Higher Sensory Cortical Energy Metabolism in Painful Diabetic Neuropathy: Evidence From a Cerebral Magnetic Resonance Spectroscopy Study

Diabetic peripheral neuropathy (DPN) is the most common chronic complication of diabetes and causes peripheral sensory loss, which is associated with a high risk of foot ulceration and subsequent amputation (1). Up to half of patients with DPN suffer with distressing painful neuropathic symptoms (painful DPN) that result in sleep interference, mood disorders, and a reduced quality of life (2). Unfortunately, current treatments provide suboptimal pain relief and are complicated by adverse effects (3). A greater insight into the pathophysiology of painful DPN is therefore urgently required.

There is emerging evidence for the involvement of the central nervous system, with alterations in both structure and function of the brain in DPN (4–6). A number of studies have reported abnormalities in patients with painful DPN, including dysfunction of the descending modulatory system (4), lower connectivity of the thalamus and primary somatosensory (S1) cortex (7), and functional reorganization (6) of the pain processing areas in patients with the painful insensate phenotype. Moreover, there is a reduction in thalamic neuronal function in patients with painless DPN, which is preserved in painful DPN and equivalent to levels in healthy volunteers (8). Recent studies also show hyperperfusion of the thalamus (9) and anterior cingulate cortex (emotional pain processing region) (10) in painful DPN, which may be normalized following the treatment of pain (10). These studies are indicative that there may be neuronal hyperactivity in patients with painful DPN in the pain processing areas of the brain. There is therefore a clear rationale to study whether there is higher energy consumption of neurons in the pain processing areas of the brain.

Under normal physiological conditions, adenosine triphosphate (ATP) is generated through mitochondrial phosphorylation whereby adenosine diphosphate (ADP) is combined with inorganic phosphate (Pi). However, during periods of increased energy expenditure or reduced ATP generation (e.g., hypoxia), phosphocreatine (PCr) donates Pi to ADP in order to maintain adequate cellular levels of ATP. PCr is considered an energy reservoir that is important for maintaining a constant level of ATP during physiological or pathological perturbations (11). The ³¹P magnetic resonance spectroscopy (³¹P-MRS) is an imaging modality that allows the detection of phosphorus-containing metabolites including ATP, PCr, and Pi, within a particular region of interest known as a voxel (11). The ³¹P-MRS is a robust technique (11) that provides an objective and reproducible measure of these metabolites and has been used in migraine (12), where higher energy consumption (lower PCr) was shown during a migraine attack. However, this technique has not been used to investigate DPN or neuropathic pain. Therefore, we conducted a ³¹P-MRS study in well-characterized cohorts of patients with painful and painless DPN. We hypothesized that there will be greater S1 cortical energy usage (i.e., lower PCr:ATP) in painful compared with painless DPN.

Thirty-two right-handed adults with type 2 diabetes and DPN, aged ≥18 years, were recruited consecutively from Sheffield Teaching Hospitals diabetes outpatient clinics to the study between 2018 and 2020. Exclusion criteria included current or past history of excessive alcohol usage (>14 units per week); nondiabetic neuropathies; other neurological disorders which may confound radiological/clinical assessments, or other factors which would preclude MRI (e.g., pacemaker/claustrophobia); pregnancy; major lower limb amputation; estimated glomerular filtration rate <45 mL/min/1.73 m²; antidepressant medication use; recurrent severe hypoglycemia; and moderate-to-severe pain from causes other than DPN. All participants gave written, informed consent before participation in the study, which had prior ethics approval by the Nottingham Research Ethics Committee, Nottingham (U.K.) (reference no. 17/EM/0430).

Participants underwent detailed clinical and neurophysiological assessments; full details are available in Supplementary Methods. In brief, this included structured assessment of neuropathy symptoms and signs (Toronto Clinical Neuropathy Score [TCNS]) (13), assessment of neuropathic pain symptoms and signs (Doleur Neuropathique en 4 [DN4]) (14) and different neuropathic pain symptoms (Neuropathic Pain Symptom Inventory [NPSI]) (15), lower limb comprehensive quantitative sensory testing (16), and nerve conduction studies.

On the basis of these assessments, participants were divided into two groups (17): 1) confirmed painless DPN, consisting of participants with painless DPN, with abnormal clinical findings (TCNS >5) and at least two abnormalities on nerve conduction studies one of which had to be an attribute of the sural nerve; and 2) confirmed painful DPN, with abnormal clinical findings (TCNS >5) and at least two abnormalities on nerve conduction studies—one of which had to be an attribute of the sural nerve—DN4 score ≥4, and chronic neuropathic pain for at least 6 months (18).

