The potential of ATF5 as a prognostic marker and therapeutic target in cancer: insights from bioinformatics analysis and experimental validation

Introduction

Cancer is a persistent non-communicable disease that poses a substantial threat to human health, with increasing global incidence and mortality rates on the rise, constituting a huge public health challenge. According to recent projections from the International Agency for Research on Cancer, about 20 million new cases of cancer and 9.7 million cancer-related deaths were recorded globally in 2022. Projections indicate that the incidence of new cancer cases could reach 35 million by the year 2050 (1). The etiological aspects of cancer are complex, shaped by both intrinsic genetic factors and external environmental influences. Presently, multiple treatment options are available for cancer, with conventional therapies comprising surgery, radiation, chemotherapy, targeted therapy, and immunotherapy (2). Recent advancements in bioinformatics, leveraging public cancer datasets and repositories, have enabled the progression of extensive cancer multi-omics databases, offering enhanced understanding of tumor-associated biological traits and molecular pathways. Through pan-cancer investigations of oncogenes and tumor suppressor genes of interest, universal mechanisms of tumorigenesis can be elucidated, offering potential targets for overarching strategies in cancer therapy regimens. This method may result in more comprehensive and efficacious treatments, thereby improving patient prognosis and survival rates (3).

The activating transcription factor 5 (ATF5), a member of the ATF/CREB family, contains a basic leucine zipper domain that facilitates its participation in protein-protein interactions as well as DNA binding. Existing studies have demonstrated that ATF5 is implicated in the progression of various types of cancer, including glioma, hepatocellular carcinoma, cervical cancer, bladder cancer, and gastric cancer (4-8). ATF5 is acknowledged as a transcription factor that is highly expressed and facilitates carcinogenesis in various cancer tissues. Nonetheless, the precise activation and regulatory mechanisms of ATF5 in tumors necessitate more investigation (9). Therefore, we conducted preliminary in vitro studies to investigate whether the ATF5 protein possesses certain physiological properties. Interestingly, our research has found that ATF5 protein exhibits the ability to form aggregates in vitro, a property that has not been previously reported in earlier studies. This has established a foundation for subsequent research on the functional characteristics and regulatory mechanisms of ATF5.

This study utilizes bioinformatics analysis to explore the potential functions and fundamental mechanisms of ATF5 roles in human malignancies. We conducted an extensive analysis of ATF5 expression levels across all tumor types and association with prognosis using data from all The Cancer Genome Atlas (TCGA) cancer types, thereby developing pertinent prognostic models. This study offers profound insights into the processes of tumor immunomodulation associated with ATF5. Ultimately, we enhanced the characterization of genes and binding proteins linked to ATF5, examined the role of these related genes and binding proteins, together with their signaling pathways, and discovered probable important functions and routes by which ATF5 may affect carcinogenesis and progression. We performed targeted experimental validations in glioma, offering new insights for potential therapeutic immunotherapy of tumors. We present this article in accordance with the MDAR reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0146/rc).

Methods

Data preprocessing and differential expression analysis

The ATF5 transcriptome RNA-seq data, along with clinical data from normal and tumor tissues, were sourced from the TCGA (http://cancergenome.nih.gov) and GTEx (https://commonfund.nih.gov/GTEx) databases. We employed the SangerBox 3.0 online tool (http://sangerbox.com) to integrate and eliminate duplicates from the TCGA and GTEx datasets, thereby enabling an extensive study of ATF5 transcriptome data in both normal and tumor samples. The downloaded data were initially processed using R (version 4.4.3), where deduplication entailed averaging values, and samples with missing values were excluded. R software was used to run a Wilcoxon rank-sum test on the expression data as stratified by tumor stages to check for significance in ATF5 expression levels among the different groups, with a P value threshold of 0.05. The analytical outcomes were represented as violin plots utilizing R software.

Survival analysis

Survival data were obtained from samples acquired from TCGA. Overall survival (OS) and progression-free interval (PFI) were employed as metrics to examine the relationship between ATF5 expression and patient prognosis. We utilized R software to classify data into high-expression (> median) and low-expression (≤ median) groups according to the median expression level of the ATF5 gene, which acted as the grouping criterion. Subsequently, the log-rank test was utilized to assess the survival differences between the two groups, and the Kaplan-Meier technique was utilized to generate survival curves. Time-dependent receiver operating characteristic (ROC) curves at 1-, 3-, and 5-year intervals were calculated using time ROC analysis.

Relationship between ATF5 expression and immunity

The data was obtained from the standardized TCGA Pan-Cancer (PANCAN, N=10,535, G=60,499) dataset, retrieved from the UCSC database (https://xena.ucsc.edu/). Using R software, we acquired shared samples encompassing both ATF5 gene expression data and immune cell data by executing an intersection of sample identifiers inside the pan-cancer dataset. The values of the expressions were modified by utilizing the log₂(x+1) function. Furthermore, the analysis of Spearman correlation via CIBERSORT revealed significant connections between ATF5 expression levels and tumor immune infiltration levels. The gene expression data for glioblastoma multiforme (GBM) was obtained from the TCGA database. Target genes were discerned from the GBM gene expression data using R software. Common samples between the ATF5 expression data and immune cell data were identified by intersecting sample names, and immune cell types with a standard deviation of 0 were eliminated to assure data variability. We employed the Spearman correlation coefficient to assess the significance of relationships between ATF5 with various immune-infiltrating cells and immune checkpoints. A statistical significance was established at a threshold of significance value. Correlation coefficients were categorized into categories according to their magnitude: |r|<0.2, 0.2–0.4, >0.4–0.6, and >0.6. Additionally, Pearson correlation coefficients were computed independently for immune cells and immune checkpoints, with a consistent p-value threshold of 0.05 used for statistical significance. The stemness index (mRNAsi) for each tumor sample, computed by Malta et al. (10) using the one-class logistic regression machine learning technique on transcriptome data from TCGA datasets (table available at https://cdn.amegroups.cn/static/public/tcr-2026-1-0146-1.xlsx), was retrieved. The index indicates the activity of genes that govern stemness. We performed a Spearman correlation analysis to examine the association between ATF5 expression levels and mRNAsi, employing expression data obtained from the TCGA pan-cancer database with the use of R software. We computed the Spearman coefficient and generated a forest plot.

