Introduction

Depression is a severe, biological heterogeneity, chronic psychiatric disorder characterized by significant cognitive deficits, neural circuit dysfunction, and is projected to become the second leading cause of disease burden by 20301,2. 76.3% of patients with major depressive disorder (MDD) reported cognitive impairment (Perceived Deficit Questionnaire for Depression, PDQ-D total score ≥ 21)3. When treating patients with MDD, alleviation of cognitive symptoms should be considered as an important goal in addition to relieving depressive symptoms. Researchers proposed a cognitive biotype of depression characterized by distinct neural correlates and a functional clinical profile responsive to therapies targeting cognitive dysfunction4. Therefore, awareness, assessment and management of cognitive impairment is of great importance in biotype of depression which would lead to improved functional outcomes​​​​5.

Cognitive biotype was characterized by prominent behavioral impairments in the executive function and response inhibition domains of cognitive control. This biotype exhibited a distinct profile, including specific pretreatment depressive symptoms, poorer psychosocial functioning, lower remission rates, and reduced activation in the cognitive control circuit6. Their cognitive symptoms persisted even after the improvement of depressive symptoms, improvements in symptoms and psychosocial functioning were specifically mediated by changes in cognitive performance. Tozzi et al. also provided a new, theory-driven, clinically validated and quantitative method to identify the clinically distinct biotypes of depression through personalized brain circuit scores. They proposed six biotypes including cognitive biotype7.

Cognitive biotype in depression has long been associated with abnormalities in neural oscillations8. Neural oscillations are rhythmic electrical activity signals within the central nervous system9. They are closely associated with the pathophysiology of psychiatric disorders, particularly cognitive impairment. Based on their frequency, neural oscillations are categorized into delta, theta, alpha, beta, and gamma frequency bands, with signals in different bands reflecting distinct cognitive functions. Among these, gamma-band oscillations exhibit cross-regional coupling characteristics and are considered a “messenger-like” energy band capable of transmitting sensory information to various cortical areas10. Human electrophysiological data and behavioral outcomes demonstrate a link between cross-frequency coupling (CFC) across the cortex and working memory11. Research indicates that gamma oscillations (30–90 Hz) are critically involved in neural circuit function, behavior, and memory12,13. Impaired gamma oscillations are associated with cognitive/memory decline in neurodegenerative disorders (MCI/AD patients and AD mouse models)14. Key abnormalities include reduced spectral power and disrupted theta-gamma coupling versus healthy controls15, which correlate with deficits in working memory and episodic memory16. Clinical studies have identified gamma oscillations as a potential biomarker and endophenotype for depression17, with gamma abnormalities linked to cognitive dysfunction in patients with schizophrenia18 and bipolar disorder19. The increases in sensory responses were accompanied by induced gamma synchrony between the inferior frontal junction, depending on which object was attended20. Moreover, parietal alpha and gamma oscillations a causal for spatial attention shifts21.

Recent research has found that “entrainment of neural oscillation” techniques can modulate brain oscillations22. Rhythmic external stimuli can induce synchronized oscillatory activity in corresponding brain regions, leading to phase locking between the brain’s internal oscillations and the external rhythmic stimulation, thereby achieving synchronization. Furthermore, evidence suggests that chronic gamma entrainment provides neuroprotective effects23.Neural oscillation entrainment can be achieved through various methods, including rhythmic non-invasive brain stimulation (such as transcranial magnetic/electrical stimulation), deep brain stimulation, and rhythmic sensory stimuli (such as rhythmic auditory or visual stimuli)24.

To investigate the causal relationship between gamma oscillations and cognitive function, recent studies have employed various brain stimulation methods to induce gamma oscillations, termed “gamma entrainment.” Results showed that gamma entrainment can improve cognitive function in mouse models. Sensory-evoked gamma entrainment can induce gene expression changes in multiple cell types, including neurons and microglia24.

