Abstract
The origin, structure, and function of the claustrum, as well as its role in neural computation, have remained a mystery since its discovery in the 17th century. Assessing the in vivo connectivity of the claustrum may bring forth useful insights with relevance to model the overall functionality of the claustrum itself. Using structural and diffusion tensor neuroimaging in N = 100 healthy subjects, we found that the claustrum has the highest connectivity in the brain by regional volume. Network theoretical analyses revealed that (a) the claustrum is a primary contributor to global brain network architecture, and that (b) significant connectivity dependencies exist between the claustrum, frontal lobe, and cingulate regions. These results illustrate that the claustrum is ideally located within the human central nervous system (CNS) connectome to serve as the putative “gate keeper” of neural information for consciousness awareness. Our findings support and underscore prior theoretical contributions about the involvement of the claustrum in higher cognitive function and its relevance in devastating neurological disease. Hum Brain Mapp 36:827–838, 2015. © 2014 Wiley Periodicals, Inc.INTRODUCTION
Considerable mystery has surrounded the origin, structure, and function of the claustrum—a diminutive bilateral anatomical structure of the brain whose very name means “hidden away” [Crick and Koch, 2005]. The structure was identified in humans as early as 1672 as shown by the drawings of Thomas Willis [Bayer and Altman, 1991], though was first described (under the name “vormauer”) by Karl Friedrich Burdach in his seminal work, Von Baue und Leben des Gehirns in the early 19th century. Burdach himself, however, credited the discovery to the 1,786 drawings by Félix Vicq-d'Azyr . Despite this long history, knowledge of the claustrum, its organization, afferents, and efferents has been studied intermittently and relatively poorly explored until recently [Edelstein and Denaro, 2004; Smythies et al., 2014].In animals, extensive neuronal projections have been noted between the claustrum and numerous cortical regions [Buchanan and Johnson, 2011; Crick and Koch, 2005; Minciacchi et al., 1995; Tanne-Gariepy et al., 2002; Wilhite et al., 1986] with white matter connections to multiple cortical layers [Carey et al., 1980] (although see Crick and Koch [2005]). Likewise, the claustrum exhibits many connections to subcortical structures [Arikuni and Kubota, 1985; Berke, 1960; Buchanan and Johnson, 2011; Crick and Koch, 2005; Salerno et al., 1984; Tanne-Gariepy et al., 2002]. The structure is present in all mammals with species such as tree shrews and cats having relatively large claustra by volume in contrast to monotremes, who apparently lack claustra altogether [Butler et al., 2002].
In electrophysiological studies, the claustrum has been shown to send signals to four thalamic nuclei: the medial geniculate nucleus, the lateralis posterior, centrum medianum, and ventralis lateralis [Chachich and Powell, 2004; Cortimiglia et al., 1987, 1991; Crescimanno et al., 1989; Spector et al., 1970; see also Sherk 2013, Chapter 5]. Excitotoxic lesioning in the rat has also shown evidence of frontostriatal connectivity as well [Grasby and Talk 2013]. Both the medial ectosylvian gyrus (A2) and the anterior ectosylvian gyrus (S2) also have been reported as transmitting information to the claustrum [Edelstein and Denaro, 2004; Hassmannova, 1977]. Moreover, the claustrum is topologically organized with frontal cortex being linked to the claustrum's anterior portion, the parietal cortex with its central and posterior parts, and the occipital and temporal cortices are linked to its posterior and inferior margins [Druga, 2014; Markowitsch et al., 1984]. Nontopographic projections to other parts of the same cortical area exist, but there is complete segregation between distinct cortical areas [Minciacchi et al., 1995]. Fiber bundles such as the corona radiata, uncinate fasciculi, and inferior occipitofrontal fasciculi project to the claustrum from the superior frontal, precentral, postcentral, superior parietal, and parietooccipital regions [Fernandez-Miranda et al., 2008a, b]. Only a relatively small number of claustral cells appear to project to the contralateral hemisphere [Markowitsch et al., 1984]. Opinions have differed, however, as to whether the claustrum receive inputs via brain stem and spinal afferents [Arikuni and Kubota, 1985; Edelstein and Denaro, 2004].
