Brain levels of the neurotransmitter γ-aminobutyric acid are reduced in people with HIV-related neuropathic pain

Abstract

Previous studies suggest a dysregulation of the inhibitory γ-aminobutyric acid (GABA) and the excitatory glutamate/glutamine (Glx) neurotransmitter systems in people living with chronic pain. Here, we test this hypothesis in people with HIV (PWH) on stable antiretroviral therapy, either with or without neuropathic pain (PWHpain and PWHnopain, respectively), and people without HIV and pain (Ctrl). Fourteen PWHpain (age, mean ± SD: 59 ± 6.5, 12 males), 13 PWHnopain (55 ± 9, 12 males), and 14 Ctrl (58 ± 10, 14 males) completed a 3T ¹H-magnetic resonance spectroscopy MEGA-PRESS scan quantifying GABA and Glx in the left posterior insula. Furthermore, temporal summation was evaluated using cuff pain algometry, applied on the participants’ left calf for 120 seconds at a pressure calibrated to a subjective target pain rating of 40/100. In addition, we evaluated blood plasma levels of neurosteroids (ie, allopregnanolone) known to be endogenous modulators of GABA-A receptors. People with HIV with neuropathic pain exhibited increased temporal summation of cuff pain and decreased posterior insula GABA levels compared to Ctrl and PWHnopain (P’s < 0.05). There were no statistically significant group differences in Glx. Lower GABA levels were associated with higher average cuff pain ratings (R = −0.44, P < 0.05) and temporal summation scores (R = −0.49, P < 0.01) in PWH. In addition, lower allopregnanolone levels were associated with higher insular Glx levels in PWHpain (R = −0.64, P < 0.05). Our results provide a link between decreased GABA levels and neuropathic pain in PWHpain. These results suggest that insufficient inhibitory metabolite levels, rather than excessive excitatory metabolite levels, may be linked to neuropathic pain in PWH.

1. Introduction

More than 33 million people globally live with human immunodeficiency virus (HIV).²⁴ Approximately one-third of people with HIV (PWH) report having systemic neuropathies either as a result of the virus itself or as a side effect of HIV treatment.^18,38,65 As such, neuropathic pain (ie, pain that arises from lesions or damage to the somatosensory nervous system) is the most commonly reported and arguably the most debilitating pain symptom in PWH.⁵²

γ-Aminobutyric acid (GABA) and glutamate (Glu) are the main inhibitory and excitatory neurotransmitters in the central nervous system (CNS), respectively. These neurotransmitters are metabolically interrelated, as glutamate is the metabolic precursor of GABA,^36,50 highly correlated in the resting normal human brain,⁶¹ and coordinated activity of these systems are critical for normal neurological development and function, including pain perception.^8,27,36,61

It is well-described that HIV-1 impairs the glutamatergic system.^27,39,51,53 In the CNS, HIV-1 viral proteins (including glycoprotein 120, gp120; transactivator of transcription, Tat) infect glial cells, which in turn release neurotoxins that eventually lead to synaptic loss, disturbances in energy homeostasis, and glutamate metabolism, ultimately leading to excitatory toxicity and oxidative stresses.⁵³ More specifically, via dysregulation of astrocyte Glu clearance,⁵¹ reduced function and expression of Glu transporters, accumulation of extracellular Glu in the synaptic cleft, and overstimulation of Glu receptors.^27,39 Disruptions in the Glu-GABA homeostasis has been suggested to be a crucial factor in the formation of hyperalgesia or allodynia occurring in neuropathic pain.³⁶

Further, HIV virotoxins have the capacity to disrupt synthesis and release of circulating neuroactive steroids.⁵⁵ These include allopregnanolone, which are considered the most potent endogenous modulators of GABA-A receptors, acting through GABA-A receptor agonism,^{7,25,26,35,42} eliciting analgesic, anxiolytic, anticonvulsant, and sedative effects.⁷ Increasing the production of endogenous neuroactive steroids, in particular allopregnanolone, in the spinal cord and peripheral nerves has proven successful in modulating neuropathic, inflammatory, and postoperative experimental pain.^25,26,33,42 Such findings suggest that this molecule may not only be a candidate for neuroendocrine modulation in HIV⁵⁵ but may also be beneficial for treating neuropathic pain in PWH.^25,26,45

The main aim of this study is to evaluate the link between HIV-related neuropathic pain and brain GABA/glutamate/glutamine levels using magnetic resonance spectroscopy (MRS), a technique used to noninvasively assess in vivo a wide range of neurometabolites including GABA and Glu (the latter through a composite measure of Glu and its precursor Glutamine, Glx).¹⁷ We hypothesize to observe (1) reduced GABA levels in people with HIV with neuropathic pain (PWHpain) compared to people with HIV without neuropathic pain (PWHnopain) and Ctrl; (2) increased Glx levels in PWHpain compared to PWHnopain and Ctrl; and (3) imbalances in GABA and/or Glx levels to be associated with pain symptoms and quantitative sensory testing (QST). Quantitative sensory testing involved cuff pain algometry to measure temporal summation (ie, increased pain perception evoked by repeated noxious stimuli), which is thought to be mediated by central sensitization.^5,14 Further, we explored the association between HIV-related neuropathic pain and circulating blood plasma levels of the neuroactive steroids pregnenolone and allopregnanolone, as well as the association between these neurosteroids, neurometabolites, and behavioral pain outcomes.