The global electroencephalography (EEG) coherence in gamma bands of depressed patients was significantly higher than controls, especially in the high gamma band25.Besides, researchers found that resting-state gamma power may represent potential biomarkers of depression associated with therapeutic effects of repetitive transcranial magnetic stimulation and paroxetine treatment26,27. Clinical studies have found that gamma oscillations can be used as a biomarker and endophenotype, and abnormal gamma oscillations is associated with cognitive dysfunction in patients with depression17.

However, whether gamma oscillation decreased in cognitive biotype of depression remains unclear. Here, we distinguished cognitive biotype in stable depression by assessing cognitive function through MCCB. Moreover, gamma spectral power in resting-state was recorded through EEG. The aim was to test whether resting-state gamma power is reduced in individuals with a cognitive biotype of depression and whether it correlates with performance on cognitive tasks.

Methods and materials

Participants

The study recruited, in all, 141 males and females (Only non-pregnant individuals were included) ages 12–60 in Beijing Anding Hospital with unipolar, non-psychotic depression in remission (according to the diagnostic criteria of DSM-5 structured clinical interview) who have a Visual　Analogue　Scales (VAS) score of depressed mood with 5 or above. Participants were excluded if they had comorbid obsessive-compulsive disorder, substance use disorders, or were unable to complete the questionnaire. Participants were on stable medication. Written informed consents were provided by all participants (When participants were under 18, they and their guardians provided consent), and the study protocol was approved by the Ethics Committee of Beijing Anding Hospital and all methods were performed in accordance with the relevant guidelines and regulations. (chiCTR: MR-11-23-044810).

Cognitive function evaluation

Cognitive function of depressed patients was evaluated by the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognitive Battery (MCCB), which consists of 7 dimensions, including attention, working memory, speed of processing, verbal learning, visual learning, reasoning and problem solving, and social cognition28. T-scores adjusted for age, sex, and educated years of participants were computed29. According to the MCCB scores, we distinguish the 56-case cognitive biotype (CI) from 85-case no cognitive impairment (NCI) subgroup of depression. According to previous reports, the criteria for cognitive impairment were set as two or more MCCB dimensions lower than 4030.

Scalp gamma encephalographic data acquisition

Steady-state potentials during wakefulness were recorded at 24-electrode locations (DSI-24; Wearable Sensing, San Diego, CA) evenly distributed over the scalp at locations following the international 10–20 system. Excluding the three auxiliary electrodes, the electrodes relevant to EEG are 19 scalp electrodes and 2 mastoid electrodes. In the whole process, participants were settled casually in the electromagnetically shielded and quiet room. The subjects wore EEG recording electrode caps and earphones for about 5 min with eyes-open and closed condition, respectively. The gamma oscillation signals from 19 channels were recorded and Cz with the ground, referencing electrodes on FPz and the left earlobe, respectively. The EEG recordings had a sample rate of 300 Hz, and the EEG signal was average referenced. We extracted the gamma oscillation as low gamma 30 ~ 45 Hz, high gamma 55 ~ 90 Hz.

Gamma power spectrum analysis

Power spectral densities (PSD) of neural oscillations were analyzed using EEGLAB (https://sccn.ucsd.edu/eeglab/index.php)31, a MATLAB-based toolbox (The MathWorks, Inc., Natick, MA). The raw neural oscillation data was acquired at a sampling rate of 300 Hz. To extract spectral power in specific frequency bands, the data were first filtered with a high-pass filter at 30 Hz, a low-pass filter at 90 Hz, and a notch filter at 50 Hz to remove line noise. The EEG data were then segmented into 2-second epochs, and bad epochs were identified and removed. Bad channels were interpolated using spherical spline interpolation. The neural oscillation data were decomposed using independent component analysis (ICA), and components were visually inspected to identify and remove artifacts such as eye blinks, eye movements, body movements, and electrocardiogram (ECG) interference. Following ICA and an additional baseline correction, the data were re-referenced to the average of all EEG channels. Epochs with amplitudes exceeding the threshold range of -100 to 100 µV were rejected. EEG band power markers were used to quantify the power of the signal across different frequency bands. The average spectral power of neural oscillations in the standard frequency ranges was calculated using the Welch averaged periodogram method1 for both eyes-closed and eyes-open conditions. Spectrograms were computed using a Fast Fourier Transform (FFT) with a Rectangular Window, and the entire epoch was used to average the power within gamma frequency band. Gamma bands are divided into low gamma 30 ~ 45 Hz and high gamma 55 ~ 90 Hz (Avoid the interference range of power frequency around 50 Hz).

