Abstract

   Bladder cancer (BC) exhibits diversity in clinical outcomes and is
   characterized by heterogeneity. Anoikis, a form of programmed cell
   death, plays a crucial role in facilitating tumor invasion and
   metastasis. This study comprehensively investigated the genetic
   landscape of BC progression, identifying 300 differentially expressed
   Anoikis-related genes (DE-ARGs) through in-depth analysis of the
   [33]GSE13507 datasets. Functional enrichment analysis revealed
   associations with diverse diseases and biological processes. Employing
   machine learning algorithms, a logistic regression model based on nine
   marker genes demonstrated superior accuracy in distinguishing BC from
   normal samples. Validation in TCGA datasets highlighted the prognostic
   significance of LRP1, FASN, and SIRT6, suggesting their potential as
   cancer biomarkers. Particularly, FASN emerged as an independent
   prognostic indicator, regulating BC cell proliferation and metastasis
   through the Wnt/β-catenin pathway. The study provides crucial insights
   into altered genetic landscapes and potential therapeutic strategies
   for BC, emphasizing the significance of FASN in BC prognosis and
   progression.

   Keywords: Bladder cancer, Anoikis-related genes, FASN, Wnt/β-catenin
   pathway, Biomarker, Machine learning

1. Introduction

   With a tenfold higher risk of occurrence compared to women, bladder
   cancer (BC) is the tenth most prevalent cancer worldwide and the sixth
   most prevalent in men [[34]1]. BC has an increasing incidence with age,
   particularly in the age group of 60 and above. Elderly individuals are
   more susceptible to the influence of chronic diseases, environmental
   exposures, and genetic factors [[35]2]. Exposure to certain chemical
   substances in occupational environments, such as benzene and
   dimethylamine, is associated with the occurrence of BC [[36]3]. Certain
   industries, such as dyeing, rubber, and leather manufacturing, may be
   correlated with an increased risk of BC [[37]4,[38]5]. In addition,
   smoking is one of the primary risk factors for BC. Harmful substances
   in tobacco smoke can be transported to the bladder through the
   bloodstream, increasing the risk of developing BC. Smokers have a
   significantly higher risk of BC compared to non-smokers [[39]6].
   Despite advancements in treatment choices, effectively managing the
   disease remains challenging, given its elevated rates of recurrence and
   metastasis, along with its inherent heterogeneity. Consequently, it is
   imperative to pinpoint relevant biomarkers with predictive value for
   treatment response and clinical outcomes, aiming to forge tailored
   therapeutic strategies.

   In the process of cancer development, metastasis constitutes a crucial
   aspect of its malignant progression. Metastasis involves the intra-body
   migration and proliferation of cancer cells, with the involvement of
   the Extracellular Matrix (ECM) playing a pivotal role in this process
   [[40]7,[41]8]. The ECM not only provides structural support to cells
   but also regulates cell behavior by participating in cell signaling
   pathways [[42]9]. During the metastatic process, the interaction
   between cancer cells and the ECM plays a crucial role. However, once
   tumor cells detach from the ECM, a specific form of programmed cell
   death known as Anoikis is triggered [[43]10,[44]11]. Anoikis is a
   programmed cell death that occurs when cells lose their connection with
   the extracellular matrix, and it is of critical importance in
   maintaining tissue structure and preventing the migration of cancer
   cells to other sites. In the mechanisms of tumor invasion and
   metastasis, the inhibition or evasion of Anoikis plays a significant
   role [[45]12,[46]13]. Cancer cells demonstrate the ability to evade
   Anoikis, employing various strategies to maintain their survival and
   continue the process of metastasis. These strategies include the
   utilization of growth factors, activating survival signals through
   growth factor signaling pathways to suppress Anoikis [[47]14,[48]15].
   Additionally, the regulation of the pH levels in the cell's surrounding
   environment is a manifestation of cancer cell adaptability, mitigating
   the adverse effects of Anoikis [[49]16,[50]17]. Adapting to oxidative
   stress is also a crucial aspect of cancer cell survival and the
   metastatic process, achieved by modulating antioxidant systems.
   Interestingly, recent research suggests that anoikis-related genes
   (ARGs) may play a significant role in cancer therapy [[51]18,[52]19].
   ARGs belong to a category of genes that respond to stress in the
   extracellular environment and may be closely associated with processes
   such as maintaining cell survival and preventing apoptosis. In-depth
   studies of ARGs could provide new insights for the development of novel
   strategies in cancer treatment. This discovery offers a fresh
   perspective on cancer treatment research and applications, holding the
   potential to provide more effective therapeutic approaches for patients
   in the future.

