CNS imaging characteristics in fibromyalgia patients with and without peripheral nerve involvement

We tested the hypothesis that reduced skin innervation in fibromyalgia syndrome is associated with specific CNS changes. This prospective case–control study included 43 women diagnosed with fibromyalgia syndrome and 40 healthy controls. We further compared the fibromyalgia subgroups with reduced (n = 21) and normal (n = 22) skin innervation. Brains were analysed for cortical volume, for white matter integrity, and for functional connectivity. Compared to controls, cortical thickness was decreased in regions of the frontal, temporal and parietal cortex in the fibromyalgia group as a whole, and decreased in the bilateral pericalcarine cortices in the fibromyalgia subgroup with reduced skin innervation. Diffusion tensor imaging revealed a significant increase in fractional anisotropy in the corona radiata, the corpus callosum, cingulum and fornix in patients with fibromyalgia compared to healthy controls and decreased FA in parts of the internal capsule and thalamic radiation in the subgroup with reduced skin innervation. Using resting-state fMRI, the fibromyalgia group as a whole showed functional hypoconnectivity between the right midfrontal gyrus and the posterior cerebellum and the right crus cerebellum, respectively. The subgroup with reduced skin innervation showed hyperconnectivity between the inferior frontal gyrus, the angular gyrus and the posterior parietal gyrus. Our results suggest that the subgroup of fibromyalgia patients with pronounced pathology in the peripheral nervous system shows alterations in morphology, structural and functional connectivity also at the level of the encephalon. We propose considering these subgroups when conducting clinical trials.

The fibromyalgia syndrome (FMS) is a chronic pain disorder with a prevalence of approximately 2% in the general population 1 . Abnormalities in pain processing regions in the CNS, neurotransmitter levels, the autonomic nervous system, and in small fibers of the peripheral nervous system are frequent findings associated with FMS, but their causal connection to the manifestation and course of its symptoms is still unclear. Altered pain processing at the level of the CNS is regarded as a major pathophysiological factor 2,3 . However, structural lesions and functional deficits were also observed at the level of the PNS, where specifically small fiber pathology is a robust finding in a substantial group of patients fulfilling the established diagnostic criteria of FMS 4 . These findings of structural and functional alterations in FMS at both CNS and PNS level were reproducible: CNS structural measurements, like voxel-based-morphometry or cortical reconstruction, have revealed atrophy of the grey matter in the left prefrontal cortex and the posterior cingulate cortex 5,6 . Diffusion tensor imaging (DTI) has shown changes in white matter integrity, e.g. in the corpus callosum 7 , and functional magnetic resonance imaging (fMRI) has identified hyperactivity in many regions related to pain processing 8 , such as the left prefrontal cortex and in the posterior cingulate cortex, the insular cortex and the cerebellum. Functional connectivity was increased in the default mode network (DMN) and pain related areas, such as the insular cortex [9][10][11] . In the PNS, we and other groups described a decrease in intraepidermal nerve fiber density (IENFD) [12][13][14][15][16][17] , which was related to symptom severity 4 . Subgrouping according to intraepidermal nerve fiber density. Patients from the previous study 4 who had either normal IENFD at the lower leg (above the lower limit of normal 5.4 fibers/mm) and at the upper thigh (above the lower limit of normal 8.5 fibers/mm) or a non-length dependent abnormal IENFD, which means the IENFD was below the lower limits at both biopsy sites, were re-recruited, i.e. were contacted by H.-C. A. and invited to a follow-up appointment for MRI imaging. The first group was termed "noPNS", the second group "PNS". These cut-off values were determined based on skin biopsies of these two regions of 120 healthy women (median age = 50 years, range = 20-84 years) in our department. The cut-off values represent the lower limit of the standard deviation of the IENFD results of all the healthy controls investigated in our laboratory.
