Structural and functional hyperconnectivity within the sensorimotor system in xenomelia

Abstract Introduction Xenomelia is a rare condition characterized by the persistent and compulsive desire for the amputation of one or more physically healthy limbs. We highlight the neurological underpinnings of xenomelia by assessing structural and functional connectivity by means of whole‐brain connectome and network analyses of regions previously implicated in empirical research in this condition. Methods We compared structural and functional connectivity between 13 xenomelic men with matched controls using diffusion tensor imaging combined with fiber tractography and resting state functional magnetic resonance imaging. Altered connectivity in xenomelia within the sensorimotor system has been predicted. Results We found subnetworks showing structural and functional hyperconnectivity in xenomelia compared with controls. These subnetworks were lateralized to the right hemisphere and mainly comprised by nodes belonging to the sensorimotor system. In the connectome analyses, the paracentral lobule, supplementary motor area, postcentral gyrus, basal ganglia, and the cerebellum were hyperconnected to each other, whereas in the xenomelia‐specific network analyses, hyperconnected nodes have been found in the superior parietal lobule, primary and secondary somatosensory cortex, premotor cortex, basal ganglia, thalamus, and insula. Conclusions Our study provides empirical evidence of structural and functional hyperconnectivity within the sensorimotor system including those regions that are core for the reconstruction of a coherent body image. Aberrant connectivity is a common response to focal neurological damage. As exemplified here, it may affect different brain regions differentially. Due to the small sample size, our findings must be interpreted cautiously and future studies are needed to elucidate potential associations between hyperconnectivity and limb disownership reported in xenomelia.


| INTRODUCTION
is an unusual and rare condition characterized by the persistent and compulsive desire for the amputation of one or more physically healthy limbs. Individuals suffering from xenomelia typically report that their unwanted limb does not belong to their body and that they would feel "more complete" after its removal (Blanke, Morgenthaler, Brugger, & Overney, 2009). While it is hard to empathize with these feelings and inferences, extensive psychiatric examination of persons with xenomelia revealed a normal mental status and especially the absence of any psychotic disorder (First, 2005;Hilti et al., 2013;van Dijk et al., 2013). A recent review (Brugger, Lenggenhager, & Giummarra, 2013) covering telephone-and internetbased interview studies in well over 100 persons with the disorder summarized the main findings: (1) The majority of affected persons are men (>90%); (2) there is an association of the disorder with nonheterosexuality, and a frequently accompanying symptom involves the erotic attraction by amputees; (3) adverse feelings linked to the unaccepted limb were usually noted in early childhood or adolescence; and (4) unaccepted limbs involve more frequently the lower limbs (>80%) and, in unilateral cases, there is a marked asymmetry in favor of leftsided limbs (66%-80%; First, 2005;Hilti et al., 2013;Johnson, Liew, & Aziz-Zadeh, 2011). This latter observation, that more left-sided than right-sided limbs are the target of the amputation desire, has made some authors propose that xenomelia would primarily be a neurological disorder (McGeoch et al., 2011). In clinical neurology, transient feelings of nonbelonging of an own limb (as in hemiasomatognosia or somatoparaphrenia) or of hostility toward own body parts (as in misoplegia) are often-described peculiarities of body-self relations (Brugger, 2007;Critchely, 1953;Loetscher, Regard, & Brugger, 2006). As a matter of fact, these disorders are typically seen after right hemisphere lesions and accordingly involve the left side of the body with a much higher frequency (Brugger & Lenggenhager, 2014;Vallar & Ronchi, 2009). Stimulated by these reflections, Brang, McGeoch, & Ramachandran (2008) were the first who set out to seek empirical evidence for a neurological account of xenomelia. They applied pinprick to two individuals suffering from the disorder and found that stimulation of undesired compared to accepted parts of their body elicited an abnormally strong galvanic skin response. This was interpreted as the consequence of a pathologically exaggerated sympathetic outflow from limbic areas, notably the insula, to the parietal cortex. Such pathological responsivity might arguably have interfered, during early development, with the formation of a proper representation of the unwanted limb. The desire for amputation would thus be the consequence of an underrepresentation of a body part despite intact primary sensory and motor functions. In a subsequent study with four persons with xenomelia, the same author group recorded the magnetoencephalographic signal to tactile stimulation below and above the line of desired amputation (McGeoch et al., 2011). In contrast to touch on accepted body parts, touch on the undesired limb failed to elicit a cortical response in one particular brain area, that is, the right superior parietal lobule (SPL). This part of the parietal lobe is known to be core for establishing and maintaining a coherent sense of having a body (Berlucchi & Aglioti, 2010;Giummarra, Gibson, Georgiou-Karistianis, & Bradshaw, 2008;Moseley, Gallace, & Spence, 2012). Intriguingly, a recent structural imaging study in participants with xenomelia (Hilti et al., 2013) identified neuroarchitectural abnormalities in exactly those cortical areas implicated by empirical research: the right anterior insular cortex (Brang et al., 2008) and the right SPL (McGeoch et al., 2011).
Unexpectedly, these authors also found decreased cortical surface area in the right primary somatosensory cortex (SI) involved in the representation of the left leg, and in the right secondary somatosensory cortex (SII; Hilti et al., 2013).
To identify potential neuroarchitectural differences that could underlie the phenomenological feeling of "over-completeness" reportedly characteristic of xenomelia (McGeoch et al., 2011), we set out to investigate structural and functional connectivity of the whole connectome. For that purpose, we compared 13 participants suffering from xenomelia with a carefully matched control group with respect to structural and functional connectivity using diffusion tensor imaging (DTI) combined with fiber tractography and resting state functional magnetic resonance imaging (rsfMRI), respectively. A whole-brain connectome as well as region-of-interest (ROI) approach have been applied to the DTI and rsfMRI data using advanced network-based statistic tools. For the definition of the network nodes, we applied the commonly used automated anatomical labeling (AAL) ROI atlas (Tzourio-Mazoyer et al., 2002) consisting of 116 different brain regions covering the whole cerebrum, cerebellum, and subcortical structures (whole-brain connectome approach), but we also constructed a xenomelia-specific network comprised by brain regions formerly shown to be anatomically altered in xenomelia compared with control men (Hänggi, Bellwald, & Brugger, 2016;Hilti et al., 2013) or differentially activated in different conditions in participants with xenomelia (van Dijk et al., 2013;ROI approach). We expected to find differences in structural and functional connectivity between subjects suffering from xenomelia and control men in brain regions belonging to the sensorimotor system and in the insula. We were also interested in finding further important brain regions associated with xenomelia.