   This study conducted an analysis using the [53]GSE13507 dataset,
   identifying 300 differentially expressed ARGs (DE-ARGs) that were
   predominantly enriched in various diseases. Utilizing machine learning
   algorithms, nine potential biomarkers were determined, and a logistic
   regression model based on these markers exhibited excellent performance
   in distinguishing BC from normal samples. Our findings delved into the
   diagnostic and prognostic value of FASN in BC, demonstrating its
   independence as a prognostic indicator and constructing a diagnostic
   model for comprehensive assessment of patient survival prospects.
   Finally, we explored the potential function of FASN in BC. This
   comprehensive study provides genetic and functional insights for the
   diagnosis, treatment, and prognosis of BC, offering new directions for
   future cancer research and therapy.

2. Materials and methods

2.1. TCGA-BLCA cohort and GEO cohort

   Through the Cancer Genome Atlas (TCGA) data portal
   ([54]https://gdc-portal.nci.nih.gov/), we successfully obtained
   Level-three transcriptome RNA sequencing data from patients with
   BC(including 19 normal samples and 412 tumor samples) and collected
   relevant clinicopathological features. Additionally, we extracted RNA
   sequence information from normal and BC samples using the [55]GSE13507
   (67 normal samples and 165 tumor samples) and [56]GSE3167 (9 normal
   samples and 41 tumor samples) datasets, laying the groundwork for
   further analysis. Moreover, we employed MSigDB
   ([57]https://www.gsea-msigdb.org/gsea/msigdb) for screening and
   identified 508 Anoikis-related genes (ARGs). These ARGs may play a
   crucial role in the development and progression of BC, providing
   compelling clues for a deeper understanding of its molecular
   mechanisms.

2.2. Differential expression analysis

   We conducted a comprehensive analysis of the expression profiles of 508
   ARGs obtained from the [58]GSE13507 database, encompassing both normal
   and BC samples. Utilizing the R platform and employing Student's
   t-test, we successfully identified differentially expressed ARGs
   (DE-ARGs) between these two sample sets. Employing a threshold of a
   p-value less than 0.05, we specifically defined genes showing
   significant differential expression between the normal and BC groups as
   DE-ARGs. This stringent screening process ensured our focus on genes
   that may play crucial roles in Anoikis regulation, allowing for
   in-depth exploration of their potential functions in the development of
   BC.

2.3. Functional enrichment of DE-ARGs

   “clusterProfiler” is an R package designed for bioinformatics and
   biostatistics analyses, primarily focusing on functional enrichment and
   pathway analysis [[59]20]. The main purpose of this package is to
   assist researchers in understanding the biological significance of
   differentially expressed genes within large-scale gene expression
   datasets. The “clusterProfiler” package provides a robust toolkit that
   enables researchers to systematically analyze gene expression data,
   gaining insights into the functionality of differentially expressed
   genes and their roles in biological processes. This is crucial for
   interpreting results from biological experiments, comprehending gene
   functions, and identifying key pathways associated with diseases
   [[60]21]. We utilized the “clusterProfiler” package in the R language
   to conduct in-depth exploration of the functional and pathway
   associations of the genes through Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathways. Comprehensive
   functional annotations in biological processes, cellular components,
   and molecular functions were performed using the “clusterProfiler”
   package. Additionally, we accurately identified the key signaling
   pathways involved in the DE-ARGs. For the Disease Ontology (DO)
   enrichment analysis of DE-ARGs, we employed the “clusterProfiler” and
   “DOSE” packages in R. “DOSE” is a package designed for bioinformatics
   and biostatistics analysis in the R language. Its primary purpose is to
   facilitate functional enrichment analysis using DO, aiding researchers
   in understanding the associations between gene sets and specific
   diseases. In the analysis, we set the minimum gene set to 5, the
   maximum gene set to 5000, and established statistical significance
   criteria with a significance level of P < 0.05 and a false discovery
   rate (FDR) < 0.1.

2.4. Analysis of DEGs between FASN-high and -low expression BC groups

   In the analysis of BC groups, we aimed to identify Differentially
   Expressed Genes (DEGs) based on the expression levels of FASN. The
   “limma” package was employed for this purpose, allowing us to pinpoint
   genes with distinct expression patterns in individuals with high and
   low FASN expression. These identified DEGs constitute valuable
   candidates for further investigation. To visually represent the
   results, a heat map was generated. These graphical tools provide a
   clear and intuitive depiction of the expression differences among the
   identified DEGs.

