Effects of N-methyl-d-aspartate receptor modulation on brain glutamate, functional connectivity, and clinical symptoms in fibromyalgia

Supplemental Digital Content is Available in the Text.

A novel NMDAR modulator reduced pain-related brain glutamate levels and connectivity in patients with fibromyalgia, with clinical improvements supporting continued therapeutic investigation.

Abstract

Objective:

This exploratory, 2-site phase 2 clinical trial investigated whether neuroimaging biomarkers could demonstrate the analgesic potential of NYX-2925, a novel N-methyl-d-aspartate receptor modulator, in women with fibromyalgia.

Methods:

Twenty-two participants completed sequential 2-week treatment periods: placebo, 20 mg NYX-2925, and 200 mg NYX-2925. Resting state functional connectivity magnetic resonance imaging and proton magnetic resonance spectroscopy were performed during the second week of each period.

Results:

The 200 mg dose produced statistically significant improvements in pain, fatigue, fibromyalgia severity, and function compared with placebo. Neuroimaging revealed that NYX-2925 reduced glutamate and glutamine/total creatine ratios in the dorsal anterior cingulate cortex at rest (P = 0.032) and in the posterior insular cortex after evoked pain (P = 0.039). Notably, higher baseline pain-evoked glutamate and glutamine levels in the pIns predicted greater reductions in experimental pain sensitivity at both the 20 mg (r = 0.55; P = 0.029) and 200 mg (r = 0.51; P = 0.046) doses. The 20 mg dose also reduced resting state connectivity between the dorsal anterior cingulate cortex and both primary somatosensory cortex (P = 0.030 family-wise error) and precuneus (P = 0.031 family-wise error). NYX-2925 demonstrated good tolerability and safety.

Conclusion:

These exploratory findings suggest that NYX-2925 modulates excitatory neurometabolic activity and functional connectivity in pain-processing networks, with modest accompanying clinical improvement. A subsequent larger randomized trial did not confirm clinical benefit, so the present results should be regarded as hypothesis-generating and require independent replication. We discuss the value, and the limits, of neuroimaging measures for mechanistic drug development and personalized pain medicine.

1. Introduction

Chronic pain affects over 24.3% of Americans,⁵⁴ with more than 8.5% experiencing severe pain that significantly impairs daily functioning.⁶⁴ Existing pain treatments demonstrate limited efficacy, requiring 6 to 15 patients to be treated for one to benefit, while also carrying substantial risks of side effects and abuse potential.^{15,22,25,41,43,57} These limitations have contributed to large-scale opioid misuse,⁶ which underscores the critical need for effective, nonopioid pain medications.

Traditional drug development timelines of 10 to 15 years and costs exceeding $2 billion US dollars^19,84 are inadequate given the impact of opioid misuse and the anticipated surge in chronic pain prevalence as the population ages.^51,86 A paradigm shift in the process by which novel compounds are identified and evaluated is required. For compounds targeting the central nervous system (CNS), neuroimaging offers a promising approach to accelerate development by providing early evidence of clinical efficacy through neural biomarkers.^18,21,75 Benefits include smaller clinical trial sample sizes due to larger effect sizes in imaging outcomes and the ability to evaluate drug mechanisms directly in human subjects rather than relying solely on preclinical models.²¹

Two neuroimaging techniques show particular promise for analgesic drug development: proton magnetic resonance spectroscopy (¹H-MRS)—used to measure concentrations of brain metabolites in vivo—and resting state functional connectivity magnetic resonance imaging (rs-fcMRI). Our previous research demonstrated the predictive value of these approaches. In patients with fibromyalgia, higher baseline levels of excitatory neurotransmitters (glutamate and glutamine, collectively termed Glx) in the insula and increased insular connectivity to the Default Mode Network (DMN) predicted superior analgesic response to pregabalin.³⁴ Pain improvement correlated with normalization of these aberrant neural patterns, suggesting their direct contribution to the subjective pain experience.³⁶ Similarly, reduced resting connectivity to pain-inhibiting brain regions predicted responsiveness to milnacipran, a serotonin-norepinephrine reuptake inhibitor (SNRI) approved for fibromyalgia.⁷⁰

xxxxxxxxNYX-2925 is a novel N-methyl-d-aspartate receptor (NMDAR) positive allosteric modulator developed for the treatment of fibromyalgia. N-methyl-d-aspartate receptors serve as glutamate binding sites and play crucial roles in both normal physiology and pathological processes, including chronic pain.^58,62 NYX-2925 binds to all 4 NMDAR2A-D subtypes,⁴⁶ distinct from NMDAR agonists and antagonists previously studied as chronic pain treatments including d-cycloserine, ketamine, MK-801, and kynurenic acid.⁴⁸ In vitro studies demonstrate that NYX-2925 influences NMDAR-mediated neuronal plasticity by modulating ion channel calcium influx and long-term potentiation^9,46—processes implicated in chronic pain development.⁵⁰ Preclinical studies confirmed behavioral antinociceptive effects,^29,31,59 and a phase 1 clinical trial established that NYX-2925 was safe, well-tolerated, and crossed the blood–brain barrier in healthy humans.⁴²

