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

Abstract

Stress is a significant risk factor for the development of major depressive disorder (MDD), yet the underlying mechanisms remain unclear. Preclinically, adaptive and maladaptive stress-induced changes in glutamatergic function have been observed in the medial prefrontal cortex (mPFC). Here, we examine stress-induced changes in human mPFC glutamate using magnetic resonance spectroscopy (MRS) in two healthy control samples and a third sample of unmedicated participants with MDD who completed the Maastricht acute stress task, and one sample of healthy control participants who completed a no-stress control manipulation. In healthy controls, we find that the magnitude of mPFC glutamate response to the acute stressor decreases as individual levels of perceived stress increase. This adaptative glutamate response is absent in individuals with MDD and is associated with pessimistic expectations during a 1-month follow-up period. Together, this work shows evidence for glutamatergic adaptation to stress that is significantly disrupted in MDD.

Fig. 1. Study design and effects of stress on salivary cortisol and subjective ratings.

a Schematic diagram of the study visits and approximate timing of MRS, fMRI RL task, VAMS, and saliva measurements. Note that the Healthy Control Stress sample did not complete resting-state scans, STRAIN, or EMA. EMA Ecological Momentary Assessment, MRS Magnetic Resonance Spectroscopy, PSS Perceived Stress Scale, RL Reinforcement Learning, RS Resting State, SCID Structured Clinical Interview for DSM Disorders, STRAIN Stress and Adversity Inventory. MAST Maastricht Acute Stress Test, VAMS Visual Analog Mood Scales. b Representative MRS voxel placement. c Representative MRS spectrum (black) and LCModel fit (red) with labeled metabolite peaks. arb.units arbitrary units, Cr creatine, PCr Phosphocreatine, Glu glutamate, Glx (glutamine + glutamate), GCP glycerophosphocholine and PC, phosphocholine (choline-containing metabolites), ml myo-inositol, MRS Magnetic Resonance Spectroscopy, NAA N-acetylaspartic acid, NAAG N-acetylaspartylglutamate, ppm parts per million. d Effect of MAST acute stress task and No Stress Control (NSC) on mood. Items are coded such that higher scores indicate greater negative emotional experience and averaged across items. Data represented as mean ± standard error of the mean (N = 75 participants). e Salivary cortisol response to acute stress manipulation and no-stress control. Graph depicts percent change in salivary cortisol from the timepoint immediately prior to the onset of the MAST stressor (Pre-MAST). Data represented as mean ± standard error of the mean (N = 83 participants). f Subjective stress ratings for each group (1–5). Data represented as mean ± standard error of the mean (N = 82 participants). HC Healthy Control participants, MDD participants with Major Depressive Disorder. Source data for d–f are provided as a Source Data file.

Fig. 2. Changes in mPFC creatine-normalized glutamate in response to acute and perceived stress.

a Association between perceived stress (PSS scores) and percent change in Glu/Cr signal (r_s(23) = −0.457, p = 0.022, two-tailed, uncorrected) in healthy control stress sample. b Association between perceived stress (PSS scores) and percent change in Glu/Cr signal (r_s(20) = −0.517, p = 0.014, two-tailed, uncorrected) in the healthy control stress replication sample. c Association between perceived stress (PSS scores) and percent change in Glu/Cr signal (r_s(16) = 0.139, p = 0.581, two-tailed, uncorrected) in no-stress control sample. d Association between perceived stress (PSS scores) and percent change in Glu/Cr signal (r_s(21) = 0.115, p = 0.602, two-tailed, uncorrected) in participants with major depressive disorder. Shaded area on a–d represents 95% confidence interval, *p < 0.05. e–h Glu/Cr ratios before and after MAST in e healthy control stress sample (n = 25 participants), f healthy control stress replication (n = 22 participants), g no-stress control (n = 18 participants), and h participants with major depression (n = 23 participants). Boxplot elements for e-h indicate median (center line), first and third quartiles (box limits; 25–75th percentile), smallest observation within 1.5 times the interquartile range from the lower quartile (bottom whisker), largest observation within 1.5 times the interquartile range from the upper quartile (top whisker), and all individual participants (points). Cr Creatine-containing metabolites (Creatine and Phosphocreatine), Glu glutamate, HC healthy control, MAST Maastricht Acute Stress Test, MDD participants with major depressive disorder, NSC no-stress control, PSS Perceived Stress Scale. Source data are provided as a Source Data file.