| Subjects and study design
Thirteen men suffering from xenomelia were recruited via an Internet site. Aside from their desire for amputation, their medical history was free of any neurological and psychiatric disease, and thorough neurological and psychiatric examinations proved normal (Hilti et al., 2013; see Table 1 for details). A control group was separately recruited and did not differ from the xenomelia group with respect to handedness, footedness, age, and education. The individuals suffering from xenomelia, all desired an above knee amputation; eight of the left leg, two of the right leg and three of both legs. They completed the Zurich Xenomelia Scale (Aoyama,  To construct the connectivity matrix of the white matter pathways, the following fully automated preprocessing steps were realized: (1) In a first step, a binary brain mask was created using FSL's brain extraction tool. This mask was used in later steps to exclude non-brain tissue.
(2) Eddy current distortions and head movements were corrected using the EDDY_CORRECT tool of FDT.  connectivity matrices were then subjected to a network-based statistical analysis (see below). We also investigated a smaller and more xenomelia-specific network with 28 nodes (see below), the brain regions of which were derived from previous empirical neuroimaging findings in xenomelia (Hänggi et al., 2016;Hilti et al., 2013; van Dijk et al., 2013).

| Regions-of-interest definition for the structural and functional xenomelia-specific network analyses
A xenomelia-specific 28-node network has been constructed in addition. Sixteen nodes were derived from the eight clusters showing reduced cortical thickness or surface area in xenomelia compared with control men (Hilti et al., 2013). The 28-node xenomelia-specific network investigated in this study encompasses these 16 nodes in addition to the thalamus, caudate nucleus, putamen, and pallidum (Hänggi et al., 2016) as well as the ventral and dorsal premotor cortex (Blom et al., 2016;van Dijk et al., 2013). The eight subcortical nodes (bilateral thalamus, caudate nucleus, putamen, and pallidum) based on probability maps were derived from the Harvard-Oxford subcortical structural atlas (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases), whereas the remaining 20 nodes of the xenomelia-specific 28-node networks were spherical in nature and constructed using the MNI coordinates with a radius of 3 mm and 6 mm for the functional and structural analysis, respectively. These xenomelia-specific 28-node networks were used for both structural (with 6 mm radius spheres) and functional (with 3 mm radius spheres) network analyses. A larger radius for the spheres in the DTI analysis has been used to cover the white matter sufficiently. It is important to note that the findings reported by Hilti et al. (2013) and those reported by Hänggi et al. (2016) are based on the same experimental subjects as investigated in this study. Therefore, the construction of the xenomelia-specific 28-node network is not fully independent from the data previously reported (Hänggi et al., 2016;Hilti et al., 2013).
However, in addition to the independent AAL atlas (116-node network analyses), some of the ROIs (i.e., ventral and dorsal premotor cortex) of the 28-node xenomelia-specific network were derived from the findings reported by van Dijk et al. (2013) and the subcortical structures thalamus, caudate nucleus, putamen and pallidum were derived from another independent brain atlas.
The functional network analyses have been performed on a wholebrain basis as well as restricted to the edges that showed statistically significant differences in structural connectivity strength (DTI analysis) between the two groups. For that purpose, the strength of the functional connections between all pairs of nodes that were not involved in the structural subnetworks showing hyperconnectivity in xenomelia were set to zero in both the functional 116-node and functional 28node connectivity matrices.