Here we describe an exploratory phase 2, placebo-controlled, single-blind, clinical neuroimaging trial examining the effects of NYX-2925 in women with fibromyalgia. Fibromyalgia is considered the archetypal nociplastic pain condition,^26,44 characterized by CNS-mediated amplified and dysregulated pain and sensory perception,^32,70 elevated Glx levels on ¹H-MRS,^35,36 and increased functional connectivity within neural circuitry implicated in pain processing.^45,61 Our primary objective was to determine whether the effects of NYX-2925 on Glx levels and resting state functional connectivity support its analgesic potential and justify continued clinical development. Given the pharmacological action of NYX-2925 on NMDAR, we hypothesized, a priori, that NYX-2925 would reduce Glx levels within and modulate functional connectivity between brain areas involved in pain processing, including the insular and anterior cingulate cortices.

2. Methods

2.1. Sample

Study participants were right-handed women aged 18 to 70 meeting 2010 American College of Rheumatology fibromyalgia diagnostic criteria,⁸³ with baseline pain scores ≥4/10 on a numeric pain rating scale. Pre-existing fibromyalgia treatments required stable dosing for 14 days before and during the study. Nonsteroidal anti-inflammatory drugs were prohibited 24 hours before MRI. Exclusions included: use of opioids, narcotic analgesics, monoamine oxidase inhibitors, anticonvulsants (except gabapentinoids), benzodiazepines, sedatives, hypnotics, and NMDAR ligands; conditions causing pain or untreated endocrine disorders that may have confounded fibromyalgia pain assessment; significant psychiatric disorders, cognitive impairment, substance abuse, hepatic/renal impairment, cardiovascular disease, malignancy, or infectious disease; MRI contraindications; pregnancy or breastfeeding.

2.2. Study design

Participants were recruited from the University of Michigan (Ann Arbor, MI) and the University of Cincinnati/Cincinnati Children's Hospital Medical Center (Cincinnati, OH). After a 30-day screening period, participants entered a fixed-sequence, single-blind treatment: 2 weeks placebo, 2 weeks 20 mg NYX-2925, then 2 weeks 200 mg NYX-2925. Daily average and worst pain were collected from time of enrollment to end of study.

Data collected at screening, during week 2 of each dosing period, and at follow-up included: (1) patient-reported outcome measures (PROMs) for pain, function, fatigue, sleep, depression, and anxiety; (2) ¹H-MRS in dorsal anterior cingulate cortex (dACC), perigenual anterior cingulate cortex (pgACC), right anterior insula cortex, right posterior insula cortex (pIns; measured before and after painful stimulation), precuneus, thalamus, and medial prefrontal cortex (mPFC); (3) seed-based, region-of-interest, rs-fcMRI analyses for these same ¹H-MRS voxels; and (4) quantitative sensory testing (QST). Study design shown in Figure 1.

Figure 1.

Study design. Participants completed a baseline MRI session during a 30-day screening period before starting placebo. Placebo (matching capsules) and NYX-2925 (20 and 200 mg) were administered orally, once-daily, for 2-week (W) periods (W1–6). A follow-up visit occurred at W7. Average and worst numerical pain intensity scores were captured daily. Patient-reported outcomes, MRI (including resting state functional connectivity MRI and proton magnetic resonance spectroscopy), and sensory sensitivity measured by quantitative sensory testing, were assessed during screening, the second week of each dosing period, and at W7. Participants and data analysts were blinded to dosing sequence. MRI, magnetic resonance imaging.

The primary study objective was to determine whether NYX-2925 (20 or 200 mg) alters CNS pain processing markers. The secondary objective was to evaluate NYX-2925 safety. Exploratory objectives assessed effects on self-reported pain, function, fibromyalgia symptoms, and QST. Primary comparisons were week 4 (20 mg) and week 6 (200 mg) vs week 2 (placebo). Participants were blinded to the drug/dose sequence, and neuroimaging analysts were similarly blinded during analysis. Ethics review and approval was provided by University of Michigan (HUM00128061) and University of Cincinnati (00003152) Institutional Review Boards.

2.3. Patient-reported outcome measures

Numeric pain rating scale: 0 to 10 scale where 0 = no pain and 10 = worst imaginable pain.^13,23 Participants reported daily average and worst pain through a telephone-based system. Scoring criteria for this and all PROMs are provided in Supplementary Methods and Materials (available at https://links.lww.com/PR9/A439).

Brief Pain Inventory (BPI): 7-item pain interference subscale assessing impact on activities, mood, walking, work, relationships, sleep, and enjoyment.^16,20 Higher scores indicate greater pain interference.

