thus, our focus is on local connectivity because the InsC in consistently compromised in both groups.
To address this hitherto unexplored issue, we assessed InsC functional connectivity in patients with stroke and bvFTD. Specifically, we focused on correlations between local connectivity decays and intervoxel distance within the InsC. By considering various spatial ranges in connectivity, we aimed to reveal cause-specific patterns of regional functional connectivity of InsC at both small and large levels.
We recruited 33 participants from an ongoing project.6 The sample comprised 1 patient with hemorrhagic and 7 patients with ischemic stroke, all featuring damage to the InsC and also peripheral areas (38–66 years, 2 women: 6 men, Mini-Mental State Examination: 29 [26–30]; Figure A; Table I in the online-only Data Supplement), 11 patients with probable bvFTD (54–70 years, 6 women: 5 men, Mini-Mental State Examination: 25.11 [18–30]; Figure B), and 14 healthy subjects (33–72, 4 women: 10 men, Mini-Mental State Examination: 29.23 [25–30]; Section 1.1 in the online-only Data Supplement). The 3 groups were matched by age, sex, educational level, and premorbid intelligence quotient. All participants provided signed consent in accordance with the Declaration of Helsinki. The study was approved by the Ethics Committee of the Institute of Cognitive Neurology.
Structural images and resting functional magnetic resonance imaging were acquired from patients and healthy subjects in a-1.5 T Phillips Intera scanner (Section 1.2.1 in the online-only Data Supplement). Demographic information was compared among groups using ANOVAs and Pearson χ2 tests. The global atrophy pattern in patients with bvFTD was assessed through voxel-based morphometry (Figure B; Section 1.2.3 in the online-only Data Supplement). Functional images were slice-time corrected, realigned, and normalized on SPM8. They were subsequently band-pass filtered (0.01–0.05 Hz). We finally regressed out the 6 motion parameters estimated during realignment (Section 1.2.4 in the online-only Data Supplement). All groups were also compared in terms of long-range connectivity between InsC and regions of interest (ROIs) of 6 functionally defined networks (Section 1.2.5 in the online-only Data Supplement).
We assessed local InsC connectivity by analyzing interaction decays as a function of between-voxel distance. This approach is similar to previously reported regional homogeneity analyses,11 as both assess local correlations. However, the distance measure in regional homogeneity analyses is a fixed point rather than a dimensional parameter, and distant correlations are not even considered within a region. The correlation function analysis allowed us to study the interaction between voxels which are at a distance d (for all possible values of d), assessing connectivity variations within the InsC at different levels: at the small spatial level, we analyzed voxels in close proximity; at the large spatial level, we studied distant regional connectivity among InsC voxels. The level of functional decay indicates whether local connectivity in a ROI is affected by disease.
Blood oxygen level–dependent contrast imaging temporal series and x–y–z coordinates in Montreal Neurological Institute space were calculated for each voxel in 2 bilateral ROIs: the InsC and the cuneus (as defined in the automated anatomical label atlas). The cuneus was selected as a control area because it resembles the insula’s size and functional properties (viz, integration of sensory and high-order information), but it was not affected by bvFTD atrophy or stroke in our patients (Figure A and B).
We determined the degree of functional interaction between voxels lying at a distance d within each ROI. To this end, we calculated the average Spearman rank correlation coefficient, ρ, between blood oxygen level–dependent contrast imaging signals of voxels separated by d. We studied this measure (henceforth, correlation function) as a function of the Euclidean distance (d) between voxels, as follows:
where is the rank time series of the blood oxygen level–dependent contrast imaging signal of the k voxel, and T is the length of this signal.
This function was calculated for each subject and ROI and averaged within groups (Section 1.2.6 in the online-only Data Supplement).
Correlation functions were compared via Monte Carlo permutation tests combined with bootstrapping (Section 1.2.6 in the online-only Data Supplement). This simple method gives a straightforward solution for the multiple comparison problems and does not depend on multiple comparison corrections or Gaussian distribution assumptions.
The groups presented no significant differences in sex, age, formal education, or premorbid intelligence quotient (Table).
Figure A shows the lesion overlap of the InsC, which was the most consistently affected region across patients (Table I in the online-only Data Supplement). Voxel-based morphometry analysis revealed a pattern of global atrophy in the bvFTD group in fronto-temporo-insular structures (P<0.05; false discovery rate corrected; Figure B), which replicated previous reports4 (Section 2.2 and Table II in the online-only Data Supplement). We have also provided global volumetric measures (grey matter, white matter, and cerebrospinal fluid) of patients with bvFTD and controls (Sections 1.2.3 and 2.2 and Table III in the online-only Data Supplement).
Grey matter volume of intact/infarcted/atrophied InsC for both patient groups is described in Tables IV and V in the online-only Data Supplement (Methods Section 1.2.3 and Extended Results Section 2.2 in the online-only Data Supplement). To control for possible InsC differences among patient groups, we compared the extent of bilateral insular damage induced by stroke and bvFTD atrophy. The number of infarcted versus atrophic voxels and the volume of infarcted versus atrophied areas were comparable in both groups (Section 1.2.7 in the online-only Data Supplement).
For patients with stroke, when compared with controls, pairwise correlations between long-range voxels within bilateral InsCs significantly decreased as a function of voxel distance (P<0.05). Such hypoconnectivity appeared at a distance of ≈9 to 19 voxels (18–40 mm; Figure C). Conversely, patients with bvFTD, relative to controls, showed significant hyperconnectivity between neighbor voxels in the same area (P<0.05, at a distance of ≈4–7 voxels or 11–14 mm; Figure D). Voxel-distance analyses of the cuneus revealed no significant differences among groups (Figure C and D).
