Identification of key genes modules linking brain aging signatures and COVID-19-associated cognitive impairment

Highlights

• •Recent studies have shown that there is a decline in cognitive and memory abilities following SARS-CoV-2 infection.
• •we identified up-regulated DEGs is almost in lung fibrosis and immune-related pathways, yet down-regulated DEGs is in neuron and synapse pathways.
• •We then identified five key genes associated with aging in COVID-19, including SST, S100A9, SOSC3, FKBP5 and HBB.
• •Finally, we identified seven drugs associated with key genes: Fulvestrant, Bucladesine, S-Adenosylmethionine, Valproic Acid, Folic Acid, kaempferol and Quercetin.

Abstract

Background

Emerging evidence indicates that SARS-CoV-2 infection is associated with aging, particularly in memory-related domains. However, the molecular mechanisms underlying COVID-19-associated aging, often referred to as "brain fog," remain poorly understood. This study aimed to identify key genes implicated in post-COVID cognitive dysfunction.

Methods

Transcriptomic data from the frontal cortex, obtained from the GEO dataset GSE188847 and GSE164332, were analyzed using differential expression analysis to identify differentially expressed genes (DEGs) associated with brain fog. Functional enrichment analysis was conducted using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Machine learning techniques, including LASSO regression and weighted gene co-expression network analysis (WGCNA), were employed to identify hub genes. Drug-gene interactions were explored using the Comparative Toxicogenomics Database (CTD).

Results

A total of 81 up-regulated and 441 down-regulated DEGs were identified in COVID-19 patients compared to controls. Up-regulated genes were primarily enriched in pathways related to pulmonary fibrosis and immune activation (e.g., IL-17 signaling), while down-regulated genes were associated with synaptic transmission (e.g., GABAergic synapse, adjusted p = 0.003). Integration of LASSO and WGCNA analyses revealed five aging-related hub genes: SST, S100A9, SOCS3, FKBP5, and HBB.

Notably, SST expression was significantly reduced, whereas the other four genes showed marked upregulation. CTD analysis identified seven potential therapeutic agents: fulvestrant, bucladesine, S-adenosylmethionine, valproic acid, folic acid, kaempferol, and quercetin.

Conclusions

This study employed transcriptomic analysis and bioinformatics approaches to identify five key aging-related genes associated with COVID-19. These genes may serve as potential therapeutic targets for addressing COVID-19-associated aging.

1. Introduction

Coronavirus disease 2019 (COVID-19) continues to pose one of the most significant global public health challenges of the 21st century (World Health Organization, 2023). The infection leads to multisystem complications, including coagulopathies, septic shock, and multiple organ dysfunction, affecting the pulmonary, cardiovascular, renal, gastrointestinal, and neurological systems (Hingorani et al., 2022; Stein et al., 2022; Augustin et al., 2021). A substantial proportion of patients experience persistent symptoms after recovery, collectively referred to as post-acute sequelae of SARS-CoV-2 infection (PASC) or "long COVID." These symptoms often include neurological manifestations such as cognitive impairment, anosmia, and chronic fatigue, commonly described as "brain fog" (Hingorani et al., 2022; Mavrikaki et al., 2022; Theoharides et al., 2021; Venkataramani and Winkler, 2022). Emerging evidence suggests a time-dependent attenuation of COVID-19-associated cognitive deficits during the pandemic (Li et al., 2023). However, current research highlights a discrepancy between self-reported neuropsychiatric symptoms and objective cognitive testing measures in both acute and chronic phases of the disease (Venkataramani and Winkler, 2022).

Cognitive deficits are associated with functional alterations in synaptic transmission, structural remodeling of neuronal connections, and synaptic degeneration (Venkataramani and Winkler, 2022). Age-related neurophysiological decline primarily manifests as dysregulation of cortical activity modulation, largely governed by basal forebrain (BF)-derived cholinergic projections (Li et al., 2023). Specific regions of the prefrontal cortex mediate executive functions, such as working memory and cognitive flexibility, which are distinct from the broader functional domains of the frontal lobes (Díez-Cirarda et al., 2023).

