Abstract

   Ulcerative Colitis (UC) is an inflammatory disorder characterized by
   chronic intestinal inflammation and immune dysregulation. Despite a
   clear association between cellular senescence and chronic inflammation
   and immune dysregulation, the mechanisms underlying cellular senescence
   in UC remain unclear. We screened differentially expressed genes (DEGs)
   associated with cellular senescence in multiple UC datasets, performed
   immune infiltration analysis, and constructed clinical diagnostic
   models. Additionally, we investigated the relationship between key
   genes related to cellular senescence and disease remission in UC
   patients undergoing biologic therapy, validating their expression in a
   single-cell dataset. We identified six DEGs associated with cellular
   senescence (TWIST1, IGFBP5, MME, IFNG, ME1, FOS). Immune infiltration
   results indicated strong correlations of four of these genes with
   immune cells and pathways. Through WGCNA, GO, and KEGG analyses, we
   found that gene modules strongly associated with the expression of hub
   genes in cellular senescence were enriched in inflammation-related
   pathways. In the single-cell dataset, the expression of these six key
   genes exhibited similarities with Immune infiltration results.
   Additionally, we constructed a nomogram using these six genes for
   diagnosing UC, demonstrating good diagnostic capability and clinical
   efficacy. Kaplan–Meier survival analysis revealed a significant
   association between changes in the expression levels of these cell
   genes and disease remission in UC patients undergoing biologic therapy.
   This study utilizes bioinformatic analysis and machine learning to
   identify and analyze features associated with cellular senescence in
   multiple UC datasets. It provides insights into the role of cellular
   senescence in the premature onset of intestinal aging in UC and offers
   new perspectives for exploring novel therapeutic targets.

   Keywords: Ulcerative colitis, Cellular senescence, GEO dataset,
   Biologics, Immune, Inflammatory bowel disease

   Subject terms: Immunology, Gastrointestinal diseases, Inflammatory
   bowel disease, Ulcerative colitis

Introduction

   Ulcerative Colitis (UC) is a predominant inflammatory bowel disease
   (IBD) affecting the colon and rectum, characterized by persistent and
   recurrent chronic intestinal inflammation with symptoms of mucous
   bloody stools^[38]1. The clinical course of UC is unpredictable,
   influenced by various factors such as genetic inheritance, dietary
   habits, environmental factors, immune system dysregulation, and gut
   microbiota imbalance^[39]2. In recent years, the global incidence of UC
   has been steadily increasing, posing a serious threat to public health,
   especially in developing countries^[40]2,[41]3. Currently, UC has no
   cure, relying mainly on long-term use of immunosuppressive drugs or
   steroids to control disease activity, which often entails significant
   side effects^[42]4. Thus, in-depth investigation into the pathogenesis
   of UC is essential for discovering new therapeutic targets and
   diagnostic markers, significantly impacting the alleviation of clinical
   symptoms in UC patients and improving the efficacy of UC drug
   treatments.

   UC is an immune system dysregulation-related inflammatory disease of
   the intestines. Disease progression in UC is closely associated with
   cellular senescence, although their interplay remains largely
   unexplored^[43]5. Cellular senescence is a cellular state linked to
   various physiological processes and a broad spectrum of age-related
   diseases. It is triggered by diverse stress signals throughout the
   cell’s lifespan, characterized by cell cycle arrest, a
   senescence-associated secretory phenotype (SASP), metabolic imbalance,
   and macromolecular damage^[44]6,[45]7. In a healthy immune system,
   aging cells within the intestines are effectively cleared to maintain
   gut homeostasis^[46]8. However, in UC, disrupted immune system
   homeostasis leads to the accumulation of senescent cells, which promote
   an inflammatory state associated with aging and exacerbate intestinal
   inflammation^[47]9,[48]10. Chronic inflammation in the intestines
   results in the continuous production of inflammatory factors, depleting
   adaptive immune responses, eventually leading to immune senescence.
   Immune senescence and intestinal inflammation can operate
   simultaneously, forming a detrimental feedback loop^[49]11.
   Additionally, aging epithelial cells weaken the intestinal mucosal
   barrier, allowing bacterial translocation, further worsening local and
   systemic inflammation^[50]12,[51]13. Although aging is commonly
   considered a major issue for the elderly population, patients with UC
   experiencing recurrent intestinal inflammation and immune system
   dysregulation may accelerate the onset of intestinal aging. Therefore,
   evaluating the involvement of genes related to cellular senescence in
   UC intestinal inflammation and immunity helps elucidate the mechanistic
   role of cellular senescence in UC progression.