2. Methods

2.1. Study design

The current study was conducted at the Athinoula A. Martinos Center for Biomedical Imaging at Massachusetts General Hospital (MGH), Boston, MA. Mass General Brigham (MGB) Institutional Review Board and the Radioactive Drug Research Committee at MGH approved the research protocol. Before study participation, all participants provided written informed consent according to the Declaration of Helsinki. At visit 1, a trained medical professional collected the participants’ medical/surgical history, a list of their current medications, and venous blood and assessed their eligibility. In addition, participants completed questionnaires and performed cuff algometry and neurocognitive testing. At visit 2, participants underwent simultaneous [¹¹C]PBR28 positron emission tomography (PET)-MR scanning, filled out questionnaires such as PainDETECT and brief pain inventory, and underwent von Frey QST. The PET data, von Frey QST, neurocognitive testing exploratory questionnaires are outside of the scope of the current work and will be reported elsewhere.

2.2. Participant eligibility criteria

A total of 16 PWHpain (age, mean ± SD: 59.2 ± 6.2), 15 PWHnopain (55.4 ± 10.6), and 14 Ctrl (58.6 ± 10.6) participants were enrolled in the current study. Inclusion criteria for PWH were age ≥18 and ≤80, positive result on a HIV-1/2 enzyme-linked immunosorbent assay (ELISA), and on stable antiretroviral therapy for 3 months with a HIV viral load <200 copies/mL at visit 1. In addition, PWH were grouped based on the presence (PWHpain) or the absence (PWHnopain) of chronic neuropathic pain as assessed by a combination of medical history and self-report. A review of medical records confirmed that nearly all pain subjects had documented neuropathic pain diagnoses or symptoms consistent with pain associated with neuropathy (eg, peripheral polyneuropathy, paresthesia of limbs, bilateral neuropathy with burning/vibratory sensation, HIV-associated neuropathy). Although no pain level cut off was implemented, to maximize the dynamic range in our data, all pain patients had to have perceived this symptom as sufficiently vexing to have sought medical care for it. Exclusion criteria for PWH included untreated viral or bacterial infection, current or past cancers (unless resolved or in remission for >5 years), PET/MR contraindications (eg, pacemakers, pregnancy or breast feeding, claustrophobia), history of major neurological disorders (eg, neurodegenerative disease), history (<1 year) of head trauma requiring hospitalization, and history of opportunistic infection of the CNS (eg, pneumocystis pneumonia). Potentially GABA-modulating medication taken at the time of screening included clonazepam, gabapentin, lorazepam, pregabalin, and topiramate (for more details, see supplementary table 1, http://links.lww.com/PAIN/C375). Number of participants that took medication with potential interaction effects with GABA (Ctr = 0; PWHnopain = 2; PWHpain = 7). Please see supplementary table 1 and supplementary figure 1, http://links.lww.com/PAIN/C375, for more information.

The eligibility criteria for Ctrl were identical to those for PWH, with the exception of inclusion criteria of confirmed HIV negative status using ELISA, the lack of antiretroviral therapy, and confirmed pain-free status based on self-report and medical records. The final study sample consisted of 14 PWHpain (age, mean ± SD: 59 ± 6.5, 12 males), 13 PWHnopain (55 ± 9, 12 males), and 14 Ctrl (58 ± 10.6, 14 males). Two PWHpain and 2 PWHnopain were excluded because of incomplete insular ¹H-MRS MEGA-PRESS data collection.

2.3. Brain imaging data acquisition

Brain imaging data were acquired using a 3 Tesla Siemens TIM Trio whole-body MRI with a dedicated BrainPET head camera insert equipped with an 8-channel head coil. A high-resolution multiecho MPRAGE (MEMPRAGE: T1-weighted structural MRI) volume was acquired (TR = 2530 ms; TE1 = 1.63 ms; TE2 = 3.49 ms; TE3 = 5.35 ms; TE4 = 7.21 ms, flip angle = 7°) to aid in the anatomical placement of the MRS voxel and to correct for partial volume effects of cerebrospinal fluid (CSF) across all metabolites and alpha correction of GABA metabolites (described in detail below).²⁹ A ¹H-MRS MEGA-PRESS sequence (TR = 1700; TE = 68; NA = 160, voxel size = 20 × 25 × 40 mm³) was used to measure GABA concentrations in the left posterior insula, among other regions that will be published elsewhere. The water unsuppressed spectrum was acquired from the same voxel placed in the left posterior insula to estimate absolute concentrations in institutional units (i.u.). Before data acquisition, the MRS volume of interest (VOI) underwent an automatic shim routine based on gradient double acquisition (GRE-shim) followed by manual shimming.