| Network-based statistical analyses
The network-based statistical analyses of the structural (DTI) and functional (rsfMRI) connectivity data were performed using the network-based statistic (NBS) tool (Zalesky et al., 2010; https://www. nitrc.org/projects/nbs/) using MATLAB (version R2013b, http:// www.mathworks.com/). NBS is a method to control the family-wise error rate when mass-univariate testing is performed at every connection comprising the graph (network). NBS exploits the extent to which the connections comprising the contrast or effect of interest are interconnected. It is based on the principles underpinning traditional cluster-based thresholding of statistical parametric maps (Zalesky et al., 2010).
On the basis of the general linear model approach, we used the t-test module in NBS to compare the number of reconstructed streamlines of each connection (structural connectome) or the Fisher's z-transformed correlation coefficient between two brain regions (functional connectome) of the 116-and 28-node networks between the two groups. Comparisons were done with the component extent option in NBS that is suited for detecting an experimental effect that is relatively weak, but extends to encompass many connections. The numbers of streamlines of the significant connections (structural connectome) and the mean correlation (functional connectome) were extracted and averaged up over all significantly altered connections to correlate these values with each other.
For the NBS analyses based on the 116-node network, we reported two network solutions using two different sensitivity thresholds (here, t-thresholds), that is, a solution at a lower threshold (solution with many edges) and a solution at a higher threshold (solution with only few edges) for both contrasts (xenomelia < controls and xenomelia > controls). These t-thresholds are not the actual alpha error probabilities. Rather they represent sensitivity (set) thresholds (Zalesky et al., 2010) and are used to determine which edges of the connectivity matrix form the largest subnetwork (suprathreshold component) that is subjected to the permutation statistic.
These sensitivity thresholds must be determined by exploration and are therefore indeed chosen in an arbitrary way. However, this does not affect the false-positive rate of the actual permutation statistic of the alpha error probability. Reporting findings based on selective t-thresholds does not represent some sort of "cherry picking" because statistically significant subnetworks emerged at almost all set (sensitivity) thresholds explored. Therefore, we simply chose a lower threshold that provided a subnetwork of a size that is still interpretable, that is, not possessing too many connections at low thresholds. At high thresholds, we reported the threshold at which a statistically significant single subnetwork emerged, that is, we did not report subnetworks that disintegrated into different components at higher thresholds. It is important to note that in all analyses reported, we controlled the alpha error probability (p < .05) for multiple comparisons using 5,000 permutations of the group label. The permutation testing method implemented in NBS (Zalesky et al., 2010) is not new and is synonymous with conventional cluster-based thresholding procedures of statistical parametric maps (Bullmore et al., 1999;Hayasaka & Nichols, 2004;Nichols & Holmes, 2002

| Other statistical analyses
Demographic, xenomelia-related, psychiatric as well as global brain

| Demographic, xenomelia-related, psychiatric, and global brain measures
Demographic, xenomelia-related, psychiatric as well as global brain measures are summarized in

| Structural 116-node whole-brain connectome analyses
The network-based statistical analyses of the 116-node network revealed a subnetwork with statistically significantly increased structural connectivity in participants with xenomelia compared with control men ( Figure 1, Table 2). With the lower sensitivity (set) threshold (t = 2.30), a subnetwork comprised by 35 connections distributed over 33 nodes has been revealed (p = .019, corrected for multiple comparisons, Figure 1a, Table 2). Most of these connections were realized among nodes belonging to the sensorimotor system such as the paracentral lobule (PCL), postcentral gyrus (PoCG), supplementary motor area (SMA), and parts of the cerebellum and these findings were, except for the cerebellum, clearly lateralized to the right hemisphere ( Figure 1a, b, Table 2).
With the higher sensitivity threshold (t = 2.49), a subnetwork comprised by 11 connections distributed over 12 nodes has been found (p = .039, corrected for multiple comparisons; Figure 1b, Table 2). All nodes involved belong to the sensorimotor system and except for one node (cerebellum 4, 5 left), all nodes were within the right hemisphere.
Animations of the subnetworks presented in Figure 1a, b can be found online in the Supplementary Animations S1 and S2, respectively.
No subnetworks with reduced structural connectivity in xenomelia compared with control men have been found.

