Default mode network changes in fibromyalgia patients are largely dependent on current clinical pain.

Two independent cohorts of participants were assessed. The ‘scanner pain-free cohort’ included 27 FM patients (25 females, 2 males; mean age ±SD: 42.3 ± 13.1 years) who had no current clinical pain at the time of scan and 27 individually age-matched (±3 years) healthy controls (25 females, 2 males; 41.8 ± 12.3 years, p = 0.84). The ‘scanner pain cohort’ included 16 FM patients (all females; mean age ± SD: 47.8 ± 7.3 years) who experienced some level of clinical pain during the scan and 16 individually age-matched (±3 years) healthy controls (all females; 48 ± 6.7 years, p = 0.92). Groups were also matched on level of physical activity (International Physical Activity Questionnaire, IPAQ ( Craig et al., 2003 ). For both cohorts, patients were included if they had experienced chronic widespread pain for at least 1 year with an average daily intensity of at least 4 out of 10. The diagnosis of FM and exclusion of other (pain) disorders was confirmed by medical records or directly by the treating physician (scanner pain-free cohort), or by the collaborating rheumatologist (scanner pain cohort). Exclusion criteria for all subjects in either cohort included use of recreational drugs, use of opioid medication, pregnancy or breastfeeding, pain conditions (other than FM for patients), major medical, neurological, or current psychiatric conditions, including severe depression and generalized anxiety disorder, and MRI contraindications, alcohol consumption of >7 drinks per week, and smoking >10 cigarettes per week (scanner pain-free cohort)/smoking (scanner pain cohort). An additional exclusion criterion for the scanner pain-free cohort for this paper was the presence of any current pain at the time of scan (this led to the exclusion of 5 out of 32 initially enrolled FM patients in this cohort). Healthy controls were further excluded if they had taken any pain medication other than NSAIDs within the last month or for more than one month on a continual basis within the last 6 months. Patients in either cohort were on stable preventive or as needed medication ( Table 3 ). All procedures were approved by the Institutional Review Board (IRB) of NIH (scanner pain-free cohort) or McGill University (scanner pain cohort). Written informed consent was obtained from all subjects in both cohorts according to the Declaration of Helsinki. For each cohort, the investigation presented here was part of a larger study comprising two visits to the lab (scanner pain-free cohort: manuscript submitted, ( Case and Ceko, 2016 ); scanner pain cohort ( Ceko et al., 2013 ; Čeko et al., 2015). For each cohort, questionnaire data were collected during the first visit and MRI data was collected during the second visit, with visits less than a week apart. In this manuscript we report only measures that were available for each cohort allowing for direct comparisons between cohorts. For all subjects, duration of FM symptoms (in years) was assessed through chart reviews and patient history. Depressive and anxiety symptoms were assessed using the Hospital Anxiety and Depression Scale (HADS ( Zigmond and Snaith, 1983 )) during the first visit. The intensity of current clinical pain experienced during the resting state scan was assessed immediately after the scan (scanner pain-free cohort) or immediately after the scanning session (8 min after the resting state scan, scanner pain cohort) on a 0–10 numerical rating scale (0–10 on a numerical rating scale; 0 = no pain, 1 = pain threshold, 10 = worst bearable pain). In both cohorts, subjects were asked to report the average pain they experienced during the resting state scan. All subjects completed an anatomical MRI scan, and a resting-state fMRI scan with eyes open, instructed to fixate on a black cross-hair on white background (6 min for scanner pain-free cohort; 8 min for scanner pain cohort but only the first 6 min used for analysis to match the other cohort). Throughout the session, subjects wore earplugs and their heads were immobilized. Brain images were acquired using a 3 T S TIM Trio MRI scanner (Siemens, Erlangen, Germany) with12-channel head coil (scanner pain cohort) or a 3 T S Skyra (Siemens, Erlangen, Germany) with a 20-channel head and neck coil (scanner pain-free cohort). Structural (T1-weighted) MRI data were acquired using a 3D MP-RAGE sequence (scanner pain cohort: TR = 1900 ms, TE = 2.07 ms, flip angle = 9°, resolution 1 × 1 × 1 mm, image matrix = 256 × 256; scanner pain-free cohort: TR = 2300 ms, TE = 2.98 ms, flip angle = 9°, resolution 1 × 1 × 1 mm, image matrix = 256 × 256). FMRI data were acquired using a blood oxygenation level-dependent (BOLD) protocol with a T2*-weighted gradient echo planar imaging (EPI) sequence (scanner pain-free cohort: TR = 2000 ms, TE = 29 ms, 180 vol, flip angle = 70°, resolution 3.5 × 3.5 × 3.5 