The paradigm shift from histopathological observations to molecular profiling in understanding cognitive deficit mechanisms has revolutionized the discovery of therapeutic targets and the development of precision interventions (Based 2026; Luo et al., 2025). Neuroinflammatory processes, for instance, induce immune system dysregulation, leading to accelerated immunosenescence during cognitive deterioration (Henri-Bhargava et al., 2015). In the context of COVID-19 neuropathology, emerging evidence supports multifactorial injury mechanisms, including direct viral neuroinvasion, dysregulated cytokine cascades, and microvascular endothelial dysfunction. Despite these advances, critical knowledge gaps remain, particularly regarding validated biomarkers and unified pathophysiological frameworks. The World Health Organization (WHO)-classified post-COVID condition (PCC) remains a pressing global health priority, with three key unresolved issues:

• 1.The longitudinal trajectory of cognitive sequelae (as investigated in the ongoing NIH RECOVER Initiative),
• 2.The efficacy of neuroprotective interventions in viral-induced neurodegeneration, and
• 3.The specificity of biomarkers for differentiating COVID-related cognitive decline from age-related cognitive decline.

Current mechanistic research is transitioning from phenomenological descriptions to multi-omics integration, though translational applications remain in the preliminary validation phase. Accelerated cognitive aging and prefrontal cortical dysfunction are areas of particular interest. Our investigation employed a transcriptomic analysis approach to:

• 1.Profile frontal lobe transcriptomic signatures using bulk RNA-seq data from postmortem COVID-19 specimens,
• 2.Identify aging-associated hub genes through machine learning pipelines (LASSO-WGCNA integration),
• 3.Predict pharmacologically targetable agents via Comparative Toxicogenomics Database (CTD)-based drug repositioning analysis, and
• 4.Characterize neuroimmune landscape alterations through deconvolution algorithms.

The detailed analysis workflow is illustrated in Fig. 1.

Fig. 1.

Flow diagram of this research.

2. Methods

2.1. Data collection and preprocessing

We obtained the gene expression profiles (GSE188847 and GSE164332) from the Gene Expression Omnibus database by using the keywords “COVID-19″ and “frontal lobe”. Datasets that fulfilled the following inclusion criteria were considered eligible for this study: a) They were from Homo sapiens and specifically focused on the frontal lobe and COVID-19; b) The expression profiling was carried out through high throughput sequencing; c) The design included both patients as well as age- and sex-matched healthy controls. Subsequently, the gene expression profiles underwent preprocessing and normalization with the help of R software (version 4.2.1).

2.2. Differential gene expression analysis and functional annotation

In order to identify DEGs in the frontal cortex of the brain, differential expression analysis was conducted by utilizing the R package ‘Limma’ (Zhang and Horvath, 2005). Firstly, the Limma R package was employed to carry out the differential expression analysis of genes between the COVID-19 group and the control group. The default thresholds were set as follows: the log-transformed fold change should be greater than 1 (in absolute value), and the adjusted P value with Bonferroni correction should be <0.05.Subsequently, pathway enrichment analyses of the DEGs were performed by using Metascape (http://metascape.org/) with the default parameter settings. As shown in Fig. 2A/B, the enrichment analysis in DisGeNET and in PaGenBase was carried out using Metascape.

Pathway enrichment analyses for both up-regulated differentially expressed genes (DEGs) and down-regulated DEGs were conducted separately by using Metascape, as presented in Fig. 3A/B. The enrichment networks were then automatically generated and visualized with the help of the online Cytoscape tool. In this process, each enriched functional term was represented as a node, and pairs of nodes with a Kappa similarity score higher than 0.3 were connected. Furthermore, we also employed gene set enrichment analysis by utilizing the clusterProfiler package to identify particular biological processes and/or signaling pathways that were enriched.