| Structural 28-node xenomelia-specific network analyses
The network-based statistical analyses of the 28-node xenomeliaspecific network (Hilti et al., 2013;van Dijk et al., 2013) revealed a subnetwork with increased structural connectivity in participants with xenomelia compared with controls ( Figure 2, Table 3). With a sensitivity threshold of t = 1.295, a subnetwork comprised by 28 connections distributed over 19 nodes has been found ( Figure 2, Table 3), albeit only on a trend level toward statistical significance (p = .087, corrected for multiple comparisons).
An animation of the subnetwork presented in Figure 2 can be found online in the Supplementary Animation S3. No subnetworks with reduced structural connectivity in xenomelia compared with controls have been found.

| Functional 116-node whole-brain connectome analyses
The functional network analyses have been performed using the full connectivity matrix information as well as using only the connections that showed statistically significant differences in structural connectivity strength between the two groups (see 3.2 above), setting all other connections in the connectivity matrix to zero.
The network-based statistical analyses of the 116-node network using the full connectivity matrix information revealed a subnetwork with increased functional connectivity in participants with xenomelia compared with control men (Figure 3, Table 4). With the lower sensitivity threshold (t = 3.60), a subnetwork comprised by 24 connections distributed over 20 nodes has been revealed (p = .037, corrected for multiple comparisons, Figure 3a, Table 4). Fourteen of these 24 connections were found among nodes belonging to the sensorimotor system with a preponderance of cerebellar nodes, but also including the right SMA, right PCL, right PoCG, and bilateral precentral gyrus (Figure 3a, Table 4). With the higher sensitivity threshold (t = 4.20), a subnetwork comprised by eight connections distributed over seven nodes has been found (p = .022, corrected for multiple comparisons, Figure 3b, Table 4). In this solution, the bilateral middle cingulate cortex encompassed most connections ( Figure 3b, Table 4).
Animations of the subnetworks presented in Figure 3a, b can be found online in the Supplementary Animations S4 and S5, respectively.
No subnetworks with statistically significantly reduced functional connectivity in participants with xenomelia compared with healthy control men have been revealed.

| Functional connectivity within the structurally hyperconnected subnetworks
The network-based statistical analyses of the 116-node network using only those connections of the connectivity matrix that were structurally altered between groups revealed a subnetwork with increased functional connectivity in participants with xenomelia compared with control men (Figure 4, Table 5).
With a sensitivity (set) threshold of t = 0.40, a subnetwork comprised by 23 connections distributed over 23 nodes has been revealed (p = .035, corrected for multiple comparisons, Figure 4, Table 5). Almost all of these connections were realized among nodes belonging to the sensorimotor system such as the PCL, SMA, thalamus, caudate nucleus, pallidum, and parts of the cerebellum and these findings were, with the exception of the subcortical structures, clearly lateralized to the right hemisphere ( Figure 4, Table 5). No subnetworks with statistically significantly reduced functional connectivity in participants with xenomelia compared with healthy control men have been revealed.
Also, no statistically significant differences in functional connectivity within the 28-node xenomelia-specific network have been revealed between groups when focusing only on the connections of the 28-node xenomelia-specific subnetwork that already showed structural hyperconnectivity between groups (see 3.3 above).