mm, image matrix = 64 × 64, 38 slices; scanner pain cohort: TR = 2260 ms, TE = 30 ms, 160 vol, flip angle = 90°, resolution 3.5 × 3.5 × 3.5 mm). Axial slices were oriented 30° from the line between the anterior and posterior commissures, covering the entire brain, and excluding the eyes. During the fMRI scans, heart rate, blood oxygenation and respiration were monitored in both cohorts. Data are expressed as means ± SD unless otherwise noted. Outcome measures were compared between groups in SPSS 21 (IBM) using independent samples two-tailed t-tests, and a significance level of p < 0.05 was used in all analyses. Correlations between behavioral measures and fMRI data were investigated using linear regression analyses. FMRI data were preprocessed in FSL https://fsl.fmrib.ox.ac.uk/fsl/ . Briefly, preprocessing included removal of non-brain tissue, six-parameter (3 translations and 3 rotations) rigid body correction for head motion, high-pass filtering at 0.01Hz, co-registration to the individual’s T1-weighted anatomical image and spatial normalization to MNI space using a 12-parameter non-linear registration, and smoothing with a 5 mm Gaussian kernel. Additional steps included masking of non-brain voxels, voxel-wise de-meaning of the data, normalization of the voxel-wise variance, and pre-whitening to account for auto-correlations in the data. Single-session probabilistic independent component analysis (ICA) of the resting state fMRI data was conducted with MELODIC (Multivariate Exploratory Linear Decomposition into Independent Components) in FSL (version 3.14) and set to automatic dimensionality estimation. FMRIB’s ICA-based Xnoisefier (FIX 1.06 ( Griffanti et al., 2014 ); was used to automatically detect and regress out artifacts from each single-session resting-state ICA. The “Standard.RData” training dataset included in the FIX toolbox was used to classify (spatially and temporally) the components that corresponded to artifacts including motion, white matter and cardiac signals, or MRI acquisition-related issues (classification threshold = 20). Artefactual components were found to be correctly classified upon inspection of randomly selected FIX results. The denoised resting-state data were then used for the subsequent seed-based functional connectivity analyses and ICA dual regression analysis described below. We tested for group differences in the overall head motion because such differences could affect the connectivity group comparisons. We compared head motion, calculated as the square root of the sum of squares of the six motion parameters (x, y, z translations and rotations, in mm) ( Van Dijk et al., 2012 ; Power et al., 2012 ) between groups for each cohort. In the scanner pain-free cohort, head motion differed significantly between patients (mean ± SD 0.299 mm ± 0.215) and controls (0.202 mm ± 0.106, p = 0.043, unpaired t -test) because of two patients with outlier values (>2 SD above group mean). After excluding these two patients, head motion no longer significantly differed between patients (0.254 mm ± 0.149) and controls (p = 0.158). Excluding these two patients from group analyses (seed-based connectivity, ICA) did not affect the results, therefore we are presenting the results of the full cohort. In the scanner pain cohort, head motion did not differ significantly between patients (0.188 mm ± 0.097) and controls (0.146 mm ± 0.082; p = 0.196, unpaired t -test). The ICA dual regression analyses were conducted separately for each cohort. The group-average ICA included additional steps to prepare the denoised resting state data for dual regression, i.e., masking of non-brain voxels, voxel-wise de-meaning of the data, normalization of the voxel-wise variance, and pre-whitening to account for auto-correlations in the data. The resting-state data were then temporally concatenated to create one 4-dimensional dataset. Probabilistic ICA was applied to the 4D dataset with a restriction of 25 components (IC map threshold = 0.66). The group spatial map of the DMN network was visually determined in each cohort by identifying the characteristic regions (posterior cingulate cortex (PCC), medial prefrontal cortex (MPFC), bilateral inferior parietal lobule (IPL)). Next, FSL’s dual regression ( Nickerson et al., 2017 ) was applied, using the DMN map as a spatial regressor to extract the associated fMRI time series of each subject. The extracted fMRI time-series of each subject was used as a temporal regressor to back-construct each subject’s spatial map matching the DMN. Significant differences between the groups were determined with FSL’s Randomise (permutation tests) using voxel-based z > 2.3, p < 0.05 cluster-correction for spatial extent (to match the threshold commonly used in resting state fMRI studies in chronic pain and decrease the probability of a Type II error, given the hypothesis of no group differences in patients with no pain during scan). The