Moreover, relative to participants with stroke, patients with bvFTD presented significant hyperconnectivity for all distances (from 4–20 voxels or 8 a 40 mm) in the InsC but not in the cuneus (Figure I in the online-only Data Supplement).
Analysis of global connectivity in patients with stroke relative to controls revealed hypoconnectivity in insulo-frontal hubs. Compared with controls, patients with bvFTD exhibited disconnection in salience-network hubs4,8 (Section 2.3 and Figure II in the online-only Data Supplement).
Each patient group exhibited a distinct pattern of connectivity decay as a function of neighboring-voxel distance in the InsC. Patients with stroke showed hypoconnectivity, as correlations decreased with increasing distance. Conversely, patients with bvFTD were characterized by hyperconnectivity, showing the opposite pattern of correlations. Lack of differences in the cuneus, which proved intact in both diseases, suggests that local connectivity abnormalities are partly dependent on the underlying ischemic and neurodegenerative processes.
Regional connectivity has been successfully used to examine local connectivity patterns in healthy individuals.12 Previous research has also shown increased and decreased long-range connections in patients with stroke.7 Our study extends available findings by demonstrating that stroke induces abnormal local connectivity among distant InsC voxels (Figure C).
Studies about patients with bvFTD have revealed alterations in specific large-scale networks, manifested as both reduced and enhanced long-range connectivity.8 Only one study assessing local connectivity in this pathology showed hyperconnectivity in the InsC and interpreted it as a compensatory response in bvFTD.9 These findings align with our results, which showed that bvFTD, as opposed to stroke, features significant hyperconnectivity changes in nearby InsC voxels (Figure D).
Local hypoconnectivity in ischemic stroke could be the consequence of hypometabolism mediated by tissue hypoxia (eg, reduced adenosine triphosphate, mitochondrial damage, and release of DNA fragmentation proteins).13 Conversely, hyperconnectivity in bvFTD can reflect the effect of protein aggregation and neurotoxicity,4 which would activate compensatory mechanisms or disrupt the excitatory–inhibitory balance of damaged networks.8,14
Beyond these tentative explanations, our results suggest that topographically similar brain lesions with different causes may yield dissimilar aberrant changes in local connectivity. This highlights the importance of comparing diseases with different underlying mechanisms.6 Importantly, our findings challenge the common assumption that damage to a given area results in similar connectivity disturbances irrespective of its causes. At the least, this evidence calls for a re-evaluation of current proposals that postulate that specific brain hubs may be similarly compromised by different pathological conditions.15
Our study has some limitations. Similar to previous studies,9 our sample size was low. However, this limitation was considerably counteracted by the strict control of several relevant variables (eg, diagnosis, lesion cause, age, sex, intelligence quotient, education, and volume of insular affectation among patients with stroke and bvFTD). The detection of significant effects in target regions, and their absence in control areas, further attests to the adequacy of our samples.
Stroke damage is not restricted to the InsC. Nevertheless, all subjects had lesions exclusively caused by stroke and no comorbidity with other diseases. The main overlap among several compromised regions was found in the target region (InsC). Finally, images spatially normalized to allow for joint analysis of different pathological brains could provoke matching problems (especially in and around the stroke area). Yet, we largely minimized these negative effects of standard preprocessing by omitting spatial smoothing. This avoids shift of activation peaks and the reduction of spatial resolution.
Both patient groups evinced global hypoconnectivity, as previously reported separately for stroke7 and neurodegeneration9,10 (Section 3 and Figure II in the online-only Data Supplement). Nevertheless, global connectivity results should be taken with reserve. Extra-insular areas are not consistently damaged across patients with stroke and bvFTD. The InsC has widespread connections, and such diversity makes it difficult to find homogeneous samples warranting analysis of long-range connections. Thus, we cannot determine the extent to which deficits are associated with damage in the (1) InsC, (2) other areas, or (3) a combination of (1) and (2).
Previous reports supported the idea that different neural disorders with different causes involve similar disconnection patterns in characteristic brain hubs.15 In contrast, we suggest that different causes and physiopathological processes may influence cerebral connectivity in dissimilar and specific ways. Such idiosyncratic patterns may reflect differences in molecular and temporal dynamics between cerebral ischemia and neurodegeneration.
Sources of Funding
This study was supported by grants from Fondo Nacional de Desarrollo Científico y Tecnológico Regular (FONDECYT Regular 1130920/1140114); Proyecto de investigación de Ciencia y Tecnología (PICT 2012-0412/2012-1309), Ineco Foundation, and Innovation and Dissemination Center for Neuromathematics (FAPESP Research #2013/07699-0, S.Paulo Research Foundation).
Guest Editor for this article was Sean Savitz, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.009598/-/DC1.
- Received April 1, 2015.
- Revision received May 22, 2015.
- Accepted June 11, 2015.
- © 2015 American Heart Association, Inc.
- Rascovsky K,
- Hodges JR,
- Knopman D,
- Mendez MF,
- Kramer JH,
- Neuhaus J,
- et al
- Simmons WK,
- Avery JA,
- Barcalow JC,
- Bodurka J,
- Drevets WC,
- Bellgowan P.
- Carmichael ST.
- Crossley NA,
- Mechelli A,
- Scott J,
- Carletti F,
- Fox PT,
- McGuire P,
- et al