   In this study, we downloaded three datasets containing gene expression
   data from UC colonic mucosa ([52]GSE107499, [53]GSE87466, [54]GSE59071)
   from the GEO database. After merging and screening for differentially
   expressed genes (log|FC|> 1, p-value < 0.05), we intersected these
   genes with those related to cellular senescence, obtaining 10
   differentially expressed genes (all upregulated). Subsequently, we
   utilized lasso regression to further narrow down the genes obtained in
   the previous step, ultimately identifying six upregulated genes related
   to cellular senescence (Fig. [55]1). Additionally, we observed the
   immune infiltration landscape in the UC intestines and constructed a
   nomogram with these six key genes for diagnosing UC, assessing the
   clinical efficacy of the diagnostic model. Finally, we downloaded
   dataset [56]GSE73661 to validate the performance of these six genes in
   UC biologic therapy (vedolizumab (VDZ) and infliximab (IFX)), and
   further validated the expression changes of these six genes in UC
   single-cell RNA dataset ([57]GSE221987).

Fig. 1.

   [58]Fig. 1
   [59]Open in a new tab

   The flow chart of the study, including experimental grouping design and
   research process. UC Ulcerative colitis, CS Cell senescence.

Materials and methods

Data source and clinical validation experiments of key genes

   The microarray data used for our analysis was obtained from the Gene
   Expression Omnibus (GEO) database
   ([60]https://www.ncbi.nlm.nih.gov/geo/), with the accession numbers
   [61]GSE107499, [62]GSE87466, [63]GSE59071, [64]GSE73661, and
   [65]GSE221987. A total of 312 samples of intestinal mucosal tissue were
   selected, including 236 samples from UC active patients and 76 samples
   of normal intestinal mucosal tissue. The gene expression data was
   normalized and merged for subsequent analysis. The [66]GSE73661 dataset
   was chosen to evaluate the impact of vedolizumab and infliximab on the
   expression of hub genes, identifying the utility of hub genes in
   predicting disease remission after biologic therapy. The [67]GSE221987
   dataset, which includes single-cell RNA sequencing results of colonic
   mucosa from four UC patients, was used to validate the expression
   patterns of key genes across various cells.

   Clinical specimens from UC patients undergoing colonoscopy were
   collected at the Second Xiangya Hospital of Central South University
   for experimental detection. This procedure was approved by the Ethics
   Committee of the Second Xiangya Hospital of Central South University
   (Approval No: 2022-155). We confirm that all methods were carried out
   in accordance with relevant guidelines and regulations. Informed
   consent was obtained from all subjects. We collected intestinal mucosal
   tissues from 10 UC patients in the active phase, with basic information
   shown in Supplementary Table [68]1. Normal tissues were taken from the
   normal colonic mucosa of the same UC patients. At week 8 after patients
   received regular vedolizumab monoclonal antibody treatment, we
   performed colonoscopy on these UC patients to assess the presence of
   drug remission. Response was defined as endoscopic mucosal healing
   (Mayo endoscopic subscore 0 or 1) and was assessed for vedolizumab at
   W8. The qRT-PCR primer sequences are shown in Supplementary Table 2.

Acquirement of differentially expressed cell senescence related genes

   We downloaded 432 and 949 genes associated with cellular senescence
   from the MSigDB (Molecular Signatures Database) genetic database^[69]14
   and the CellAge genetic database^[70]15 , respectively. After taking
   the intersection of the two sets, we obtained a total of 134 genes
   related to cellular senescence for subsequent analysis. Using the
   criteria of fold change (log|FC|) > 1 and p < 0.05 for differential
   gene expression, we applied the “limma” R package to identify genes
   with differential expression in the intestinal mucosa of UC patients.
   We then compared the differentially expressed genes with the cell
   senescence related genes and found a total of 10 upregulated genes and
   we created an Upset diagram to illustrate the results.

Diagnostic model built based on Lasso regression

   The ten upregulated genes identified previously underwent Lasso
   regression for further screening. Utilizing a tenfold cross-validation
   method, we iteratively developed a high-performance model with minimal
   variables. From these, six genes were selected and used to construct a
   diagnostic model (Nomogram). The expression matrix was randomly divided
   into training and validation sets in a 7:3 ratio. The training set was
   employed for model establishment, while the validation set assessed
   model performance.

Validation of the cell senescence related genes signature

   A predictive model was developed using six hub genes associated with
   cell senescence to distinguish UC patients from healthy individuals.
   The “pROC” package in R software was utilized to generate a receiver
   operating characteristic (ROC) curve and compute the area under the
   curve (AUC), which validated the performance of the predictive feature.
   Calibration curves were plotted to evaluate the calibration of the
   nomogram prediction model. Decision curve analysis (DCA) and clinical
   impact curve were performed to determine the clinical usefulness of the
   nomogram. DCA compared the net benefits of each prediction model at
   different threshold probabilities. Net benefit calculations considered
   the trade-off between missed interventions and the adverse effects of
   unnecessary interventions.