Fibromyalgia related symptoms. Results of the questionnaire and clinical examination data of the FMS patients have already been published 4 . To evaluate pain severity, two pain scores were used (Graded Chronic Pain Scale (GCPS) and Neuropathic Pain Symptom Inventory (NPSI)). In order to assess the depressiveness of the patients, the "Allgemeine Depressionskala" (ADS) was used, which is a German version of the Center for Epidemiological Studies-Depression scale questionnaire 19 . To evaluate catastrophizing, the Pain Catastrophizing Scale (PCS) 20 , which is a self-report measure, consisting of 13 items scored from 0 to 4, resulting in a total possible score of 52, was assessed. To test the anxiety level, the State-Trait Anxiety Inventory (STAI) was used 21 , which is a commonly used measure of trait and state anxiety. In order to assess the influence of the disease on daily experience, the Fibromyalgia Impact Questionnaire (FIQ) 22 was used. Also, the Symptom Severity Scale (SSS) was used to query other FMS-associated symptoms 18 . It measures three key symptoms during the past week: Fatigue, unrefreshed wakening and cognitive impairment. The O'Leary-Sant Symptom and Problem Index assesses the impairment by bladder dysfunction 23 and was selected, as FMS patients frequently report abdominal pain and problems with urination. Data collected in the context of the clinical diagnostics, such as the conduction studies of the sural nerve and the blood values, for example HbA1c and vitamin D, were also analyzed.
MR imaging and analysis. Data acquisition. Magnetic resonance imaging was performed on a Siemens MAGNETOM Prisma fit Scanner (Siemens Healthcare GmbH, Erlangen, Germany), operating at 3 T, equipped with a 64-channel head coil at the Department of Neuroradiology, University Hospital Würzburg. For each participant we included a structural T1-weighted (T1w) sequence, diffusion weighted imaging (DWI), fieldmap data and resting-state functional MRI (rs-fMRI) series. The T1w gradient echo MPRAGE sequence (repetition time (TR) 2400 ms, echo time (TE) 3.17 ms, flip angle (FA) 8°, inversion recovery (IR) 1000 ms) contained 176 sagittal slices with an isotropic voxel size of 1 × 1 × 1 mm. The visual examination of the T1w-structural images revealed no gross morphological abnormalities for any patient or subject. DWI was obtained using multiband echo-planar imaging (EPI) with the following parameters: TR = 3100 ms, TE = 89 ms, FA = 90°, isotropic voxel size of 2 × 2 × 2 mm. Diffusion data were collected with reversed phase-encode blips, resulting in pairs of b0images with distortions in opposite directions for further susceptibility induced distortion correction. Resting state fMRI data was acquired using a T2*-weighted multiband EPI sequence with TR = 1610 ms, TE = 30 ms, FA = 70°, isotropic voxel size of 2 × 2 × 2 mm, 69 slices. During the 9-min resting state fMRI acquisition period www.nature.com/scientificreports/ with 300 volumes the subjects were told to lie still and remain awake with their eyes open. Participants' motion was minimized using tight foam pads around the head, their physiology was monitored.
Structural analysis. Cortical reconstruction and volumetric segmentation was performed with the FreeSurfer image analysis suite v6.0.0 (Martinos Center for Biomedical Imaging, Boston, MA, USA) using the 3D T1w data. The technical details of these procedures are described in prior publications 24,25 . Parcellations were classified according to the Desikan-Killiany Atlas 26 . The exact listing of all ROIs used can be found under supplementary material 1a. Volume was measured in mm 3 . In addition to the exploratory whole-brain approach, hypothesisdriven group comparisons were also performed with volumes of cortical regions that had been shown in a meta-analysis to be specifically affected in FMS 5 (namely, the left medial frontal cortex and the right posterior cingulate cortex). Since the factor age has been shown to be associated with differences in white and grey matter volume 27 , we decided to include this factor as a covariate. We also included the pain intensity score of the GCPS as a covariate.