Functional analysis of ATF5-interacting proteins and co-expressed genes

To examine the proteins that associate with ATF5, we obtained pertinent data from the STRING database (https://cn.string-db.org/) and subsequently visualized it using Cytoscape (version 3.10.0). Additionally, we employed the ’similar Genes Detection’ option in the GEPIA2 tool (http://gepia2.cancer-pku.cn/) to select the top 100 genes exhibiting the most robust co-expression patterns with ATF5 across diverse datasets, based on the absolute values of Pearson correlation coefficients. The selected genes were further examined and analyzed for Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways utilizing the DAVID database (https://davidbioinformatics.nih.gov/). Ultimately, we produced bubble plots and chord diagrams with the aid of Sangerbox 3.0 (http://sangerbox.com/home.html), a program specifically developed for the analysis of biomedical data.

Establishment of tumor prognostic model

The TCGA database supplied gene expression and clinical information pertinent to low-grade glioma (LGG) and GBM. Using R software, we integrated survival time, survival status, and gene expression data based on sample IDs. The Cox proportional-hazards model was utilized to evaluate the correlation between gene expression levels in the samples and survival time/status, thence determining prognostic significance. Genes with P values below 0.05 were found by the log-rank test, subsequently followed by downstream regression analysis with the least absolute shrinkage and selection operator (LASSO)-Cox method. The best model was established by 10-fold cross-validation, with Lambda fixed at 0.117, finally discovering 21 genes whose expression alterations significantly influenced prognosis. We conducted Kaplan-Meier curve charting, ROC analysis, and produced expression heatmaps for these 21 genes.

Cell culture

The human glioma U87MG cell line was acquired from the Shanghai Cell Bank, Chinese Academy of Sciences. U87MG cells were grown in Advanced MEM (12492013, Gibco, Carlsbad, CA, USA). It is augmented with 10% fetal bovine serum (FBS, Gibco) and incubated at 37 ℃ in a 5% CO₂ atmosphere. In this work, the cells were regularly grown and passaged, and their mycoplasma-free status was verified.

Flow cytometry

ATF5-targeting small interfering RNA (siRNA) was transfected into U87MG cells. Subsequent to a designated duration of culture, the U87 cells were dissociated from the culture flask by using trypsin supplemented with ethylenediaminetetraacetic acid (EDTA). For additional examination, we used phosphate-buffered saline to resuspend the cells and enumerated them with a TC20 automatic cell counter. Subsequently, cells were suspended in buffer SR-VAD-FMK+ and 7-AAD+ and incubated at 25 ℃ for 20 minutes. The Guava Easycyte flow cytometer was employed for the flow cytometry analysis.

Plasmid construction

The coding region of the ATF5 gene was amplified from the U87 cell genome via RT-qPCR. Following restriction enzyme digestion and T4 ligase-mediated joining, the prokaryotic expression vector PET28a-EGFP-ATF5 was constructed. The completed recombinant plasmid was then transformed into BL21(DE3) competent cells (Escherichia coli, EC0114, Thermo Scientific, Massachusetts, USA) for in vitro expression of the EGFP-ATF5 fusion protein.

In vitro protein expression and purification

Induce the expression of EGFP-ATF5 recombinant protein in vitro using 0.1 mM IPTG. The His-tagged EGFP-ATF5 recombinant protein was purified using a Ni-NTA column. The purified protein was concentrated in an ultrafiltration tube and exchanged into phase separation buffer (containing 50 mM Tris-HCl, 150 mM NaCl, and 1 mM DTT) for subsequent fluorescence microscopy observation.

Tissue samples and immunohistochemistry (IHC)

This study utilized paraffin-embedded human glioma tissue sections provided by the Department of Pathology of the Affiliated Hospital of Qingdao University Medical College. Tissue sections were incubated at 60 ℃ for 2 hours, then dewaxed in xylene and rehydrated in a series of ethanol solutions. Samples were subsequently placed in 1 liter of sodium citrate buffer (pH =6) and microwaved for 45 minutes for microwave antigen retrieval. Endogenous peroxidase activity was blocked using 1% hydrogen peroxide in methanol to eliminate endogenous interference. ATF5 antibody (ab184923, Abcam, Abcam, Cambridge, UK, diluted 1:200) was applied and incubated for 1 hour. Staining was performed with 3,3'-diaminobenzidine (DAB; DA1015, Solarbio, Beijing, China) for 2 minutes. The slides were then rinsed with tap water, dehydrated, placed in xylene, and mounted at room temperature. A positive reaction is indicated by brown granules. Examine the stained tissue sections under a microscope (at 400× magnification) and calculate the percentage of positive cells. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the Qingdao Medical College, Qingdao University (No. QDU-HEC-2025279). Individual consent for this retrospective analysis was waived.

Differential expression of ATF5 in different grades of glioma

In the structured glioma expression dataset, we employed R software to conduct the Wilcoxon rank-sum test by systematically categorizing the samples according to distinct glioma grades [World Health Organization (WHO) grades II, III, and IV], age brackets (age >60 and ≤60 years), isocitrate dehydrogenase (IDH) mutation status, and the status of 1p/19q co-deletion. The importance of ATF5 expression levels within each group was assessed, with a P value of 0.05 deemed significant. The outcomes of the investigation were later depicted as box plots using R software.

Statistical analysis

Statistical analyses were performed using R software (version 4.4.3). The Wilcoxon test was used to compare gene expression levels between normal and tumor tissues. Statistical significance was set at P<0.05. In the survival analysis, the median ATF5 expression level was used as a threshold to divide the samples into high-expression and low-expression groups. The log-rank test was used to generate Kaplan-Meier survival curves and assess differences between groups. Statistical significance was set at P<0.05. Additionally, the Spearman correlation coefficient was used to assess the significance of the relationship between ATF5 expression levels and various immune infiltrating cells and immune checkpoints. In flow cytometry apoptosis analysis, a t-test was used to compare differences in apoptosis levels among various cell types. Statistical significance was set at P<0.05.