| DISCUSSION
This study reveals, to the best of our knowledge, first evidence for alterations in structural and functional brain connectivity in persons with xenomelia. Thirteen men suffering from an intense desire to have one leg (or both legs, n = 3) amputated were compared with a carefully matched control group. Building on previous reports of
In addition to the ROI-based approach that focused on the abovementioned sensorimotor brain regions, we also applied whole-brain structural and functional connectome analyses to uncover potentially additional regions associated with the condition, but also to establish the specificity of the findings reported here.
One such region, the SPL, was found to be unresponsive to tactile stimulation of specifically those body segments that contribute to the feeling of "overcompleteness" and are therefore desired to be amputated (McGeoch et al., 2011). A consecutive morphometric study revealed reduced cortical thickness or surface area, not only in the right SPL, but also in right SI and right SII as well as in the right insula (Hilti et al., 2013). Evidence for the involvement of subcortical brain structures in xenomelia, hitherto not in the focus of neurologically motivated investigations, can be derived from a recently published study (Hänggi et al., 2016) conducted at our department and based on the same subjects as investigated in the present work. This study applied a vertex-wise shape analysis that showed positive and negative F I G U R E 4 Increased functional connectivity within the structurally hyperconnected subnetwork. Shown are the results of the 116-node functional network analysis restricted to the connections that already showed structural hyperconnectivity (see solution 1 in Figure 1 and Table 2). Blue circles represent nodes of the sensorimotor system, whereas turquoise circles represent all other nodes. The nodes are presented in Montreal neurological institute (MNI) space using the centroids of the regions of interest of the automated anatomical labeling (AAL) atlas. Red lines connect the nodes (brain regions) showing enhanced functional connectivity (p = .035, corrected for multiple comparisons) in xenomelia compared with healthy control men tissue displacements (shape differences) in the form of thinning (hypotrophy) of bilateral dorsomedial putamina, left ventromedial caudate nucleus, and left medial pallidum in xenomelia compared with controls, whereas shape differences in the form of thickening (hypertrophy) of bilateral lateral pallida and left frontolateral thalamus in participants with xenomelia were also evident (Hänggi et al., 2016). This study provides strong evidence that in addition to the anatomical alterations already reported in the literature for cortical regions (Hilti et al., 2013;McGeoch et al., 2011), the subcortical structures thalamus, putamen, caudate nucleus, and pallidum might be related to xenomelia as well (Hänggi et al., 2016). A preliminary fMRI study based on five subjects suffering from xenomelia extended the above-mentioned findings by suggesting that the ventral as well as dorsal premotor cortex are also brain regions importantly involved in generating the clinical picture of xenomelia, both from a functional (van Dijk et al., 2013) as well as structural point of view (Blom et al., 2016).
Although xenomelia has initially been conceptualized as a personality disorder and denoted body integrity identity disorder (BIID; First, 2005;First & Fisher, 2012) or even conceptualized as a paraphilia and denoted apotemnophilia (Brang et al., 2008;Ramachandran & McGeoch, 2007), we suggest that xenomelia can be conceptualized as a neurological disorder that is free of any of the negative connotations associated with BIID and apotemnophilia as has been already suggested by others elsewhere (McGeoch et al., 2011). However, in contrast to McGeoch et al. (2011), who suggested that xenomelia might be a right parietal lobe syndrome, we were able to show that in addition to the right parietal lobe and right SI, SII, and the insula (Hilti et al., 2013), regions which have been shown to be hyperconnected in this study, several further key players of the sensorimotor system such as the PCL (housing SI leg representations), SMA, basal ganglia, cerebellum, and premotor cortex also revealed structural and functional hyperconnectivity in xenomelia compared with healthy control men.