Structural connectivity: diffusion tensor imaging. The Oxford Centre for Functional Magnetic Resonance Imaging of the Brain software library (FSL, Oxford, UK, https:// www. fmrib. ox. ac. uk) 28 was used for DTI data analysis and preprocessing. Our diffusion data, recorded in reversed phase-encode blips, were preprocessed using the FSL tools "topup" 29 , "eddy (correction)", "BET" 30 , and "FNIRT". FA images and eigenvalue images were created by fitting a tensor model to the preprocessed diffusion data using the FSL FDT toolbox (Functional MRI of the Brain Diffusion Toolbox, DTIFIT). For ROI specific evaluation of the FA data we created a mask with the ICBM-DTI-81 white-matter labels atlas (Laboratory of Brain Anatomical MRI, Johns Hopkins University 31 ) in the same space and calculated the average FA value of all voxels in 48 ROIs. The exact listing of all tracts used as ROIs can be found under supplementary material 1b.These data were analyzed for group comparisons with ANCOVAs including post-hoc testing (Tukey) and correlated with clinical data and questionnaires using a spearman Rho correlation for non-normally-distributed z-standardized clinical data analysis (significance level of 0.01, twotailed, confidence interval 0.95). In addition to the exploratory whole-brain approach, hypothesis-driven group comparisons were also performed with white matter tracts that had been shown to be affected in FMS (namely the thalamus 32 , the corpus callosum 7 , the cingulum and the white matter adjacent to the insula (anterior limb of the internal capsula 33 )). We collected fieldmaps and undistorted the EPI images using the Fieldmap Toolbox (SPM). Motion parameters from realignment were evaluated, and a motion artefact threshold (translation > 3 mm, rotation > 1°) was employed for exclusion. Participant motion parameters were included as first-level covariates. No participants displayed gross movements to require total exclusion. Slices with motion parameters outside of the threshold were discarded. After denoising, quality control measurements (mean motion and max motion) were correlated and plotted with the functional connectivity values to control for influences (QC-FC correlations). To remove blood-oxygen-level-dependent (BOLD) signal from the cerebral white matter and ventricles, each participant's T1-weighted MPRAGE image was automatically segmented into grey matter, white matter, cerebrospinal fluid, normalized and transformed to MNI space using the Computational Anatomy Toolbox (CAT12; http:// www. neuro. uni-jena. de/ cat/) running in SPM12. BOLD data were bandpass filtered (0.008-0.09 Hz) to reduce low-frequency drift and noise effects. We then generated seed-to-seed connectivity maps for each individual using 164 seeds. These seeds are provided in the CONN software 35 . The exact classification of all seeds and the MNI coordinates of all network hubs are documented in supplementary material 1c. Individual correlation maps were generated. These results were subsequently used for second-level analysis of relative functional connectivity using an ANCOVA, implemented in the CONN toolbox, to investigate differences in seed-to-seed connectivity between groups. We applied a seed-to-seed analysis to investigate which brain areas show hyper-or hypoconnectivity between patients and controls and between subgroups. In addition to the exploratory whole-brain approach, hypothesis-driven group comparisons were also performed with seed regions that had been shown to be affected by FMS (namely the insular cortex 36 , the frontoparietal network 37 , the default mode network 10 and the somatosensory network 38 ). Pain intensity (GCPS) and ADS (depression) scores were included as second-level covariates. The influence of the IENFD data on the FC-values was analyzed using a linear regression model. False discovery rate (FDR) correction was applied at the cluster level (p < 0.05).
Statistical analysis. Data were analyzed with IBM SPSS Statistics for Windows, Version 25.0 (IBM Corp. Armonk, NY, USA) and JASP (JASP Team (2021) (Version 0.14.1, Windows 10). We tested the clinical data for normal distribution with a Shapiro-Wilk test and then, depending on the result, examined for group differences with a two-tailed t-test or a Mann-Whitney-U test. Data are given as mean ± SD or median/range unless otherwise specified. We used the Levene test with a significance threshold of 0.05 to check the data for equivalence of variance. The confidence interval was set at 95%. ROI group means of the structural, DTI and functional connectivity data were compared using an ANCOVA after controlling for interactions between the covariate and fixed factor and Tukey-tests for post-hoc comparisons. For the ANCOVA, effect sizes are displayed as ώ 2 , Data availability. The raw, skull stripped, data used to analyze the following results can be obtained upon request from the corresponding author. The processing and statistical analysis of the data was done using established neuroimaging software, as described in the methods. The STROBE Statement-Checklist was used for the quality control of our case-control study.