Analyses of immune landscape and immune correlation

   Single-sample gene set enrichment analysis (ssGSEA) is a machine
   learning algorithm used to analyze the enrichment of specific pathways
   in gene expression within individual samples^[71]16. The “CIBERSORT”
   algorithm is a gene expression-based deconvolution method that
   determines the cellular composition of tissues using the LM22 reference
   set^[72]17. We performed separate “CIBERSORT” and “ssGSEA” analyses on
   the selected 6 cell senescence related differentially expressed genes.
   We assessed the correlation between these genes and immune cell type
   enrichment scores, considering a p-value < 0.05 as statistically
   significant. Based on the criteria of having a correlation score
   |(Ps)|≥ 0.60 in CIBERSORT results and ssGSEA results, we identified 4
   cell senescence related genes that exhibited a statistically
   significant association with immune functions, and these 4 genes were
   used for subsequent WGCNA analysis.

WGCNA analysis

   WGCNA (Weighted Gene Co-expression Network Analysis) was applied to
   construct a gene co-expression network based on four genes identified
   through immune infiltration^[73]18. The module Preservation function
   within WGCNA assessed the preservation of network modules across
   cohorts and evaluated the influence of missing values. Z summary
   composite preservation scores were computed using 200 permutations,
   comparing each cohort or missing value threshold to the consensus
   network^[74]19.

Functional annotation and pathway enrichment analysis

   Gene Ontology (GO) enrichment analysis was performed to annotate gene
   sets into the biological process (BP), molecular function (MF), and
   cellular component (CC) terms^[75]20. The Kyoto Encyclopedia of Genes
   and Genomes (KEGG) pathway enrichment analysis was conducted to
   identify the signaling pathways related to DEGs. p < 0.05 was set as
   the criterion to identify the significantly enriched GO terms and KEGG
   pathways^[76]20.

Single-cell RNA-seq (scRNA) analysis

   Single-cell data analysis was conducted using the ‘Seurat’ R package.
   After importing the data, standardization was performed, filtering out
   cells with UMI counts exceeding 2500 or below 200, as well as cells
   with > 5% mitochondrial content^[77]21. From each sample, 2000 highly
   variable genes were selected for downstream analysis. The
   FeatureScatter function was employed to observe the correlation between
   different variables. Following clustering, 12 cell clusters were
   obtained, and manual cell annotation was performed. Figures
   illustrating the clusters, as well as marker expression in the
   low-dimensional space (UMAP), were generated using Seurat visualization
   functions^[78]22. Code examples for scRNA-seq can be found in the
   supplementary files.

Statistical analysis

   R software (version 4.2.0) was used for data processing and statistical
   analysis. Expression differences between patients with UC and controls
   were assessed using the Wilcoxon signed-rank and Kruskal–Wallis tests.
   Pearson and Spearman correlation analyses were employed to assess the
   statistical correlation between parametric and non-parametric
   variables. The p-value < 0.05 was considered statistically significant.
   (*p < 0.05, **p < 0.01, ***p < 0.001).

Results

Identification of cell senescence related DEGs

   Firstly, we merged three UC datasets and removed batch effects
   (Fig. [79]2A,B). The PCA plot demonstrates the intra-group clustering
   of UC and control groups (Fig. [80]2C). Filtering under the criteria of
   log|FC|> 1 and p < 0.05, we identified 489 upregulated genes and 220
   downregulated genes (Fig. [81]2D). The heatmap illustrates the
   differential expression of all genes between the two groups
   (Fig. [82]2E). Next, we downloaded genes associated with cellular
   senescence from the MSigDB and CellAge databases, resulting in 134
   cellular senescence genes (Fig. [83]3A). Intersection of these genes
   with 709 differentially expressed genes in UC yielded 10 differentially
   expressed genes, all upregulated (Fig. [84]3A). To further refine our
   target genes, we subjected the expression data of these 10 genes to
   Lasso regression. After selecting a model with excellent performance
   and the minimum number of variables, we identified six key genes
   (Fig. [85]3B,C). The heatmap depicts the upregulation of these six
   genes in UC (Fig. [86]3D). Correlation analysis reveals internal
   associations among these six differentially expressed genes
   (Fig. [87]3E), and the box plot illustrates the expression level
   variations of target genes between UC and normal control groups
   (Fig. [88]3F).

Fig. 2.