Patient-Reported Outcomes Measurement Information System Fibromyalgia Fatigue Profile: 16-item instrument with four 4-item short forms examining the impact of fatigue in social, cognitive, and motivation subdomains.⁴⁷ All 16 items use a 5-point Likert scale (1–5), with different categorical responses. Lower scores indicate less fatigue.

Revised Fibromyalgia Impact Questionnaire: 21 items assessing physical functioning, work status, depression, anxiety, morning tiredness, pain, stiffness, fatigue, and wellbeing over 1 week through 21 questions using an 11-point Likert scale (0–10), with lower scores representing a better quality of life.⁸

Fibromyalgia Survey Criteria: 31 item instrument composed of 2 complementary subscales. Widespread Pain Index (0–19, extrapolated from the Michigan Body Map¹¹) assesses pain distribution across 19 body locations. Symptom Severity score (0–12) evaluates 3 cardinal symptoms (fatigue, cognitive difficulties, unrefreshing sleep) plus 3 somatic symptoms (abdominal pain, depression, headache).⁸²

Patient-Reported Outcomes Measurement Information System Sleep Disturbance-Short Form: 8 items assessing sleep disturbance (1–5).⁷⁷ Higher scores indicate greater sleep interference.

Hospital Anxiety and Depression Scale: self-assessment scale for depression (7 items, 0–21) and anxiety (7 items, 0–21).⁸⁵ Each item scored 0 to 3 based on past week experience. Higher scores indicate greater depression/anxiety.

2.4. Magnetic resonance imaging

Participants completed a 90-minute scanning session at each MRI visit (Fig. 1). Heads were secured using foam padding and a forehead strap. The protocol included (Fig. S1, available at https://links.lww.com/PR9/A439): (1) 6-minute rs-fcMRI scan with eyes open, fixating on a crosshair presented on a presentation screen³³; (2) T1 spoiled gradient echo/magnetization-prepared rapid gradient echo or fast Fourier echo structural scan; (3–5) ¹H-MRS of the dACC (midline), right anterior insula cortex, and right pIns (pre-evoked pain); (6) 12-minute evoked pressure pain paradigm at the left gastrocnemius using an MRI-compatible cuff; (7–11) ¹H-MRS of the right pIns (post-evoked pain), pgACC (midline), precuneus (midline), mPFC (midline), and thalamus (midline); (12) 3.5-minute evoked visual stimulus fMRI scan.³⁸ Voxel dimensions, locations, and coordinates shown in Figure S2 (available at https://links.lww.com/PR9/A439). Only ¹H-MRS and rs-fcMRI data for placebo, 20 and 200 mg NYX-2925 visits are reported here. Magnetic resonance imaging data acquisition, site qualification, harmonization, and quality control protocols are detailed in Supplementary Methods and Materials (available at https://links.lww.com/PR9/A439).

2.4.1. Proton magnetic resonance spectroscopy preprocessing

Data were preprocessed with LCMmodel (Stephen Provencher, Oakville, ON, Canada).⁶⁷ Raw spectra were fit with a linear combination of premetabolite spectra (basis set simulated by Stephen Provencher included: Ala, Asp, Cr, PCr, GABA, Glc, Gln, Glu, GPC, PCh, Ins, Lac, NAA, NAAG, Scyllo, and Tau). Because patients with fibromyalgia show increased Glx levels,³⁵ we focused on these metabolites. Glutamate and glutamine levels were calculated as ratios to creatine (Cr) as an internal standard and as concentrations rescaled using the water peak. Concentrations were cerebrospinal fluid-corrected using voxel-based morphometry in SPM12³⁵ and reported in arbitrary institutional units. Spectra with poor quality or Glx Cramer-Rao bounds >20% were excluded (Fig. S3, available at https://links.lww.com/PR9/A439).

2.4.2. Functional magnetic resonance imaging preprocessing

Data were checked for artifacts and motion was assessed by examining 3 translations (x, y, and z) and 3 rotations (pitch, roll, and yaw). Participants with mean absolute motion exceeding ±2 mm translation or ±1° rotation in any direction were excluded, resulting in 18 participants for resting state analyses (Fig. S3, available at https://links.lww.com/PR9/A439). Resting state fMRI data were preprocessed using FSL (http://www.fmrib.ox.ac.uk/fsl/) and SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software) on MATLAB R2017a (http://www.mathworks.co.uk/products/matlab/). Slice timing correction was performed using FSL. Motion correction, coregistration of structural and functional images to baseline T1, normalization to Montreal Neurological Institute space, and smoothing with 8 mm full width at half maximum Gaussian kernel were performed in SPM12. Preprocessed images were entered into Conn Toolbox 18a,⁸¹ and nuisance regression using aCompCor⁷ was performed with 6 participant-specific realignment parameters, white matter and cerebrospinal fluid signals, and their first-order derivatives included as confounds. Temporal filtering (0.008–0.09 Hz) was applied to focus on low-frequency fluctuations²⁸ by removing linear drift artifacts and high-frequency noise.