Fig. 3. Relationship between perceived stress and MRS metabolites.

a Relationship between PSS and change in creatine-normalized glutamate in all participants who completed the acute stress manipulation. Group-level linear trends (colors) and combined quadratic effect (black) are overlaid. b Maladaptive glutamate response was calculated for the healthy control stress replication sample (gray) and participants with major depressive disorder (green), defined as the residual between the observed percent change glutamate and expected percent change glutamate, estimated using the linear function of the healthy control stress sample, shown in black. Percent change glutamate expected = 35.647 − 3.093*PSS. c Partial effect size (Pearson’s r; controlling for age and sex, two-tailed, uncorrected) between PSS and percent change glx (glutamate and glutamine; r_partial(43) = −0.382, p = 0.010), percent change glutamate (r_partial(43) = −0.346, p = 0.020), and percent change choline-containing metabolites (r_partial(43) = −0.141, p = 0.356) in all healthy controls who completed the acute stress manipulation (n = 47 participants). Significance indicated as *p < 0.05, **p < 0.01. Estimates and error bars (95%CI) were estimated using bootstrapping with 1000 samples. Cho Choline-containing metabolites (primarily glycerophosphocholine and phosphocholine), Glu Glutamate, Glx Glutamate + Glutamine, HC Healthy Control, MDD Participants with major depressive disorder, MRS Magnetic Resonance Spectroscopy, PSS Perceived Stress Scale. Source data are provided as a Source Data file.

Fig. 4. Ecological momentary assessment and associations with maladaptive glutamate response.

a Differences in ecological momentary assessment (EMA) ratings between healthy control (HC; n = 20 participants) and participants with major depressive disorder (MDD; n = 18 participants) with significance indicated as *p < 0.05, ***p < 0.001 (t-test, two-tailed, uncorrected). Statistical comparisons were t₃₆ = −3.98, p < 0.001 for expected outcomes, t₃₆ = −3.74, p = 0.001 for experienced outcomes, t₃₅ = 0.96 p = 0.342 for optimistic expectations, t₃₆ = −5.44, p < 0.001 for positive affect, t₃₆ = 2.45, p = 0.019 for expectation inaccuracy, t₃₆ = 5.62, p < 0.001 for negative affect, and t₃₆ = 1.89, p = 0.066 for pessimistic expectations. Data represented as mean ± standard error of the mean. b Association between maladaptive glutamate response (MGR) and pessimistic expectations from EMA, r₃₆ = 0.515, p < 0.001 (two-tailed, uncorrected). Shaded band represents 95% confidence interval. c Effect sizes (partial Pearson’s r, two-tailed, uncorrected) between MGR and EMA variables, controlling for age, sex, and diagnostic group (n = 38 participants), r_partial(33) = −0.225, p = 0.193 for expected outcomes, r_partial(33) = −0.181, p = 0.299 for experienced outcomes, r_partial(32) = −0.085, p = 0.631 for optimistic expectations, r_partial(33) = 0.123, p = 0.483 for positive affect, r_partial(33) = 0.330, p = 0.053 for expectation inaccuracy, r_partial(33) = 0.341, p = 0.045 for negative affect and r_partial(33) = 0.441, p = 0.008 for pessimistic expectations. Effect sizes and error bars (95%CI) were estimated using bootstrapping with 1000 samples and estimated separately for variables with different numbers of observations. Note that the optimistic expectations comparisons in a and c had one less degree of freedom than other variables, as one participant had no surveys with optimistic expectations. Source data are provided as a Source Data file.