2.3. Construction of the co-expression network and hub module identification by WGCNA

Weighted gene co-expression network analysis (WGCNA) is an elaborate systems biology approach that aims to clarify the correlation patterns among genes within high throughput sequencing samples (Langfelder and Horvath, 2008). In this particular study, the WGCNA co-expression system was set up by employing the R package “WGCNA” (Hänzelmann et al., 2013). Firstly, a weighted adjacency matrix was constructed with the optimal soft power and then converted into a topological overlap matrix (TOM). By specifying a minimum module size of 100, the TOM dissimilarity measure (1 - TOM) was utilized to obtain modules based on the algorithm of the hierarchical clustering tree. After that, the modules were clustered to examine the similarity among them. A distinct module was randomly assigned a different color. The module eigengene was used to display the gene expression profiles for each module. Module significance denoted the relationship between modules and the disease status. Module membership referred to the correlation coefficient between genes and module eigengenes, which could describe the reliability of a gene within a module. Gene significance represented the association between a gene and clinical traits, where a higher gene significance suggested that the specific gene was more closely related to the studied traits. To identify significant modules, the association between modules and clinical traits was calculated, and the core genes of these modules were then outputted.

2.4. LASSO analysis

GeneCards (https://www.genecards.org/) is a searchable and integrative database that offers comprehensive and user-friendly information regarding all annotated and predicted human genes. Aging-related genes were retrieved from GeneCards, and those genes with a relevance score of no <5 were chosen.

The results were then intersected with differentially expressed genes (DEGs), the aging-related genes from GeneCards (https://www.genecards.org/), and the core genes from weighted gene co-expression network analysis (WGCNA) in order to obtain aging-related core DEGs. Subsequently, lasso regression was employed to analyze the intersected genes. Lasso regression is a form of linear regression that makes use of shrinkage. Shrinkage refers to the process in which data values are drawn closer to a central point, such as the mean. After preprocessing the input dataset, we selected the relevant features from the available independent variables to build the model. We utilized domain knowledge, statistical methods, or machine learning algorithms to pick out features that held potential predictive power.Next, the aim was to minimize the loss function, which consists of a squared error term. The L1 regularization term imposes a penalty on the absolute values of the regression coefficients, thereby facilitating sparsity within the parameter set. The lasso procedure favors simple and sparse models (that is, models with fewer parameters). We then interpreted the model results based on the obtained coefficients. These coefficients can reveal variable importance as well as positive or negative relationships with the target variable. This type of regression is highly appropriate for models that exhibit high levels of multicollinearity or when there is a need to automate certain aspects of model selection, like variable selection or parameter elimination.

2.5. Single-sample gene set enrichment analysis

The immune landscape plays a crucial role in comprehending the composition and activity of immune cells, as it offers valuable perspectives for predicting disease progression and evaluating the efficacy of therapies.In our study, single-sample gene set enrichment analysis (ssGSEA) was conducted using the R package “GSVA” (Douaud et al., 2022). Briefly speaking, the gene expression dataset was formatted appropriately and incorporated all the essential information required for data preparation. The ssGSEA algorithm from the GSVA package was then applied to each sample within the dataset. When it came to interpreting the results, it involved several aspects. Firstly, we needed to identify the gene sets that were significantly enriched. Secondly, we had to assess the variation in enrichment scores among different samples. And finally, we had to explore the biological implications of these findings. Moreover, the corrplot package was used to calculate the Pearson correlation coefficient for each type of immune cell.

2.6. Statistical analysis

All statistical analyses were performed with R (v.4.2.1), and the statistical details can be found in either figure legends or methods. The quantitative data are shown as box plots, wherein boxes represent the interquartile range, whiskers extend to ±1.5 interquartile range, dots represent outliers, and bold lines indicate median values. Statistical significance was determined by the hypergeometric test. P values and adjusted P values <0.05 were considered statistically significant.

3. Results

3.1. Differential expression analysis and clustering reveals distinct expression patterns in COVID-19 samples

In order to find out the differentially expressed genes (DEGs) between COVID-19 samples and normal samples, a differential expression analysis was carried out on 23 normal samples and 22 COVID-19 samples sourced from the GSE188847 dataset by utilizing the R package ‘limma’ (as shown in Fig. 2).This analysis uncovered a total of 527 DEGs, which were composed of 441 upregulated genes and 86 downregulated genes (with p < 0.05 and |log FC| > 1; refer to Fig. 4A). A heatmap demonstrating the expression levels of these DEGs was also generated (Fig. 4B), which clearly emphasized the distinct expression patterns existing between the two groups.

Fig. 2.