INS ROI was included because changes in resting-state DMN-INS connectivity are a prominent finding in chronic pain ( Table 2 ), and have also been reported in FM patients ( Napadow et al., 2010 ; Ichesco et al., 2014 ). DLPFC was included as it was the region that had less gray matter in patients vs. controls across the two cohorts (scanner pain cohort ( Ceko et al., 2013 ), scanner pain-free cohort, manuscript in preparation), is implicated in pain regulation, and both DLPFC and the fronto-parietal network which contains the DLPFC have shown altered connectivity with DMN in chronic pain ( Table 2 ), including FM ( Fallon et al., 2016 ; Kong et al., 2019 ; Napadow et al., 2010 ; Flodin et al., 2016 ). S2 was included because it is an important region implicated in pain processing and altered DMN-S2 connectivity have been reported across chronic pain ( Table 2 ), including FM ( Napadow et al., 2010 ). Bilateral INS (23,016 mm 3 , 125 resolution elements (RESELs)), S2 (48,712 mm 3, 390 RESELs) and DLPFC (middle frontal gyrus; 99,680 mm 3 , 797 RESELs))masks were derived from the Harvard-Oxford cortical structure atlas included in FSL. Each ROI was sufficiently large to allow for cluster-level inference (>25 RESELs), based on guidelines in the Handbook of fMRI Data Analysis. (( Poldrack et al., 2011 ), page 126). Directed-search analyses were performed using FLAME with a voxel-based threshold of z > 2.3 and cluster-corrected for spatial extent at p < 0.05 across each ROI. This analysis was performed because structural differences in GM could potentially influence resting-state functional connectivity ( Guo et al., 2013 ; Farb et al., 2013 ; Baliki et al., 2014 ). We first performed a voxel-based morphometry (VBM) analysis on the scanner pain cohort to test if the sample used in the current study showed similar GM differences as observed in the larger original sample in Ceko et al., 2013 ). VBM analysis was performed using the CAT12 toolbox in SPM12 ( http://www.neuro.uni-jena.de/cat/ ) with a voxel-based threshold of z > 2.3 and cluster-corrected for spatial extent at p < 0.05, as well as at p < 0.001 uncorrected to detect any subtle GM differences that could potentially influence resting-state connectivity. As in the original study ( Ceko et al., 2013 ), the VBM group comparison was controlled for age. For any observed GM differences located in one of the above ROIs, we then extracted the GM values and used them as a covariate of no interest in the above group analyses of functional connectivity in DMN-ROI. To compare more directly the DMN connectivity between the two cohorts, we extracted the connectivity values from significant DMN-ROI clusters, calculated Patients-Controls difference scores for each cohort and performed a two-sample t-test on the difference scores ((Patients-Controls) scanner pain cohort – (Patients-Controls) scanner pain-free cohort ). In addition to the ICA approach, the other most common approach to testing group differences in functional connectivity analysis in studies of chronic pain has been seed-based analysis, which is why we here also report results from seed-based analyses of DMN connectivity. Functional connectivity analysis was conducted with FSL (version 3.14). Seed ROIs were created from the cohort-specific DMN spatial maps derived from the ICA analysis. A set of three seed regions commonly used in DMN studies in chronic pain were derived from the DMN spatial map of each cohort: the PCC, the MPFC and the two regions combined (‘DMN’). The seeds were converted to native space of each subject to extract the mean time-series of each region for each subject. Using the time-series of each seed ROI, functional connectivity was calculated, at the subject level, as the correlation of time-series with all other voxels in the brain, resulting in connectivity maps showing Z-scores for each voxel indicating how well they are correlated with the seed. Whole-brain group level comparisons were carried out using FMRIB’s Local Analysis of Mixed Effects (FLAME) with a voxel-based threshold of z > 2.3 and cluster-corrected for spatial extent at p < 0.05 across the whole brain, as well as in directed search analyses in INS, DLPFC and S2. To assess the influence of current pain in the scanner on DMN connectivity, voxel-wise regression analysis with current pain as independent variable was performed in patients in ROIs showing group differences in DMN connectivity. As above, the results were considered significant at z > 2.3, p < 0.05 cluster-corrected. Separate voxel-wise regression analyses in ROIs where we observed significant group differences were performed to examine in patients of either cohort the effect of the following independent variables on DMN connectivity: (1) chronic pain duration, (2) depression scores (HADS), (3) anxiety scores (HADS). As above, the results were considered significant at z > 2.3, p < 0.05 cluster-corrected.