Results

ATF5 expression is significantly upregulated across various tumor types

The TCGA database indicates that ATF5 mRNA was markedly upregulated in fifteen human tumor types, including GBM, bladder urothelial carcinoma (BLCA), and breast invasive carcinoma (BRCA), relative to their neighboring normal tissues across all specified tumor types, with the most pronounced upregulation observed in GBM. In contrast, ATF5 expression was markedly reduced in three tumor types: cholangiocarcinoma (CHOL), liver hepatocellular carcinoma (LIHC), and pheochromocytoma and paraganglioma (PCPG). Additionally, no statistically significant differences in expression were observed in six tumor types, including kidney renal papillary cell carcinoma (KIRP), pancreatic adenocarcinoma (PAAD), and rectum adenocarcinoma (READ). Furthermore, there was an absence of ATF5 expression data from normal surrounding tissues available for nine tumor types, including adrenocortical carcinoma (ACC), LGG, and myeloid leukemia (LAML) (Figure 1A). Subsequently, acknowledging the constraints of the TCGA database, we utilized the Sangerbox 3.0 tool to amalgamate the TCGA and GTEx datasets to investigate ATF5 expression across six tumor types. The figure demonstrates that the expression of ATF5 was markedly increased in four types of tumors when compared to the corresponding adjacent normal tissues: LAML, LGG, ovarian serous cystadenocarcinoma (OV), and uterine carcinosarcoma (UCS). Conversely, ATF5 expression was significantly decreased in testicular germ cell tumors (TGCT). Nonetheless, no statistically significant variation in expression was seen in ACC (Figure 1B).

Figure 1 Expression of ATF5 in various normal and tumor tissues. (A) Differential expression of ATF5 mRNA in diverse human cancer types and their corresponding neighboring normal tissues according to the TCGA database. Statistical significance was assessed via the Wilcoxon test. (B) Comparison of ATF5 expression levels in ACC, LAML, LGG, OV, TGCT, and UCS. Statistical significance was assessed via the Wilcoxon test. (C) The expression of ATF5 has a substantial correlation with various pathological stages of tumors (P<0.05). *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001. (D) Comparison of ATF5 expression level differences between LGG and GBM (P<0.05). ACC, adrenocortical carcinoma; ATF5, activating transcription factor 5; GBM, glioblastoma multiforme; LAML, acute myeloid leukemia; LGG, low-grade glioma; OV, ovarian serous cystadenocarcinoma; TCGA, The Cancer Genome Atlas; TGCT, testicular germ cell tumors; TPM, transcripts per million; UCS, uterine carcinosarcoma.

Additionally, we assessed the correlation between ATF5 expression and different clinical stages of malignancies. In kidney renal clear cell carcinoma (KIRC) and KIRP, ATF5 exhibited a substantial correlation with pathological phases (P<0.05) (Figure 1C). Considering that the TCGA database categorizes gliomas into two categories—LGG and GBM—while the WHO categorizes both high-grade gliomas (WHO grades III and IV) and low-grade gliomas (WHO grades I and II) as central nervous system tumors (11), we additionally analyzed ATF5 expression levels between LGG and GBM using data sourced from TCGA. The results revealed a significant difference in expression levels between the two groups (P<0.05) (Figure 1D).

The level of ATF5 expression has a strong correlation with the prognosis in tumor patients

Considering the heightened expression of ATF5 in various tumors, we sought to examine its correlation with prognosis across different cancer types. We acquired clinical data for different tumor types from the TCGA database and conducted a correlation analysis of ATF5 expression levels with OS of patients. The findings demonstrated a significant association between higher ATF5 expression and poor OS in LGG, ACC, READ, uveal melanoma (UVM), colon adenocarcinoma (COAD), and KIRC (P<0.05) (Figure 2A-2G). Moreover, we examined the correlation between ATF5 expression levels and another survival metric: PFI across diverse tumors. The results indicated that increased ATF5 expression was markedly correlated with unfavorable progression-free periods in LGG, ACC, BLCA, and KIRC (P<0.05) (Figure 2H-2L). This finding further substantiates the idea that increased ATF5 expression correlates with adverse clinical features in patients with particular tumors.

Figure 2 The relationship between ATF5 expression levels and prognosis in cancer patients. (A) The Cox proportional hazards model was utilized to evaluate the relationship between OS and ATF5 expression. (B-G) The correlation between ATF5 expression levels and OS across different cancer types was illustrated using Kaplan-Meier curves. (H) The correlation between ATF5 expression and PFI was examined using the Cox proportional hazard model. (I-L) The relationship between ATF5 expression levels and PFI across different types of cancer was illustrated using Kaplan-Meier curves. ATF5, activating transcription factor 5; CI, confidence interval; HR, hazard ratio; OS, overall survival; PFI, progression-free interval.

Analyzing the immunological characteristics of ATF5 in the pan-cancer context, with specific analysis conducted in GBM

We evaluated the relationship between immunological checkpoints, immune-infiltrating cells, tumor stemness scores, and the ATF5 gene in tumors so as to investigate the immunomodulatory function of ATF5. First, we examined the correlation between ATF5 gene expression and immune-infiltrating cells. The ATF5 gene expression data in various tumor samples was obtained from the TCGA Pan-Cancer (PANCAN, N=10,535, G=60,499) dataset. A clustered heatmap was then employed to reassess the correlation between ATF5 expression levels and immune-infiltrating cells (such as B-cells, CD4⁺ T-cells, macrophages, etc.) in each patient across various tumor types based on gene expression patterns. The findings indicated that ATF5 gene expression levels in 33 tumor types showed substantial associations with the infiltration scores of the majority of tumor-infiltrating immune cells. ATF5 expression levels in both GBM and LGG exhibited strong associations with the immune infiltration of macrophages, encompassing M1 and M2 subtypes, in addition to mast cells, astrocytes, B-cells, CD4⁺ Tcm, eosinophils, fibroblasts, and monocytes (P<0.05) (Figure 3A). Notably, previous research has revealed that macrophages are the predominant immune-infiltrating cells in the glioma tumor microenvironment (TME), broadly categorized into M1 and M2 subtypes. The inflammatory mediators and chemokines prevalent in the glioma TME can recruit glioma-associated macrophages, facilitating tumor cell proliferation, invasion, and angiogenesis (12). Subsequently, we examined the influence of ATF5 on immune infiltration in GBM comprehensively. We collected gene expression data for GBM from the TCGA database, extracting expression values for the ATF5 gene alongside 47 widely recognized immune checkpoint genes, including BTLA, CD200, CD244, LAG3, ICOS, CD40LG, and CTLA4.