The insular cortex is a key region in pain processing⁵⁹ and is strongly linked to central sensitization mechanisms in various chronic pain conditions,^32,37,44 highlighting its importance in studying neurometabolic alterations associated with pain perception. The posterior insula, in particular, has been associated with nociceptive processing that is specific to modality, location, and intensity of the stimuli.^10,46,59 In addition, neuroinflammation within the insula has been observed in people with HIV, as evidenced by elevated TSPO PET signal,⁵⁶ making it a relevant region of interest for investigating neurochemical alterations in individuals with HIV-related neuropathic pain.

2.4. ¹H magnetic resonance spectroscopy data processing

The MEGA-PRESS data were analyzed using Gannet toolkit (v.3.3.1)¹⁷ running under SPM12⁶⁰ in Matlab (Version R2022b). The time-domain data were imported from the scanner and processed into the frequency-domain using a fast Fourier transformation. Nonlinear least-squares fitting of the spectra was used to integrate the edited GABA and Glx peak at 3 ppm and 3.75 ppm, respectively, to produce GABA and Glx concentration estimates in i.u. relative to water.¹⁷ MEGA-PRESS data were assessed for quality, and those exceeding a fit error >20% were excluded from final analysis. Next, participants’ native space MRS voxel masks were coregistered to their individual native space T1 images and segmented for gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF).⁶

Tissue correction for GABA-edited MRS, also known as alpha correction (ie, assuming twice as much GABA in GM compared to WM),²⁹ was performed as follows: signal integrals for GABA and water were applied to the tissue specific relaxation and water visibility constants for each tissue fractions to quantify GABA. The differences in GABA concentration were corrected across compartments using alpha = 0.5.²⁹ This step generated the compartment-corrected GABA estimate. Finally, the estimates were normalized to the group-average voxel tissue fractions.²⁹ Average voxel placement maps were generated for visualization purposes in SPM12 and normalized to the Montreal Neurological Institute (MNI) space. Specifically, we first calculated the nonlinear transformation from the T1 to MNI. This transformation was then applied to each individual voxel mask. Finally, all binary masks were summed to generate a probability map of the voxel placement and coregistered to the T1 in MNI space. Of note, all GABA concentrations are reported as GABA+ as most editing techniques to measure GABA result in some contamination of the GABA signal with macromolecules.

2.5. Quantitative sensory testing

At visit 1, we assessed temporal summation of pressure pain by placing a rapid cuff inflation system on the participants’ left calf (CPA, Hokanson E20/AG101). In 2 separate trials, the inflatable cuff pressure was slowly ramped up (10 mm Hg/second) until the participants reported a moderately painful sensation, corresponding to ~40 on a 0 to 100 numerical rating scale (NRS; 0 = no pain; 100 = worst pain imaginable). Next, the target pressure levels from these 2 trials were averaged and continuously applied to the participant for 120 seconds. Participants were asked to report their current pain every 30 seconds during the continued cuff application using the same 0 to 100 NRS and to rate any ongoing pain 15 seconds and 30 seconds after cuff deflation, so called painful aftersensations (according to Ref. 57).

2.6. Pain questionnaires

PainDETECT is a self-report assessment for neuropathic pain, originally developed by Freynhagen et al.²³ The higher the score, the higher the likelihood that an individual has neuropathic pain. A sum score of ≤12 or ≥19 indicate, respectively, a low (ie, <15%) or high (ie, >90%) probability that the pain has a neuropathic component, whereas a sum score of between these indicates an “unclear” classification.

Brief pain inventory is a self-report questionnaire designed to assess both the intensity of pain and its impact on daily functioning. Patients are asked to rate their pain at its worst, least, average, and current levels by combining shading painful regions in a diagram, and by using a 10-point Likert scale from 0 = “No pain at all” to 10 = “Pain as bad as you can imagine” as well as 0 = “does not interfere” to 10 = “completely interferes.” Higher score indicate more severe pain and disability.¹³