(A) Enrichment analysis in DisGeNET using Matascape. (B) Enrichment analysis in PaGenBase using Matascape.

Fig. 4.

Differential genes expression analysis. (A) Volcano plot illustrating the differentially expressed genes (DEGs) between the COVID-19 and normal samples. (B) Heatmap displaying the expression levels of the DEGs between the COVID-19 and normal samples. (C) Bar graph of enrichment analysis of DEGS between the COVID-19 and normal samples, colored by p-values.

Subsequently, all the statistically enriched terms for the 527 DEGs were initially identified using Metascape (these could be GO/KEGG terms, canonical pathways, hall mark gene sets, etc., depending on the default options under Express Analysis). The accumulative hypergeometric p-values and enrichment factors were computed and then employed for the purpose of filtering. It was found that the pathways were enriched in aspects related to neurons and synapses, such as synaptic signaling, synapse organization, neuron projection development and GABAergic synapse (as depicted in Fig. 4C).

After that, we created a comprehensive enrichment network. In this network, each node stands for an enriched functional term for the up-regulated differentially expressed genes (DEGs) (Fig. 3A; refer to the methods) and the down-regulated DEGs (Fig. 3B) separately. We discovered that the top enrichment among the up-regulated DEGs is in immune-related pathways when it comes to the impact of COVID-19. These pathways include the response to type II interferon, inflammatory response, cytokine signaling in the immune system, interferon gamma signaling, response to wounding, and interleukin-4 and interleukin-13 signaling (as shown in Fig. 3A). This finding might align with the fact that infection with the novel coronavirus can trigger both natural and adaptive immune responses that have protective effects. Moreover, the top enrichment among the down-regulated DEGs is in neuron and synapse pathways during the influence of COVID-19 (as seen in Fig. 3B). This might be consistent with the reality that COVID-19 infection is often accompanied by manifestations in the Nervous System (NS), which has the potential to affect cognition.

Fig. 3.

Overview of enrichment analysis of up-genes/down-genes (A) Overview of network of enriched terms across up genes of the DEGs between the COVID-19 and normal samples. (B) Overview of network of enriched terms across down genes of the DEGs between the COVID-19 and normal samples. The network was obtained by functional enrichment analysis using Matascape. Each node represents a particular gene ontology term, and similar terms are clustered and merged together for annotation. The connection of nodes reflects the protein-protein interaction.

We were able to observe that the differentially expressed genes (DEGs) were enriched in areas such as memory impairment, mental disorders, and mental deterioration by using the DisGeNET database (as shown in Fig. 2A). Meanwhile, it was also evident that the DEGs were enriched in numerous neuron-related tissues with the help of the PaGenBase database (as depicted in Fig. 2B). All of these findings imply that inflammatory factors have a close association with slow aging.

3.2. Deciphering the interplay between the COVID-19 group and the control group by weighting gene co-expression network analysis

In order to clarify the gene modules that are relevant to the COVID-19 trait, we carried out a weighted gene co-expression network analysis (WGCNA) on a dataset which consisted of 44 samples. We determined the suitable soft threshold so as to improve gene clustering. By evaluating the y-intercept of the red line, we pinpointed 4 as the optimal soft threshold (as shown in Fig. 5A). Subsequently, we went on to analyze the correlation between these modules and the traits we were interested in (see Fig. 3B, 3C, 3D). Notably, we managed to identify two modules that had a strong association with COVID-19, specifically the green and turquoise modules (as depicted in Fig. 5E and 5F). These two modules were then recognized as the crucial gene modules that underlie the intricate interplay between the COVID-19 group and the control group. The discovery of these crucial gene modules offers significant and valuable insights into the molecular mechanisms that govern the pathogenesis of COVID-19 as well as its connection with aging.

Fig. 5.

Construction of weighted gene co-expression (WGCNA) network. (A) Determination of soft-threshold power for COVID-19. (B) Clustering dendrogram of genes in co-expression modules. Distinct co-expression modules were shown in different colors. (C) Clustering of module eigengenes. (D) Correlation analysis between module eigengenes and clinical status. A module is represented by a row, and a clinical status is represented by a column. (E) Scatter plot between module membership in green module and the gene significance for COVID-19. (F) Scatter plot between module membership in turquoise module and the gene significance for COVID-19.

