Suppressed prefrontal cortex oscillations associate with clinical pain in fibrodysplasia ossificans progressiva

A total of 17 consecutive patients diagnosed with FOP between 7 and 61 years of age (11 females and 6 males) gave consent and were enrolled in the current study (Table 1). Of the 17 FOP patients, 13 individuals were confirmed carriers of the classic R206H mutation. It is estimated that 97% of FOP patients harbor a gain-of-function R206H mutation in ACVR1 [43]. One FOP patient (patient 16) had a clinical and genetic variant of FOP (personal communication with Dr. Fredrick S. Kaplan at Center for Research in FOP and Related Disorders at University of Pennsylvania), while 3 individuals had not undergone genetic testing, but possessed other FOP diagnostic features such as malformation of the big toes. Following enrollment, FOP patients provided medical history, analgesic use, demographics, clinical pain intensity rating, and fNIRS data was acquired. fNIRS data was collected in multiple settings (i.e., patients’ home or community gathering) in order to accommodate patients and families. An additional clinical pain intensity rating and completion of online clinical questionnaires occurred after the fNIRS session and within a two- to three-week period. Thus, clinical pain intensity ratings were first collected at the time of fNIRS data acquisition and subsequently, during the administration of the full battery of clinical questionnaires. Clinical questionnaires were completed at a slightly later time as some patients required special equipment or software to perform online study-related task.

FOP patients harbored HO of varying degree (Table 1). While some individuals presented with known HO localized to a single site (e.g., craniofacial region or middle of the back), others possessed HO lesions in multiple upper and lower body regions. FOP patients also had varying levels of physical disability with some patients being completely ambulatory and required no assistance, while others used a cane or motorized wheelchair. Nine FOP patients utilized a device (i.e., motorized wheel chair or walking cane) to assist with mobilization. One FOP patient (Patient 1) was potentially in a flare-up state during time of fNIRS acquisition and all other patients were in a quiescent state. Patients did not report experiencing recent acute illnesses (e.g., fever or flu) or physical trauma.

We designated a 0–3 pain rating as low pain, while ratings ≥ 4 were classified as high pain. The rationale for choosing this criterion stems from prior work noting pain interference in patients’ daily life occurring at pain levels of 4 or more [6, 44] or that analgesic treatment is frequently sought when pain intensities are higher than 3 [11]. A battery of clinical questionnaires derived from the Patient-Reported Outcomes Measurement Information System (PROMIS; http://​www.​healthmeasures.​net) database were utilized to capture patient reported levels of (1) pain intensity (0–10 numerical rating scale), where 0 is no pain and 10 is the worst pain imaginable, (2) manifestations of pain, (3) quality of pain (i.e., sensory & affective elements), (4) physical stress, (5) physical ability, (6) psychological stress, (7) anxiety, and (8) depressive mood. Parents or guardians of pediatric FOP patients also completed the Adult Response to Children’s Symptoms (ARCS) questionnaire. The ACRS (29 items) was utilized to determine the parents’ or guardians’ behaviors in response to the child’s pain. Study participants also provided demographic data (age and gender). Each ACRS question was scored on a five-point Likert scale (Never: 0 to Always: 4). All questionnaires were administered and completed online using REDCap (https://​www.​project-redcap.​org). It is noted that two study participants did not fully complete questionnaires and thus the cross-correlation analysis was restricted to an N = 15 FOP patient population.

While methods such as functional magnetic resonance imaging (fMRI) remain critical for characterizing CNS pain pathways in healthy subjects and clinical populations, undergoing an MRI-based assessment can be extremely difficult and, in some cases, impossible, due to the physical limitations experienced by FOP patients. Therefore, CNS hemodynamic properties (i.e., concentration changes of oxygenated hemoglobin (HbO), deoxygenated hemoglobin (HbR) and total hemoglobin) in FOP patients were measured using a multichannel continuous wave fNIRS system (CW7 System, TechEn, Milford, MA, USA) operating at 690 nm and 830 nm wavelengths. Each patient was first comfortably positioned either in a chair or in many cases, remained in their own wheelchair. Once study participants were positioned and at rest, pain intensity ratings (0–10 numerical rating scale) were obtained. Subsequently, an fNIRS cap—one for adults and one for children—consisting of 6 light sources, 8 standard-separation detectors, and 6 short-separation detectors was placed on the participant’s head (Fig. 1). Short-separation detectors were utilized to measure and mitigate any signal changes arising from non-cerebral sources (i.e., scalp or skull). Following the completion of fNIRS cap positioning (~ 20 to 30 min), each FOP patient was asked to remain still and visually fixate on a white cross displayed on computer screening. Resting-state fNIRS data were collected over a six-minute period.