2.7. Quantification of the neuroactive steroids pregnenolone and allopregnanolone

Blood plasma levels of neuroactive steroids were collected in participants (PWHpain n = 18; PWHnopain n = 13; Ctrl n = 8. Of these, 12 PWHpain, 9 PWHnopain, and 8 Ctrl had both neuroactive steroid samples taken and successful MRS imaging). Cryopreserved EDTA plasma collected at visit 2 were subjected to triplicate liquid–liquid extractions using ethyl acetate. Combined extracts were further purified by collecting fractions of interest from an Agilent 1100 Series high-performance liquid chromatography system (Agilent Technologies, Santa Clara, CA) equipped with a 25 cm × 4.6 mm × 5 μm LiChrosorb DIOL-5 high-performance liquid chromatography column (Merck KGaA, Darmstadt, Germany) under normal phase conditions. Purified samples were subsequently derivatized with heptafluoroacetic anhydride and injected onto an Agilent 7013A gas chromatography–triple quadrupole mass spectrometer with chromatographic separations performed on the interfaced Agilent 7890B GC. A splitless injection of 2 μL was used on a 30.0 m × 0.25 mm × 0.25 μm installed HP-5MS GC column. The limit of quantification was 0.2 to 2 pg for neuroactive steroids.

2.8. Statistical analyses

Statistical analyses were performed in R (Version 4.2.0). Shapiro–Wilk tests and visual inspection of data were used to determine normality of continuous variables. In cases these assumptions were met, we used parametric statistics; otherwise, we used nonparametric statistics. Average pain ratings were obtained by averaging the pain ratings from 0 seconds to 120 seconds. Temporal summation (TS) score was calculated as NRS pain ratings at 120 seconds minus at 0 seconds and then normalized to millimeter of mercury (mm Hg). Chi square test was used to assess differences in sex distribution across the 3 groups. All metabolite concentration data met the assumption of a Gaussian distribution and were subsequently analyzed using linear models, followed by pair-wise correction for multiple comparisons using Tukey Honest Significant Difference (HSD). Non-Gaussian distributed data (eg, painDETECT scores, average pain pressure in mmHg, temporal summation score and neuroactive steroid concentrations) were analyzed using Kruskal–Wallis, followed up by Dunn post-hoc test. To determine whether sex and age should be controlled for in the statistical analyses of metabolite group differences, we compared Akaike Information Criterion (AIC)¹ and Bayesian Information Criterion (BIC)⁵⁸ values from the linear models that (1) controlled for age and sex (AIC = 118.4; BIC = 128.6), (2) controlled for age only (AIC = 122.4; BIC = 130.9), (3) controlled for sex only (AIC = 117; BIC = 125.6), and (4) included no covariates (AIC = 121.1; BIC = 127.9). Akaike Information Criterion and Bayesian Information Criterion provide means for model selection by comparing the relative statistical quality amongst selected models to reduce the likelihood of overfitting a model. Lower AIC and BIC are generally preferred, but in cases where the model shows comparable results, the simplest model is typically recommended.^1,58 Because the model fit results yielded comparable results, the low n of females in each group, and chi square test revealed no group differences in sex distribution across the 3 groups (X² = 0.34, df = 2, P = 0.85), we here on report the results from the model omitting covariates. Variables involving repeated measures and extended over time (ie, QST measures) were also assessed with mixed linear model followed by Tukey HSD correction for multiple comparisons. Akaike Information Criterion and Bayesian Information Criterion values were compared from the QST linear model assessing pain ratings over time that (1) controlled for age and sex (AIC = 285; BIC = 294.8), (2) controlled for age only (AIC = 283.4; BIC = 291.6), (3) controlled for sex only (AIC = 283.2; BIC = 291.3), and (4) included no covariates (AIC = 289.9; BIC = 296.5). Similarly, the AIC and BIC suggested comparable fits across models. Hence, we report the results from the model omitting covariates.

Associations across MRS metabolites and between metabolites and QST assessments were investigated using Pearson product-moment correlation. We used Fisher r to z transformation to compare correlation coefficients, obtaining z-difference scores to estimate the magnitude of the differences and P-values to estimate the statistical significance.

3. Results

3.1. Demographics and clinical variables

Demographic variables and patient characteristics can be found in Table 1. All participants had a HIV viral load <200 copies/mL at visit 1 except for 1 individual who, because of an oversight, was included with a viral load of 260 copies/mL. Upon verification that this participant was not an outlier in any of the measures evaluated, their data were included in our group analyses.

Table 1.

Demographics and clinical characteristics.

