Reduced adaptation of glutamatergic stress response is associated with pessimistic expectations in depression

The study investigates how acute stress affects medial prefrontal cortex (mPFC) glutamatergic function in humans, and whether recent perceived stress modulates this response. Stress, especially unpredictable and uncontrollable "toxic" stress, is linked to anhedonia and major depressive disorder (MDD). Preclinical work shows acute stress increases mPFC glutamatergic transmission but repeated stress leads to attenuated glutamate responses, potentially an adaptive mechanism. Given mPFC's role in stress regulation and reward valuation, the authors hypothesize that healthy individuals with low perceived stress would show increased mPFC glutamate after an acute stressor, whereas those with higher perceived stress would show reduced responses. They further hypothesize this adaptive attenuation would be disrupted in MDD and relate to pessimistic expectations in daily life.

Prior animal studies demonstrate that initial stress exposure increases extracellular glutamate, potentiates excitatory currents, and upregulates AMPA/NMDA receptors in mPFC, enhancing short-term learning. With repeated stress, glutamate responses habituate and dendritic complexity/spines in mPFC are reduced, possibly as protective adaptation. The mPFC coordinates behavioral/endocrine stress responses and encodes expected value and outcome probabilities, implicated in learned helplessness. Human stress reduces reward-related processing in corticostriatal circuits, with individual differences: perceived stress predicts blunted mPFC reward responses. However, mechanisms linking perceived stress, acute stress, and mPFC glutamate in humans remain unclear. The study builds on this literature to test adaptive versus maladaptive glutamatergic responses and their disruption in MDD.

Design: Four samples participated. Two healthy control (HC) stress samples underwent an acute stressor (MAST) between two mPFC MRS scans. A third HC sample completed a matched no-stress control (NSC) task. A fourth sample comprised unmedicated participants with current MDD who completed the stress manipulation. Post-scan, EMA was conducted in Emory samples for 4 weeks to assess expectations and outcomes in daily life. Participants: Adults 18–60. Exclusions for HCs: any current/past psychiatric disorder (except specific phobia) or past alcohol abuse; for MDD: SCID-confirmed MDD, with exclusions for substance use disorders, OCD, bipolar disorder, psychosis, active suicidal ideation; comorbid anxiety/PTSD allowed. All groups excluded for psychotropic/illicit drugs (urine-confirmed) and tobacco (beyond occasional use). Final analyzable sample N=88 after MRS quality control; group demographics in Table 1 (HC Stress n=25; HC Stress Replication n=22; HC No-Stress Control n=18; MDD Stress n=23). Procedures and measures: Visit 1 included consent, SCID, self-reports including Perceived Stress Scale (PSS; past-month perceived stress). Emory samples also completed the Stress and Adversity Inventory (STRAIN) for lifetime stressor exposure (counts/severity of acute and chronic stressors). Visit 2 included MRI session with two single-voxel mPFC MRS scans (pre- and post-manipulation), a reinforcement learning task, stress/NSC manipulation, salivary cortisol (baseline, ~20 and ~40 min post-manipulation), and mood ratings (adapted VAMS at multiple timepoints). Post-scan subjective ratings of stress, unpleasantness, difficulty were collected. Stress manipulations: MAST adapted for MRI: alternating blocks of cold pressor (1–8°C) hand immersion (five immersions of 30–90 s, randomized, unpredictable) and evaluative serial subtraction (from 2043 by 17; restart on errors). NSC matched timing but used warm water (26–36°C) and simple counting aloud from one without evaluation. MRS acquisition: 3T Siemens Tim Trio, 32-channel head coil at both sites; single 2×2×2 cm3 voxel placed in mPFC anterior to corpus callosum. 2D-JPRESS protocol collecting 22 TE-stepped spectra (TE 30–350 ms in 15 ms steps), TR=2 s, spectral bandwidth=2 kHz, NEX=16/TE-step, ~12 min per scan. Processing with LCModel using GAMMA-simulated J-resolved basis sets; metabolites expressed as ratios to total creatine (Cr). Quality criteria: SNR≥9, FWHM≤0.15, CRLB Glu≤20%, visual quality review. Test-retest ICC for Glu/Cr was excellent (Emory ICC=0.89). Primary outcome: Percent change in creatine-normalized glutamate (%ΔGlu) from pre- to post-manipulation: %ΔGlu = (Glu/Cr_post − Glu/Cr_pre) / (Glu/Cr_pre). Secondary: %ΔGlx (Glu+Gln), %ΔCho. EMA: 4-week protocol (every other day; six surveys/day, ~2-h spacing) via mobile Qualtrics. Measures: current positive/negative affect, expected valence of planned activity (−4 to +4), whether previous planned activity occurred and its experienced valence. Computed variables included expectation accuracy/inaccuracy, frequency/magnitude of pessimistic and optimistic expectations. Usable EMA data: HC=20 participants (1236 surveys), MDD=18 (1138 surveys). Derived metric: Maladaptive Glutamate Response (MGR) = observed %ΔGlu − expected %ΔGlu, where expected %ΔGlu was estimated from the linear PSS–%ΔGlu function in the McLean HC sample. Positive MGR indicates larger-than-expected glutamate increase given PSS. Statistics: Repeated-measures ANOVAs for mood and cortisol (Greenhouse–Geisser corrections as needed). Spearman correlations within samples for PSS–%ΔGlu relationships. Hierarchical linear regression tested interactions (PSS×Acute Stress; PSS×Diagnostic Group) controlling for age, sex, site; stepwise selection. Partial correlations (controlling for age, sex, diagnosis, and in some analyses PSS and pessimistic frequency) linked MGR to EMA variables. Bootstrapped CIs for effect sizes. Two-tailed tests unless noted.