The Hospital Anxiety and Depression Scale (HADS) was used to determine the presence of depression (HADS-D) and anxiety (HADS-A), using a cut-off score of >10 (19).

The ³¹P-MRS was performed with the patients having discontinued pain-relieving medications for over 48 h. Participants had blood glucose measured before the scan to exclude hypoglycemia. An 11-point numeric rating scale (NRS, 0 = no pain to 10 = most severe pain ever) was used to record pain severity during the scan.

Participants underwent cerebral ³¹P-MRS [echo time 0.26 ms; repetition time 4,000 ms; acquisition technique image-selected in vivo spectroscopy (20)] at 3 tesla (Ingenia 3.0T; Phillips Healthcare, Best, the Netherlands) using a dual-tuned ¹H/³¹P head coil (RAPID Biomedical GmbH Rimpaer, Germany). A two-dimensional survey scan was performed for voxel localization. The voxels (size 25 × 25 × 40 mm³) were placed over the foot region of the S1 cortex (Fig. 1A). The ³¹P-MRS spectra (Fig. 1B) were quantified using the non-least-squares algorithm function (21) within the MR user interface package (jMRUI, version 5.2) (22).

The main study outcomes were ATP, PCr, Pi, and pH. By convention, metabolite levels were expressed as ratios: PCr:ATP as a measure of cellular energetics, with lower values indicative of greater energy usage; and Pi:ATP, which measures the status of oxidative phosphorylation within the cellular mitochondrion, with lower values suggestive of dysfunction. Metabolite ratios and tissue pH were calculated for the left, right, and mean (total) S1 cortical value.

Details are provided in the Supplementary Material and legends.

As ³¹P-MRS has not previously been performed in chronic pain conditions, we used a migraine study to determine our sample size. A sample size of at least 12 provides 0.90 power (α < 0.05) to detect a 0.12 difference in Pi:ATP (12).

The data sets generated during the current study are available from the corresponding author upon reasonable request.

There were no significant differences in demographic, biochemical, and neurophysiological variables (Table 1). The TCNS was greater in the painful compared with the painless DPN group (t test, P = 0.017, 95% CI −7.2, −0.76). As expected, the DN4 and total NPSI scores were significantly higher in the painful DPN group. Objective measures of neuropathy, including nerve conduction parameters, and quantitative sensory testing z-scores (Supplementary Table 1) were not significantly different between groups.

Table 1 shows the left, right, and total (left and right combined mean) S1 cortex ³¹P-MRS results for the two groups. The left, right, and total PCr:ATP were significantly lower in the painful compared with the painless DPN group (left: P = 0.005, 95% CI 0.15, 0.80; right: P = 0.019, 95% CI 0.08, 0.87; total: P = 0.006, 95% CI 0.14, 0.81) (Fig. 2A). There were no significant group differences in Pi:ATP and pH at the right, left, and total S1 cortices.

The mean S1 cortex PCr:ATP showed significant negative correlations with the DN4 score (Pearson’s r −0.465, P = 0.007), pain intensity (NRS) during the MRI scan (r −0.428, P = 0.016), and total NPSI (r −0.543, P = 0.001). Subscores of the NPSI also correlated with the S1 PCr:ATP (burning spontaneous pain, r −0.424, P = 0.016; pressing spontaneous pain, r −0.375, P = 0.034; paroxysmal pain, r −0.594, P < 0.001; evoked pain, r −0.396, P = 0.025; paresthesia/dysesthesia, r −0.486, P = 0.005). There were no correlations seen between PCr:ATP to other clinical or neurophysiological measures.

There were no significant differences in clinical, demographic, biochemical, or neurophysiological parameters, other than pain severity between high-pain (NRS 5.6 ± 1.8) and low-pain (NRS 0.4 ± 0.9) participants with painful DPN (see Supplementary Table 2). The PCr:ATP was significantly lower in the high-pain group [median (interquartile range) 1.87 (0.35)] compared with the low-pain group [2.35 (0.67), Mann-Whitney U test P = 0.033] (Fig. 2B).