Figure 3 Immunological characteristics of ATF5 in various tumors. (A) The heatmap displayed is derived from the Spearman correlation coefficient and cluster analysis, demonstrating the association between ATF5 expression levels and diverse immune infiltrate cells across distinct cancer types. In this image, red is used to denote a positive connection, blue for a negative correlation, and white to represent no correlation. (B) Correlation analysis between ATF5 and immune checkpoints. The hue of connecting lines denotes the nature of the connection: red indicates a positive correlation, blue signifies a negative correlation, and gray symbolizes the absence of correlation. The thickness of the lines in the graphical depiction correlates directly with the absolute value of the correlation coefficient. The color gradient in cell-cell correlation represents the intensity of correlation between immune gene checkpoints, represented by the fill color within the rectangles. (C) Correlation investigation of ATF5 and immune infiltrating cells. The connecting lines are color-coded to denote correlation relationships: red signifies positive correlation, while gray shows no correlation. The thickness of the lines utilized in the graphical representation indicates a direct proportionality to the absolute value of the correlation coefficient. The cell-cell correlation color gradient depicts the strength of relationships among various immune cells, indicated by the fill color within the rectangles. (D) Differential analysis of ATF5 in 22 immune infiltrating cells. Each boxplot represents the immune infiltration level of corresponding immune cells under different ATF5 expression levels, with P values indicating statistical significance. (E) Bubble plot illustrating the Spearman correlation analysis between ATF5 expression level and tumor stemness index. The bubble size corresponds to the sample size, the numerical value on the horizontal axis denotes the Spearman correlation coefficient, and the color of the bubble signifies the P value. *, P<0.05; **, P<0.01; ***, P<0.001. ATF5, activating transcription factor 5; NK, natural killer.

Subsequently, we conducted a correlation analysis to assess ATF5 expression in relation to the expression levels of each immune checkpoint gene. The expression of the majority of checkpoint genes, such as CD200, TNFRSF14, LAIR1, TNFSE4, LAG3, CD48, CD276, LGALS9, TNFSF14, TNFSF9, TNFRSF25, TNFRSF4, CD40, TNFRSF18, and CD44, has a positive correlation with the amount of ATF5 expression. Conversely, substantial negative relationships were identified with the CD160 and ADORA2A checkpoint genes. No significant associations were seen between ATF5 expression and the CD40LG, CD80, BTNL2, or TIGIT checkpoint genes (Figure 3B). Meanwhile, we further analyze ATF5 expression levels in correlation with immune cell infiltration using GBM clinical patient’s gene expression data from the TCGA database. Based on immune cell expression values, we conducted immune infiltration scoring. A correlation study was conducted between ATF5 expression and the associated immune cell infiltration scores in the samples. Our results demonstrated a positive correlation between ATF5 expression and various immune cell types, including naive B cells, CD8⁺ T cells, and M1 macrophages, among others (Figure 3C,3D).

Finally, we examined ATF5 expression related to tumor stemness. The UCSC database was utilized to download a pan-cancer data set from TCGA to extract ATF5 expression data across samples. A correlation investigation between mRNA stemness index (mRNAsi) and ATF5 gene expression levels revealed substantial associations of the ATF5 gene exhibited significant correlations with tumor stemness indices across 20 cancer types. Among these, significant positive correlations were observed in 11 cancer types, including BRCA, gastric cancer (STES), KIRP, gastric adenocarcinoma (STAD), prostate cancer (PRAD), endometrial carcinoma (UCEC), KIRC, thyroid carcinoma (THCA), mesothelioma (MESO), PCPG, and ACC. Conversely, significant negative correlations were noted in 9 cancer types, including GBM, GBMLGG, LGG, COAD, colorectal adenocarcinoma (COADREAD), LAML, LAML, thymoma (THYM), TGCT, and UCS (Figure 3E). The examination of ATF5’s immunological features throughout pan-cancer and specifically in GBM indicates that ATF5 may influence tumor immune responses by regulating immune checkpoints, immune infiltration, or tumor stemness. This comprehension is essential for clarifying the immunoregulatory processes of malignancies and improving the efficacy of immunotherapy.

The expression of ATF5 is consistent with the expression trend of glioma poor prognosis associated genes

We used the Sangerbox tool in conjunction with R software to amalgamate the acquired LGG and GBM gene expression data with survival time, survival duration, and gene expression data for assessing the prognostic significance of specific genes. Genes exhibiting statistically significant differences were identified for downstream regression analysis. The model was confirmed by regression analysis, identifying 21 genes (the Lambda value is 0.117) most closely linked to glioma patient prognosis (Figure 4A,4B). These 21 genes underwent Kaplan-Meier curve charting (Figure 4C), ROC analysis, and development of heatmap generation. Based on patients’ survival status and survival time, we categorized them into two distinct groups: high-risk and low-risk mortality groups. The results indicated that among the 21 genes analyzed, the expression levels of EFEMP2, AC079944.2, CPQ, OSMR, FBXO17, TGIF1, SHISA5, ADPRH, CRNDE, POLR2J4, FAM86C1, EN1, GAS2L3, MGME1, and PBX3 demonstrated a positive correlation with patients’ mortality risk, exhibiting high expression in the high-risk mortality group. On the contrary, the research revealed that the expression levels of FRA10AC1, SLITRK5, GNL1, DESI1, ARHGAP12, and AC048382.5 had a negative connection, with high expression seen in the low-risk mortality group (Figure 4D). Subsequently, we employed Spearman’s correlation coefficient to examine the relationships between the expression levels of ATF5 and of 21 other genes. The findings indicate that the ATF5 expression is positively correlated with genes positively expressed in the high-risk mortality group, whereas it is negatively correlated with those genes that are primarily expressed in the low-risk mortality group (Figure 4E). This finding further substantiates that increased ATF5 expression in LGG and GBM may heighten patients’ mortality risk and is prognostically adverse.