	Ctrl	PWHnopain	PWHpain
Subjects, n	14	13	14
Age (mean ± SD)	58 ± 10.6	55 ± 9	59 ± 6.5
Sex (M/F)	14/0	12/1	12/2
Race, n (%)
White	13(93%)	11 (84.6%)	8(57.1%)
Black	0 (0%)	0 (0%)	4 (28.6%)
Asian	0 (0%)	1 (7.7%)	0 (0%)
Multiracial	1 (7%)	1 (7.7%)	2 (14.3%)
Ethnicity, n (%)
Non-Hispanic/Latino	14 (100%)	11 (84.6%)	12 (85.7%)
Hispanic/Latino	0 (0%)	1 (7.7%)	2 (14.3%)
Not reported	0 (0%)	1 (7.7%)	0 (0%)
PainDETECT
Current pain (mean ± SD)	0.1 ± 0.3	0.2 ± 0.4	5.1 ± 3.1
Average pain (mean ± SD)	0.4 ± 0.7	0.8 ± 1	6.1 ± 2.2
Strongest pain (mean ± SD)	0.9 ± 1.8	1.3 ± 1.4	7.6 ± 2.2
Total score (mean ± SD)	0.9 ± 1.5	1.6 ± 1.7	16.3 ± 6.8

PWHnopain, people with HIV without neuropathic pain; PWHpain, people with HIV with neuropathic pain.

There was a significant difference between the 3 groups’ PainDETECT final scores (F(2,36) = 64.74, P < 0.00001). Post-hoc test with Bonferroni correction for multiple comparisons revealed that PWHpain had higher scores compared to Ctrl and PWHnopain (both P’s < 0.001), whereas the Ctrl and PWHnopain groups did not differ from each other (P = 0.94). See Table 1. To assess the overall distribution of the experienced pain (ie, localized or widespread) at the day of scanning, we summarized the data from Brief Pain Inventory diagram, dividing pain distribution into body regions. The results revealed that 86% of the participants in the PWHpain group reported pain in >3 body regions, suggesting widespread pain rather than focalized symptoms.

3.2. Magnetic resonance spectroscopy voxel tissue composition

The probabilistic MEGA-PRESS voxel placement map can be seen in Figure 1A. The average tissue composition across all subjects was 53 ± 6%, 34 ± 8%, and 13 ± 6% for GM, WM, and CSF, respectively (Ctrl: 55 ± 4%, 33 ± 6%, and 12 ± 5%; PWHnopain 52 ± 6%, 35 ± 9%, and 13 ± 6%; PWHpain: 50 ± 8%, 35 ± 9%, and 15 ± 6%). There were no differences in MEGA-PRESS voxel tissue compositions between patients and controls (P’s > 0.19).

Figure 1.

(A)Average ¹H-MRS MEGA-PRESS voxel placement in the left posterior insula for all participants. (B) Example MEGA-PRESS spectrum including Glx and GABA. (C) GABA levels per group. (D) Glx levels per group. **P < 0.01. Ctrl, people without HIV and pain; GABA, γ-aminobutyric acid; Glx, glutamate/glutamine; i.u, institutional units; mmol/L, millimolar per liter; MRS, magnetic resonance spectroscopy; ppm, parts per million; PWHnopain, people with HIV without neuropathic pain; PWHpain, people with HIV with neuropathic pain.

3.3. Group differences in ¹H magnetic resonance spectroscopy metabolite levels

Example GABA and Glx spectra can be found in Figure 1B. When using linear regression to test for potential group differences in GABA levels, we observed a significant difference in GABA levels across groups (F(2,38) = 5.62, P = 0.007). Post-hoc tests revealed statistically lower GABA levels in PWHpain compared to Ctrl (P = 0.01) and PWHnopain (P = 0.02), but not between Ctrl and PWHnopain (P = 0.98) (Fig. 1C). This analysis was repeated, excluding the Ctrl and PWHnopain with a PainDETECT total score >1, yielding similar results as the aforementioned test (F(2,32) = 6.56, P = 0.0041). There were no significant differences in Glx levels across groups (F(2,35) = 1.27, P = 0.3) (Fig. 1D). For GABA/Cr ratios, see Supplementary Figure 2, http://links.lww.com/PAIN/C375.

To assess whether medication might influence GABA levels, we conducted a linear regression analysis comparing GABA levels across the 3 groups, with medication (potentially affecting GABA: yes/no) included as a factor. Consistent with our primary findings, we observed a significant difference in GABA levels between groups (F(2,36) = 5.42, P = 0.008). However, there was no significant effect of medication (P = 0.47) and no interaction between group and medication (P = 0.77).

3.4. Group differences in quantitative sensory testing

All participants received a deep-tissue pressure pain stimulus, using cuff pressure algometry over the left calf. There was no significant group difference in the average applied mmHg pressure that was subjectively calibrated to elicit a target pain intensity of ~40 on a 0 to 100 NRS (P = 0.29) (Fig. 2A).

Figure 2.