   [89]Fig. 2
   [90]Open in a new tab

   The three datasets [91]GSE107499, [92]GSE87466, and [93]GSE59071 were
   combined and normalized to create a new dataset. (A) Box plots before
   merging. (B) Box plots after merging. (C) Principal component analysis
   of the new dataset. (D) Volcano plot of differentially expressed genes
   (DEGs), where the red nodes represent the 489 significantly upregulated
   DEGs, and the blue nodes show the 220 downregulated DEGs based on log2
   |FC|> 1 and p < 0.05. (E) Heatmap illustrating the expression patterns
   of all DEGs in individual Ulcerative Colitis (UC) and normal samples.
   DEGs Differentially expressed genes.

Fig. 3.

   [94]Fig. 3
   [95]Open in a new tab

   Selection of differentially expressed genes related to cellular
   senescence. (A) UpSet plot showing the intersection quantity of
   upregulated and downregulated genes with cellular senescence genes,
   respectively. (B) LASSO coefficient distribution plot of differentially
   expressed genes related to cellular senescence. (C) Cross-validation
   plot of penalty terms, indicating the retention of 7 variables when the
   error is minimized; corresponding to the position of the left dashed
   line. To avoid overfitting and simplify the model, the selected
   standard error does not exceed 1, retaining 6 variables, corresponding
   to the position of the right dashed line. (D) Heatmap displaying the
   expression of these 6 key genes in the merged dataset. (E) Correlation
   heatmap of these 6 key genes. (F) Boxplot showing the expression level
   variation of target genes between UC and normal control groups.

The immune infiltration landscape of UC and the selection of immune-related
cell senescence DEGs

   Cellular senescence is associated with immune dysregulation, but the
   connection between changes in the immune microenvironment in the UC
   intestinal tract and cellular senescence remains unclear. Therefore, to
   further understand the pathogenesis of UC, we utilized two different
   algorithms, CIBERSORT and ssGSEA, to compare the differences in immune
   cell infiltration levels in the intestinal mucosa of UC patients.
   CIBERSORT analysis revealed that both UC and the normal control group
   showed activation of T cells in the intestinal tract. However, the
   infiltration level of NK cells in UC patients was significantly lower
   than in the control group. Additionally, there was a significant
   increase in M1-type pro-inflammatory macrophages and neutrophil
   infiltration in UC (Fig. [96]4C). In ssGSEA results, the infiltration
   level of CD8 T cells in UC patients was significantly higher than in
   the control group (Fig. [97]4D), which is also a hallmark of cellular
   senescence, indicating the presence of excessive cellular senescence in
   the UC intestinal tract. To further elucidate the connection between
   cellular senescence and the occurrence and development of UC, we
   conducted correlation analysis between the six key genes selected in
   the previous step and the abundance of immune cells. As expected, these
   six key genes showed a strong correlation with immune cells
   (Fig. [98]4A,B). Based on the correlation coefficients and p-values
   from the results of the two different algorithms, we selected four
   genes (TWIST1, IFNG, IGFBP5, MME) with the strongest correlations for
   WGCNA to construct a co-expression network.

Fig. 4.

   [99]Fig. 4
   [100]Open in a new tab

   Immune infiltration landscape of UC. (A) Correlation matrix of
   enrichment scores for cellular senescence-related DEGs calculated by
   CIBERSORT. (B) Correlation matrix of enrichment scores for cellular
   senescence-related DEGs calculated by ssGSEA. (C) Comparison of immune
   cell infiltration in UC and normal control groups calculated by
   CIBERSORT. (D) Comparison of immune cell infiltration in UC and normal
   control groups calculated by ssGSEA. (*p < 0.05, **p < 0.01,
   ***p < 0.001).

Construction of weighted gene co-expression network and functional annotation
of key modules for immune-related cell senescence DEGs

   Due to the association between cellular senescence and UC, it is
   essential to utilize WGCNA to identify genes co-expressed with target
   genes. We selected a soft-threshold of 8 (based on the scale-free
   topology criterion with R2 = 0.80) to construct a scale-free network.
   The adjacency matrix was then transformed into a TOM (Topological
   Overlap Matrix) matrix (Fig. [101]5A), which represents the similarity
   between nodes by considering weighted correlations. A total of 26 gene
   modules were obtained (Fig. [102]5B), with all modules having Z-scores
   greater than 10 (Fig. [103]5C). Correlation analysis was performed
   between all modules and hub genes, revealing strong correlations
   between the turquoise, red, pink, blue and brown modules with these 4
   genes (Fig. [104]6A). Subsequently, differential gene analysis was
   conducted on the genes belonging to these five modules, using a
   threshold of |(FC)|> 1 and p < 0.05 (Fig. [105]6C). This analysis
   resulted in the identification of 662 differentially expressed genes,
   including 435 upregulated genes and 227 downregulated genes
   (Fig. [106]6B). Finally, "GO and KEGG enrichment analysis" was
   conducted to explore the potential functions and mechanisms associated
   with UC using these DEGs. The majority of biological processes were
   found to be enriched in leukocyte migration, and the GO molecular
   function analysis revealed enrichment in cytokine activity
   (Fig. [107]6D). KEGG enrichment analysis of these DEGs demonstrated
   significant enrichment in inflammation-related pathways (Fig. [108]6E),
   aligning with the known pathogenesis of UC.