Figure 4 Establishing a prognosis model for glioma based on prognostic-related genes. (A) The plot of the LASSO coefficient path illustrates how the log-transformed values of the regularization parameter λ are represented on the x-axis, with the matching regression coefficients displayed on the y-axis. Each curve represents the trajectory of a feature variable’s coefficient in relation to variations in λ, with distinct colors employed to distinguish among the variables. The optimal value of λ is indicated by a vertical dashed line. (B) Partial likelihood deviance versus λ plot: the x-axis represents the log-transformed values of λ, and the y-axis illustrates the partial likelihood deviation. The solid black line denotes the deviation curve, whereas the red dashed line depicts the 1 − standard error (1 − SE) line. Additionally, the optimal value is marked by a dashed vertical line at λ=0.12. (C) Kaplan-Meier curves for high-risk and low-risk cohorts.(D) Top picture: patients are classifed into high-risk and low-risk categories using the median value. Central image: scatter plot depicting survival durations. Bottom image: heatmap of gene expression profiles for genes involved in constructing the risk model. (E) Spearman correlation analysis indicates a substantial association of ATF5 with various genes implicated in the development of the risk model, where red denotes a positive correlation and blue indicates a negative correlation. ATF5, activating transcription factor 5; CI, confidence interval; H, high; HR, hazard ratio; L, low; LASSO, least absolute shrinkage and selection operator.

Functional clustering analysis of ATF5 interacting proteins and co-expressed proteins

We performed a systematic investigation of proteins interacting with ATF5 to investigate its potential signaling pathways and regulatory mechanisms in diverse malignancies. Initially, we utilized the STRING web tool to discover proteins that interact with ATF5 (interaction confidence threshold is 0.4) (Figure 5A), which were then subjected to GO enrichment analysis. The findings demonstrated that these proteins participate in various biological functions, encompassing the positive regulation of cytokine synthesis, negative regulation of cell death, generation of neurons, regulation of neuronal death, negative regulation of the cell cycle, negative regulation of the apoptotic process, negative regulation of programmed cell death, cell cycle activities, positive regulation of neuronal death, regulation of intrinsic apoptotic mechanisms, regulation of neuronal apoptotic processes, and positive regulation of cellular proliferation (Figure 5B).

Figure 5 Functional network of ATF5-interacting proteins and related genes. (A) The image depicts an interaction network of proteins linked with ATF5, which is illustrated in the figure. Each node in the network symbolizes a protein, with node size reflecting the number of interacting proteins (bigger nodes signify a greater number of interaction partners). The node color of the node denotes the clustering coefficient, where redder hues indicate higher coefficients, signifying more central positions within the network. Connecting lines between nodes signify the strength of connections, with thicker lines denoting more robust protein-protein interactions. (B) GO analysis of ATF5-interacting proteins, accompanied by a bubble chart illustrating select functionalities. (C) KEGG enrichment analysis of ATF5-interacting proteins, accompanied by a chord diagram illustrating many signaling pathways involving these interacting proteins. (D) KEGG enrichment analysis of genes significantly correlated with ATF5 expression levels, accompanied by a chord diagram illustrating selected signaling pathways involving these genes. (E) Bubble plot depicting the partial GO functions of genes highly correlated with ATF5 expression levels. (F-H) Flow cytometry was employed to assess cell apoptosis across three groups: the control group (control), 24 hours post-siATF5 (siATF5 24H), and 48 hours post-siATF5 (siATF5 48H). The cells in turquoise are living cells, the cells in blue are early apoptotic cells, the cells in pink are late apoptotic cells, and the cells in purple are deceased. (I) The apoptotic rates in the control group, siATF5 24H (24 hours post siATF5 therapy), and siATF5 48H (48 hours post siATF5 treatment). Three biological replicates were performed, with significant differences observed among groups (P<0.0001), indicating statistical significance. ***, P<0.01; ****, P<0.001. ATF5, activating transcription factor 5; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Next, we conducted a KEGG enrichment analysis on the pathways linked to ATF5-interacting proteins. The results indicated that these proteins were intricately associated with various signaling pathways and biological functions, including the tumor necrosis factor signaling pathway, viral carcinogenesis, human cytomegalovirus infection, the AMPK signaling pathway, the PI3K-Akt signaling pathway, human immunodeficiency virus 1 infection, pathways in cancer, and focal adhesion (Figure 5C). Subsequently, we analyzed 100 genes (PCC >0.3) demonstrating expression patterns analogous to ATF5 across samples using the GEPIA2 tool. A selection of these genes was selected for KEGG and GO enrichment analyses. The KEGG enrichment analysis results indicated that ATF5 may affect tumorigenesis via pathways such as pathways in cancer, the Ras signaling pathway, human cytomegalovirus infection, viral carcinogenesis, signaling pathways regulating stem cell pluripotency, and the peroxisome proliferator-activated receptor signaling pathway (Figure 5D).

Meanwhile, GO enrichment analysis revealed that most ATF5-related genes are implicated in apoptosis, proliferation, differentiation, metabolism, immunity, signal transmission, signal transduction, and protease inhibition (Figure 5E). In summary, through the expression and functional analysis of ATF5-interacting proteins and associated genes, we conclude that ATF5 is probably able to promote tumorigenesis by influencing the proliferation, differentiation, and apoptosis of tumor cells and the body’s immune response.

We performed pertinent experimental validation in GBM. We designed siRNA to reduce ATF5 expression in the glioma cell line U87MG and conducted flow cytometry analysis to assess the survival status of glioma cells across three groups: the control group, 24 hours post-siATF5 treatment, and 48 hours post-siATF5 treatment (Figure 5F-5I). We employed t-tests to analyze differences between groups. Results indicated that siATF5 treatment significantly increased the number of apoptotic cells over time. The apoptosis rate in the siATF5 48-hour group (39.36%) was approximately 2.5 times that of the siATF5 24-hour group (15.66%) and 5 times that of the control group (8.18%). The apoptosis rate in the siATF5 48-hour group was higher than that in the siATF5 24-hour group and the control group (P<0.05). These findings demonstrate that diminished ATF5 expression triggers apoptosis in glioma cells.