2.5. Quantitative sensory testing

2.5.1. Pressure pain sensitivity

Participants underwent QST after familiarization training. Thumbnail pressure sensitivity was measured using a multimodal automated sensory testing system³⁹ with computer-controlled ascending pressures, yielding 3 outcomes: Thumb-PPT (pressure pain detection threshold), Thumb-Pain50 threshold (pressure eliciting moderate pain), and Thumb-tolerance threshold, as previously reported.^32,40,63 Additional pressure pain thresholds were assessed bilaterally at patella, wrist, and trapezius using a handheld digital algometer. Large-volume deep tissue pain was evaluated at the left gastrocnemius using an MRI-compatible pressure cuff to determine Cuff-Pain50 thresholds.^52,65,66

2.5.2. Visual sensitivity

The Michigan Visual Aversion Stress Test assessed nonsomatosensory sensory processing pathways that bypass peripheral receptors and the spinal cord—mechanisms characteristically amplified in nociplastic chronic pain conditions including fibromyalgia.^{30,38,53,56,78} Complete QST methodology and fMRI-adapted protocols are detailed in Supplementary Methods and Materials (available at https://links.lww.com/PR9/A439).

2.6. Statistical analysis

2.6.1. Clinical and behavioral data

Participant characteristics, PROMs, and QST data were summarized using descriptive statistics (n, mean, SD, median, minimum, maximum). Between-treatment differences were assessed using paired-sample t tests, comparing pre- and post-treatment values. Analyses were conducted using SAS 9.4+ (Cary, NC) and SPSS 25+ (IBM, Armonk, NY). Significance was set at P ≤ 0.05 (2-tailed). The sample size for this pilot study was based on practical and clinical considerations rather than a formal power determination. Multiple comparison corrections were not applied to the clinical/behavioral data, which were exploratory. The correlation analyses described below were likewise exploratory and were not corrected for multiple comparisons. Given the modest sample size, these exploratory analyses should be regarded as hypothesis-generating. Missing visits were not imputed. Each paired comparison used all participants with data available at both timepoints. Because the comparisons were within-participant and few observations were missing, imputation offered limited benefit and risked introducing bias. All analyses were prespecified in the Statistical Analysis Plan.

2.6.2. Proton magnetic resonance spectroscopy analysis

Glutamate and glutamine was calculated as concentrations and Glx/Cr ratios, with pIns Glx reduction as the a priori primary outcome. Paired-sample t-tests compared change from placebo with 20 mg and 200 mg NYX-2925. For all voxels except pIns, 1 value per timepoint was analyzed. For pIns, we tested evoked pain Glx (post- minus precuff pain) at placebo, 20 mg, and 200 mg, plus change scores across treatment transitions. Tests were Bonferroni-corrected for multiple comparisons. Pearson partial correlations (age-adjusted) assessed relationships between: (1) changes in pIns Glx (post- minus precuff) and changes in evoked cuff pain (from QST) for both treatment transitions; (2) resting Glx levels and QST metrics (Thumb-PPT, -Pain50, -Tolerance; Cuff-Pain50; bilateral patella-PPT) across all timepoints.

2.6.3. Functional magnetic resonance imaging analysis

Seed-to-whole brain region of interest (ROI) functional connectivity maps were generated in Conn Toolbox for each ¹H-MRS voxel (Fig. S2, available at https://links.lww.com/PR9/A439). First-level maps correlated ROI time series with whole-brain voxel signals, producing bivariate Fisher z-transformation correlation maps (1 per ROI/visit/individual). Second-level paired t-tests used general linear models with scanning site as a covariate of no interest in SPM12, comparing: (1) placebo vs 20 mg, and (2) placebo vs 200 mg. Correlations between connectivity and clinical outcomes were performed. Results were deemed significant at a cluster-level family-wise error (FWE) corrected P-value ≤0.05, derived from an uncorrected voxel level P-value ≤0.001.

3. Results

3.1. Sample characteristics

Twenty-two female participants (mean age 47.3, range 20–66 years) were enrolled; 86.4% were White, 4.5% Hispanic/Latino (Fig. S3 and Table S1, available at https://links.lww.com/PR9/A439). The median fibromyalgia duration was 8.8 years (range 1.4–52.3 years). Medication use is reported in Tables S2–S4 (available at https://links.lww.com/PR9/A439).