Fig. 5.

   [109]Fig. 5
   [110]Open in a new tab

   WGCNA analysis. (A) Relationship between the scale-free fitting index
   and various soft-thresholding powers, as well as the relationship
   between average connectivity and various soft-thresholding powers. (B)
   Hierarchical clustering dendrogram of genes, with different colors
   representing different modules. (C) Module preservation calculated
   using a composite Zscore. The blue dashed line indicates a Zscore of 2,
   and the dark green dashed line indicates a Zscore of 10. Modules
   preserved above this threshold are considered to have high statistical
   significance.

Fig. 6.

   [111]Fig. 6
   [112]Open in a new tab

   Four immune-related cellular senescence-associated differentially
   expressed genes (TWIST1, IGFBP5, MME, IFNG). (A) Pearson correlation
   analysis results between merged modules and immune-related cellular
   senescence DEGs. (B) Differential analysis of gene modules (5 in total)
   with Ps > 0.6, visualized using a heatmap (|log2 FC|> 1, p < 0.05). (C)
   Volcano plot showing the expression of differentially expressed genes.
   (D) Gene Ontology (GO) analysis. (E) Kyoto Encyclopedia of Genes and
   Genomes (KEGG) functional enrichment analysis.

Constructing and validating a nomogram for the diagnosis of UC

   We employed ROC analysis to assess the ability of these six key genes
   to distinguish between UC patients and healthy individuals. The results
   demonstrated good predictive value for diagnosing UC, with AUC values
   exceeding 0.75 for each gene (Fig. [113]7A,B). To further enhance the
   diagnostic capability, we randomly divided the expression data of these
   six genes into training and validation sets at a 7:3 ratio. In the
   training set, we constructed a nomogram for diagnosing UC
   (Fig. [114]7C). Calibration curves of the generated model showed
   minimal deviation from the ideal line, indicating good diagnostic
   consistency and high calibration accuracy (Fig. [115]7D). The c-indices
   for the nomogram in the training and validation sets were 0.9551 and
   0.9973, respectively (Fig. [116]8A,B). Decision Curve Analysis (DCA)
   was used to evaluate the clinical utility of the model by assessing the
   net benefit of interventions based on model predictions. DCA indicated
   that, as the threshold probability varied, the net benefit of
   intervention based on model predictions exceeded that of complete
   intervention or non-intervention, suggesting greater net benefit from
   clinical interventions guided by this decision curve (Fig. [117]8C).
   Clinical impact curves were generated to analyze the number of patients
   classified as high-risk and the true positive patients at various risk
   thresholds (Fig. [118]8D). To strike a balance between high net benefit
   and a low false-positive rate, DCA was combined with clinical impact
   curves. Calibration suggested that setting the risk threshold for UC at
   0.70 or higher could provide higher clinical benefits and a lower
   false-positive rate for the entire population under consideration.

Fig. 7.

   [119]Fig. 7
   [120]Open in a new tab

   Construction of the nomogram for diagnosing UC. (A) ROC analysis
   demonstrating the diagnostic capability of the six key genes for UC.
   (B) AUC values. (C) Nomogram construction, with red dots representing
   randomly selected samples for demonstration. (D) Model calibration
   curve indicating a small deviation between the line and the calibrated
   line from the ideal line, indicating good diagnostic consistency and
   high calibration accuracy.

Fig. 8.

   [121]Fig. 8
   [122]Open in a new tab

   Validation of the nomogram for diagnosing UC. (A) Diagnostic capability
   of the model for UC in the training set (AUC = 0.9551). (B) Diagnostic
   capability of the model for UC in the test set (AUC = 0.9973). (C)
   Decision curve analysis of the nomogram prediction model, with the
   light red curve representing the 95% confidence interval; the net
   benefit of intervention based on the model’s predicted values is higher
   than full intervention or no intervention as the threshold probability
   changes. (D) Clinical impact curve of the nomogram prediction model,
   where the red solid line represents the number of UC patients
   classified by the nomogram at each risk threshold among every 1000
   patients, and the blue dashed line represents the actual number of UC
   patients at each risk threshold.

Differential expression and predictive value of 6 cell senescence genes in
the intestinal mucosa of UC patients before and after biologic therapy

   The [123]GSE73661 dataset comprises gene expression data from the
   intestinal mucosa of UC patients before and after treatment with
   vedolizumab (VDZ) and infliximab (IFX). The dataset includes 12
   individuals in the healthy control group, 23 individuals receiving IFX
   treatment, and 37 individuals receiving VDZ treatment.