Gamma neural oscillations asymmetry analysis

Gamma neural oscillations asymmetry were calculated in resting state with eyes open and closed, respectively. These recordings was re-referenced to Cz and asymmetry scores were computed by the normalized power difference between homologous right- and left-side locations, (R − L)/(R + L)32. Moreover, gamma oscillation asymmetry was analyzed in following five areas: prefrontal [FP: (FP2-FP1)/(FP2 + FP1)], frontal [F: (F4-F3)/(F4 + F3)], central [C: (C4-C3)/(C4 + C3)], Temporal [T: (T4-T3)/(T4 + T3)], parietal [P: (P4-P3)/(P4 + P3)], and occipital [O: (O2-O1)/(O2 + O1)] cortex. The index represents the relative activation of the left locations over right locations in depression.

Statistical analysis

All data were presented as the means ± standard deviation (SD). Statistical analyses were performed with SPSS software (SPSS, version 28.0) and a two-sided p value less than 0.05 was considered statistically significant. Independent-sample T-test was used to analyze differences of MCCB scores and gamma power spectral density between the CI and NCI subgroup. In detail, we conducted independent-sample T-test (FDR adjusted) to test differences in gamma power spectral density between two subgroups. Pearson correlation analysis was used to explore the relationship of the gamma power spectral density of averaged over regions and cognition in all participants, including CI and NCI biotype of depression.

Results

Demographic and clinical characteristics of depressed patients

Depressed patients were divided into cognitive biotype and non-cognitive impairment subgroups. And their demographic and clinical characteristics are presented in another related article33.

The P3 gamma power Oscillation decreased in cognitive biotype of depression at rest

To comprehensively evaluate the gamma oscillation in cognitive biotype of depression, we calculated the power spectral density in the states of both open and closed eye conditions. In the eyes-closed condition (EC), CI biotype showed decreased low-gamma power in P3 channel (t =-3.267, FDR-corrected P = 0.026) (Fig. 1C) and no other channels reached significance after FDR correction (Table 1、Figure 1A). However, in the eyes-open condition (EO), we found no significant differences (FDR-corrected P > 0.05) in gamma oscillation between the two subgroups (Table 2、Figure 1B). Thus, this is a region-specific finding, potentially related to parietal cortex involvement in working memory or attentional control.

Table 1 In the EC condition, statistical results across all analyzed electrodes.
Full size table
Table 2 In the EO condition, statistical results across all analyzed electrodes.
Full size table
Fig. 1
figure 1

The characteristics of gamma power spectral density in cognitive biotype of depression at rest. (A) In the eyes-closed condition, CI biotype decreased low-gamma power. (B) In the eyes-open condition, there was no significant differences in gamma oscillation between CI and NCI biotype. (C) In the eyes-closed condition, CI biotype’s gamma power spectral density decreased in P3 channel. CI: cognitive impairment group, NCI: non-cognitive impairment.

Full size image

The asymmetrical of gamma oscillation in cognitive biotype of depression

Furthermore, we analyzed the gamma asymmetry in cognitive biotype of depression. Our results showed that in the eyes-closed condition, there was no statistical difference in low-gamma and high-gamma power in Fp, F, C, T, P, O between CI and NCI biotype in depression (Fig. 2A). Similarly, in the eyes-open condition, there was also no statistical difference in low-gamma and high-gamma power in these channels (Fig. 2B). These results suggest that there is no significant difference in the asymmetry of gamma oscillation between cognitive and non-cognitive subtypes.

Fig. 2
figure 2

The asymmetry of gamma power values in CI and NCI biotype in MDD. (A) In the eyes-closed condition, there was no statistical difference in gamma power between two biotypes. (B) In the eyes-open condition, there was no statistical difference in gamma power between two biotypes. CI: cognitive impairment group, NCI: non-cognitive impairment.