Differential expression of ATF5 in gliomas of different grades

Based on the above analysis, the expression of ATF5 exhibits highly significant differences between normal surrounding tissues and GBM, as well as notable discrepancies between LGG and GBM. Our research team earlier identified ATF5 as a transcription factor strongly linked to gliomas, with its expression level considerably impacting glioma progression (13,14). Therefore, we performed immunohistochemical analysis on the collected tissue sections to evaluate ATF5 expression in gliomas of different grades. Among these, the ATF5 positivity rate in the normal group was 1.40±1.02, while it was 25.80±0.98, 60.80±2.04, and 78.80±2.86 in the low grade, anaplastic, and GBM groups, respectively. Through analysis of variance, we determined that the ATF5 positivity rate in the GBM group was significantly higher than that in the LGG group (P<0.05). Results confirmed that ATF5 expression levels are positively connected with the pathological grade of gliomas (Figure 6A), as demonstrated by our online database (Figure 6B). Simultaneously, we also explored the correlation between ATF5 expression and other critical indications in gliomas. The results demonstrated that the expression of ATF5 was much higher in tumors with IDH wild-type compared to those with IDH mutant (Figure 6C). Furthermore, ATF5 expression was significantly increased in tumors lacking 1p/19q co-deletion relative to those with 1p/19q co-deletion (Figure 6D). Moreover, we analyzed the correlation between ATF5 expression and age, revealing that ATF5 expression was found to be significantly higher in the older age category (age >60 years), compared with the younger age group (age ≤60 years) (Figure 6E). In conclusion, ATF5 expression is intricately linked to various stages of glioma and has a notable association with patient age.

Figure 6 ATF5 expression differences in different grades of gliomas. (A) Expression levels of ATF5 in tissue sections across various glioma grades. In comparison to LGG and anaplastic, the immunohistochemical staining of ATF5 in GBM exhibited greater uniformity and intensity. Magnification: 400×. (B) Differential expression of ATF5 in gliomas across WHO grading classifications. (C) Differential expression of ATF5 in IDH wild-type and IDH mutant gliomas. (D) Differential expression of ATF5 in gliomas with and without 1p/19q co-deletion. (E) ATF5 expression differences among age groups. ***, P<0.001. ATF5, activating transcription factor 5; GBM, glioblastoma multiforme; IDH, isocitrate dehydrogenase; LGG, low-grade glioma; Mut, mutant; TPM, transcripts per million; WHO, World Health Organization; WT, wild type.

The ATF5 protein exhibits droplet-like aggregation in vitro

Through prior research and analysis, we have concluded that high expression of ATF5 is associated with multiple types of tumors. To investigate whether the ATF5 protein possesses specific characteristics that enable its physiological functions, we conducted relevant in vitro studies. First, the coding sequence (CDS) of the ATF5 gene was amplified from glioblastoma cells and cloned into the pET28a-EGFP vector, successfully constructing a prokaryotic expression recombinant plasmid. Following protein expression and purification, a high-purity recombinant ATF5 protein was obtained. We observed under fluorescence microscopy that ATF5 exhibits droplet-like aggregation in vitro. Building upon this foundation, we further investigated the effects of different in vitro conditions on the droplet-like aggregation characteristics of the ATF5 protein.

We conducted three replicate experiments on the same set of samples under different environmental conditions and performed quantitative statistical analysis of droplet count and droplet area based on the average values, with EGFP protein serving as the control group. We investigated the effect of protein concentration on phase separation by establishing different protein concentration gradients (protein concentrations of 5, 10, and 15 µM). As protein concentration increases, the number of droplets formed rises (with the fewest droplets at 5 µM and the most at 15 µM), the total droplet area expands (with the smallest area at 5 µM and the largest at 15 µM), and protein aggregation capacity enhances (Figure 7A). However, ATF5 protein failed to form droplet-like aggregates in vitro and showed no dependence on protein concentration. Experimental results indicate that ATF5’s capacity for aggregate in vitro is influenced by protein concentration. To determine whether crowding agents could promote the aggregation capacity of ATF5 protein, at an ATF5 protein concentration of 15 µM, we set up different polyethylene glycol (PEG) concentration gradients (concentrations of 0, 5% w/v, 10% w/v, and 20% w/v). Experiments revealed that at increasing PEG concentrations, protein aggregation became more pronounced across all PEG levels. However, only at 5% PEG did ATF5 form a significant number of droplets; at other PEG concentrations, irregular aggregates were predominantly observed (Figure 7B). However, EGFP protein fails to form droplet-like aggregates under different PEG concentration gradients. These results indicate that the aggregation process of the ATF5 protein in vitro is significantly influenced by PEG concentration, with low PEG concentrations (5% w/v) promoting the formation of droplet-like aggregates. In contrast, EGFP protein lacks the ability to form droplet-like aggregates in vitro. To investigate whether salt ions affect the aggregation of the ATF5 protein, we established different NaCl concentration gradients (0, 50, 100, 200, and 300 mM) at a protein concentration of 15 µM. As NaCl concentration increases, both the number of droplets formed by ATF5 protein aggregation and the droplet area decrease. High NaCl concentrations inhibited the aggregation of ATF5 protein (Figure 7C). Subsequently, we investigated the influence of hydrophobic interactions on in vitro protein aggregation by establishing 1,6-hexanediol concentration gradients (0, 5% w/v, 10% w/v). The results showed that the addition of 1,6-hexanediol weakened the aggregation capacity of ATF5, with this phenomenon becoming more pronounced as the 1,6-hexanediol concentration increased (Figure 7D).

Figure 7 Investigating the effects of different in vitro conditions on the aggregation of ATF5 protein in vitro. (A) Prepare three different ATF5 protein concentration gradients (5, 10, and 15 μM) using a dilution buffer of 50 mM Tris-HCl, 150 mM NaCl, and 1 mM DTT. Under the same conditions, EGFP protein fails to form droplet-like aggregates in vitro and is unaffected by protein concentration. (B) Add 50% PEG-8000 reagent to the ATF5 protein solution (15 μM concentration) to prepare protein mixtures with final PEG-8000 concentrations of 0%, 5%, 10%, and 20%. Incubate at room temperature for 30 minutes. Under the same conditions, EGFP protein fails to form droplet-like aggregates in vitro and is unaffected by PEG concentration. (C) Add 500 mM NaCl solution to the ATF5 protein solution (15 μM concentration) to prepare protein mixtures with final NaCl concentrations of 0, 50, 100, 200, and 300 mM. Incubate at room temperature for 30 minutes. (D) Add 50% 1,6-HD reagent to the ATF5 protein solution (15 μM concentration) to prepare protein mixtures with final 1,6-HD concentrations of 0%, 5%, and 10%; incubate at room temperature for 30 minutes. ATF5, activating transcription factor 5.