3.2. NYX-2925 improved pain, fatigue, and quality of life

At 200 mg, NYX-2925 significantly improved average daily pain (19%, P = 0.007), worst daily pain (12%, P = 0.036), quality of life (16%, P = 0.014), and fatigue (9%, P = 0.005) vs placebo (Fig. 2, Table S5, available at https://links.lww.com/PR9/A439). Fibromyalgia Symptom Severity decreased (−0.9, P = 0.026), with a marginal reduction in pain extent on the Widespread Pain Index (−1.3, P = 0.054). The 20 mg dose showed nonsignificant improvements in these outcomes. No significant changes occurred in BPI, Patient-Reported Outcomes Measurement Information System Sleep, or Hospital Anxiety and Depression Scale scores after either dose (Table S6, available at https://links.lww.com/PR9/A439).

Figure 2.

Clinical effects of NYX-2925. Compared with placebo, 200 mg NYX-2925 resulted in significant improvements on (A) average and (B) worst daily pain, (C) fibromyalgia symptom impact as measured by the FIQR, and (D) fatigue. Smaller mean improvements in these outcomes from placebo were generally observed after treatment with 20 mg NYX-2925, but these effects did not reach statistical significance. On each of these symptom measures, the clinical benefits observed at week 6 with 200 mg NYX-2925 were diminished at the follow-up visit at week 7. Error bars reflect SEM. Placebo: n = 22; 20 mg: n = 22; 200 mg: n = 21; follow-up: n = 20. P < 0.05 considered significant (uncorrected). FIQR, Revised Fibromyalgia Impact Questionnaire; PBO, placebo; PROMIS_FM, Patient-Reported Outcomes Measurement Information System Fibromyalgia.

3.3. NYX-2925 decreased resting dorsal anterior cingulate cortex glutamate and glutamine levels

The 20 mg dose of NYX-2925 significantly reduced resting dACC Glx levels (P = 0.017 absolute), with a marginal decrease when creatine-normalized (P = 0.032), relative to placebo (Fig. 3, Tables S7 and S8, available at https://links.lww.com/PR9/A439). The 200 mg produced no significant changes in dACC Glx levels. No significant Glx changes occurred in other brain regions.

Figure 3.

NYX-2925 decreased resting levels of Glx in dorsal anterior cingulate cortex. Proton magnetic resonance spectroscopy revealed decreased resting levels of Glx (measured as ratios to creatine [left panel] and as concentrations [right panel]) in dorsal ACC after treatment with 20 mg NYX-2925 (NYX) relative to placebo. P < 0.025 considered significant (Bonferroni-corrected). ACC, anterior cingulate cortex; AIU, arbitrary institutional unit; Glx, glutamate plus glutamine.

3.4. NYX-2925 decreased pain-evoked posterior insular cortex glutamate and glutamine levels

During placebo treatment, painful cuff pressure applied to the gastrocnemius significantly elevated pIns Glx levels (both P ≤ 0.013; Fig. 4A). Both NYX-2925 doses eliminated this response (Tables S9 and S10, available at https://links.lww.com/PR9/A439). Direct comparison with placebo revealed that neither NYX-2925 dose produced significant reductions in pain-evoked Glx changes (calculated as post- minus prepain levels) within the pIns (Fig. 4B). However, the 200 mg dose marginally reduced postpain pIns Glx vs placebo (P = 0.039, creatine-normalized only; Fig. 5).

Figure 4.

NYX-2925 decreased pain-evoked levels of Glx in posterior insula. (A) At the placebo visit (week 2) before drug administration, painful cuff pressure applied to the gastrocnemius muscle during fMRI significantly increased posterior insula Glx levels (measured as ratios to creatine and as concentrations). (B) When compared with placebo, there were marginal decreases in pain evoked Glx levels in posterior insula with both doses of NYX-2925. P < 0.017 considered significant (Bonferroni-corrected). AIU, arbitrary institutional unit; fMRI, functional magnetic resonance imaging; Glx, glutamate plus glutamine; tCR, total creatine.

Figure 5.

NYX-2925 reduced postpain Glx levels in the posterior insula. Compared with placebo, there was a marginally significant decrease in posterior Glx levels (ratios to creatine only) following evoked pressure pain with 200 mg NYX-2925. P < 0.025 considered significant (Bonferroni-corrected). Glx, glutamate plus glutamine; tCR, total creatine.

3.5. Relationship between pain sensitivity and posterior insular cortex glutamate and glutamine differed on placebo vs NYX-2925

NYX-2925 did not directly affect evoked pain or visual sensitivity (Tables S11 and S12, available at https://links.lww.com/PR9/A439). However, on placebo, lower pressure pain thresholds correlated with increased pain-evoked pIns Glx levels. NYX-2925 reversed this relationship: higher pain thresholds associated with increased Glx changes at 20 mg, with progressively stronger effects at 200 mg (Fig. 6, Table S13, available at https://links.lww.com/PR9/A439). No dACC correlations were observed (Table S14, available at https://links.lww.com/PR9/A439).

Figure 6.