In humans, the volume of the claustrum is one quarter of 1% of the volume of the cerebral cortex [Crick and Koch, 2005; Edelstein and Denaro, 2004]. As illustrated in the Talairach and Tournoux [1988] human brain atlas, the claustrum extends 22 mm inferior-to-superior and 38 mm anterior-to-posterior. The right claustrum has an average volume of 828.83 mm3, while the left has a volume of 705.82 mm3 and lies approximately 1 mm from the insular cortex and about 1 mm from the putamen [Kapakin, 2011]. Bilaterally, it displays asymmetry in shape and anisotropy between the hemispheres [Cao et al., 2003] (as shown schematically in Fig. 1A). It projects downward and crosses the rhinalis fissure—though it may be interrupted by fibers from the uncinate fasciculus in some places—before extending to the lateral rhinencephalon. A thorough review of claustral gross anatomy can be found in Druga [2014]. Historically, the claustrum was often considered to be the innermost layer of the insula [Landau, 1919], though has also been associated with the basal ganglia as a potential pallial derivative [Puelles, 2014]. The claustrum contains both fusiform cells and pyramidal somata, which are indicative of cortical areas, but also contains subcortical cell types. Due to the variation in cell morphometry, the generally nonlayered structure of the claustrum cannot formally be pronounced as strictly cortical or subcortical [Mathur et al., 2009]. This might explain its only modest description in early as well as modern neuroscience texts and resources (e.g., http://neurolex.org/wiki/Category: Claustrum).
Despite its consideration in a modest number of largely descriptive neuroimaging studies, however, the relative importance or strengths of the diverse and wide-ranging pattern of claustral connectivity have not, to date, been thoroughly explored using large human samples. Quantitative methods for the assessment of white matter networks offer an opportunity to explore the connectivity of the claustrum in vivo and to measure its relative influence on brain subnetworks. Indeed, graph theoretical analytic techniques have been widely applied to connectomics data from human neuroimaging studies to characterize the novel properties of structural and functional brain networks [Bassett and Bullmore, 2009; Bullmore and Sporns, 2009]. Graph-wide as well as nodal connectivity characteristics provide insights into the integrity and efficiency of information flow through a network and can shed light on the effects of injury or disease. They can also highlight the macrolevel importance of certain regions on the overall structure of a network and how those influence the connectivity of brain subnetworks.
With this in mind, we sought to quantitatively examine the relative importance of putative subnetworks, which involve the claustrum—assessed using DTI, structural neuroimaging, and graph theoretical analysis—in a cohort large enough to establish a reliable and comprehensive model of the full macroscale connectivity of this so often over-looked brain structure.
METHODS
Subjects and Data Acquisition
The study cohort included N = 100 healthy adult subjects (47 males and 53 females) with ages between 18.6 and 61.1 (mean and SD: 32.71 ± 11.6 years). Subjects provided informed written consent as required by the Declaration of Helsinki, U.S. 45 CFR 46, and neuroimage volume acquisition was conducted with the approval of the local ethics committees at the respective research institutions where data were acquired. Participants were recruited by advertisements in local newspapers and campus flyers. Subjects were all healthy and had no history of neurological or psychiatric illnesses. No participant had a current or past psychiatric diagnosis (including substance abuse) or was taking medications for any medical reasons. Additional exclusion criteria for all participants included left-handedness, hypertension, neurological illness, metal implants, and a history of head trauma with loss of consciousness for more than 5 min. Neuroimaging datasets were fully anonymized, and no linked coding or keys to subject identity were maintained. For these reasons, in compliance with the U.S. Health Insurance Portability and Accountability Act (HIPAA; http://www.hhs.gov/ocr/privacy), this study does not qualify as involving human subjects' materials.Structural T1-weighted magnetic resonance imaging (MRI) and DTI volumes were acquired from each patient using a Siemens Trio Tim 3.0 Tesla whole body scanning system with a 12-channel head coil. For MRI, a Turbo MP-RAGE sequence (repetition time [TR] = 20 ms, echo time [TE] = 3 ms, flip angle = 25°, slice thickness = 1 mm, acquisition matrix = 256 × 256 × 256) was obtained. A DTI protocol having the following parameters was obtained: TR = 9.4 s, TE = 88 ms, flip angle = 90°, slice thickness = 2 mm, number of gradient directions = 68, acquisition matrix = 128 × 128 × 128. Two nondiffusion-weighted volumes were acquired for each subject (b-values: 0 and 1,000).