Discussion

The ATF family of transcription factors, a significant group of bZIP transcription factors, engages in cell proliferation, apoptosis, differentiation, and inflammation-related pathological processes through their homodimerization or heterodimerization with other bZIP factors. Numerous members of this family have been associated with the emergence of malignant tumors, as indicated by current research (15). ATF1 is significantly upregulated in lung cancer, promoting the migration and invasion of cancer cells (16). ATF2 promotes the progression of clear cell renal cell carcinoma. while exhibiting tumor-suppressive effects in skin cancer (17,18). In various malignancies, ATF3 has a range of actions that may even be contradictory. For example, in colon and hepatocellular carcinoma, it functions as a tumor suppressor (19-21), whereas it facilitates cancer progression in breast cancer and ovarian tumor (22,23). Similar to ATF3, ATF4 demonstrates many functions in cancer. For example, in gastric cancer, LXRβ suppresses proliferation and promotes apoptosis by upregulating ATF4 expression (24), while overexpression of ATF4 leads to increased cell viability, migration, and invasion in colorectal cancer cells (25). ATF6 functions as an unconventional biomarker in colon cancer (26), while simultaneously diminishing proliferation and stemness in colorectal cancer through the activation of UPR effector proteins (27). Overall, the role of the ATF family in cancer progression is significant, demonstrating notable variability among various cancer types. Existing research indicates that ATF5 participates in the progression of multiple cancers. It acts as an oncogene in most tumors, promoting tumor cell survival and proliferation. ATF5 exhibits high expression in pancreatic cancer cells, and targeting ATF5 significantly enhances paclitaxel-induced apoptosis in human pancreatic cancer cells (4). ATF5 promotes tumorigenicity in cervical cancer through the Wnt/β-catenin signaling pathway (6). It enhances tumorigenicity in bladder cancer and activates the Wnt/β-catenin pathway (7). METTL14 attenuates cancer stemness by inhibiting the ATF5/WDR74/β-catenin axis in gastric cancer (8). However, recent studies reveal that ATF5 exhibits tumor-suppressive functions in hepatocellular carcinoma. ATF5 inhibits autophagy by increasing mTOR expression and suppressing Wnt/β-catenin pathway activation, thereby inhibiting tumor growth and cancer stemness in vivo (5). This suggests that ATF5 exhibits distinct functional roles across different tumor types. When investigating ATF5’s function in various tumors in future studies, a comprehensive analysis is needed.

Pan-cancer analysis using public databases is constrained by the limited timeliness of existing online databases and variations in sample sizes collected, which may lead to discrepancies in analytical results and potential errors in some findings. We note that in their published work, Ishihara et al. (9) and Chen et al. (15) utilized the GEPIA2 database to demonstrate that ATF5 expression is upregulated in PAAD, SKCM, and THYM, while downregulated in LAML. In contrast, our integrated analysis of TCGA data using R software revealed no significant differences in ATF5 expression levels between PAAD, SKCM, THYM, and adjacent normal tissue. Furthermore, we observed that ATF5 expression was upregulated in LAML compared to adjacent normal tissue by integrating ATF5 expression data from both TCGA and GTEx databases. These differences may be due to variations in the analytical tools and methods employed, as well as differences in the sample sizes collected. This remains an issue requiring further investigation.

Previous studies have demonstrated that biomolecular aggregates are closely implicated in tumor pathogenesis (28), and multiple transcription factors have been shown to form biomolecular aggregates for gene regulation (29,30). The ability of proteins to form aggregates is the result of the synergistic interaction between intrinsic molecular factors (including the protein’s amino acid sequence, multivalency, charge, and conformation) and external environmental factors (including pH, temperature, ionic strength, and crowding agents) (31). This study preliminarily demonstrates that ATF5 can form aggregates in vitro, providing new insights for subsequent investigations into whether ATF5 also forms aggregates within tumor cells to regulate specific pathways and perform distinct functions.

Conclusions

In summary, this study systematically assesses ATF5 expression characteristics and prognostic value in cancer across multiple cancers by pan-cancer bioinformatics analysis, providing a theoretical basis for future functional research and clinical application.

Acknowledgments

We extend our gratitude to the Bioinformatics Platform at Qingdao University for providing server and computational resources.

Footnote

Reporting Checklist: The authors have completed the MDAR reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0146/rc

Data Sharing Statement: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0146/dss