Clinical data and questionnaires. We included patients with normal skin innervation and patients with reduced IENFD both at the lower leg and the upper thigh from the cohort described in 4 . In patients with reduced www.nature.com/scientificreports/ distal and proximal IENFD (PNS group), FMS symptoms were more severe (p = 0.02) and quality of life was lower compared to FMS patients with normal distal and proximal IENFD (p = 0.01) as reflected by the values of the GCPS pain intensity and the FIQ questionnaire (Table 2). There was no difference between the subgroups regarding parameters evaluating how widespread the pain was (WPI or tender points).

Structural analysis.
With the values of the cortical volume per parcellation calculated by Freesurfer, we performed an ANCOVA with post-hoc testing including all patients (n = 43), PNS patients (n = 21) and noPNS patients (n = 22). Cortical volume differed between the FMS and control groups in 10 cortical regions (see Table 3). Cortical volume differed between the subgroups (PNS versus noPNS) in the left pericalcarine cortex (F = 4.1, p-adjusted = 0.049, ώ2 = 0.06) and the right pericalcarine cortex (F = 7.2, p-adjusted = 0.03, ώ2 = 0.13) (see Fig. 1). Except for the left pericalcarine cortex, all cortical regions of FMS patients showed lower volumes than those of healthy controls (see supplementary material 2). To examine possible influences of clinical data including the severity of pain and depression on cortex volume, correlation analyses between questionnaire data for pain and depression and cortical volumes were calculated. This was a-priori restricted to the 10 ROIs, which showed significant alterations in the FMS group compared to the control group. We found no significant influence of clinical data including the severity of pain and depression on cortex volume in the correlation analysis after FDR correction.

Diffusion tensor imaging.
In the ROI-based analysis comparing patients and controls, a significant increase in FA was found in 14 out of 48 ROIs in FMS patients (after FDR-correction). This was evident in corticospinal pathways such as the corona radiata, but also in regions of the limbic systems such as the fornix and cingulum.  www.nature.com/scientificreports/ A detailed list of these regions with the respective FA values can be found in Table 4. Scatter plots to check for the distribution of the data can be found in the supplementary material 3a/b. The ROI-based comparison of the two subgroups PNS and noPNS showed elevated FA levels in the left posterior limb of the internal capsule and the posterior thalamic radiation (after FDR-correction) (Fig. 2). The Pearson correlation analysis, a-priori restricted to the 14 regions that revealed differences in the group comparison, showed a negative association with the anxiety questionnaire (STAI-S) and the FA of the fornix (Pearson's r = −0.4, p = 0.006), the posterior thalamic radiation (Pearson's r = −0.4, p = 0.006) and the right posterior corona radiata (Pearson's r = −0.4, p = 0.005). This means that higher anxiety scores were associated with lowered FA in the respective areas.
Diffusion tensor imaging of FMS "specific" regions. White matter tracts that had already shown changes in patients with FMS in the literature are the corpus callosum, the thalamus, the cingulate, and the insular cortex connecting tracts (anterior limbs of the internal capsule). Our data indicated also an increased FA in the FMS group compared to controls in the cingulum (p < 0.001, η 2 = 0.13), in the body of the corpus callosum (p < 0.001, η 2 = 0.14), in the genu of the corpus callosum (p = 0.004, η 2 = 0.1), and in the posterior thalamic radiation (p = 0.002, η 2 = 0.12). No significant differences were found in the anterior limb of the left (p = 0.2, η 2 = 0.02) and right (p = 0.62, η 2 = 0.003) internal capsule and the splenium of the corpus callosum (p = 0.53, η 2 = 0.005).