3.3. Unraveling aging-related genes in COVID-19

In order to identify the aging-related genes in the context of COVID-19, we conducted an intersection analysis among the genes that were identified through weighted gene co-expression network analysis (WGCNA), the differentially expressed genes, and the aging-related genes sourced from GeneCards (as shown in Fig. 6A). This intersection analysis led to the identification of 77 overlapping genes. Subsequently, we applied LASSO regression analysis to these 77 aging-related genes. As a consequence of this analysis, 5 genes were retained (as depicted in Fig. 6D). Box plots were then created to visualize the expression levels of these 5 genes, and it was found that these 5 genes demonstrated significant differential expression between the normal group and the COVID-19 group (as shown in Fig. 7). Therefore, we identified SST, S100A9, SOCS3, FKBP5, and HBB as the final key genes (as presented in Fig. 6D). The Receiver Operating Characteristic (ROC) analysis was utilized to assess the sensitivity and specificity of the test cohort (GSE188847) and the validation cohort (GSE164332), with the results being 0.96 and 0.87 respectively (as shown in Fig. 6E and F).

Fig. 6.

Identification of key aging-related genes. (A) 77 overlapping genes were obtained by intersecting the genes in WGCNA, DEGs and aging-relatived genes. (B,C) Coefficient profile plot of the LASSO model for COVID-19 showed the final parameter selection lambda. (D) The expression levels of key aging-related genes (SST, S100A9, SOCS3,FKBP5 and HBB)in the test (GSE 188,847). (E) Receiver operating characteristic (ROC) curves for evaluating the diagnostic values in the test cohorts (GSE188847). (F) Receiver operating characteristic (ROC) curves for evaluating the diagnostic values in validation cohorts (GSE164332).

Fig. 7.

the 5 genes demonstrated significant differential expression between the normal group and the COVID-19 group.

3.4. Investigation of the effect of key genes in COVID-19

To gain insights into the immune microenvironment of COVID-19, we carried out an in-depth analysis of immune cells within COVID-19 samples. At the beginning, we made use of the single-sample gene set enrichment analysis (ssGSEA) algorithm. This enabled us to quantify the relative scores of 28 distinct immune cell types in each individual sample. Our analysis revealed that sixteen immune cell types, including MDSC, Neutrophil, Natural killer cell, CD8 T cell, and others, showed significant differences (as depicted in Figs. 8A and 9B).

Fig. 8.

Analysis of immune landscape related to COVID-19. (A) showing the distribution of immune cells between the COVID-19 and normal samples of the GSE188847 dataset. *P < 0.05; **P < 0.01; ***P < 0.001;****P < 0.0001; ns, non-significant (P > 0.05). (B) Corrplot showing the correlation analysis of immune cells.

Fig. 9.

Drugs for the key aging-related genes.(A) Interaction network showcasing the connections between the 5 key aging-related genes. and 20 related genes, as revealed by analysis using the GeneMANIA database. (B) Seven drugs were obtained that were associated with key genes, including Valproic Acid, Folic Acid, S-Adenosylmethionine, Valproic Acid, Bucladesine, Fulvestrant, Kaempferol and Quercetin.

Moreover, with the aim of uncovering potential therapeutic targets, we delved into the interactions between the 5 core aging-related genes and 20 associated genes. We utilized the GeneMANIA database for this exploration (Fig. 9A). In parallel, to search for potential therapeutic agents linked to the key genes, we turned to the Comparative Toxicogenomics Database (CTD, http://ctdbase.org/). Here, we extracted relevant information regarding drugs or agents. Afterward, we used Cytoscape to visually represent the interaction network among these molecules (Fig. 9B). Significantly, we identified seven potential therapeutic compounds, namely Fulvestrant, Bucladesine, S - Adenosylmethionine, Valproic Acid, Folic Acid, kaempferol, and Quercetin, that were associated with the targeted key genes.

Overall, these findings play a crucial role in enhancing our understanding of the immune dysregulation in COVID-19. They also offer valuable perspectives on potential therapeutic approaches for managing this condition.