   In UC patients undergoing IFX treatment, ROC analysis revealed a good
   predictive ability of the six genes for the effectiveness of biologic
   therapy. Except for FOS, the AUC values for the other genes exceeded
   0.70 (Fig. [124]9A,B). Consistent with previous results, the expression
   of these six genes was upregulated in these UC patients
   (Fig. [125]9C–H). In non-responsive patients after IFX treatment, the
   expression of these six genes showed no significant change compared to
   pre-treatment levels. However, in patients responding to IFX treatment,
   the expression of the other five genes, except for FOS, was
   downregulated compared to pre-treatment levels, with no significant
   difference from the expression levels in the normal control group.
   Similarly, in UC patients receiving VDZ treatment, ROC analysis showed
   a good predictive ability of the six genes for the effectiveness of
   biologic therapy, with AUC values exceeding 0.70 for all genes except
   FOS (Fig. [126]10A,B). Again, these six genes were upregulated in these
   UC patients (Fig. [127]10C–H). In non-responsive patients after VDZ
   treatment, the expression of these six genes showed no statistically
   significant change compared to pre-treatment levels. However, in
   patients responding to VDZ treatment, the expression of the other five
   genes, except for FOS, was downregulated compared to pre-treatment
   levels, but only MME and TWIST1 expression levels dropped to normal
   levels. Due to limited data and the lack of specific follow-up time
   data for patients undergoing IFX treatment, we only used data from
   patients undergoing VDZ treatment for survival analysis. In the
   survival analysis, outcomes were defined as disease remission, with
   gene expression in the post-treatment matrix classified as ‘Yes’ if it
   was below the average pre-treatment expression level; otherwise, it was
   classified as 'No.' KM analysis showed that, except for FOS, the
   expression decrease of the other five genes was time-dependent with
   disease remission after VDZ treatment (Fig. [128]11A–F).

Fig. 9.

   [129]Fig. 9
   [130]Open in a new tab

   Expression changes of these 6 key genes in UC patients before and after
   IFX treatment. (A) ROC analysis showing the diagnostic capability of
   these genes for predicting treatment effectiveness. (B) AUC value.
   (C–H) Changes in the expression of these genes before and after
   treatment.

Fig. 10.

   [131]Fig. 10
   [132]Open in a new tab

   Expression changes of these 6 key genes in UC patients before and after
   VDZ treatment. (A) ROC analysis showing the diagnostic capability of
   these genes for predicting treatment effectiveness. (B) AUC value.
   (C–H) Changes in the expression of these genes before and after
   treatment.

Fig. 11.

   [133]Fig. 11
   [134]Open in a new tab

   Kaplan–Meier analysis depicting the probability of an effective
   response to VDZ treatment over time based on the expression changes of
   the 6 central genes. In the survival analysis, outcomes were defined as
   disease remission. Genes with post-treatment expression levels below
   the pre-treatment average were classified as "Yes," otherwise as "No."
   The shaded area in the graph represents the 95% confidence interval.

   To further explore the indicative role of cellular senescence in VDZ
   treatment, we used the ‘ConsensusClusterPlus’ package to cluster
   patients undergoing VDZ treatment based on the expression of key genes.
   The results showed that patients could be clustered into two groups
   (Fig. [135]12A–E), identified as the high-risk and low-risk groups.
   KM-plot results suggested that over time, patients in the high-risk
   group had a significantly lower response rate to VDZ biologic therapy
   compared to the low-risk group (Fig. [136]12F).

Fig. 12.

   [137]Fig. 12
   [138]Open in a new tab

   Consensus cluster analysis results. (A) Cluster matrix heatmap (K = 2).
   (B) Cluster matrix heatmap (K = 9). (C) Delta Area Plot. (D) Cumulative
   Distribution Function (CDF) plot, showing optimal clustering at K = 2.
   (E) Tracking Plot. (F) Stratification of all patients into high-risk
   and low-risk groups, with the high-risk group demonstrating
   significantly lower responsiveness to VDZ biological therapy compared
   to the low-risk group.

   Additionally, we detected the expression of these six key genes in the
   colonic mucosa of 10 UC patients treated with VDZ. Consistent with the
   bioinformatics analysis results, the expression of these six genes was
   significantly elevated in the inflammatory mucosa of UC patients
   (Supplementary Fig. [139]1). In the colonic mucosa of the six UC
   patients who achieved endoscopic remission after treatment, the
   expression of TWIST1, IGFBP5, MME, and ME1 significantly decreased,
   while the expression of FOS and IFNG did not show a significant
   decrease. In the colonic mucosa of the four UC patients who did not
   show a noticeable response to treatment, the expression of these six
   genes did not change significantly compared to before treatment.