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0146/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0146/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The Affiliated Hospital of Qingdao University and the School of Basic Medical Sciences at Qingdao University do not have independent ethics committees. Both institutions are guided by the Ethics Committee of the School of Medicine at Qingdao University. The study was approved by the Ethics Committee of the Qingdao Medical College, Qingdao University (No. QDU-HEC-2025279). Individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Duma N, Santana-Davila R, Molina JR. Non-Small Cell Lung Cancer: Epidemiology, Screening, Diagnosis, and Treatment. Mayo Clin Proc 2019;94:1623-40. [Crossref] [PubMed]
3. Liu J, Lichtenberg T, Hoadley KA, et al. An Integrated TCGA Pan-Cancer Clinical Data Resource to Drive High-Quality Survival Outcome Analytics. Cell 2018;173:400-416.e11. [Crossref] [PubMed]
4. Hu M, Wang B, Qian D, et al. Interference with ATF5 function enhances the sensitivity of human pancreatic cancer cells to paclitaxel-induced apoptosis. Anticancer Res 2012;32:4385-94.
5. Jiang J, Li X, Zhang H, et al. ATF5 inhibits autophagy and the Wnt/β-catenin pathway by upregulating mTOR to suppress the stemness of liver cancer stem cells. Tissue Cell 2025;97:103094. [Crossref] [PubMed]
6. Shi F, Wei Y, Huang Y, et al. Activating Transcription Factor 5 Promotes Tumorigenic Capability in Cervical Cancer Through the Wnt/β-Catenin Signaling Pathway. Cancer Manag Res 2025;17:131-43. [Crossref] [PubMed]
7. Zhou J, Tian H, Zhi X, et al. Activating transcription factor 5 (ATF5) promotes tumorigenic capability and activates the Wnt/b-catenin pathway in bladder cancer. Cancer Cell Int 2021;21:660. [Crossref] [PubMed]
8. Zhang P, Xiang H, Peng Q, et al. METTL14 attenuates cancer stemness by suppressing ATF5/WDR74/β-catenin axis in gastric cancer. Cancer Sci 2025;116:112-27. [Crossref] [PubMed]
9. Ishihara S, Enomoto A, Sakai A, et al. Stiff extracellular matrix activates the transcription factor ATF5 to promote the proliferation of cancer cells. iScience 2025;28:112057. [Crossref] [PubMed]
10. Malta TM, Sokolov A, Gentles AJ, et al. Machine Learning Identifies Stemness Features Associated with Oncogenic Dedifferentiation. Cell 2018;173:338-354.e15. [Crossref] [PubMed]
11. Louis DN, Perry A, Wesseling P, et al. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol 2021;23:1231-51. [Crossref] [PubMed]
12. Fu T, Dai LJ, Wu SY, et al. Spatial architecture of the immune microenvironment orchestrates tumor immunity and therapeutic response. J Hematol Oncol 2021;14:98. [Crossref] [PubMed]
13. Hu M, Li H, Xie H, et al. ELF1 Transcription Factor Enhances the Progression of Glioma via ATF5 promoter. ACS Chem Neurosci 2021;12:1252-61. [Crossref] [PubMed]
14. Hu M, Wang B, Qian D, et al. Human cytomegalovirus immediate-early protein promotes survival of glioma cells through interacting and acetylating ATF5. Oncotarget 2017;8:32157-70. [Crossref] [PubMed]
15. Chen M, Liu Y, Yang Y, et al. Emerging roles of activating transcription factor (ATF) family members in tumourigenesis and immunity: Implications in cancer immunotherapy. Genes Dis 2021;9:981-99.
16. Cui J, Yin Z, Liu G, et al. Activating transcription factor 1 promoted migration and invasion in lung cancer cells through regulating EGFR and MMP-2. Mol Carcinog 2019;58:1919-24. [Crossref] [PubMed]
17. Lu Z, Xu J, Li J. The Transcription Factor ATF2 Accelerates Clear Cell Renal Cell Carcinoma Progression Through Activating the PLEKHO1/NUS1 Pathway. Mol Carcinog 2025;64:617-28. [Crossref] [PubMed]
18. Zhang X, Li Z, Zhang X, et al. ATF family members as therapeutic targets in cancer: From mechanisms to pharmacological interventions. Pharmacol Res 2024;208:107355. [Crossref] [PubMed]
19. Rohini M, Haritha Menon A, Selvamurugan N. Role of activating transcription factor 3 and its interacting proteins under physiological and pathological conditions. Int J Biol Macromol 2018;120:310-7. [Crossref] [PubMed]
20. Inoue M, Uchida Y, Edagawa M, et al. The stress response gene ATF3 is a direct target of the Wnt/β-catenin pathway and inhibits the invasion and migration of HCT116 human colorectal cancer cells. PLoS One 2018;13:e0194160. [Crossref] [PubMed]
21. Lan T, Wang W, Zeng XX, et al. Saikosaponin A triggers cell ferroptosis in hepatocellular carcinoma by inducing endoplasmic reticulum stress-stimulated ATF3 expression. Biochem Biophys Res Commun 2023;674:10-8. [Crossref] [PubMed]
22. Dhamdhere SG, Bansal A, Singh P, et al. Hypoxia-induced ATF3 escalates breast cancer invasion by increasing collagen deposition via P4HA1. Cell Death Dis 2025;16:142. [Crossref] [PubMed]
23. Li X, Liu P, Sun X, et al. Analyzing the impact of ATF3 in tumorigenesis and immune cell infiltration of ovarian tumor: a bioinformatics study. Med Oncol 2021;38:91. [Crossref] [PubMed]
24. Liu J, Qi YB. Activation of LXRβ inhibits proliferation, promotes apoptosis, and increases chemosensitivity of gastric cancer cells by upregulating the expression of ATF4. J Cell Biochem 2019;120:14336-47. [Crossref] [PubMed]
25. Zhou Z, Ye S, Chen J, et al. ATF4 promotes glutaminolysis and glycolysis in colorectal cancer by transcriptionally inducing SLC1A5. Acta Biochim Biophys Sin (Shanghai) 2025;57:1093-105. [Crossref] [PubMed]
26. Hanaoka M, Ishikawa T, Ishiguro M, et al. Expression of ATF6 as a marker of pre-cancerous atypical change in ulcerative colitis-associated colorectal cancer: a potential role in the management of dysplasia. J Gastroenterol 2018;53:631-41. [Crossref] [PubMed]
27. Spaan CN, Smit WL, van Lidth de Jeude JF, et al. Expression of UPR effector proteins ATF6 and XBP1 reduce colorectal cancer cell proliferation and stemness by activating PERK signaling. Cell Death Dis 2019;10:490. [Crossref] [PubMed]
28. Ruan K, Bai G, Fang Y, et al. Biomolecular condensates and disease pathogenesis. Sci China Life Sci 2024;67:1792-832. [Crossref] [PubMed]
29. Ji D, Shao C, Yu J, et al. FOXA1 forms biomolecular condensates that unpack condensed chromatin to function as a pioneer factor. Mol Cell 2024;84:244-260.e7. [Crossref] [PubMed]
30. Frank F, Liu X, Ortlund EA. Glucocorticoid receptor condensates link DNA-dependent receptor dimerization and transcriptional transactivation. Proc Natl Acad Sci U S A 2021;118:e2024685118. [Crossref] [PubMed]
31. Villegas JA, Heidenreich M, Levy ED. Molecular and environmental determinants of biomolecular condensate formation. Nat Chem Biol 2022;18:1319-29. [Crossref] [PubMed]