PROMIS questionnaires: For each individual patient, the summed raw scores derived from short-form, PROMIS-based questionnaires were converted to a T-score metric. The ensuing score was represented by a T-score, a standardized score with a mean ± standard deviation (SD) of 50 ± 10. Thus, a t-score of 40 is one SD below the mean. Higher scores for symptom measures (i.e., depressive symptoms) suggest worse symptom presentation, while higher scores for functional measures (i.e., cognitive function) correspond with higher functioning. Spearman's rank correlation coefficients were calculated to explore the association between each PROMIS-based measure. A false discovery rate (FDR) adjusted p-value (α = 0.05) was subsequently calculated. Statistical comparisons between low and high pain FOP study cohorts was performed using a two-tailed t-test (α = 0.05).

fNIRS data: An open-source MATLAB (Mathworks, Natick, MA, USA) toolbox, HomER2 (https://​homer-fnirs.​org), was utilized to preprocess all fNIRS datasets. Briefly, the raw optical intensity data were first transferred to changes in optical density by taking the logarithm of the signal. Optical density time courses were then visually inspected for quality assurance/control (QA/QC) across the 6-min acquisition period for any signal drifts or spikes, indicative of head motion (Additional file 1). During QA/QC of prefrontal cortex signals, 4 subject datasets were excluded due to excessive head motion or low signal to noise ratio, yielding 13 subject datasets for power-spectral analysis (Pain Level: 0–3; N = 7 & Pain Level: 4–10; N = 6). Separate QA/QC of somatosensory cortex signal resulted in 10 viable datasets (Pain Level: 0–3; N = 5 & Pain Level: 4–10; N = 5). Filtered (0.01–0.5 Hz) optical density time courses were further converted to concentration changes in HbO, HbR and total hemoglobin using the modified Beer-Lambert Law with a partial pathlength factor of 6. Finally, a linear regression model was employed to regress out superficial noises from the time series of each cortical fNIRS channel (i.e., having a distance of 3 cm) by including the hemoglobin time course of the short separation channel (i.e. having a distance of 8 mm) with the highest correlation as a covariate in the linear model. A constant regressor with the value of 1 was also added to the model.

Power spectral analysis of low-frequency oscillations on the recorded fNIRS hemodynamic signals was performed to investigate the changes in neuronal oscillatory dynamics with pain. Here, the HbO data were utilized rather than HbR data, based on the availability of higher signal-to-noise ratio of HbO compared to HbR time courses [23]. For each patient, the 6-min HbO time courses of each channel were first transformed to the frequency domain using the fast Fourier transform. The power of each frequency component was obtained by taking the square of its absolute amplitude value. Furthermore, prior work has shown that cortical oscillations occur in distinct frequency bands, where each range is associated with unique physiological functions [3]. Thus, we further subdivided the low-frequency oscillations range into five sub-bands: slow-5 (0.01–0.027 Hz), slow-4 (0.027–0.073 Hz), slow-3 (0.073–0.198 Hz), slow-2 (0.198–0.25 Hz), and slow-1 (0.25–0.5 Hz). For each sub-band, we conducted two-sample t-tests on the mean power values of each normal fNIRS channel between the high pain group and low pain group to test the null hypothesis that patients with high pain levels do not have a statistically significant reduction of the sub-band power of a particular channel (and therefore the underlying brain region). Channel-wise FDR correction was then applied to control the number of false positives. FDR adjusted (α = 0.05) p-values were utilized to determine significance. Resting-state functional connectivity was also explored (Additional file 1).