Neuroimage Data Processing
Data processing workflows were created using the Laboratory of Neuro Imaging (LONI) Pipeline Workflow Environment (http://pipeline.loni.ucla.edu; version 5.9.1). The LONI Pipeline is a graphical environment for construction, validation and execution of advanced neuroimaging data analysis protocols [Dinov et al., 2010]. It enables the creation of heterogeneous tool interactivity, automated data format conversion, allows grid utilization, facilitates data provenance, and provides a significant library of computational tools. It is built as a distributed grid computing environment and permits efficient tool integration, protocol validation, and broad resource distribution. Further details are available via the LONI Pipeline website.A set of customized LONI Pipeline workflow was created to perform basic preprocessing of all subject data as a prelude to the application of regionally focused processing, atlas creation, and connectomics estimation. This initial workflow performed FreeSurfer 5.3 reconstruction [Dale et al., 1999; Fischl et al., 1999], stripped the skulls and other nonbrain tissues out of the image volumes, bias-corrected the T1-weighted structural images, performed eddy current correction on the DTI, and then reconstructed the diffusion image into 3D tracts in both .vtk and .trk file formats so that the white matter can be visually examined using commonly available tractography software. Specifically, DTI fractional anisotropy (FA) computation and fiber reconstruction was performed using the TrackVis Diffusion Toolkit (http://trackvis.org) via streamline tractography using a subject-specific B0 white-matter mask and a turning angle threshold of 35°. The workflow also carried out the spatial registration of the T1 and DTI images, and likewise applied the transformation to the cortical parcels of the FreeSurfer labelmap. Finally, connectivity matrices were computed to quantify the relative number of extracted DTI fibers and average FA values over them which connected each pair of cortical regions. These matrices are then combined with the cortical region morphometric analyses to generate whole-brain connectograms (see below). The LONI Pipeline workflows described herein, along with additional supporting workflows, are available as Supporting Information.
Identification and Segmentation of the Claustrum
While the claustrum has been assigned labelmap values in the context of the FreeSurfer implementation of the Destrieux et al. [2010] brain atlas, the version of FreeSurfer we used (version 5.3) does not, in fact, assign voxels to this structure. So, following the application of the initial image processing workflow, 3D Slicer (http://slicer.org) was used to manually delineate left and right claustrum masks in the sample of healthy subjects. For each subject, an axially resampled T1 anatomical was examined in a superior-to-inferior manner to identify the body of the claustrum. As described above and illustrated in Figure 1A, the claustrum is located between the external and internal capsules, medially from the insular cortex and laterally from the body of the putamen. Separate left and right binary masks of this region were manually drawn (by Carinna M. Torgerson (CMT)) in the axial plane and then edited in the sagittal plane for accuracy, before being visually reviewed by trained raters (by Andrei Irimia (AI) and John D. Van Horn (JDVH)), and the regional delineation refined as necessary.Pooling and Applying Claustral Masks
Each subject's pair of claustral masks were then used as input into an “Atlas Creation” LONI Pipeline workflow, which spatially warped the claustrum label masks into Montreal Neurological Institute (MNI) Atlas space and, using FSL modules, spatially averaged them, before returning the average back into the native space of each individual subject via inverting the spatial warping transform. This averaging was performed to make claustral delineation consistent with the FreeSurfer mapping technique, which uses averaged data to generate accurate labelmaps. These newly applied masks of the claustrum (now in native space for all 100 subjects), the “Adding Labelmaps” workflow was used to spatially merge the left and right claustrum labels into