We confirmed that only patients in the scanner pain cohort experienced current pain during the scan (mean ± SD 4.4 ± 2.1; scanner pain-free cohort mean 0). Duration of FM symptoms was comparable across the two patient cohorts (scanner pain-free cohort 10.66 ± 7.72 years, scanner pain cohort 11.5 ± 10.04 years; comparison between cohorts p = 0.816), and so were participants’ age (p = 0.194) and level of physical activity (p = 0.154). Patients in both cohorts had similar levels of anxiety (p = 0.371) and depressive symptoms (p = 0.589). These levels were in the sub-clinical range but significantly elevated compared to the respective control groups (see Table 3 for details). We used ICA to identify and compare DMN network connectivity between groups of each cohort during the resting state fMRI scan. In FM patients who were pain-free at the time of scan, we observed no differences in DMN connectivity compared to controls in corrected whole brain or directed-search analyses in the INS, DLPFC and S2 (lenient corrected threshold: z > 2.3, p < 0.05 cluster-corrected). Because altered resting-state connectivity between DMN and INS seems to be the most reproducible DMN finding in chronic pain patients (see Table 2 ), and has been reported for FM patients ( Napadow et al., 2010 ; Ichesco et al., 2014 ) and correlated with patients’ current clinical pain ( Napadow et al., 2010 ), we also examined the INS ROI using a lenient uncorrected significance threshold (voxel-based threshold z > 3.1 uncorrected for spatial extent) to detect any subtle group differences. In both patients and controls of the scanner pain-free cohort, we observed DMN connectivity to the bilateral INS, with a small cluster in the aINS ( Fig. 1 , inset) showing stronger DMN connectivity in patients vs. controls (10 voxels, z > 3.1 uncorrected). FM patients who had current pain during the scan showed increased DMN-right aINS, DMN-left aINS and DMN-right DLPFC connectivity compared to their matched controls in corrected directed search analyses ( Fig. 1A , Table 4 ). No group differences were observed in corrected whole-brain analysis. The DMN-DLPFC connectivity group difference was no longer significant when controlling for GM volume in the DLPFC ( Table 4 ). The correction for GM volume was performed because patients showed a subtle decrease of GM volume in the DLPFC (p < 0.001 uncorrected in location overlapping with connectivity results). Supplementary analyses in each cohort using seed-based functional connectivity were performed to match previous studies in FM and other chronic pain populations that have used this approach instead of the ICA. For these analyses the PCC and MPFC were combined as seed (PCC + MPFC, additional analyses with PCC and MPFC as separate seeds gave similar results and are therefore not mentioned further). In the scanner pain-free cohort, these analyses showed no group differences in aforementioned regions implicated in DMN changes in chronic pain, in whole brain or directed search analyses. Outside of the regions implicated in DMN changes in chronic pain, we observed one significant cluster in the contrast controls > patients in the left putamen (peak MNI coordinates, –20,16,4; cluster size 1365 voxels, peak z = 4.06, p < 0.001). In the scanner pain cohort we observed increased DMN-vermis cerebellum connectivity in patients vs. controls (peak MNI coordinates, 35,–58, –30; cluster size 1192 voxels, peak z = 4.30, p < 0.001). No other regions showed group differences in seed-based ROI analyses in either cohort. To provide a more direct comparison of the two cohorts we performed an interaction analysis of DMN-INS connectivity values extracted from significant aINS clusters. In the scanner pain-free cohort, the DMN-INS connectivity values did not differ significantly between patients and controls (p 0.435, Fig. 1B ). The group difference in DMN-INS connectivity of the scanner pain cohort was significantly larger than the group difference in DMN-INS connectivity in the scanner pain-free cohort (two-sample t -test on group difference scores, p < 0.001, Fig. 1B ). Next, we investigated in patients with current clinical pain whether DMN connectivity was related to the level of their current pain by performing regression analyses in patients in brain areas where we observed group differences (bilateral INS ROI). A significant positive relationship was observed between DMN-left midINS connectivity and the level of current pain ( Fig. 1B ), and the cluster was located adjacent to the group difference cluster ( Fig. 1C ). In addition, when controlling for current pain in the group comparison, the insula cluster was no longer significant. Together, these results indicate that at least in part group differences observed in this cohort are related to the level of clinical pain that patients experienced during the fMRI scan. No further significant regression results between DMN connectivity and current pain were observed in the S2 ROI, DLPFC ROI or in whole-brain regression analyses (voxel-based z > 2.3, p < 0.5 cluster-corrected). In seed-based whole-brain regression analysis with current clinical pain as independent variable, patients showed a negative correlation between DMN-left DLPFC connectivity and the level of their ongoing pain (peak MNI coordinates, –44, 24, 32; cluster size 408 voxels, peak z = 4.81, p = 0.006). To test if DMN connectivity was associated with disease burden (chronic pain duration (yrs), anxiety scores (HADS), depression scores (HADS)), we performed separate voxel-wise regression analyses of ICA-derived DMN connectivity with these variables in patients of each cohort. In patients with current pain during the scan, regression analyses were performed in the area showing significant group differences (bilateral INS ROI) and showed no significant clusters for any of the three variables tested (voxel-based z > 2.3, p < 0.05 cluster-corrected across the INS ROI). Of the three disease burden variables tested, chronic pain duration and anxiety scores were not correlated significantly with level of current pain (r = –0.149, p = 0.584 for pain duration, r = 0.089, p = 0.761 for anxiety scores), but there was a significant positive correlation between depression scores and current pain (r = 0.560, p = 0.037). Since patients who were pain-free at the time of scan showed a subtle increase in DMN connectivity to the dorsal aINS (>3.1 uncorrected in the INS ROI), we also performed regression analyses with chronic pain duration, anxiety and depression scores in these patients. As in the scanner-pain sample, DMN connectivity of patients who were pain-free at the time of scan was not related to anxiety or depression scores (voxel-based z > 2.3, p < 0.05 cluster-corrected across bilateral INS ROI). However, in patients pain-free at the time of scan we observed a significant positive relationship between the DMN-left mid/pINS connectivity and chronic pain duration (voxel-based z > 2.3, p < 0.05 cluster-corrected across bilateral INS ROI; Fig. 2 , Table 3 ), indicating that DMN connectivity to the INS is at least in part associated with disease duration in this cohort.