| Structural hyperconnectivity in xenomelia
Our structural whole-brain connectome analysis using network-based statistics (Zalesky et al., 2010) revealed a subnetwork with enhanced connectivity in xenomelia compared with control men (Figure 1).
When focusing on the hyperconnected nodes, it is quite impressive that most of the affected nodes belong to the sensorimotor system (see Table 2). Of the 35 connections distributed over 33 nodes shown in solution 1 of the 116-node network analysis (Figure 1a), 17 connections have both nodes within the sensorimotor system, 16 connections have at least one node within the sensorimotor system, whereas T A B L E 5 Enhanced functional connectivity in participants with xenomelia compared with control subjects derived from the 116-node network analysis when restricting it to the structurally hyperconnected subnetwork (solution 1) shown in Figure 1 and only two connections were between regions not predicted a priori.
Furthermore, most of the cortical nodes affected are located within the right hemisphere, whereas the subcortical nodes pallidum, caudate, putamen, and thalamus were more often affected in the left than the right hemisphere. This finding is in line with our subcortical shape analyses that revealed a preponderance of tissue displacements in left compared with right subcortical structures (Hänggi et al., 2016). The cerebellum, PCL, and SMA are the nodes with the highest degree of affected connections when inspecting solution 2 of the 116-node network analysis (Figure 1b), and all these nodes except one are within the right hemisphere (connections printed in bold in Table 2).
In contrast to the two subnetworks found with the 116-node network analyses that showed clear lateralization to the right hemisphere (see above), the subnetwork found in the 28-node xenomelia-specific network analysis is less lateralized to the right and, more important, showed only a statistical trend toward significance. However, when looking at the three most important cortical brain regions that have been found to be altered in xenomelia with respect to cortical thickness and surface area (Hilti et al., 2013), that is, the SI (four connections) and SII (two connections) and the SPL (two connections), structural connectivity of the 28-node xenomelia-specific network analysis was enhanced only in the right-hemispheric representations of these three xenomelia-related core brain regions.
Taken together, our structural connectome analyses provide strong evidence that in addition to the structural and functional anomalies of specific brain regions in xenomelia already reported in the literature (Blom et al., 2016;Hänggi et al., 2016;Hilti et al., 2013;McGeoch et al., 2011;van Dijk et al., 2013), the number of reconstructed streamlines from/to these brain regions is strongly increased in men with compared to without xenomelia, suggesting that these regions are hyperconnected to other brain regions inside and outside of the sensorimotor system. The network solution 1 and 2 of the 116node network analyses revealed that a lot of the connections showing structural hyperconnectivity have nodes located within the cerebellum, including both the cerebellar hemispheres as well as the vermis.
With respect to the specificity of the findings, the nodes found to be hyperconnected in xenomelia in network solution 2 ( Figure 1b and connections printed in bold in Table 2) showed high specificity because 21 of 22 nodes were within the sensorimotor system. One node, the right frontal operculum is considered to be a premotor region in the broader sense as well and could therefore also be accounted for the sensorimotor system. The frontal operculum is importantly involved in the rubber hand illusion (Ehrsson, Spence, & Passingham, 2004), which consists in a distorted sense of self of one particular body part (Lenggenhager, Hilti, & Brugger, 2015).
More broadly speaking, the same region is involved in the construction and maintenance of a coherent representation of the entire body (Moseley et al., 2012;Tsakiris, Hesse, Boy, Haggard, & Fink, 2007), a representation arguably disturbed in xenomelia. The right PCL showed three pathways to/from the cerebellum that were hyperconnected, and a cluster with an anomaly small cortical surface area in xenomelia has already been reported in this brain region for the same participants (Hilti et al., 2013). It is important to realize that the primary somatosensory and primary motor representation of the left leg, that is, the body part most frequently desired to be amputated in xenomelia in general and specifically in our sample, are exactly represented in this particular brain region of the right PCL.
One of the most interesting nodes related to xenomelia within the sensorimotor network is a certain area in the right SPL (Hilti et al., 2013;McGeoch et al., 2011;van Dijk et al., 2013) and this region has been suggested to be the highest region of integration with respect to the concepts of body scheme, image, or even body matrix (Moseley et al., 2012). In the 116-network analyses, neither the right nor the left SPL showed to be hyperconnected to any other nodes in the whole-brain network in xenomelia. This might be related to the fact that the SPL is a relatively large brain region and the area of differences reported within the SPL is rather small (Hilti et al., 2013;McGeoch et al., 2011). In the xenomelia-specific network analysis (28 nodes), however, where the node SPL has been constructed using a sphere around the coordinates reported in the literature (Hilti et al., 2013), two pathways from/to the right SPL were hyperconnected, the stronger connection is running to/from the right caudate nucleus and the weaker connection is running to/from the left caudate nucleus (Table 2).

| Functional hyperconnectivity in xenomelia
Our functional network analyses followed the same approach as reported above for the structural ones; hence, we first report the findings from the 116-node network analyses followed by the xenomeliaspecific 28-node network analysis. In addition, we constructed a functional subnetwork restricted to the connections that have been shown to be structurally hyperconnected (see Figure 1a, Table 2). When T A B L E 6 Associations between functional and structural connectivity strength of the altered networks focusing on the hyperconnected nodes, a predominant involvement of the sensorimotor system is less impressive (see Table 4) compared to the structural network findings. Of the 24 connections distributed over 20 nodes shown in solution 1 of the 116-node network analysis ( Figure 3a), 14 connections were found among nodes belonging to the sensorimotor system with a preponderance of cerebellar nodes, but also including the right SMA, right PCL, right PoCG as well as the left and right precentral gyrus.
In contrast to the findings from the structural network analyses (see Figures 1 and 2, Tables 2 and 3), the connections with increased functional connectivity derived from the 116-node functional network analyses are less lateralized to the right hemisphere (see Figure 3, Table 4). The nodes with the highest degrees in the 116-node functional network analyses are located in the middle part of the cingulate cortex bilaterally (10 connections) and in the cerebellum (9 connections). Although we did not formulate any a priori hypothesis with respect to the cingulate cortex, we note that the cingulate motor area, located ventrally adjacent to the SMA, is housed within the middle part of the cingulate cortex (Brazdil, Kuba, & Rektor, 2006;Liberg et al., 2014;Ullsperger & von Cramon, 2003). This cingulate motor area is also part of the sensorimotor system, and it should therefore not come as a surprise that it evidenced an increased functional connectivity in our xenomelia group.