the FreeSurfer parcellation and segmentation labelmap. At this stage, new connectivity matrices were generated for each subject based upon whole-brain labelmaps, which now included the spatially averaged and unwarped claustra, using the “Connectivity Calculation Workflow.”Subsequently, the fully integrated labelmaps, which include the left and right claustra, were used as input into three different workflows. First, the “Whole Brain Plus Claustrum” workflow took the full labelmaps and generated connectograms depicting the whole, macroscale pattern of DTI connectivity across the complete collection of segmented brain regions. Finally, the “Only Claustrum” workflow used the connectivity matrices from the whole-brain plus claustrum analysis resulting in new files created for each subject depicting the specific connectivity patterns of the left and right claustrum in each subject. 1 Values of all connections originating or terminating in the insular cortex an putamen were set to zeroes to avoid misattribution due to proximity, as well as double-counting errors.
Connectogram Generation
Cortical parcellations were represented as a circular arrangement of 165 positioned elements representing the connectivity matrix of the left (quadrant IV of Fig. 1D) and right (quadrants I of Fig. 1D) cerebral hemispheres, and their jointly connecting fiber pathways (quadrants II and III), each positioned symmetrically with respect to the vertical axis. This graphical representation—termed a “connectogram”—has been described extensively elsewhere [Irimia et al., 2012a, b; Van Horn et al., 2012]. The connectogram is becoming widely used as a method for showing patterns of connectivity in a variety of research domains, as is demonstrated by the variety of topics mentioned on its Wikipedia page (http://en.wikipedia.org/wiki/Connectogram). Briefly, the connectograms represent cortical region labeling and morphometric measurements as color-coded rings and illustrate inter-regional connectivity as lines between lobar and regional segments [Irimia et al., 2012a, b]. Line opacity is proportional to connection density, while in these figures, specifically, red indicates high average FA (aggregated over fibers connecting those regions), green represents moderate FA, and blue represents low FA. FA measures the degree of preferred water diffusivity, ranging between zero (no preferred diffusion) and unity (strong directional diffusion) [Basser et al., 2000]. A complete list of parcellations, their abbreviations and associated Red-Green-Blue (RGB) codes are provided in the supplemental materials of Van Horn et al. [2012]. Parcellations were arranged within each lobe in the order of their location along the anteroposterior axis of the cortical surface associated with the published FreeSurfer normal population atlas [Destrieux et al., 2010].Network Theoretical Analyses
To examine the network contributions of the claustrum, we used graph theoretical methods, which have been demonstrated to provide useful constructs in characterizing neuroimaging-derived brain networks [Sporns, 2012]. Graph theoretical metrics are useful for describing the properties of network architecture and have been successfully applied to patterns of inter-regional brain connectivity. Additionally, using a method for systematic removal of white matter connections in conjunction with graph theoretical analysis [Irimia and Van Horn, 2014], the left and right claustra were analyzed for their role in major brain networks (Fig. 1D). Such methods have provided a convenient approach for the computational modeling of the effects of brain injury [Van Horn et al., 2012] and for the assessment of nodal clustering and assortativity which can illustrate the presence of various brain sub-networks [van den Heuvel et al., 2008].All connectivity metrics were determined using the Brain Connectivity Toolbox (https://sites.google.com/site/bctnet/). Specifically, the graph-wide metrics used here included network density, diameter, and assortativity, as well as the nodal metric betweeness centrality. To simplify interpretation, particularly short (<5 a="" accompanied="" analyses.="" analysis="" and="" as="" being="" by="" claustra="" claustrum="" close="" connections="" considered="" determine="" disregarded="" errors="" excision="" graph-wide="" i="" in="" included="" increase="" influential.="" insula="" leave-one-out="" likelihood="" manova="" metrics="" mm="" modeling="" most="" neighborhood="" not="" notably="" of="" our="" p="" performed.="" proximity="" putamen="" s="" statistical="" statistically="" strategy="" subnetwork="" the="" their="" then="" this="" to="" topological="" treatment="" type="" was="" were="" which="" with="" would="">