Differential relationship between posterior insula Glx levels and pressure pain thresholds. Pain evoked Glx concentrations in posterior insula shift from pronociceptive action to antinociceptive action with NYX-2925 treatment. After PBO administration (gray line and circles), higher pain evoked Glx concentration changes are associated with lower pain thresholds (increased pain sensitivity) at the leg (top panel) and thumb (bottom panel). An opposite relationship emerges after NYX-2925 administration (blue line and circles): higher pain evoked Glx concentration changes are associated with higher pressure pain thresholds (decreased pain sensitivity) at both sites. When Glx concentration is plotted as a continuum across all treatments, the highest pain thresholds center around an intermediate level of Glx, whereas higher and lower levels of Glx (shaded areas) are associated with lower pressure pain thresholds. Each participant contributed 3 data points (placebo, 20 mg NYX 2925, and 200 mg NYX 2925). The Gaussian curve is fitted for descriptive purposes to illustrate the apparent nonlinear trend and was not formally tested for significance. P < 0.05 considered significant (uncorrected). AIU, arbitrary institutional unit; Glx, glutamate plus glutamine; PBO, placebo; tCR, total creatine.

3.6. Predrug pain induced glutamate and glutamine changes predicted NYX-2925 analgesic effect

Higher levels of pain-evoked pIns Glx changes during placebo were associated with increases in cuff-Pain50 thresholds after both NYX-2925 doses (all r ≥ 0.51, P ≤ 0.046; Fig. 7).

Figure 7.

Predrug Glx level predicted subsequent NYX-2925 analgesic effect. Pain evoked change in Glx levels (measured as ratios to creatine and as concentrations) in posterior insula during placebo treatment were correlated with changes in cuff Pain50 thresholds from placebo to 20 mg and to 200 mg NYX-2925. Higher pain induced Glx changes during placebo were associated with increased pain thresholds (decreased pain sensitivity) during NYX-2925 treatments. P < 0.05 considered significant (uncorrected). AIU, arbitrary institutional unit; Glx, glutamate plus glutamine; tCR, total creatine.

3.7. NYX-2925 reduced resting-state brain connectivity

Dorsal anterior cingulate cortex: 20 mg NYX-2925 (placebo >20 mg) decreased connectivity with right superior temporal gyrus, left premotor/primary somatosensory cortex, right precuneus, and left precentral gyrus/premotor cortex (all P ≤ 0.049 FWE; Fig. 8, Table S15, available at https://links.lww.com/PR9/A439). Reduced dACC-precuneus connectivity correlated with improved BPI pain interference (r = 0.472, P = 0.048).

Figure 8.

Changes in resting state functional connectivity magnetic resonance imaging after treatment with NYX-2925. Decreased functional connectivity between a dACC seed and several brain regions were observed between week 2 (placebo) and week 4 (20 mg NYX-2925) visits (placebo >20 mg NYX-2925). These decreases were found between dACC and: right STG; left premotor cortex/primary somatosensory cortex (Premotor/SI); right precuneus; and left precentral gyrus/premotor cortex. P < 0.05 considered significant (family wise error corrected). dACC, dorsal anterior cingulate; STG, superior temporal gyrus.

Perigenual anterior cingulate cortex: 20 mg NYX-2925 decreased connectivity with left precuneus/precentral gyrus (P < 0.001 FWE); at 200 mg (placebo >200 mg) with right inferior frontal gyrus (P = 0.008 FWE; Table S16, available at https://links.lww.com/PR9/A439).

Medial prefrontal cortex: 200 mg decreased connectivity with right anterior prefrontal cortex (P = 0.003 FWE) and right midcingulate cortex (P = 0.036 FWE; Table S16, available at https://links.lww.com/PR9/A439).

3.8. NYX-2925 was safe and well-tolerated

No discontinuations or serious adverse events occurred. Common adverse events: abdominal distension (n = 3); anxiety, diarrhea, eructation, tension headache, upper respiratory infection (n = 2 each). See Safety and Adverse Events Data in Supplementary Methods and Materials (available at https://links.lww.com/PR9/A439) for additional details.

4. Discussion

This exploratory study revealed several antinociceptive effects of NYX-2925 on neuroimaging measures. NYX-2925 showed dose-specific effects on glutamate signaling: 20 mg reduced basal Glx levels in dACC, while both doses attenuated pain-evoked Glx in pIns. Participants with higher pain-evoked pIns Glx during placebo showed greater pressure pain threshold increases following NYX-2925. These findings suggest ¹H-MRS measurements could serve as neural markers for personalizing analgesic therapy, particularly for identifying chronic pain patients with glutamatergic dysregulation who may be more treatment-responsive. NYX-2925 also modulated resting-state connectivity, decreasing pain-associated functional connectivity from dACC, pgACC, and mPFC. Clinical assessments revealed modest improvements in daily pain, fatigue, and fibromyalgia severity after 200 mg NYX-2925. These convergent findings provided rationale for advancing to phase 2b trials.