In sum, in the present study altered connectivity of the DMN in FM patients was related to processing of currently experienced pain, more so than to the diagnosis of chronic pain. Thus, DMN alterations observed in chronic pain patients who are experiencing pain during scanning might sometimes more closely reflect transient disruption of intrinsic brain networks due to the current experience of pain (state) than more permanent functional consequences of having the chronic pain condition (diagnosis/trait). Although these processes likely influence each other, individual and specific contributions of pain state vs. trait need to be considered when studying functional brain connectivity in chronic pain patients.

Here we investigated the effects of chronic pain condition (trait) vs. currently experienced clinical pain (state) on resting-state functional connectivity of the DMN in FM patients. To this end, we compared DMN connectivity between FM patients and their matched controls in two separate cohorts: (1) a scanner pain-free cohort (FM patients had no clinical pain at the time of scan), and (2) a scanner pain cohort (FM patients had some level of clinical pain during the scan). In areas of differential findings between cohorts we also directly tested for interaction effects across cohorts. Scanner pain-free FM patients largely showed no differences in resting-state DMN connectivity compared to matched controls. In contrast, FM patients experiencing current clinical pain during the scan showed significant cluster-corrected DMN differences compared to matched controls in the INS. Importantly, the interaction analysis of DMN-INS connectivity values compared the two cohorts directly and confirmed that only in the scanner pain cohort, but not in the scanner pain free cohort, there was a significant group difference in DMN-INS connectivity. In FM patients with current pain during the scan, we observed a significant correlation between DMN-INS connectivity and level of their current clinical pain, in line with previous reports on FM patients experiencing pain at the time of scan. Thus, in this study, the diagnosis of chronic pain (trait) did not significantly affect patients’ DMN connectivity as compared to controls, but current clinical pain (state) did. Altered DMN connectivity has repeatedly been shown in FM and other chronic pain patients. The literature is mixed, but two types of findings are frequently observed. First, patients show altered functional connectivity compared to healthy controls within the DMN (i.e. between the core regions PCC and MPFC), whereby both connectivity increases ( Baliki et al., 2014 ; Cauda et al., 2009 ; Čeko et al., 2015; Flodin et al., 2016 ; Kucyi et al., 2014 ; Loggia et al., 2013 ; Ichesco et al., 2014 ; Kim et al., 2019 ) and decreases ( Alshelh et al., 2018 ; Baliki et al., 2014 ; Ceko et al., 2013 ; Li et al., 2014 ; Martucci et al., 2015 ; Tian et al., 2016 ; Ichesco et al., 2014 ) within the DMN have been reported. Second, patients show increased functional connectivity of DMN regions with the rest of the brain, specifically with the INS and several other regions associated with processing and regulation of pain, including the DLPFC ( Cauda et al., 2009 ; Baliki et al. 2011 , 2014 ; Flodin et al., 2016 ; Fallon et al., 2016 ; Kong et al., 2019 ; Tagliazucchi et al., 2010 ; Wu et al., 2016 ; Napadow et al., 2010 ; Martucci et al., 2015 ; Ichesco et al., 2014 ; As-Sanie et al., 2016 ; Li et al., 2014 ; Čeko et al., 2015; Kucyi et al., 2014 ; Loggia et al., 2013 ). Common to many of these studies is that patients have pain at the time of scanning. In fact, several studies report significant correlations between the observed DMN connectivity changes and patients’ current clinical pain, most consistently changes with the INS ( Baliki et al. 2011 , 2014 ; Loggia et al., 2013 ; Martucci et al., 2015 ; Napadow et al., 2010 ). Moreover, several studies have reported partial reversals of these DMN changes with reductions in clinical pain (i.e. after treatment) ( Li et al., 2014 ; Napadow et al., 2012 ) or augmentation of DMN changes with exacerbation of clinical pain ( Loggia et al., 2013 ; Kim et al., 2019 ; Letzen and Robinson, 2017 ). These findings have variably been interpreted as tracking the subjective momentary experience of clinical pain (current pain experience state) and as an indication of the impact of chronic pain (general chronic pain condition trait) on functional connectivity of the DMN. In the current study, we also observed DMN connectivity differences to bilateral aINS and right