Functional resting state imaging of FMS "specific" network hubs. Network hubs that had already shown changes in FMS patients in previous publications are the default mode network, the somatosensory network, the frontoparietal network, and the insular cortex. Even after restricting the analysis to these regions of interest, we could  www.nature.com/scientificreports/ not find any connectivity cluster differences between the FMS and the control group or between the PNS and noPNS subgroups in our data.

Discussion
In this study, two group comparisons were conducted using structural, DWI and functional MRI data: Firstly, we compared FMS patients to healthy controls, secondly, we divided the FMS group into two subgroups with and without PNS pathology (PNS and noPNS groups) and compared these subgroups with each other. While the structural and functional differences in MRI studies of FM patients have been described in the literature, so far no study has investigated the possible interaction between the peripheral nervous system and the brain of FMS patients. We show that in FMS (1) cortical volume is decreased in the left and right frontal/temporal cortices and the left insula, (2) FA is generally increased in corticospinal tracts and regions of the limbic system and (3) functional connectivity is reduced between the right midfrontal gyrus and the posterior cerebellum as well as the right crus cerebelli.
Comparison of the noPNS and PNS subgroups showed (1) lower volumes in the bilateral pericalcarine cortex in the PNS group, (2) lower FA in the left posterior limb of internal capsule and in the posterior thalamic radiation in the PNS group and (3) a hyperconnectivity cluster between the bilateral inferior frontal gyri, the angular gyrus and the posterior parietal cortex in the PNS group. In summary, the noPNS group showed greater deviations from healthy controls in structural MRI measures than the PNS group.
Comparison of the present findings with published data. Our results on the cortical volume are for the most part (regarding the alterations in the temporal, parietal and insular cortices) in line with the results of a meta-analysis which pooled structural and functional MRI studies comparing FMS patients to healthy controls 40 . These regions also appear to change their cortical thickness as the disease progresses 41 . Decreased gray matter in the left fusiform and prefrontal cortex was also found in FMS patients in another voxel morphometry-based meta-analysis 42 . In our hypothesis-driven analysis restricted to regions that showed lower cortex volumes in a meta-analysis of structural FMS data (left medial frontal cortex and right posterior cingulate cortex), we were able to reproduce the results of the meta-analysis 5 . However, in our subgroup comparisons, these regions showed no significant differences. The prefrontal cortex is a known site of pain modulation. Indeed, a dual role has been described including antinociceptive effects by modulating sensory afferent influx, as well as the furthering of chronic pain via corticostriatal projections. Interestingly, decline of prefrontal cortex volume in chronic pain can be reversed with successful biopsychosocial therapy, be it cognitive behavioral therapy, exercise or transcranial magnetic stimulation 43 .
Our subgroup comparison of cortex volume data showed a bilateral decrease in the volume of the pericalcarine cortex in the PNS group. Interestingly, in our results, the pericalcarine cortex is the only region that shows larger volumes in the FMS patients compared to the healthy controls. Thus, the noPNS group has a greater change in pericalcarine volume compared to the healthy controls. The pericalcarine cortex is part of the visual cortex. In our literature research, this region has not yet been associated with FMS symptoms. A magnetoencephalography study showed that the visual cortex in FMS patients has decreased connectivity to other brain regions 44 . This hypoconnectivity was also demonstrated in another study using resting state fMRI 45 and was associated with decreased resiliency towards pain 46 . However, the pericalcarine cortex is also involved in other pain disorders, for example, its volume changes during acute migraine attacks and normalizes in post-ictal phases 47 . Our results do not allow us to determine whether the pericalcarine cortices decrease in volume during the course of the disease in the PNS group or whether the difference exists at the onset of the disease. Longitudinal studies are needed to explore the role of the pericalcarine cortex in pain development.