The expression of 6 cell senescence genes in scRNA analysis

   We downloaded a single-cell RNA dataset containing four UC patients
   ([140]GSE221987) and performed single-cell data analysis using the
   ‘Seurat’ R package. The quality control plot displayed gene expression
   in each cell, including mitochondrial gene expression (Fig. [141]13A).
   The FeatureScatter function was employed to observe correlations
   between different variables (Fig. [142]13B). After clustering, 12 cell
   clusters were obtained, and manual cell annotation was performed
   (Fig. [143]13C). Similar to the previous results on immune
   infiltration, the expression of these six genes showed significant
   differences across various immune cell types. Additionally, TWIST1 and
   IGFBP5 were highly expressed in stem cells (Fig. [144]14A,C), MME
   showed high expression in neutrophils (Fig. [145]14E), IFNG exhibited
   high expression in CD4 T cells (Fig. [146]14B), ME1 was highly
   expressed in epithelial cells and macrophages (Fig. [147]14D), and FOS
   was expressed in various cell types (Fig. [148]14F).

Fig. 13.

   [149]Fig. 13
   [150]Open in a new tab

   Single-cell RNA analysis results. (A) Quality control plot (QC). (B)
   Feature Scatter plot displaying the correlation between different
   variables. (C) Cell annotation plot.

Fig. 14.

   [151]Fig. 14
   [152]Open in a new tab

   Validation of the expression of 6 cellular senescence genes in the
   single-cell dataset, consistent with immune infiltration results.

Discussion

   UC is a chronic inflammatory bowel disease characterized by recurrent
   inflammation primarily associated with immune dysregulation in the
   intestinal tract^[153]23. Recently, the association between cellular
   senescence and immune-related inflammation has gained clarity.
   Senescent cells, which cease division but secrete damaged and
   metabolically altered molecules, often resist apoptosis^[154]24. These
   cells frequently release cytokines via paracrine signaling, impacting
   tissue remodeling, promoting tumorigenesis, and sustaining chronic
   inflammation^[155]25. Identifying senescent cells molecularly presents
   challenges, as the senescent state induced by various triggers exhibits
   heterogeneity across tissues^[156]26. Therefore, understanding the role
   of cellular senescence in UC development is pivotal.

   In this study, we downloaded and analyzed UC-related datasets. Through
   differential analysis and Lasso regression, we identified six key
   cellular senescence genes, all of which were upregulated (TWIST1,
   IGFBP5, MME, IFNG, ME1, FOS). Using two algorithms, CIBERSORT and
   ssGSEA, we comprehensively investigated the immune infiltration
   landscape of these differentially expressed cellular senescence genes.
   Additionally, we constructed a diagnostic nomogram for UC using these
   six genes and validated its diagnostic capability and clinical
   application value. Furthermore, by assessing their expression levels,
   we evaluated their performance in predicting the response to biological
   therapies, particularly infliximab (IFX) or vedolizumab (VDZ). Finally,
   we validated the expression of these six genes in various cell types in
   UC single-cell data, and the results were consistent with the immune
   infiltration results.

   Twist1, a basic helix–loop–helix transcription factor, is implicated as
   a key mediator of epithelial–mesenchymal transition and is
   overexpressed in various cancer tissues^[157]27. TWIST1 regulates
   cellular senescence and energy metabolism in mesenchymal stem cells and
   plays a crucial role in epithelial-mesenchymal transition^[158]28. In
   osteoarthritis, the elevated expression of TWIST1 in cartilage
   exacerbates the catabolic response of human chondrocytes by
   upregulating MMP3 expression^[159]29. Furthermore, nuclear
   translocation of Twist1 represents a novel mechanism involved in the
   progression of osteoarthritis^[160]30. In IBD, the elevated expression
   of TWIST1 in intestinal tissues is associated with Th1-mediated
   intestinal immune responses, playing a crucial role in regulating the
   production of inflammatory cytokines by intestinal epithelial
   cells^[161]31,[162]32.

   Insulin-like growth factor (IGF) signaling plays a pivotal role in the
   regulation of cell growth, differentiation, apoptosis, and aging, with
   IGFBP5 being a significant member of the IGF axis^[163]33. IGFBP5 is
   induced in the cellular senescence process by the tumor suppressor
   p53^[164]34. In IBD, the heightened expression of IGFBP5 in
   inflammatory colonic tissues is linked to neutrophil
   gelatinase-associated lipocalin-mediated intestinal
   inflammation^[165]35,[166]36.

   Membrane metalloendopeptidase (MME) is a transmembrane glycoprotein
   with a distinctive extracellular protease domain capable of degrading
   various substrates^[167]37. MME can regulate the inflammatory response
   of adipocytes and insulin signaling^[168]38. Moreover, MME has been
   identified as a downstream effector of PI3K and can induce cellular
   senescence, a process that may be related to its glycosylation^[169]39.