• Manipulation checks: Acute stress (MAST) increased negative mood (Timepoint×Group interaction p=0.002). Cortisol showed Timepoint×Group interaction (p=0.023); in HCs, cortisol rose after MAST (t(42)=3.33, p=0.002), while NSC showed slight decrease (t(17)=-2.18, p=0.044). Subjective ratings of stress/unpleasantness/difficulty were highest in stress groups and lowest in NSC (ps<1×10^-13); no differences between HC and MDD within stress condition on these ratings.
• Primary effect in healthy controls: Perceived stress (PSS) inversely predicted mPFC %ΔGlu after acute stress in two independent HC samples: • HC Stress (McLean, n=25): rs=-0.457, p=0.022. • HC Stress Replication (Emory, n=22): rs=-0.517, p=0.014. Individuals with low PSS showed increases in Glu/Cr after stress; higher PSS showed little change or decreases. No overall pre–post main effect on Glu/Cr in HCs (paired t ps>0.7), but among low PSS (<10), stress increased Glu/Cr (t(22)=2.39, p=0.026).
• Specificity to stress: HC No-Stress Control (n=18) showed no association between PSS and %ΔGlu (r=0.139, p=0.581). Hierarchical regression across HC stress and NSC showed a significant PSS×Acute Stress interaction predicting %ΔGlu (β=-0.40, t(60)=-2.15, p=0.035) and an age effect (β=0.29, t(60)=2.43, p=0.018); adjusted R²=0.196.
• Objective stress (STRAIN): No significant associations between STRAIN indices (lifetime or past-year acute/chronic counts or severity) and %ΔGlu (|r|≤0.148, ps>0.5), suggesting perceived stress (PSS) rather than objective exposure predicted glutamate response.
• MDD sample: In MDD (n=23), PSS did not correlate with %ΔGlu after stress (r=0.115, p=0.602). Across all stress-exposed participants, regression showed significant PSS main effect and a significant PSS×Diagnostic Group interaction predicting %ΔGlu (model adjusted R²=0.093; ΔR² for interaction p=0.016), indicating disrupted adaptation in MDD.
• Baseline and main effects: No differences in baseline Glu/Cr or Glx/Cr between HC and MDD, and no Timepoint or Timepoint×Diagnosis effects (all ps≥0.658). Baseline Glu/Cr negatively correlated with age across all participants (r=-0.237, p=0.026).
• EMA results: Compared to HCs, MDD participants reported higher negative affect (t=5.62, p<0.001), lower positive affect (t=-5.44, p<0.001), lower expected (t=-3.98, p<0.001) and experienced outcomes (t=-3.74, p=0.001), fewer accurate expectations (MDD M=0.38 vs HC M=0.53; t=-2.35, p=0.024), and greater expectation inaccuracy (t=2.45, p=0.019). MDD had a higher proportion of pessimistic expectations (0.38 vs 0.28; t=2.11, p=0.042).
• Linking glutamate response to daily-life expectations: MGR positively associated with pessimistic expectations (partial r=0.457, p=0.006; remained significant controlling for PSS: partial r=0.353, p=0.040) and with negative affect (partial r=0.367, p=0.030; ns after controlling for PSS). No association with optimistic expectations (partial r=-0.096, p=0.590). The association with pessimistic vs optimistic expectations differed significantly (Steiger’s Z=2.11, p=0.035).