Full size image

The correlation of gamma oscillation with cognitive function in depression

To further explore the relationship between gamma oscillation and cognitive function in depression, we did correlation analysis. Modest but statistically significant associations between cognitive function and gamma power were observed. In the eyes-closed condition, low-gamma oscillation was correlated with MCCB scores in working memory (r = 0.205, P = 0.015) (Fig. 3), however, it is no longer significant after FDR correction (q = 0.240). Also, in the eyes-open condition, low- and high-gamma oscillation was correlated with MCCB scores in social cognition (r = -0.175, P = 0.038; r = -0.241, P = 0.004) (Fig. 4). Also, it is no longer significant after FDR correction (q = 0.305; q = 0.063). However, we still believe that our findings are instructive for clinical exploration and for subsequent studies with larger samples34.

Fig. 3
figure 3

In the eyes-closed condition, low-gamma oscillation was correlated with working memory. CI: cognitive impairment group, NCI: non-cognitive impairment. Green: low-gamma oscillation; Blue: high-gamma oscillation.

Full size image
Fig. 4
figure 4

In the eyes-open condition, low- and high-gamma oscillation was correlated with social cognition. CI: cognitive impairment group, NCI: non-cognitive impairment. Green: low-gamma oscillation; Blue: high-gamma oscillation.

Full size image

Discussion

Here, a class of cognitive biotype in depression was distinguished, and cognitively related electrophysiological indicators were assessed. The low-gamma power of the left parietal lobe decreased in the cognitive biotype. Also, the frontal gamma oscillation was asymmetrical in cognitive biotype of depression. Moreover, exploratory and preliminary correlations between cognitive function and gamma power were observed. In the eyes-closed condition, low-gamma oscillation was statistically correlated with working memory. Also, in the eyes-open condition, low- and high-gamma oscillation was statistically correlated with social cognition. Thus, gamma neural oscillations may serve as a biomarker and develop potential therapeutic measures for cognitive biotype in depression.

Some depressed patients in remission may still have cognitive symptoms and changes in electroencephalogram. In the acute phase, there are prominent impairments in processing speed, learning, and memory. Follow-up studies revealed less pronounced deficits in remission35. Semkovska et al. found that working memory, selective attention, and long-term memory persist in the remission stage after a major depressive episode and worsen as the number of episodes increases36.

Oscillatory activity in neuronal cells has been associated with a variety of perceptual, motor, and cognitive functions37,38. Consistent with our results, decreased gamma oscillation has been reported that is associated with cognitive declines in Alzheimer’s disease, such as memory dysfunction22,39. Moreover, increases in frontotemporal gamma, frontal and parietal theta connectivity were related with increased cognitive impairment in Parkinson’s disease40. As for depression, gamma oscillations contribute to its pathogenesis mechanisms have been proposed. Researchers reported that depressed patients showed not only significantly decreased gamma powers in the left temporal and the bilateral occipital regions but also weakened gamma connectivity between the left hemisphere and the right frontal region41. Importantly, the deficits in left temporal beta-mid-gamma phase-amplitude coupling (PAC) and beta-high gamma PAC negatively correlated with cognitive disturbance41. Furthermore, this gamma-band hyper-connectivity was positively correlated with depression severity42. Several brain regions, including the left inferior parietal differed significantly between MDD43. Parietal alpha and gamma oscillations a causal for spatial attention shifts21. Not only correlatively, alpha oscillations can also be used to decode attention on a trial by trial level, as shown by the work of Ingmar de vries for example. Helen et al. reported severity of depressive symptoms in old patients who presented a memory loss was associated with decreased left/right temporal-parietal cerebral blood volumes ratios, hinted at the importance of parietal lobe regions in cognitive function44. Similarly, left inferior parietal lobe engagement in social cognition and language45. And the parietal cortex is central to numerical cognition, the right parietal region is primarily involved in basic quantity processing, while the left parietal region is additionally involved in precise number processing and numerical operations46. Interestingly, the work by Orhan Soyuhos on resting state has shown that dorsal areas including the parietal cortex communicate more dominantly in beta oscillations, and gamma oscillations were found more in the ventral stream (although also they report some gamma band interactions for parietal cortex)47. These findings suggest that the potential usefulness of the parietal P3 event-related potential as a marker of Transcranial direct current stimulation-induced effects during task performance48.