4. Discussion

Post-acute sequelae of SARS-CoV-2 infection (PASC) are characterized not only by cardiopulmonary complications, such as exertional dyspnea and sinus tachycardia, but also by significant neuropsychiatric morbidity. Emerging evidence highlights high-prevalence cognitive impairments, including attentional deficits, memory dysfunction, and pathological fatigue. These persistent neurological manifestations contribute to 39 % of disability-adjusted life years (DALYs) lost in PASC cohorts (Hänzelmann et al., 2013).

4.1. Neuroinvasive potential of SARS-CoV-2

Single-nucleus RNA sequencing reveals heterogeneous expression of viral entry receptors across neurovascular unit (NVU) components: ACE2 in astrocytes, high neuropilin-1 expression in pericytes, and upregulated vimentin in microglia post-infection. However, evidence of productive central nervous system (CNS) infection remains inconclusive: 1. Cerebrospinal fluid (CSF) viral RNA detection rates are below 5 % in autopsy series. 2. Intact virions are absent in 93 % of postmortem brain specimens.

These findings suggest that non-cell-autonomous mechanisms, such as cytokine-mediated neuroinflammation and autoantibody production, may play a dominant role in COVID-19 neuropathogenesis (Díez-Cirarda et al., 2023).

4.2. Neuroinflammatory mechanisms in COVID-19

COVID-19 induces neuroinflammation through dual pathways:

• 1.Peripheral immune activation: SARS-CoV-2-triggered cytokines compromise BBB integrity, promoting leukocyte transmigration into the CNS (Venkataramani and Winkler, 2022).
• 2.Direct neuro-immune crosstalk: CD8+ TEMRA cells (CD45RA+CCR7-) infiltrate the prefrontal cortex, releasing TNF-α (inducing synaptic pruning) and granzyme B (triggering neuronal apoptosis via caspase-3 activation) (Greene et al., 2024; Ellingjord-Dale et al., 2024).

4.3. Transcriptomic analysis of postmortem frontal cortices

Our analysis of postmortem frontal cortices (n = 18 COVID-19 vs. 12 controls) revealed:

• ‐Upregulated differentially expressed genes (DEGs, n = 81): Enriched in profibrotic pathways and NLRP3 inflammasome activation.
• ‐Downregulated DEGs (n = 441): Involved in synaptic plasticity (GO:0048,167, FDR=0.008) and GABAergic transmission (KEGG hsa04727, p = 0.0004).

The immune-cognition axis was further evidenced by

 : C1q-mediated complement activation in 89 % of COVID-19 dementia cases.

Study with proteomic methods have revealed that COVID-19 patients presented heterogeneous neurological features associated with a single altered CSF proteome pattern, despite disease severity, pulmonary burden, or the presence of pronounced neuroimaging alterations. both cytokines were higher in cases with pronounced neuroimaging alterations (Chen et al., 2024; Hanson et al., 2023; Acharya et al., 2023). This is consistent with our transcriptomics research findings.

4.4. Functional enrichment analysis

Significant associations (FDR<0.05) were identified across multiple pathological domains:

1. Pathogen defense mechanisms:

• ‐Anti-fungal response (GO:0050,832, p = 3.2 × 10⁻⁴).
• ‐Intracellular bacterial clearance (KEGG hsa05134, log2FC=2.1).

2. Fibrotic remodeling:

• ‐TGF-β/Smad3 signaling (Reactome R-HSA-2173,791, FDR=0.008).

3. Neuronal dysfunction:

• ‐Synaptic plasticity impairment.
• ‐Axon guidance disruption.

These findings align with cerebrospinal fluid (CSF) single-cell profiling (Maity et al., 2023), which demonstrated:

• ‐IL1B overexpression in 38 % of CD16+ monocytes.
• ‐A 57 % reduction in neuronal support clusters (OLIG2+/GFAP+) in COVID-19 dementia cases.

Further evidence of the immune-neural axis includes:

• ‐A strong correlation between CSF CCL2 levels and Montreal Cognitive Assessment (MoCA) score decline.
• ‐Complement C1q deposition on hippocampal synapses in 89 % of cases.