DLPFC, but mainly in patients with current clinical pain at the time of scan vs. their matched controls, and not in patients who were pain-free at the time of scan. The group difference in DMN-right DLPFC connectivity was no longer significant when we controlled for GM volume in the right DLPFC where patients had less GM volume than controls. Controlling for GM volume is not standard practice in resting-state studies in chronic pain (with the exception of ( Baliki et al., 2014 )), but any existing GM volume group differences in areas under investigation could influence resting-state connectivity group results as they did in our current study. In patients with current clinical pain we observed a significant relationship between the level of their current clinical pain and connectivity between the DMN and mid portion of the left INS, a region prominently implicated in pain processing and regulation ( Bushnell et al., 2013 ; Villemure and Bushnell 2009 ; Ceko et al., 2013 ; Segerdahl et al., 2015 ; Reddan and Wager, 2018 ; Wager et al., 2013 ; Fazeli and Büchel, 2018 ; Geuter et al., 2017 ). Additionally, using seed-based analysis, we observed a negative relationship between DMN-left DLPFC connectivity and current clinical pain, possibly indicating a (successful) mounting of endogenous top-down regulation of the currently experienced pain. We interpret the differential findings in the two cohorts investigated here as indicating that current clinical pain experienced during the scan (i.e. pain state) might influence the connectivity of the DMN separately from the influence of the underlying chronic pain diagnosis (i.e. pain trait). Why should current clinical pain at the time of scan (pain state) in fluence the DMN? Pain engages cognitive mechanisms, and in healthy volunteers both experimental pain and cognitive tasks have been associated with decreased activation (‘deactivation’) of the DMN ( Villemure and Bushnell 2009 ; Alshelh et al., 2018 ; Anticevic et al., 2012 ; Coghill et al., 1999 ; Kong et al., 2010 ; Kucyi et al., 2013 ; Legrain et al., 2009 ; Loggia et al., 2012 ; Seminowicz and Davis, 2007a ; Seminowicz et al., 2004 ). In addition to the effect of pain on DMN activation, two recent studies using experimentally-induced pain in healthy volunteers have also shown effects on DMN connectivity. In these studies, a prolonged pain stimulus (5–20 min of intra-muscular hypertonic saline infusion) was associated with decreased connectivity of the PCC with MPFC and IPL ( Alshelh et al., 2018 ), decreased oscillatory power in the PCC, precuneus, MPFC, and IPL, ( Alshelh et al., 2016 , 2018 ) and increased regional homogeneity (a measure of local functional connectivity) of the MPFC (Zhang et al., 2014). In addition, the DMN changes observed with experimental pain in healthy volunteers were similar to DMN changes observed in a group of chronic pain patients in the same study ( Alshelh et al., 2018 ), prompting the authors to conclude that ‘this clearly indicates that DMN functional changes are not a unique characteristic of chronic pain but instead likely represent the presence of pain itself’. Another study compared DMN connectivity before and after experimental pressure pain induction between FM patients and controls and showed increased PCC/Precuneus connectivity to the thalamus in the patient group. This DMN-thalamus connectivity in crease was significantly correlated with the increase in patients’ reported pain due to pressure stimulation ( Ichesco et al., 2016 ). Additionally, pain-related cognitive regulation training has been shown to increase functional connectivity of the PCC/precuneus and MPFC with the ventrolateral PFC, indicating dynamic changes in DMN connectivity because of learned cognitive control over acute painful stimulation ( Kucyi et al., 2016 ). Together, these findings show that processing and regulation of the current pain experience can significantly alter the activation and connectivity of the DMN. These effects on the DMN are similar to the abovementioned DMN alterations in chronic pain patients experiencing spontaneous ( Baliki et al. 2011 , 2014 ; Loggia et al., 2013 ; Martucci et al., 2015 ; Napadow et al., 2010 ) or induced ( Loggia et al., 2013 ; Kim et al., 2019 ) current clinical pain at the time of scan. Given that pain experience (provoked or spontaneous clinical) can influence the connectivity of the DMN, considering other common measures of disease burden (i.e. chronic pain duration and impact, average pain over the past week/month, hyperalgesia, pain coping) can be especially relevant in