(A) Average calibrated pressure corresponding to a subjective pain rating of 40 on a 0 to 100 Numerical Pain Rating Scale. There were no significant group differences. (B) Group differences in average pain ratings (between 0 seconds and 120 seconds) during inflatable cuff temporal summation. (C) Group differences in pain ratings over time in response to mechanical cuff placed on participants’ left calf. (D) Significant association between GABA and temporal summation scores (normalized by cuff pressure, ie, mm Hg) for all PWH. mm Hg, millimeter mercury; s, time in seconds; #, significant differences between PWHpain and PWHnopain after adjustment for multiple comparisons; ##, significant differences between PWHpain and Ctrl after adjustment for multiple comparisons; *P < 0.05 surviving Bonferroni correction for multiple comparisons; **P < 0.01 surviving Bonferroni correction for multiple comparisons; GABA, γ-aminobutyric acid; PWH, people with HIV; PWHnopain, people with HIV without neuropathic pain; PWHpain, people with HIV with neuropathic pain; TS, temporal summation.

There were significant group differences in the average pain ratings throughout the cuff stimulation (P = 0.0065), with PWHpain rating their pain levels significantly higher than PWHnopain (P = 0.006). However, the group difference in the average pain ratings between PWHpain and Ctrl did not survive correction for multiple comparisons (P = 0.11), possibly because of an outlier data point in the Ctrl group. There were no significant group differences in the average pain ratings between PWHnopain and Ctrl (P = 1) (Fig. 2B).

Analyzing the pain ratings per group over time, we observed a significant effect of Group (F(2,36) = 5.54, P = 0.008), Time (F(4,143) = 4.04, P = 0.004), and a Group × Time interaction (F(8,143) = 4.56, P < 0.0001). Post-hoc tests did not reveal a significant difference between the 3 groups at 0 seconds (all P’s > 0.75) or 30 seconds (all P’s > 0.54). However, at 60 seconds, the groups started to diverge, revealing a significant difference between PWHpain and PWHnopain (P = 0.036), but not between PWHpain and Ctrl or PWHnopain vs Ctrl (all P’s > 0.24). At 90 seconds and 120 seconds, there was a significant difference between PWHpain and Ctrl and between PWHpain and PWHnopain (P’s < 0.004), but not between PWHnopain and Ctrl (P’s > 0.55) (Fig. 2C).

In a separate analysis, we investigated group differences in painful aftersensation (ie, the residual sensation of pain after the termination of the cuff stimulus). This analysis revealed significant Group (F(2,36) = 5.79, P = 0.0066) and Time effects (F(1,36) = 8.9, P = 0.005), whereas the Group 3 × Time interaction did not reach statistical significance (F(2,36) = 3.0, P = 0.061). Post-hoc tests revealed a difference that the 15 seconds post after-sensation pain ratings were significantly higher for PWHpain compared to Ctrl (P = 0.0015) and PWHnopain (P = 0.02), whereas the difference between PWHnopain and Ctrl was not statistically significant (P = 0.59). The difference between PWHpain and Ctrl in aftersensation pain ratings persisted for 30 seconds (P = 0.047), whereas there was no difference in the 30-second aftersensation pain ratings between PWHpain and PWHnopain (P = 0.13) or between Ctrl and PWHnopain (P = 0.87) (Fig. 2C).

Temporal summation scores normalized by mm Hg pressure (TS score/mm Hg) exhibited a significant effect of group (F(2,36) = 6.17, P = 0.005). Post-hoc test revealed a significant difference between PWHpain and Ctrl (P = 0.008) and between PWHpain and PWHnopain (P = 0.02), but not between Ctrl and PWHnopain (P = 0.88) (Fig. 2D).

3.5. Associations between metabolites, quantitative sensory testing, and pain reports

There were no significant correlations between either GABA or Glx and average pain during cuff stimulation, average cuff pain pressure (mm Hg), TS score (120s-0s), TS score normalized by mm Hg pressure (120s-0s)/mm Hg, or average painful after sensation within either PWH groups (P’s > 0.15). However, when collapsing the data of all patients into a single PWH groups (ie, irrespective of the presence or absence of pain), we observed significant associations between GABA and average pain ratings during cuff stimulation (R = −0.44, P = 0.02) (Fig. 3A), GABA and TS score (R = −0.49, P = 0.0096) (Fig. 3B), as well as GABA and TS score normalized by mm Hg (R = −0.52, P = 0.005).

Figure 3.

(A) Significant association between GABA levels and average pain ratings during inflatable cuff for all PWH. (B) Significant association between GABA and temporal summation scores for all PWH. GABA, γ-aminobutyric acid; PWH, people with HIV.

There were no significant correlations between neither GABA nor Glx and PainDETECT final, PainDETECT current pain, PainDETECT strongest pain past 4 weeks, PainDETECT average pain past 4 weeks within PWHpain (P’s > 0.34). Likewise, within PWHnopain, we observed no associations between GABA or Glx for the above-mentioned pain questionnaires (P’s > 0.45), potentially because of the narrow distribution of the PainDETECT scores in the PWHnopain group.