Traditional analgesic development relying on animal models has largely failed, with many promising preclinical compounds failing phase 3 trials.⁵⁵ This led us and others to propose using smaller phase 2 mechanistic trials for “go/no-go” decisions.²¹ Molecular neuroimaging techniques like ¹H-MRS are ideal for this approach, providing noninvasive measurement of molecules influencing neural activity, including glutamate and GABA—both implicated in fibromyalgia pain.^1,27,35,36 Our previous pregabalin ¹H-MRS/rs-fcMRI study established that analgesics could modify neural markers.³⁴ The present work tests whether this approach works for novel compounds, showing NYX-2925 largely reduces dACC and pIns Glx as hypothesized, with predrug insular Glx predicting analgesic action on evoked-pain.

Despite similar effects on Glx, NYX-2925 and pregabalin have very different mechanisms of action. Pregabalin binds with high affinity to α2δ-1 subunit of voltage-gated Ca²⁺ channels, where it presumably reduces presynaptic Ca²⁺ influx and glutamate release.⁷³ By contrast, NYX-2925 augments NMDAR signaling.⁹ Animal studies showed NYX-2925 improved novel object recognition and emotional learning, reversible by the NMDAR antagonist CPP.⁴⁶ Interestingly, the NMDAR antagonist ketamine elevated glutamate in animal mPFC¹⁴ and glutamine in human ACC,⁶⁸ suggesting NMDAR antagonism increases excitatory transmission. Consistent with these findings, the NMDAR antagonist memantine increased insular levels of Glx and improved cognition and depression, but not pain, in fibromyalgia.²⁴ Glutamate and glutamine reduction by NYX-2925 is therefore consistent with NMDAR activation.

The pain-evoked Glx relationship with pain sensitivity reversed during NYX-2925 vs placebo. During placebo, pain-induced Glx changes negatively correlated with pressure pain thresholds—increased Glx levels during painful pressure were associated with lower thresholds (ie, increased pain sensitivity), consistent with research showing increased insular glutamate relates to increased pain perception.^35,37,79 However, after 200 mg NYX-2925, this relationship flipped: increased Glx correlated with higher pain thresholds or reduced pain sensitivity. Intermediate pain-evoked Glx levels were associated with the greatest reductions in pain sensitivity, whereas higher and lower levels were associated with lower thresholds. This Gaussian-like distribution suggests optimal excitatory neurotransmitter balance in pIns is required for antinociception.

The underlying mechanism for such a relationship remains unknown, but at least 2 possibilities exist. First, at low glutamate levels, there may be inadequate neurotransmitter to activate descending pain inhibition pathways. A minimal pain-evoked increase in Glx may be required to activate these inhibitory pathways in the presence of an NMDAR sensitizing compound such as NYX-2925. Second, NYX-2925 may shift NMDAR function from long-term potentiation to homeostatic plasticity, as observed in animal models,^9,46 whereby the abnormally elevated pain-evoked Glx levels typical in fibromyalgia become normalized rather than simply enhanced or suppressed.⁴² Given that disrupted homeostatic plasticity has been identified in chronic pain patients,⁷⁴ restoring this balance may be central to NYX-2925's analgesic effects. Regardless of the precise mechanism, this centering effect on Glx levels aligns with existing evidence that homeostatic balance of key metabolites is essential for normal biological function,¹⁷ as both excessive and insufficient Glx may disrupt homeostasis and increase pain sensitivity. To the best of our knowledge, this is the first time such a homeostatic relationship has been demonstrated in humans with chronic pain.

Seed-based rs-fcMRI was performed for the same brain regions used to acquire ¹H-MRS data. The dACC, a pronociceptive brain region,⁶⁹ showed marked connectivity decreases to precuneus (a DMN region) between placebo and 20 mg NYX-2925, associated with decreased patient-reported pain interference. The dACC also showed Glx decreases with 20 mg. Connectivity between dACC and 2 other pronociceptive regions, premotor cortex/primary somatosensory cortex and superior temporal gyrus, also decreased with 20 mg NYX-2925. At 200 mg, NYX-2925 reduced mPFC connectivity to anterior prefrontal cortex and pronociceptive midcingulate cortex. These latter 2 areas show reduced pain induced activation in chronic back pain patients successfully treated with reprocessing psychotherapy.⁵ The mPFC is also a component of the DMN and may contribute to psychological comorbidities associated with chronic pain.⁴⁹ In rat, intracerebral microinjection of NYX-2925 into mPFC restored NMDAR signaling and produced antinociceptive effects, further suggesting an important role of mPFC in the action of NYX-2925.^31,59 It is notable that in this study, only 200 mg NYX-2925 reduced mPFC connectivity and this was also the dose that showed clinical effects. These results are consistent with previous reports showing decreased connectivity between pronociceptive regions, as well as a decoupling of pronociceptive regions to the DMN, after analgesic treatment.^34,60,70 The DMN is a constellation of brain regions engaged in self-referential cognition, which are deactivated during various externally focused tasks.^12,28 We previously reported increased connectivity between the DMN and insula—a region critical for multimodal pain and sensory integration¹⁰—in patients with fibromyalgia. Greater clinical pain was correlated with greater DMN-insula connectivity,⁶¹ which was diminished after administration of pregabalin.^34,60 Overall, findings presented here suggest NYX-2925 decreases excitatory neurotransmitter levels in pronociceptive regions with concomitant connectivity reductions.