Decreased volumes or cortical thickness in the prefrontal cortex, several temporal and parietal regions, hippocampus were associated with depression and these structural neuroimaging abnormalities were also associated with cognitive dysfunction, which is a prominent clinical feature in depression49. It has also been explained that inhibition of the medial prefrontal cortex leads to reduced emotional and cognitive engagement50,51,52.

Gamma neural oscillations play an important role in social cognition and working memory in depression. Moreover, patients with first-episode schizophrenia showed alterations in gamma synchrony during both conscious and nonconscious emotion perception. This pattern of altered synchrony predicted poor performance on each measure of social cognition among these patients53. Gamma frequency connectivity in frontostriatal networks associated with social preference reduction with traumatic brain injury54. As for working memory, executive control acts via interplay between network oscillations in gamma in superficial cortical layers and alpha and beta in deep cortical layers55. Our finding of a negative correlation between gamma oscillations and MCCB scores contrasts with conventional views linking gamma activity to cognitive efficiency. Gamma increased in frontal and temporal regions in a study in which subjects with depression performed spatial and arithmetic tasks56. Activation of fast-spiking interneurons at varied frequencies selectively amplifies gamma oscillations and disrupt focused attention57. This may reflect pathological GABAergic dysfunction, where excessive gamma synchrony impairs neural flexibility, consistent with depression-related cognitive deficits58. Alternatively, heightened gamma could indicate compensatory overactivation or metabolic trade-offs that disrupt cross-frequency coupling. While task-related gamma typically supports cognition, resting-state hyperactivity may signal network inefficiency in depression. These results challenge the “more gamma equals better cognition” paradigm, suggesting an optimal oscillation range for cognitive function59. Future studies should examine whether gamma abnormalities drive or compensate for cognitive impairment. Rigorous artifact removal and medication controls strengthen these novel findings.

While we provide valuable insights into the cognitive biotype of MDD, there exist some shortcomings. Firstly, the research is constrained by its cross-sectional design, the small effect sizes and unequal group sizes. To better understand the nature of cognitive biotype, longitudinal, larger, better-controlled samples studies, including following recurrence are needed. Next, we have not done correction for multiple comparisons in correlation analysis, age and education lack of matching in two subgroups. This is because we use a continuous sampling method, and adolescent depressed patients are prone to cognitive symptoms. Also, we did not distinguish between periodic and aperiodic components, changes in 1/f spectral characteristics may confound the results of traditional band-specific power analysis. In the future we will conduct a more in-depth and precise analysis60. For this first exploratory study, we prioritized broad sampling to detect cross-developmental patterns, with plans for targeted age-cohort studies in Phase 2. Lastly, future investigations should explore the specific stimulus parameters and target location for cognitive biotype diagnosis and treatment. Research into biomarkers for cognitive biotype has demonstrated potential promise, a deeper comprehension of the biological underpinnings of MDD and classification of its cognitive biotype based on biological criteria would prove highly beneficial for treatment.

Our findings tentatively indicate that clinicians may need to consider the potential relevance of the cognitive biotype in depression, where appropriate. Further development of cognitive screening protocols and exploration of targeted intervention strategies could be valuable future directions. Many studies have found that gamma entrainment is involved in cognitive function. Wang et al. reported gamma entrainment rescues cognitive impairment by decreasing postsynaptic transmission after traumatic brain injury61. Moreover, gamma entrainment using sensory stimulation is regarded as a novel non-pharmacological therapeutic intervention applied to cognitive symptoms associated with schizophrenia62. We also recommend future opportunities that gamma entrainment for using such strategies in cognitive biotype in depression.