3.6. Group differences in circulating neuroactive steroid levels

One Ctrl was excluded from statistical analysis of allopregnanolone levels because of an out-of-range (ie, physiologically implausible) estimated concentration exceeding 100 pg/nL. We observed significant group differences in allopregnanolone levels (P = 0.024). Dunn post-hoc test revealed higher allopregnanolone levels in Ctrl compared to PWHpain (P = 0.044), whereas the comparison between PWHpain and PWHnopain did not survive correction for multiple comparisons (P = 0.13) (Fig. 4A). There were no group differences between Ctrl and PWHnopain (P = 1). There were no significant group differences in pregnenolone levels (P = 0.137).

Figure 4.

(A) Group differences in allopregnanolone levels. (B) Associations between allopregnanolone and Glx levels in PWH. Glx, glutamate/glutamine; PWH, people with HIV.

3.7. Associations between metabolites and neuroactive steroid levels

We observed a significant relationship between allopregnanolone and Glx in PWHpain (R = −0.64; P = 0.03), but not in PWHnopain (R = −0.03, P = 0.92) (Fig. 4B). No significant differences were observed when we statistically compared these correlation coefficients (Zdiff = 1.6, P = 0.11). No associations were observed between allopregnanolone and GABA in PWHpain or PWHnopain (R = −0.08, P = 0.8; R = 0.16, P = 0.68, respectively). Regarding pregnenolone, we observed a positive association with GABA in Ctrl (R = 0.8, P = 0.018), but not in PWHpain (R = 0.28, P = 0.38) or PWHnopain (R = 0.14, P = 0.72). There were no significant associations between Glx and pregnenolone in PWHpain, PWHnopain, or Ctrl (all R’s between −0.18 and 0.59, all P’s between 0.44 and 0.91).

4. Discussion

The current study used ¹H-MRS to investigate the concentrations of GABA and glutamate/glutamine in PWH with and without neuropathic pain compared to pain-free people without HIV. Our results demonstrate lower levels of GABA levels while Glx levels remained normal, in PWHpain compared to PWHnopain and Ctrl. This work further shows that lower levels of GABA were associated with increased sensitivity to evoked deep-tissue pain, and temporal summation score in PWH. In addition, we observed that the levels of allopregnanolone appear to decrease progressively from Ctrl to PWHnopain to PWHpain, with statistically significant allopregnanolone levels differences between Ctrl and PWHpain. Moreover, lower allopregnanolone levels were associated with higher glutamate/glutamine levels in PWHpain. Together, these findings support the hypothesis that a deficiency in the brain’s inhibitory neurotransmission is associated with HIV-related pain.

Homeostasis of the major inhibitory and excitatory (GABA and Glutamate, respectively) neurotransmitters ensures normal functioning of the brain and deviation of either system or their interaction are associated with a number of neurological diseases.²² Disruptions in the balance between inhibitory and excitatory neurotransmission have been suggested as a crucial factor in the formation of hyperalgesia or allodynia occurring in neuropathic pain.^27,36 Postmortem studies of PWH confirm significantly decreased GABAergic markers in the majority of the neocortex, neostriatum, and cerebellum, including pain-relevant brain regions such as the dorsolateral prefrontal cortex, anterior cingulate cortex, superior temporal lobe, and somatosensory cortex.¹¹ GABAergic interneurons constitute 20% of the total number of neurons in the neocortex.⁴⁰ We specifically chose to place the MRS volume of interest in the left posterior insular cortex as the insular cortex is known for its involvement in bodily interoception by integrating multimodal salient information.^9,15 With regard to the pain experience, the anterior part of the insula has been recognized for its association with affective-motivational aspects of sensory stimuli,² whereas the posterior part of the insula has been shown to encode sensory-discriminative aspects specific to the modality, location, and intensity of incoming sensory stimuli.^10,46,59 Preclinical studies have shown that creating an excitatory–inhibitory molecular imbalance within the insular cortex, either by increasing the levels of Glx or decreasing the levels of GABA, is sufficient to elicit mechanical allodynia.⁶² Similarly, electrical stimulation to the posterior insula can induce mechanical analgesia via modulation of GABAergic signaling in experimental neuropathic pain.³ In line with these observations, we observed reduced GABA concentration in the insula in PWHpain, which were proportional to the participants’ temporal summation scores and average pain ratings.