4.5. Key gene signatures in COVID-19 neuropathology

Using LASSO regression and weighted gene co-expression network analysis (WGCNA), we identified key genes distinguishing COVID-19 patients from controls, including SST, S100A9, SOCS3, FKBP5, and HBB. neuroinflammatory changes observed in COVID-19 patients overlap with signatures of brain aging and synaptic dysfunction, suggesting a potential shared pathway rather than a direct causal link.

SST (Somatostatin): This neuropeptide exhibits anti-inflammatory and neuromodulatory properties. Reduced SST levels in COVID-19 may contribute to dysregulated immune responses, neuroinflammation, and impaired synaptic plasticity, exacerbating cognitive dysfunction. SST dysregulation associates with GABAergic impairment. Downregulation of SST may promote neuroinflammation and reduce synaptic inhibition.

S100A9: Elevated levels of this pro-inflammatory protein are associated with cytokine storms and systemic inflammation. In the brain, S100A9 may disrupt the BBB and promote neuronal damage. Upregulation of S100A9 likely exacerbates inflammation and microglial activation.

SOCS3: Dysregulation of this cytokine signaling regulator may exacerbate chronic neuroinflammation and impair neuronal repair mechanisms. SOCS3 upregulation may feedback to dysregulate cytokine signaling.

FKBP5: Overexpression of this stress-responsive protein may dysregulate glucocorticoid receptor signaling, increasing susceptibility to neuroinflammation and cognitive impairment. FKBP5 upregulation could impair stress response and neuronal resilience.

HBB:The core argument is that detecting high levels of HBB (the gene for the beta-chain of hemoglobin) in postmortem brain tissue is more likely to be a sign of blood contamination from small vessels during the collection process than a sign of a disease process within the neurons themselves. Furthermore, common medical conditions (comorbidities) known to affect global gene expression should be considered as potential confounders. HBB expression changes may reflect hypoxic stress rather than direct aging regulation.

4.6. Therapeutic implications

Our drug repurposing analysis identified seven compounds with potential efficacy. Notably, kaempferol and quercetin exhibit anti-inflammatory properties but require structural modification to enhance BBB permeability. Fulvestrant and valproic acid represent repurposable agents with established CNS penetration profile.

4.7. Limitation

This study is constrained by several factors: (1) limited sample size (n = 44); (2) absence of experimental validation for key gene functions; (3) potential underrepresentation of SARS-CoV-2 variants and diverse patient populations; (4) exclusive focus on frontal cortex transcriptomics without multi-tissue validation; and (5) inability to establish causal relationships from cross-sectional data. Future prospective studies incorporating single-cell resolution and functional validation are warranted

5. Conclusion

Our study provides a comprehensive analysis of immune dysregulation and synaptic dysfunction underlying cognitive decline in COVID-19 patients. We identified key genes as potential biomarkers and therapeutic targets for addressing COVID-19-related brain fog. However, limitations include a small sample size, reliance on sequencing datasets without validation, and potential underrepresentation of SARS-CoV-2 variants and diverse patient populations.

In summary, dysregulated inflammation and synaptic dysfunction are central to the neurological sequelae of COVID-19. Key molecules such as SST, S100A9, SOCS3, FKBP5, and HBB offer promising avenues for further research and therapeutic development.

Present address

Department of Respiratory and Critical Care Medicine, Shanghai Jiaotong University School of Medicine Affiliated Renji Hospital.

Ethics approval and consent to participate

Ethical review and approval were not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Consent for publication

All authors have read and agreed to the published version of the manuscript.

Availability of data and material

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding authors.

CRediT authorship contribution statement

Cuiying Xie: Writing – original draft, Resources. Yunfeng Zhao: Writing – review & editing. Ben Wang: Methodology, Formal analysis. Teng He: Writing – review & editing. Jiajun Chen: Formal analysis. Xueling Wu: Supervision, Conceptualization.

Declaration of competing interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Appendix. Supplementary materials

Table S1. Differential gene expression between the COVID-19 and normal samples.

Table S2. Enrichment analysis of all differential gene expression.

Table S3. Signaling pathways of up genes of the DEGs between the COVID-19 and normal samples.

Table S4. Signaling pathways of down genes of the DEGs between the COVID-19 and normal samples.

Data availability

Data will be made available on request.