| Associations between structural and functional connectivity strength of the altered networks
The correlations between structural and functional connectivity strengths of the hyperconnected networks were all positive suggesting that the two different types of connectivity alterations are related, that is, the stronger the structural connections the stronger the functional ones. Although this evidence is based on correlations only that do not provide any information with respect to causality, it is conceivable that functional hyperconnectivity results rather from structural hyperconnectivity than the other way around. However, if structural hyperconnectivity has the same potential to break down a network's functioning as has a structurally hypoconnected network, then functional hyperconnectivity might be interpreted as a sign of functional compensation in response to the structural network impairments. Exactly this pattern of connectivity dissociation has been reported previously in healthy subjects notably in frontoparietal networks (Eickhoff et al., 2010) and in pathological conditions such as major and late life depression Steffens, Taylor, Denny, Bergman, & Wang, 2011;Wu et al., 2011) and stroke (Zhang et al., 2014). Alternatively, however, an increase in functional connectivity may reflect a pathological loss of inhibitory neural activity within structurally damaged cortical networks as has been shown in patients suffering from amyotrophic lateral sclerosis (Douaud, Filippini, Knight, Talbot, & Turner, 2011).
For clarifying the causal relationship of the connectivity differences, further research must be performed using longitudinal study designs.

| Neurobiological mechanisms of hyperconnectivity
The desire for healthy limb amputation might be caused by a disturbed representation of the body at different levels of sensorimotor integration. Among the factors with a perturbation potential are cortical regions showing reduced cortical thickness or surface area of (Hilti et al., 2013) or subcortical regions showing altered striatal, pallidal, and thalamic shape (Hänggi et al., 2016), but also altered connectivities in networks responsible for the construction and maintenance of a coherent body image (findings of this study). Altered structural and functional network connectivities have also been reported for a variety of psychiatric diseases (Cao, Wang, & He, 2015) and it has been suggested that, at least for functional brain networks, hyperconnectivity is a fundamental response to (focal) neurological disruption (Hillary et al., 2014(Hillary et al., , 2015.
However, the question of when and how the structural and functional hyperconnectivity uncovered in the present investigation has been established in the first line is not easy to answer. We may only speculate about potential neurobiological mechanisms underlying the hyperconnectivity patterns observed in xenomelia in this study.
One such potential mechanism may be a failure in neural pruning, that is, an insufficient elimination of unneeded neural populations and connections. Such pruning processes are driven either by apoptosis (Cusack, Swahari, Hampton Henley, Michael Ramsey, & Deshmukh, 2013) or by experience (Eyding, Schweigart, & Eysel, 2002;Yu et al., 2013), but in all likelihood by a combination of both factors. It is well known from neurophysiological studies that the rapid phase of synaptogenesis in early childhood is followed by subsequent longer periods of pruning, during which synapses are eliminated by about 40% to reach near-mature levels at 11 years of age (Huttenlocher, 1984(Huttenlocher, , 1990Huttenlocher & de Courten, 1987;Huttenlocher, de Courten, Garey, & Van der Loos, 1982). For example, it has been suggested that sensory deprivation during early childhood might cause a substantial reduction in pruning of the exuberant cortico-cortical and/or corticothalamo-cortical connections that exist in early infancy. This reduced pruning due to sensory deprivation will result in a greater survival of the exuberant connections and, consequently, a thicker visual cortex.
This has been suggested for early or congenitally blind subjects in whom visual cortex was found to be thicker than in sighted control subjects (Anurova, Renier, De Volder, Carlson, & Rauschecker, 2015). It is conceivable that reduced pruning, whatever its ultimate causes may be, results in a white matter hyperconnectivity in participants suffering from xenomelia.
In what the deprivation during early development could consist in the case of xenomelia, is unclear. We would expect a lack of regular somatosensory input from and/or an underuse of the limb later considered not belonging, but persons with xenomelia do not systematically report such things. Some unspecific impoverishment during early childhood has occasionally been mentioned (Riordan & Appleby, 1994), but seems an unlikely candidate to account for the rejection of one particular limb as a consequence of aberrant neurogenesis.
The nodes of the sensorimotor system are structurally hyperconnected to each other and to nodes outside of the sensorimotor system as shown in this study, for which structural connectivity was operationalized as the number of reconstructed streamlines between two nodes.
This number is related to the number and/or volume of the real axonal pathways. Because streamlines are more abundant among sensorimotor nodes in xenomelia compared with the control group, this increase in structural connectivity can explain the functional hyperconnectivity among those sensorimotor nodes. If there are more axons, or axons have a larger diameter between sensorimotor nodes in xenomelia, either more information is propagated between those nodes or information transfer is faster, which in turn has the potential to result in increased functional connectivity. Such enhanced functional connectivity is capable of producing a breakdown in the normal functioning (Hillary et al., 2014(Hillary et al., , 2015 of the sensorimotor system, may by inference also hamper establishing a proper body image, and may ultimately lead to the feeling that a body part is not belonging to the rest of the body.