RESULTS
In our large sample of N = 100 healthy subjects, the claustrum was found to possess the highest density of fiber connections per unit volume out of all examined brain regions (Table 1; listed by name according to the FreeSurfer aparc.a2009 naming scheme, which can be found at https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/AnatomicalROI/FreeSurferColorLUT following the list of aparc.a2005 labels). The normalization by regional volume is commonly recommended practiced when comparing fiber tract densities [Bassett et al., 2011; Cahalane et al., 2012; Varkuti et al., 2011]. The complete pattern of inter-regional connectivity between all regions is illustrated as a connectogram in Figure 1E, while the specific pattern of claustral connectivity is depicted in Figure 1F. The claustra had their densest DTI fiber tract connections to frontal cortices, with more modest degrees of connectivity to parietal, temporal, and occipital lobes, respectively. Connections appeared the fewest involving limbic structures in agreement with the findings of Markowitsch et al. [1984]. Evidence was also found for fiber tracts between the claustra and the brain stem, in contrast to prior reports which hypothesized that the claustrum communicates with the brain stem and autonomic spinal neurons via intermediaries, such as the medial prefrontal cortex [Edelstein and Denaro, 2004; Hatam et al., 2013].Our statistical analyses of the claustrum's topological neighborhood subnetwork, as measured using graph theoretical metrics with claustrum excision being considered as a level of treatment, were as follows.
Analysis of Graph-Level Metrics
The three specific network feature variables included in a MANOVA statistical model were assortativity, density, and graph diameter. We note that of the variety of available graph-based metrics, these were found to be only ones which were negligibly intercorrelated. But, importantly, they are particularly useful for assessing overall network structure and sensitivity to network alteration [Sporns, 2011]. These are defined as follows: (a) density is the ratio of present connections to all possible connections; (b) assortativity measures the correlation coefficient between the degrees of all nodes on two opposite ends of a link. A positive assortativity coefficient indicates that nodes tend to link to other nodes with the same or similar degree; while (c) graph diameter refers to the maximum network eccentricity, where eccentricity is the maximal shortest path length between a node and any other node. Under the MANOVA, it was found that the removal of the claustrum led to a statistically significant change in the network measure feature vector at the omnibus level, F(3,196) = 3.49, P ≤ 0.017 (FDR corrected). Systematic removal of individual network measures in the MANOVA was performed in order to identify their relative contribution to the overall model. The results, indicated by which variable had been removed, are as follows: assortativity—F(2,197) = 2.62; P ≤ 0.07; diameter—F(2,197) = 4.73, P ≤ 0.001; and density—F(2,197) = 1.87, P ≤ 0.16.Analysis of Node-Level Metrics:
Betweenness centrality assesses the fraction of all shortest paths in the network which contain a given node. Of particular interest at the nodal level is the notion of the betweenness centrality of brain regions connecting to the claustrum and how betweenness centrality of nodes is altered when the connectivity to the claustrum is removed. Nodes with high values of betweenness centrality participate in a large number of shortest paths. Assessing this measure of a node's relative “position” in a network is often a more informative measure than measuring connectivity density alone [Sporns, 2011]. Measurement of betweenness centrality has become a useful strategy for understanding the elements of complex networks, notably in the brain [Mirzasoleiman and Jalili, 2011]. It was found that the systematic removal of claustral connections led to significant changes (P ≤ 0.05, FDR corrected) in the ANOVA main-effects models of nodal betweenness centrality involving the structures listed in Table 2.| With claustrum | Claustrum removed | |||||||