establishing the relative contribution of chronic pain condition (trait) vs. current experience of (clinical) pain (state) to observed DMN connectivity differences. In the present study, measures of anxiety and depressive symptoms, and pain duration were available for both cohorts and thus the influence of those on DMN connectivity was investigated in each cohort. Anxiety and depression were not significantly correlated with DMN connectivity in either cohort. In the cohort experiencing clinical pain during the scan, DMN connectivity was not related to their chronic pain duration. In the cohort without clinical pain during the scan, where we observed only a small cluster of increased DMN-INS connectivity (uncorrected ROI analysis, z > 3.1, 10 voxels in right dorsal aINS), we found a significant positive relationship between DMN connectivity in the left mid/pINS and chronic pain duration (z > 2.3, p < 0.05 cluster-corrected), indicating that pain chronification (trait) may also have an effect on DMN connectivity. Such an effect may be more evident when variance related to current state-related pain is absent. Having additional measures of disease burden in our study might have provided more clues on the relationship between DMN connectivity and FM in the patients of both cohorts. For example, other studies, regardless of whether patients were in pain during the scan or not, have shown significant correlations between DMN changes and chronic pain duration ( Fallon et al., 2016 ), altered mood ( Fallon et al., 2016 ), hyperalgesia ( Fallon et al., 2016 ; Flodin et al., 2016 ; Ichesco et al., 2014 ), which can be indicative of central sensitization processes ( Woolf, 2011 ), as well as maladaptive coping with pain ( Coulombe et al., 2017 ) in FM patients, and other chronic pain disorders ( Kucyi et al., 2014 ; Baliki et al., 2014 ; Martucci et al., 2015 ). However, often DMN changes observed in pain patients are not correlated with collected measures of disease burden in a particular study, in FM ( Napadow et al., 2010 ; Ceko et al., 2013 ) or other chronic pain disorders ( Alshelh et al., 2018 ; Baliki et al., 2011 ; Flodin et al., 2016 ; Otti et al., 2013a ; Tian et al., 2016 ; Huang et al., 2016 ) This underlines the need to more comprehensively capture (or exclude) the influence of indices of pain chronification in addition to the likely more dynamic measures of the current state-related clinical pain experience. Two recent studies reported differential contributions of pain trait (average pain over the last month ( Cheng et al. (2018) ), pain during menses ( Wu et al., 2016 )) vs. pain state (current level of clinical pain at the time of scan) to changes in DMN and several other brain networks relevant for pain and cognitive processing in patients with neuropathic pain ( Cheng et al., 2018 ) and primary dysmenorrhea ( Wu et al., 2016 ). Therefore, both trait and state pain features are important to consider in resting state studies of chronic pain, and might have differential contributions to functional connectivity of the DMN and other networks (comprehensively reviewed in ( Davis and Cheng., 2019 )). Multivariate network approaches in larger samples, as applied in Cheng et al., 2018 , or investigation of chronic pain disorders with clear delineation of painful vs. not painful episodes ( Wu et al., 2016 ), including within-subject designs where chronic pain patients are scanned once during a painful episode and once during a non-painful episode might be particularly well suited to capture the relative contribution of state vs. trait pain on brain connectivity. Another worthy approach is to have chronic pain patients provide continuous ratings of their spontaneous pain fluctuations and apply dynamic measures of connectivity to test if the DMN-INS connectivity tracks pain over time ( Baliki et al., 2011 ; Kucyi and Davis, 2015 ). Given that pain evaluation is a cognitive task that can influence the DMN in and of itself, a particular challenge in these studies is to disentangle the influence of current clinical pain vs. that of continuous cognitive evaluation of pain. Additionally, future work is needed to determine to what extent increased attention to pain common in chronic pain patients may effect stable trait-like changes in DMN connectivity ( Kucyi and Davis, 2015 ). Finally, the inconsistency in findings of significant relationships between DMN connectivity and disease (trait) burden across the literature, including our own study, calls for a more systematic assessment of a broader set of measures of chronic pain impact.