Elevated glutamate levels in the insula has been reported in various pain conditions.^21,28,34 Contrary to our hypothesis, however, we found that Glx metabolite levels remained unaltered in PWHpain. Although elevated extracellular levels of glutamate have been observed in cerebrospinal fluid and plasma in PWH,^4,12,20 the MRS literature on Glx in PWH is inconclusive. In PWH, some studies have reported decreased levels of glutamate in the frontal gray and white matter,^19,43,54 whereas others have reported increased levels of glutamate in the frontal gray matter.⁶⁶ A meta-analysis of 27 studies using ¹H-MRS PRESS sequence in PWH compared to Ctrl, where Glx were only reported in half of the studies, observed a trend towards elevated glutamatergic metabolite levels in the frontal white matter.¹⁶ Regarding neuropathic pain, previous studies have reported reduced Glx in the anterior cingulate cortex of people with neuropathic pain after spinal cord injury (SCI) with high pain compared to low pain⁶³ and lower thalamic Glx/Cr ratios in SCI with high pain compared to healthy controls.⁶⁴ On the other hand, neither of our PWH subgroups (with or without neuropathic pain) showed elevated Glx compared to Ctrl. Taken together, our results rather suggest a lack of inhibitory metabolites, rather than an excess of excitatory metabolites, may drive the pain sensitivity in HIV-related neuropathic pain.

A meta-analysis of 35 studies using ¹H-MRS suggest distinct GABA and Glx neurometabolic signatures across different pain conditions.⁴⁹ Migraine was linked to elevated glutamate and GABA levels possibly associated with cortical spreading depression, whereas chronic widespread pain syndromes like fibromyalgia exhibited an increase in Glx but no significant change in GABA. Musculoskeletal pain, on the other hand, demonstrated elevated glutamate levels, although further validation is required because of methodological inconsistencies. Unfortunately, neuropathic pain was included as part of the miscellaneous pain group and not extensively analyzed. Overall, across all pain conditions, Glx levels were consistently elevated, whereas GABA levels were only elevated in the migraine pain group.⁴⁹

Numerous studies have demonstrated the importance of steroid hormones in modulating neuronal excitability.³³ In the current study, we observed reduced levels of allopregnanolone in PWHpain compared to Ctrl, which were associated with levels of Glx in the posterior insular cortex. Although allopregnanolone primarily modulates GABA-A receptors, research suggests that its influence on excitatory neurotransmission may be indirect. For example, by regulating glutamate release through presynaptic GABA-A receptors³¹ and suppressing evoked glutamate release primarily via L-type calcium channels.³⁰ As these neurotransmitters are crucial for excitation and inhibition of sensory information, it is perhaps not surprising that allopregnanolone exhibits neuroprotective and analgesic properties and could be a valuable future target for treating neuropathic pain.^25,26,33 Indeed, increasing the production of allopregnanolone in the spinal cord and peripheral nerves have proven successful in several neuropathic, inflammatory, and postoperative experimental chronic pain models.^25,26,33,42 For example, allopregnanolone has been found to successfully alleviate mechanical and thermal hyperalgesia in rodent models of neuropathic pain by modulating T-type calcium channels and GABA_A receptors.⁴⁷ In the dorsal horn, there is evidence of direct inhibition of allopregnanolone synthesis with an increase of the potent pronociceptive neuropeptide substance P.⁴⁸ Mechanistically, downregulated production of allopregnanolone reduces GABA-A receptor activity and increases primary afferent release of substance P.³³ Hence, it is perhaps not surprising that in our study we observed a significant association between the lower levels of allopregnanolone and levels of the excitatory neurotransmitter glutamate in PWHpain. Increasing the production of endogenous neuroactive steroids such as allopregnanolone in the spinal cord and peripheral nerves deserves further attention, as it might offer interesting opportunities to develop effective novel neuroactive steroid-based therapies against neuropathic pain. Although the current study focuses solely on neuropathic pain, additional work is needed for headache, chronic musculoskeletal pain, and “fibromyalgianess” in PWH.^38,41,52

When interpreting the results of the current work, several limitations should be considered. These include a relatively small sample size in each cohort, as well as the small representation of females in general (and the lack of females in the Ctrl group). Therefore, future, larger studies will be needed to confirm and extend the results of the current work. As we were only able to investigate the metabolite alterations in the insula, we cannot comment on whether the observed changes are regionally specific or reflective of broader neurochemical alterations. In addition, although we did our best to target the posterior part of the insular cortex, we acknowledge that the size of the voxel has likely encompassed neighboring regions. However, larger voxels result in better SNR and were necessary to get reliable GABA level estimates.

In conclusion, our results demonstrate lower levels of the inhibitory neurotransmitter GABA in the left posterior insula in PWHpain. Further, the lower levels of inhibitory metabolites were associated with increased pain sensitivity in PWH. Speculatively, our findings may suggest an imbalance of less inhibitory relative to excitatory metabolites, which may drive the system to hyperexcitability, possibly augmenting neuropathic pain symptoms in PWH. Future studies should aim to investigate whether PWH with neuropathic pain may benefit from pharmacologically restoring the Glu-GABA homeostasis.

Supplementary Material

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C375.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (www.painjournalonline.com).

Data availability statement:

The data that support the findings of this study are available from the corresponding authors, M.L.L. and E.M.R., upon reasonable request.