|---|---|---|---|---|---|---|---|---|
| Structure | Hemisphere | Mean | SD | Mean | SD | F | Student's t-test | P |
| G_front_inf-Opercular | Right | 59.98 | 17.42 | 64.89 | 15.57 | 4.4168 | 2.1016 | 0.0369 |
| G_precentral | Right | 37.95 | 12.94 | 44.17 | 9.38 | 3.8984 | 1.9744 | 0.0497 |
| G_postcentral | Right | 42.98 | 15.82 | 47.90 | 15.23 | 5.0053 | 2.2373 | 0.0264 |
| G_front_middle | Right | 47.47 | 11.91 | 51.49 | 14.95 | 4.4243 | 2.1034 | 0.0367 |
| G_front_middle | Left | 45.92 | 10.89 | 49.41 | 11.78 | 4.7387 | 2.1769 | 0.0312 |
| G_postcentral | Left | 40.52 | 14.21 | 44.57 | 13.34 | 4.2994 | 2.0735 | 0.0394 |
| G_precentral | Left | 39.92 | 11.27 | 42.79 | 8.74 | 4.0488 | 2.0122 | 0.0456 |
| G_front_inf-Opercular | Left | 55.62 | 14.76 | 65.25 | 17.65 | 4.1853 | 2.0458 | 0.0421 |
DISCUSSION
As examined using tract-tracing and other methods [Berke, 1960; Markowitsch et al., 1984], claustral neuroanatomy is noteworthy across multiple species for receiving input from almost all regions of cortex and then directly projecting back to them [Buchanan and Johnson, 2011; Edelstein and Denaro, 2004; Kowianski et al., 1999; Smythies et al., 2012a, b]. Yet, while prior descriptive or small-sample human neuroimaging examinations of the claustrum's white matter architecture have been performed, to our knowledge, ours is the first population-level study to quantitatively examine the structural connectivity of the claustrum using in vivo brain mapping and graph analytic methods.Here, using DTI tractography and graph theoretical analytics, we computationally verify in a large human sample (N = 100) that the human claustrum is widely connected as assessed by its contribution to graph-wide diameter. Dense connectivity exists to most cortical areas although the structure is only modestly connected to limbic and occipital regions. While many contemporary examinations of the claustrum subdivide the claustrum [Fernandez-Miranda et al., 2008a, b, 2012; Gattass et al., 2014; Smith and Alloway, 2014; Smith et al., 2012], we did not feel that any such divisions would be necessary for our study, as boundaries are generally drawn according to perceived function or previously reported cortical region connectivity. In contrast, this study sought to solidify understanding of the entire connectome of the claustrum and attempt to better comprehend how this connectivity shapes networks; theoretically, one could use our network analysis in the future to subdivide the region according to network participation. Careful and accurate delineation of the claustrum itself, however, is already difficult at standard imaging resolution, and therefore, research that seeks to understand single networks that the claustrum participates in ought to be carefully designed to achieve fine resolution.
In our results, the claustrum appears to play a central role in linking multiple disparate structural brain networks. Conversely, the nodal betweenness centrality of predominantly frontal regions connected to the claustrum change significantly more if their connections to the claustrum are systematically removed than any other region. Traditionally, betweenness centrality increases as the relative contributions of shorter pathways come to dominate a network. Increases in betweenness centrality following removal of claustrum connectivity indicate that the white matter networks associated with them are, more heavily influenced by shorter—presumably local—pathways. This illustrates that the claustrum occupies a unique, and presumably critical, “location” in the overall architecture of network connectivity in the brain. Given the traffic converging on this hub, damage to the claustrum may be difficult to compensate for, since efficiency would be so greatly compromised by distributing such a large number of connections over a larger area.