A clear limitation of the current study is that we tested our hypotheses using two cohorts of participants tested in different fMRI centers. The magnet strength and scanning sequences were similar, however there likely were differences in signal-to-noise ratio and other scanner- and head-coil specific parameters between the two Siemens 3T models (Trio and Skyra), which is a general concern in multi-center studies ( Suckling et al., 2012 ; Teipel et al., 2017 ). To circumvent this, we adopted two strategies. First, we analysed each patient group as compared to their own matched control group. Second, we performed a direct comparison of the cohorts by testing the difference between difference scores (interaction) in the DMN-INS connectivity. Another important consideration in our and other studies testing brain differences between two groups of participants is that any existing structural differences might potentially influence the analysis and interpretation of functional brain differences. Here, our DMN-DLPFC connectivity analysis was motivated by known GM alterations in the DLPFC of chronic pain patients observed in our and other studies. However, these structural GM decreases (or increases) could influence functional connectivity analyses via partial volume effects or other factors ( Dukart and Bertolino, 2014 ). This is likely more pertinent in studies investigating cohorts with more pronounced brain changes (such as dementia or aphasia ( Guo et al., 2013 ; Farb et al., 2013 ). In such studies controlling for the amount of GM seems to be a more standard practice, unlike in resting-state studies of chronic pain (cf.( Baliki et al., 2014 ). But some type of control might also be warranted in studies of chronic pain and similar disorders that report more subtle functional and structural brain changes. In the current study, for example, controlling for GM volume (which showed significant group difference spatially overlapping with our functional connectivity findings) abolished the significant group difference in DMN-DLPFC connectivity, which changes the interpretation of those results. Therefore, caution might be warranted when interpreting any group differences in functional brain connectivity against the backdrop of structural brain differences or vice versa.

Chronic pain has been associated with functional alterations in the default mode network (DMN) ( Table 1 for fibromyalgia (FM) patients and Table 2 for other chronic pain conditions). Differences in resting-state functional connectivity between patients and controls are typically observed within the DMN ( Alshelh et al., 2018 ; Baliki et al., 2011 ; Baliki et al., 2014 ; Cauda et al., 2009 ; Ceko et al., 2013 ; Čeko et al., 2015; Flodin et al., 2016 ; Huang et al., 2016 ; Kim et al., 2019 ; Kucyi et al., 2014 ; Li et al., 2014 ; Loggia et al., 2013 ; Tian et al., 2016 ) and between DMN and several other areas implicated in attentional and somatosensory processing and regulation, primarily the insula (INS), fronto-parietal networks, including the dorsolateral prefrontal cortex (DLFPC), and somatosensory cortices including the secondary somatosensory cortex (S2) ( Baliki et al., 2011 , 2014 ; Cauda et al., 2009 ; Čeko et al., 2015; Flodin et al., 2016 ; Loggia et al., 2013 ; Martucci et al., 2015 ; Napadow et al., 2010 ). Several studies have reported significant correlations between altered DMN connectivity and patients’ current clinical pain, i.e. the pain that patients experience at the time of the fMRI scan, sometimes referred to as ‘ongoing pain’ ( Baliki et al., 2011 , 2014 ; Loggia et al., 2013 ; Martucci et al., 2015 ; Napadow et al., 2010 ). Across these studies, current clinical pain has been interpreted as a proxy for disease burden, and the alterations in DMN connectivity (e.g. with the INS) as indicating brain reorganization due to living with chronic pain. In fact, since increased DMN-INS connectivity is often observed across pain disorders with a significant central sensitization component, such as FM, it has been proposed as a key candidate biomarker of central sensitization ( Basu et al., 2018 ). However, current clinical pain at the time of scanning could also influence the DMN because pain engages cognitive mechanisms. Engaging in a cognitive task alters the activity and connectivity within the DMN and between DMN and attentional networks ( Anticevic et al., 2012 ; Gordon et al., 2014 ; Hampson et al., 2006 ). Similarly, pain demands attention and stimulus evaluation, and engages frontal regulatory mechanisms ( Villemure and Bushnell 2009 ; Kucyi et al., 2013 ; Legrain et al., 2009 ; Loggia et al., 2012 ; Seminowicz and Davis, 2007a ; Seminowicz et al., 2004 ). Consequently, experimental pain has been shown to affect the DMN in a similar way to a cognitive task ( Alshelh et al., 2018 ; Kong et al., 2010 ; Coghill et al., 1999 ; Seminowicz and Davis, 2007a ; Kucyi et al., 2016 ). Therefore, alterations of the DMN observed in patients with chronic pain could be related to the property of the current clinical pain (i.e. ‘state’ pain) - as an attention-demanding stimulus - to acutely perturb the DMN, in addition to indicating a more permanent functional reorganization of brain networks related to the chronic pain condition (diagnosis/trait). To explore the potential differential contributions of the current clinical pain (state) and of chronic pain condition (diagnosis/trait) on the connectivity of the DMN in FM we examined FM patients with and without current clinical pain during scanning, with each group compared to individually matched healthy control subjects. We hypothesized that DMN connectivity of patients not experiencing current clinical pain during the scan would not differ significantly from controls in the aforementioned regions associated with DMN in chronic pain, whereas DMN connectivity in those regions would be significantly altered in patients experiencing pain during scanning. Further, we hypothesized that the degree of connectivity disruption in patients experiencing current clinical pain during the scan would be related to the magnitude of the experienced pain.