Regarding FA, a marker for the integrity of the white matter, our whole brain analysis showed an increase in FA in the corona radiata and regions of the limbic system (e.g. fornix and cingulate cortex) in the FMS group compared to controls. The previous results of diffusion imaging in FMS patients are not consistent, and the results here vary widely. Regions that frequently showed changes in FA in the literature were the corpus callosum, the cingulum, the thalamus, and the anterior limb of the internal capsule adjacent to the insular cortex 7,33 . Except for the anterior limbs of the internal capsule, we were able to reproduce these results in our hypotheses driven analyses. Regarding our subgroups analyses, two regions showed a significant decrease of FA in the PNS group compared with the noPNS group (left posterior limb of internal capsule and the posterior thalamic radiation). Increased FA of these regions has already been found in studies with FMS patients or other chronic pain disorders and was associated with pain severity 48 . It has also been shown in FMS patients that white matter pathways with increased FA after a prolonged period of increased activity 49 , in this case in pain processing regions, show decreased FA again after pain chronification and show lower values than healthy controls 33 . Longitudinal study designs are needed to clarify the extent to which FA changes over the course of chronic pain disorders and the influences of a reduction or increase in FA on symptoms.
Regarding functional connectivity, even after limiting the regions of interest included in the analysis to network hubs already published in the FMS literature (default mode network, somatosensory network, frontoparietal network, insular cortex) 10,36,38,50 we could not reproduce alterations in these hubs with our data. The reason for this could be the lack of control for depression or pain intensity in other studies or different methods of analysis. The cluster found in our subgroup analysis has not been described in the FMS literature before. All involved regions (inferior frontal gyrus, angular gyrus and posterior parietal cortex) are involved in attention and evaluation of external and internal stimuli. Overactivation of the angular gyrus in fMRI has been associated with a stronger negative evaluation of pain 51 , while the inferior frontal gyrus seems to be involved in the regulation of www.nature.com/scientificreports/ emotions 52 . The posterior parietal gyrus with its connections to the somatosensory cortex appears to have an important role in the spatial perception of pain stimuli 53 .
Are the findings specific for FMS? Most of our findings have been described in other publications about chronic pain imaging 54 . For example, it has already been suggested that a lower activity of the prefrontal cortex, a well-known pain modulation area, could lead to a failure in the elimination of subcortically driven fear behaviors, thereby resulting in pain chronification 55 . It is currently unclear whether these processes are adaptive, maladaptive or cause some of the symptoms. In order to better understand the pathophysiology of FMS, it is therefore important to first understand the role of brain neuroplasticity in chronic pain, as a brain signature of pain appears to be found across various pain syndromes 56 . Neuroimaging studies with multiple pain syndromes as comparison groups are needed here before finding brain regions specific to FMS that could potentially trigger some of the symptomatology.
Limitations of our study. Our study has some limitations. Because our study was designed as a cross-sectional study, the question of the reasons for and the effects of our detected group differences cannot be answered. By including individual pain intensity as a covariate in our group statistics, we attempted to account for a possible influence of pain intensity on our MRI results. However, because none of the MRI modalities showed a significant association with IENFD scores after FDR correction, we cannot rule out the possibility that subgroup differences were driven by other factors not captured in our clinical examinations. Furthermore, even structural MRI markers, such as cortical volume, are subject to temporal variations, depending, for example, on acute stimulus severity 57 . This emphasizes the need for longitudinal studies.
The healthy controls in our study did not receive a skin biopsy, so we cannot rule out that some persons with reduced IENFD might have been in this group. However, in our previous study 4 , only 2% of normal controls had reduced IENFD at the lower and upper leg, so that it is highly unlikely that a large number of our present controls would have had this finding.

Conclusions
While structural and functional MRI changes in FMS patients have already been investigated, our study first demonstrated differences between FMS subgroups with and without peripheral nerve involvement. The study design obviously does not allow any conclusions to be drawn about the reasons for and effects of these subgroup differences. While most clinical trials on FMS therapy included only patients diagnosed according to current diagnostic criteria, one has to consider that FMS is a heterogeneous condition with potentially different underlying pathophysiological processes within subgroups. These subgroups might respond differentially to specific treatments. Psychiatric comorbidities, such as depression and anxiety, also affect the brain structure in FMS and thus influence the results in MRI imaging. We therefore advocate that future studies should take into account the different subgroups of patients both on the basis of small nerve fiber pathology, symptom severity, and psychiatric comorbidities.