Findings demonstrate an adaptive modulation of mPFC glutamatergic response to acute stress in healthy individuals: higher recent perceived stress is associated with attenuated %ΔGlu to a new stressor, consistent with preclinical evidence of habituation/adaptation under repeated stress and with allostatic load models. This adaptation appears specific to an acute stressor and not merely engagement in a matched cognitive/sensory task. In contrast, individuals with current unmedicated MDD do not exhibit this inverse PSS–%ΔGlu relationship, indicating disrupted glutamatergic adaptation. Moreover, the degree to which an individual's glutamate response deviated from the expected adaptive pattern (MGR) predicted more pessimistic expectations in daily life over a month, linking neurochemical stress responses to anticipatory aspects of anhedonia/negative bias. Baseline mPFC glutamate and gross pre–post changes did not differ by diagnosis, suggesting that context-dependent neurochemical dynamics, rather than static metabolite levels, may be more informative for stress-related psychopathology.

This work identifies attenuation of mPFC glutamate response to acute stress as an adaptive process modulated by recent perceived stress in healthy adults and shows that this adaptation is disrupted in unmedicated MDD. The maladaptive deviation from this response predicts pessimistic expectations in daily life, providing a potential neurochemical mechanism contributing to anticipatory anhedonia and negative bias. These insights suggest that dynamic glutamatergic responses under stress may serve as targets for intervention and treatment monitoring in stress-related disorders. Future research should examine medication effects, symptom dimensions (e.g., anhedonia), illness course (number of episodes), finer timescales of stress appraisal, broader age ranges (adolescents and older adults), and improved temporal resolution linking glucocorticoids and glutamatergic dynamics.

• MRS glutamate primarily reflects intracellular pools, limiting direct inference about synaptic transmission; temporal resolution may not capture rapid stress-related changes.
• Moderate sample sizes and exclusion due to MRS quality reduce power and generalizability; however, effects replicated across two HC samples.
• PSS distributions did not fully overlap between HC and MDD, complicating attribution of maladaptive response to diagnosis vs stress severity.
• Age range was limited to young-to-middle adulthood; findings may not generalize to adolescents or older adults; baseline Glu/Cr declined with age and %ΔGlu increased with age in this sample.
• STRAIN collected only in Emory samples; lack of association with %ΔGlu may reflect timescale or construct differences from PSS.
• Only unmedicated MDD participants were included; medication effects and chronicity may influence glutamatergic measures.
• No group differences in basal metabolites; static measures alone may not serve as biomarkers, underscoring need for context-dependent paradigms.