Patients with fibromyalgia have limited pharmacological options. FDA-approved treatments include pregabalin, cyclobenzaprine, and SNRIs duloxetine and milnacipran, although these are often insufficient for a large proportion of patients.⁷⁶ Although preliminary, NYX-2925 reduced daily pain intensity, fatigue, and fibromyalgia impact, suggesting NMDAR modulation may be a viable option. It also improved pain spatial distribution and was safe and well-tolerated without widespread Glx changes beyond dACC and pIns.

NYX-2925 did not improve sleep, anxiety, or depression, although subjects were not selected for these symptoms and a selective action of NYX-2925 on specific symptoms would not without precedent. Pregabalin improves pain, sleep, and function,² whereas SNRIs only affect pain.⁸⁰ Both show mixed effects on depression and anxiety.^3,4 NYX-2925 also did not alter pressure and visual stimuli responsivity on QST, contrasting with preclinical studies showing thermal and mechanical antinociception.³¹

4.1. Limitations

Participants were predominantly White women, limiting generalizability. Modest sample size constrains interpretability and may explain absent effects and inconsistent dose–response patterns. Brain region selection was hypothesis-driven; NYX-2925 may affect unexamined circuits. Several target regions, notably the dACC, are engaged by salience detection and cognitive processing beyond nociception, so the observed changes may not be specific to pain.^71,72 Treatment duration was brief (2 weeks per dose). The fixed-sequence, single-blind administration of placebo followed by ascending doses of NYX-2925 is an important limitation. Because dose was confounded with treatment order and time, period and expectancy effects cannot be fully separated from pharmacological effects, and the week 6 (200 mg) comparison in particular should be interpreted with this in mind. Blinding of participants and imaging analysts to the dose sequence lessens, but does not eliminate, this concern, and a randomized, fully blinded design would provide more robust causal inference. The clinical and correlational findings are exploratory and require independent replication. Proton magnetic resonance spectroscopy focused on excitatory systems, not inhibitory neurotransmitters like GABA.

A subsequent 12-week, double-blind phase 2b study in 300 fibromyalgia participants (NCT04147858) addressed some of these limitations but failed to demonstrate significant placebo separation on primary end points, terminating NYX-2925 development. Beyond the increased sample size and randomization in the phase 2b trial, differences in dosing strategy may explain the conflicting results. The phase 2b study tested 50 mg and 100 mg doses—not the 20 mg and 200 mg used here—and critically excluded the 200 mg dose showing the most robust clinical effects.

4.2. Conclusion

Despite promising neuroimaging findings, NYX-2925's phase 2b failure underscores chronic pain therapeutic development complexity and the importance of trial design consistency. The discrepancy between studies, particularly excluding the 200 mg dose from phase 2b, highlights how methodological variations can profoundly affect outcomes. Although these results raise questions about the predictive value of neuroimaging markers for clinical success, they should not diminish the potential utility of ¹H-MRS and rs-fcMRI in drug development. Mechanistic insights including target engagement through Glx modulation remain valuable for understanding neural pain mechanisms and identifying responsive patient subpopulations. Future programs would benefit from maintaining neuroimaging-informed treatment parameters that show early-phase promise while also considering heterogeneous mechanism–response relationships for personalized pain medicine approaches.

Disclosures

S. E. Harte has consulted for Aptinyx, Indiana University, Wayne State University, and Dana Farber Cancer Institute, and received research funding from Aptinyx, Arbor Medical Innovations (AMI), and SPIRE paid to the University of Michigan. He is coinventor of the MAST System (US9307906B2) used in this study and is a member of AMI, the current licensee of the MAST System from the University of Michigan. R. E. Harris received research funding from Aptinyx paid to the University of Michigan. L. M. Arnold has consulted for Aptinyx, Axsome Therapeutics, and Ono Pharmaceutical Co., and received research funding from Aptinyx paid to the University of Cincinnati. DCJ has consulted for Eli Lilly and Company, Tonix Pharmaceuticals, Merck, Virios, Axsome, and Saulda, received research funding from Aptinyx paid to the University of Michigan, and is a co-inventor of the MAST System. DCJ also testified in vaccine litigation on the safety of the HPV vaccine and is a member of the Tonix Advisory Board. TMM was a full-time employee at Aptinyx, and owned stock in the company. The remaining authors have no conflicts of interest to declare.