Indeed, disruption to corticoclaustral connectivity—forcing a greater dependence among local regional connections—appears to result in a variety of neurological symptoms (Supporting Information, Table 3). Studies from a variety of human neurological and psychiatric syndromes [Smythies et al., 2014] have identified altered claustral morphometry or have found disrupted patterns of white matter connectivity.
Disorders such as Wilson's Disease, for example, are noteworthy for their specificity of claustral involvement. Wilson's Disease is an autosomal recessive genetic disorder in which copper accumulates in brain tissues [Lorincz, 2010]. Copper deposition in the claustrum is notable using T1-weighted neuroimaging and is often considered a neurological hallmark of the disease, and can lead to general executive control difficulty, as well as various symptoms of memory dysfunction [Sener, 1993]. Other neurological conditions have been examined in the claustrum as well.
Negative correlations between anhedonia and metabolism in the claustrum have been shown in both patients with unipolar depression and bipolar disorder [Chen et al., 2011]. Claustral amyloid plaques accumulation has been implicated in the outcomes of Alzheimer's disease and aging [Fernandez-Miranda et al., 2008a, b; Morys et al., 1994]. Severity of delusions in schizophrenia is correlated with the reduction in left claustral volume [Cascella et al., 2011] and schizophrenia patients with hallucinations show signs of white matter excesses [Shapleske et al., 2002].
See Supporting Information for further summary of clinical syndromes associated with claustrum dysfunction.
Our identification of the strongest network dependencies existing between the claustrum, frontal, and cingulate cortices is a particularly noteworthy result. The simulated excision of the claustrum's representation in the overall network forces these regions to effectively bias their connectivity via more spatially local pathways. Such changes do not appear in other brain areas.
This finding supports previous findings of frontal lobe connectivity across many methodologies, including electrophysiology [Berke, 1960; Chachich and Powell, 2004; Cortimiglia et al., 1991; Grasby and Talk, 2013], anterograde, and retrograde tracers [Arikuni and Kubota, 1985; Bayer and Altman, 1991; Pearson et al., 1982; Sadowski et al., 1997], and neuroimaging [Fauvel et al., 2014; Milardi et al., 2013; Park et al., 2012] and highlights the importance of widening the scope of functional observations of the claustrum beyond the primary sensory and motor regions.
Some authors have suggested that the claustrum may act to synchronize cortical subnetworks, which are responsible for a variety of coordinated behaviors [Smith and Alloway, 2010] possibly serving to counterbalance spatially over-represented cortical regions [LeVay and Sherk, 1981; Minciacchi et al., 1995]. The claustrum may also be particularly useful in sensory cross-modal matching [Arnow et al., 2002] and possibly modulate the cortical neuron receptive field properties [Shima et al., 1996].
Such putative roles emphasize the importance of inhibitory claustral connectivity in the regulation of conscious behaviors. Furthermore, it lends credence to the suggestions of Smythies et al. [2012a, b] that, unlike the thalamus, the claustrum may not be a strict multisensory processor, per se, but may be indispensable in organizing the information used by multisensory processors as relevant to the brain's executive functions.
In this way, the claustrum may well serve to filter signals about the relationships of all the thousands of sensorimotor inputs from the outside world to and from frontal and cingulate processing subnetworks.
Our study is, however, not without several limitations. Due to the paucity of large sample, in vivo human neuroimaging studies of the claustrum, we chose to focus more broadly on the macroscale structural connectomics properties of the structure, rather than examine factors which influence between-group differences.
For instance, we pooled male and female subjects together to get a picture of the connectivity of the average claustrum. We did not consider handedness or age-related effects. Handedness, gender, age, ethnicity, and a number of other phenotypic as well as genetic factors, will be very important for understanding the role of the claustrum in developmental, mental health, and age-related disorders.
Additionally, we made no attempt to identify subdivisions of the claustrum. Already exploring one of the smaller structures of the brain, it is arguable whether conventional imaging sequences provide sufficient spatial resolution or image contrast to permit the identification of tissue subclasses or reliable spatial landmarks thereof.