The main finding of this first-ever ³¹P-MRS study in clinical DPN was that the PCr:ATP, a measure of cellular bioenergetics, was significantly lower in painful compared with painless DPN, indicating higher energy consumption. The PCr:ATP was also significantly lower in participants with moderate/high pain compared with low pain within the painful-DPN group. During periods of high energy demand, PCr acts as a buffer, donating Pi to ADP to maintain cellular ATP levels. Hence the lower levels of PCr are due to increased S1 cortical cellular activity and greater energy usage in painful DPN, specifically in those with high self-reported real-time pain scores. However, the pH and Pi:ATP were similar between groups, indicating that cellular and mitochondrial functions were not different.

The S1 cortex is one of the main components of the cerebral network for processing pain, receiving ascending impulses from important pain pathways. Increased brain activity in this region during experimental pain has been shown to be coupled with increased blood flow (23) and relates to the severity of stimulus intensity (24). These studies support our findings of depleted S1 PCr in painful DPN, likely due to the continuous barrage of afferent impulses. Moreover, our results are consistent with other studies that show evidence of increased cerebral neuronal activity in painful DPN in other brain regions, including increased cerebral blood flow in the thalamus (9) and anterior cingulate cortex (10).

This study also found a relationship between S1 cortical PCr:ATP and measures of neuropathic pain. This further strengthens the link between cerebral energy consumption and neuropathic pain. In particular, the NPSI paroxysmal pain subscore (stabbing and electric shock–type pains) was most significantly correlated with PCr:ATP, whereas there were weaker correlations with other NPSI subscores such as burning and pressing spontaneous pains. A post hoc cluster analysis of the COmbination vs Monotherapy of pregaBalin and dulOxetine in Diabetic Neuropathy (COMBO-DN) painful DPN clinical found that those patients with an overall higher pain intensity were also characterized by higher paradoxical pain scores (25). Therefore, paradoxical pain may be associated with greater S1 cortical energy usage and higher overall pain intensity.

PCr:ATP levels in patients with painless DPN were significantly higher compared with those with painful DPN, despite a similar severity of neuropathy. This signifies the relatively lower resting-state neuronal activity in the S1 cortex compared with patients with painful DPN. Other studies demonstrate that painless DPN may be associated with reduced cerebral neuronal activity, including a reduction in neuronal function and blood flow (9) in the thalamus (8). Moreover, previous studies demonstrate a reduction in S1 cortex peripheral gray matter volume in participants with DPN, but the potential pathophysiological mechanisms have been unexplained (5). To our knowledge, this is the first study to show reduced bioenergetics in patients with painless DPN, suggesting reduced neuronal activity, which may be associated with cortical atrophy or a dying-back axonopathy secondary to reduced sensory input.

The strengths of this study include the use of ³¹P-MRS, a unique MR modality that does not require the use of an ionizing radiation or administration of contrast/radioactive tracers (11). Pain is an individual and subjective experience, and this technique offers an objective measure of tissue metabolism to gain insight into the cerebral mechanisms of painful DPN. The sample size was adequate for a preliminary, proof-of-concept study and is similar to other cerebral imaging studies in DPN (9). The groups were well characterized using a battery of validated neurological clinical scoring systems and neurophysiological tests. Now, larger cohort studies are necessary to analyze different pain phenotypes (e.g., irritable and nonirritable nociceptors) with the inclusion of deeper assessments of neuronal function such as conditioned pain modulation and spinal inhibitory function (17,26). The limitations of this study include the cross-sectional nature of the study, which does not allow a causal association to be determined.

In conclusion, this preliminary study of ³¹P-MRS in DPN demonstrates increased cortical energy metabolism in the S1 cortex in participants with painful DPN—specifically in those with greater neuropathic pain severity. Our findings may have clinical relevance, as ³¹P metabolites have the potential to act as an objective biomarker of painful DPN or neuropathic pain in future clinical trials. Further larger studies are required to confirm these findings, including examination of other brain regions involved with nociceptive processing.

Acknowledgments. The authors acknowledge the hard work, skills, and contributions of the radiographers at the University of Sheffield Magnetic Resonance Imaging Department. The authors also greatly appreciate the study volunteers who spent considerable time participating in this study. I.W. is deceased.

Funding. S.T. thanks Sheffield Teaching Hospitals Diabetes Charitable Trust for funding the study.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. G.S. recruited participants, undertook clinical and neurophysiological assessments, researched and analyzed clinical and MR data, and wrote the manuscript. A.A., I.W., S.C., and D.S. made substantial contributions to the design, acquisition, analysis, and interpretation of data, and drafting of the article, and gave final approval of the version published. S.T. made substantial contributions to conception and designs, reviewed and revised the manuscript, and gave final approval of the version published. S.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.