   IFNG (IFN-γ) is a well-known pathway crucial in the immune response to
   inflammatory diseases^[170]40. IFNG activates innate immune responses,
   fostering the production of inflammatory cytokines and augmenting
   antigen presentation and natural killer cell function^[171]41. In IBD,
   IFNG contributes to disease pathogenesis by disrupting the vascular
   barrier through the upregulation of VE-cadherin^[172]42. The
   pro-inflammatory cytokine IFNG is essential for inducing cellular
   senescence^[173]43, and in turn, senescent cells can activate IFN
   signaling^[174]44.

   Malic enzyme 1 (ME1) is a cytoplasmic protein that catalyzes the
   conversion of malate to pyruvate, concurrently producing NADPH from
   NADP^[175]45. ME1 can also promote lipid synthesis and the onset of
   colorectal cancer^[176]46. In our analysis, although ME1 exhibited high
   expression in UC, it did not demonstrate a robust correlation in immune
   infiltration results and did not yield favorable outcomes in KM
   analysis after VDZ treatment (p > 0.01).

   FOS belongs to the cell cycle-related genes implicated in cell
   proliferation, growth, and differentiation, serving as a regulatory
   factor in cellular proliferation, differentiation, and
   transformation^[177]47. In specific contexts, the expression of the FOS
   gene is also associated with cellular senescence^[178]48,[179]49. In
   IBD, FOS mRNA levels are markedly expressed in UC but not in Crohn’s
   disease^[180]50. Likewise, high expression of FOS is observed in
   colonic tissues of DSS-induced colitis mice^[181]51.

   Diagnosing ulcerative colitis (UC) in patients can be challenging and
   often requires a combination of clinical history, endoscopic
   examination, pathological assessment, and imaging studies, which may
   lead to potential misdiagnosis^[182]1,[183]4,[184]23. The nomogram
   developed in this study demonstrated strong predictive capability and
   clinical utility for diagnosing UC, offering valuable guidance for IBD
   physicians in clinical decision-making. Biological therapies, such as
   vedolizumab and infliximab, are increasingly utilized in the treatment
   of moderate to severe UC due to their favorable therapeutic
   effects^[185]52,[186]53. Presently, biological agents are relatively
   expensive, with no definitive indicators available to accurately
   predict patient responsiveness to these drugs before use. Hence, we
   endeavored to utilize these six cellular senescence-related genes to
   forecast the response of UC patients to biological therapy.
   Intriguingly, our research results indicate that these six key genes
   are highly expressed in UC before treatment, and their post-treatment
   downregulation effectively predicts disease remission, demonstrating a
   robust correlation. Conversely, the expression changes of these six
   genes in UC patients who did not respond to treatment after therapy
   were not statistically significant. Additionally, single-cell RNA
   analysis further validated the expression of these six genes in UC.

   This study identified and analyzed features related to cellular
   senescence across multiple datasets for UC through bioinformatics
   analysis and machine learning. The alterations in expression observed
   in the intestinal mucosa of UC patients were confirmed in both a
   single-cell RNA dataset and a dataset derived from biological treatment
   for UC. As the era of an aging population approaches, aging is not only
   manifested as degenerative changes in the macroscopic world but also as
   dysregulation of cellular senescence in the microscopic world, which is
   closely associated with various diseases, especially those in the
   elderly. UC is a lifelong disease that cannot be completely cured.
   Under the stimulation of recurrent, chronic intestinal inflammation,
   the continuous disruption of the immune system in the intestine leads
   to abnormal expression of multiple genes controlling cellular
   senescence. By detecting the expression of these genes, we can predict
   the extent of cellular senescence changes in the UC intestine, which,
   to some extent, reflects local inflammation changes in the UC intestine
   and can predict the response to biologic therapy. In summary, our
   findings suggest that, unlike general population aging, intestinal
   aging in UC patients may occur prematurely. Through bioinformatics
   analysis, we have deepened our understanding of the role of cellular
   senescence in the development of UC and provided new avenues for
   exploring novel therapeutic targets for UC.

Supplementary Information

   [187]Supplementary Information 1.^ (71.9KB, tsv)
   [188]Supplementary Information 2.^ (638.5KB, jpg)
   [189]Supplementary Information 3.^ (43.5KB, xls)
   [190]Supplementary Information 4.^ (101.2KB, r)
   [191]Supplementary Information 5.^ (3.1MB, tif)
   [192]Supplementary Information 6.^ (18.7KB, docx)
   [193]Supplementary Information 7.^ (53.3KB, xlsx)

Acknowledgements