While earlier studies have suggested a compelling approach to parcellating claustral subdivisions using Magnetic Resonance (MR) [Pathak and Fernandez-Miranda, 2014], further validation and testing of such approaches are likely required before they can be of practical utility. Thus, here, we chose to focus on the claustrum as a single entity, which likely limits our ability to assign claustro-cortico projections with high specificity. Advances in ultrahigh field, high-resolution MRI will make such studies possible.
All in all, the high-resolution imaging of the claustrum and the exploration of phenomic variables is encouraged in future studies to determine how the structure and connectivity of the claustrum varies between genders, how it develops, and is altered due to aging and age-related disease.
While we have discussed how our results may support some of the functional findings of previous investigations, further research using functional imaging—such as fMRI—would be necessary to confirm that these functions are in fact correlated with these putative connections. Moreover, traditional functional imaging methods may not provide sufficient spatial resolution to accurately assess such a small structure; technological advancements may be necessary to obtain an accurate picture of the behavior of the claustrum. Hagmann et al.
[2008] applied community detection or modularity analysis to demonstrate a close relationship between structural connections and functional connections. Therefore, an analysis of the structural connections of the claustrum may help generate hypotheses regarding its functional role in the networks to which it contributes. Despite these caveats, our results are in line with previous neuroimaging and non-neuroimaging findings.
Given the results of our large-scale analysis—and in concurrence with previous insightful research [Crick and Koch, 2005]—we concur with the hypothesis that the claustrum is likely more than a simple neural relay station in the overall context of inter-regional white matter networks. Its differential influence on brain region subnetworks typically associate with high cognitive activity, attention, and action suggest that the claustrum plays a central role in linking multiple sensory networks with those regions, which can interpret and take action on such information. In support of previous hypotheses on the role of consciousness [Smythies et al., 2012a, b], our results show that the claustrum's position in the macroscale network would be structurally capable of allowing salient sensory information to enter brain regions associated with conscious thought while keeping superfluous input out of awareness.
In conclusion, the claustrum embodies a highly connected yet still curious brain structure whose in vivo examination in humans using neuroimaging has remained challenging due to its small, thin, irregular shape, and location. Determining the functions of the claustrum in humans has formed the basis of recent empirical and theoretical exploration [Smythies et al., 2014]. Our study highlights the claustrum as a critical and “ideally located” component in brain connectomic architecture upon which frontal and cingulate regions appear particularly dependent.
These results have particular relevance for neurological disorders and diseases that affect the structure and connectivity of the claustrum and provide further support for the claustrum as a putative neural filter for conscious thought.
ACKNOWLEDGMENTS
The authors wish to acknowledge the artistic talents of the late Mr. Carlos Mena, his creativity, friendship, and, for this article, his contribution to the graphical illustration of claustral anatomy. Finally, the authors thank the dedicated staff of the Institute for Neuroimaging and Informatics based at the University of Southern California, Los Angeles, CA.CONFLICT OF INTEREST STATEMENT
This research was conducted in absence of any commercial or financial relationships which could be construed as a potential conflict of interest.CONTRIBUTIONS
CMT pre-processed all neuroimaging data, helped to prepare the manuscript and supplemental materials. AI conducted all graph theoretical connectivity analyses. SYMG helped to prepare figures and review the manuscript text for technical correctness. JDVH conceived of the examination, guided the necessary data processing and analyses, and helped to author the manuscript and supplemental materials. All authors had the opportunity to discuss the results, participated in writing/editing, and made comments on earlier versions of the manuscript.Footnotes
- 1 Note: In our examination, due to issues of image resolution and the uncertainty in the literature over the precise intraregional boundaries for so doing, we purposefully did not attempt to segment the claustrum into putative substructures. However, see the chapter by Pathak and Fernandez-Miranda (2014), for a potential approach for subdividing the claustrum based on MRI data.
No comments:
Post a Comment