| Limitations
Several limitations of this study are worth mentioning. First, the sample size (13 participants with xenomelia plus an equal number of control participants) is rather small. However, we have gone a long way to find enough persons presenting with the condition, yet belonging to an adequately homogenous group of nonpsychotic individuals, who do not display any self-injurious behavior, do not suffer from bodydysmorphic disorder, and whose amputation desire is not primarily sexually motivated. Second, because of the small sample size, it was not possible to build subgroups to investigate, for instance, differences with respect to whether the amputation desire targeted the left leg, the right leg, or both legs. Whether the pattern of hyperconnectivity observed in this study changes if more individuals with a rightsided amputation desire are included remains to be shown in future studies that should recruit larger samples to address the issue of lateralization explicitly. It might be difficult, however, to recruit enough subjects with a right leg amputation desire due to the preponderance of the left leg as the amputation target. Third, we investigated men only; again, given the rarity of the condition and the fact that an overwhelming majority of persons affected are male (only about 10% of people affected are female; Blom, Hennekam, & Denys, 2012) this gender bias may be excused. Fourth, although we tested the networks across a range of set thresholds and corrected each single threshold for multiple comparisons, we did not additionally adjust the alpha error probability for the number of set thresholds tested. However, a Bonferroni correction would be too conservative in this context. Fifth, the analysis of the structural 28-node xenomelia-specific network revealed an effect only significant on a trend level (p = .087, corrected) and some of the brain regions of the 28-node network were selected nonindependently, that is, based on the same experimental subjects as investigated in previous studies (Hänggi et al., 2016;Hilti et al., 2013). This finding must be interpreted cautiously, therefore, and our study should be taken as a springboard to more extended examinations. Finally, this study is based on cross-sectional data and therefore it remains unclear whether the structural and functional hyperconnectivity evident within the sensorimotor system are the causes of xenomelia or rather the consequence of the lifelong persistent desire for limb amputation and its associated experiences.

| CONCLUSIONS
Proponents of a neurological nature of xenomelia have suggested the right SPL as the key node affected within a brain network responsible for body ownership (Brang et al., 2008;Hilti et al., 2013;McGeoch et al., 2011;van Dijk et al., 2013). This network of body ownership or body image (Moseley et al., 2012) encompasses brain regions of sensorimotor relevance, such as the SI, SII, basal ganglia, thalamus, and premotor cortex (Blom et al., 2016;Hänggi et al., 2016;Hilti et al., 2013;van Dijk et al., 2013). This study extends these previous findings by showing that, in addition to the focal structural alterations in the above mentioned brain structures, structural and functional connectivity between the single parts of this network are markedly increased in xenomelia. Our study provides strong empirical evidence of structural and functional hyperconnectivity in the sensorimotor system in individuals with xenomelia, comprising those regions that are core for the experience of a unified bodily self. Structural and functional brain hyperconnectivity is a common response to focal neurological disruption and affects brain regions differentially (Hillary et al., 2014(Hillary et al., , 2015 as observed here within the somatosensory system in xenomelia. Brain hyperconnectivity might account for the limb disownership commonly reported by xenomelia subjects. We speculate that one phenomenal correlate of this hyperconnectivity might be a pathological "over-attention" to specific body parts and the feeling of "over-completeness", culminating, in some individuals, in a desire for healthy limb amputation. We emphasize, however, that the present findings must be interpreted cautiously due to the small sample size.

ACKNOWLEDGMENTS
We thank the subjects for their participation, and are generally grateful to those suffering from the condition under study for their willingness to contribute to research. We acknowledge the assistance of Michael Linnebank, Ilijas Jelcic, and Katarina Sabova in the neurological examinations and Simone Hobi in the psychiatric assessments.
We also thank Hansruedi Baetschmann for the usage of his MATLAB scripts that support the network-based statistic (NBS) tool.