Of the 17 enrolled FOP patients (Table 1), 11 individuals (6 males, 5 females) were categorized in the low pain (Pain Level: 0–3/10) cohort at the time of fNIRS data acquisition, while 6 female patients were allotted to the high pain (Pain Level: ≥ 4) group. In order to compare patients’ gender with low pain vs. high pain, the Fisher’s Exact test was utilized. We observed a significant association between gender and pain (p = 0.043). Pain ratings between participants less than 18 years of age vs. adult patients was moderately significant (p = 0.046). A comparison of pain levels between the first (fNIRS data acquisition) and second (clinical questionnaire completion) evaluation time points demonstrated that while pain intensity remained relatively stable for many FOP patients, others showed increases or decreases in pain (Fig. 2). Moreover, no relationship among gender and pain was present at the second clinical pain intensity assessment (p = 0.14). Similarly, a significant difference between FOP patients less that 18 years compared to individuals over the age of 18 was not observed at the subsequent time point (p = 0.18). Analgesics, including prescription opioids, were used by both study populations with some patients utilizing treatments on an intermittent or as needed basis, while others had more consistent usage. A Fisher’s Exact test was utilized to assess the association between pain (≥ 4 vs < 4) and analgesic usage (Any usage = 1 vs. Otherwise = 0). No association between pain and analgesics usage was observed with the p-value being highly non-significant (p = 1.0). Clinical questionnaire data acquired from FOP patients showed that individuals in the high pain cohort (N = 7) relative to the low pain group (N = 9) had altered emotional functioning in domains such as anxiety in conjunction with physical symptoms, mainly physical stress (Table 2). The interactions between pain intensity, the quality of pain (affective vs. sensory), mental health and physical health in FOP, all of which were self-reported assessments, were also characterized (Fig. 3). Of note, the intensity of pain was significantly correlated magnitude of self-reported depressive symptoms (r = 0.74, p = 0.0016) in addition to the magnitude physical stress experienced (r = 0.65, p = 0.009). Interestingly, significant relationships were further observed between level of anxiety experienced by patients, sensory aspect of pain and physical stress, suggesting an interaction among these three domains in FOP. Finally, across pediatric FOP patients, the mean ± SE ARCS questionnaire response was 2.31 ± 0.26. Low ARCS scores were likely reflective of low pain levels experienced by pediatric FOP patients.

As shown in Fig. 4a, relative to the low pain cohort (N = 7), the amplitude of low-frequency oscillations localized to the prefrontal cortex and within the slow-5 (0.01–0.027 Hz) sub-band were significantly lower in high pain FOP patients (N = 6). A robust suppression of slow-5 power was observed in the high pain group across medial (channels C2 (FDR-p_corr = 0.012) and C7 (FDR-p_corr = 0.013)) as well as more lateral (channel C5 (FDR-p_corr = 0.015)) components of the prefrontal cortex (Cohen’s d: C2 = 1.51; C5 = 1.45; C7 = 1.46). Across all study participants, self-reported pain levels were negatively correlated with prefrontal cortex, slow-5 power (Fig. 4b). Pain score obtained from the second pain evaluation were also negatively correlated with the magnitude of prefrontal cortex, slow-5 power (C2: r = − 0.60, p = 0.039; C5: r = − 0.28, p = 0.39; C7: r = − 0.73, p = 0.007). A significant correlation was not detected between slow-5 power and severity of depressive symptoms (r = − 0.36 to − 0.41; p > 0.05). The age of the patient did not correlate with the magnitude of prefrontal cortex, slow-5 power (C2: r = − 0.06, p = 0.84; C5: r = 0.16, p = 0.60; C7: r = − 0.27, p = 0.37). As low-frequency oscillations at the slow-5 range are involved with long-range interaction of brain regions, these finding suggests that FOP patients experiencing higher levels of pain likely possess a greater magnitude of dysfunctional long-range communication involving the prefrontal cortex [3, 5, 40]. (See also resting-state functional connectivity analysis results in Additional file 1). In other words, higher self-reported pain in FOP was associated with greater disruption of large-scale CNS networks encompassing the prefrontal cortex. Figure 5 demonstrates that slow-5 amplitude alterations were not significantly different over the primary and secondary somatosensory cortices (pre- and post-central gyrus) when comparing low (N = 5) vs. high pain (N = 5) FOP patients. Calculation of effect sizes (Cohen’s d (Range): 0.03–0.79) indicated a low to moderate pain-dependent effect on slow-5 power in somatosensory cortices. A correlation between clinical pain intensity ratings and slow-5 power for any somatosensory channel, channels C9-C16, was not observed (r = − 0.04 to 0.23; FDR-adjusted p > 0.05). Significant differences between low and high pain FOP patients were not detected in other frequency sub-bands (i.e., slow-1 to slow-4) and in either prefrontal or somatosensory cortices.