2.5. Construction of the LASSO model and SVM-RFE feature selection process

   Firstly, the LASSO regression method was employed using the “glmnet”
   package, with the response type set to binomial and an alpha value of
   1, to screen genes exhibiting significant differences in the diagnosis
   of BC. This step distinctly identified potential biomarkers,
   contributing to precise BC diagnosis. Secondly, the Support Vector
   Machine Recursive Feature Elimination (SVM-RFE) algorithm was
   introduced as a supervised machine learning approach. SVM-RFE is a
   machine learning algorithm used for feature selection. It is a variant
   of Support Vector Machine (SVM) that focuses on iteratively eliminating
   less relevant features from a dataset. The goal is to find the subset
   of features that contributes the most to the predictive performance of
   the SVM classifier. Through iteratively eliminating irrelevant
   features, it gradually identified crucial biomarkers contributing to BC
   diagnosis. SVM-RFE utilized the SVM classifier from the R package at
   [61]http://github.com/johncolby/SVM-RFE, incorporating k-fold
   cross-validation to ensure model robustness and generalizability.
   Throughout the entire selection process, the number of features was
   halved in each round until fewer than 100 remained, ensuring that the
   selected biomarkers possessed better interpretability and
   applicability. Finally, by overlapping the genes obtained from both
   algorithms, a final set of candidate genes was determined, considered
   to have higher reliability and importance in the diagnosis of BC. To
   further validate these candidate genes, researchers turned to the
   [62]GSE3167 dataset, conducting an examination of the expression levels
   of these genes. The overarching aim of this study design was to
   integrate the strengths of LASSO regression and SVM-RFE to enhance the
   accuracy of BC diagnosis. Through validation on the [63]GSE3167
   dataset, the study reinforced the reliability and robustness of the
   candidate genes across multiple experimental datasets.

2.6. Construction and evaluation of the nomogram model

   Firstly, we conducted a thorough univariate Cox regression analysis to
   explore the potential prognostic value of critical diagnostic genes and
   clinical-pathological features in patient outcomes. This step aimed to
   identify factors with significant impacts in survival analysis,
   encompassing gene expression levels and clinical-pathological
   information. Subsequently, for a comprehensive understanding of the
   independent contributions of each factor to patient prognosis, we
   performed a multivariate Cox regression analysis. Through this step, we
   could further filter out key factors that maintained prognostic
   significance even when considering other variables, thereby obtaining a
   more reliable predictive model. Following that, we constructed a
   nomogram model using the “rms” package and the “survival” package in
   the R language. This model synthesized multiple factors influencing
   patient prognosis, including critical diagnostic genes and
   clinical-pathological features. The establishment of the nomogram aimed
   to provide an intuitive and effective tool to assist clinicians in
   evaluating the individualized prognosis of patients. Finally, to
   comprehensively assess the performance of the nomogram model, we
   employed metrics such as the concordance index (C-index), ROC curves,
   calibration curves, and decision curve analysis (DCA).

2.7. Cell culture and transfection reagents

   We utilized the SV40 immortalized uroepithelium cell line SVHUC-1 and
   human bladder cell lines T24, SW780, 5637, RT4, and J82. The selection
   of these cell lines was based on their widespread application and
   reliability in relevant research areas. All cell lines were sourced
   from the American Type Culture Collection (ATCC), with suppliers
   claiming regular testing for mycoplasma contamination to ensure the
   reliability of experiments and the credibility of data. To investigate
   the role of FASN in cell growth and metabolism in detail, we employed
   short hairpin RNAs (shRNAs) synthesized by Jikai Gene to suppress FASN
   expression. These shRNAs were meticulously designed and successfully
   cloned into the Lentivirus plasmid pcDNA-EF2-puromycin. Using this
   system, we were able to selectively silence the FASN gene, further
   exploring its regulatory mechanisms in cellular functions. To ensure
   effective delivery and action of shRNAs, Lipofectamine 2000
   transfection reagent was chosen for transfection experiments due to its
   efficient transfection performance, facilitating high-level expression
   of shRNAs within cells. Following transfection, the effectiveness of
   the process was assessed through molecular biology techniques such as
   Western blotting and RT-qPCR. In terms of culture conditions, we opted
   for Dulbecco's modified Eagle medium supplemented with 10 % fetal
   bovine serum, maintained at 37 °C and 5 % CO[2]. These culture
   conditions provided an optimal growth environment for cells, promoting
   the maintenance of their normal physiological state and ensuring the
   repeatability and stability of experimental results.

2.8. RNA extraction and QPCR

   In the pivotal steps of this study, the successful application of
   TRIzol reagent not only facilitated cell lysis but also enabled
   efficient purification of RNA through sequential addition of
   chloroform, 2-propanol, and 75 % ethanol. This ensured the acquisition
   of high-quality, purified RNA samples, laying a reliable foundation for
   subsequent experiments. The rigorous examination of the purity and
   concentration of purified RNA samples is a crucial step to ensure the
   accuracy of experimental data. The utilization of the Nanodrop 2000
   spectrophotometer allows researchers to precisely assess the quality of
   RNA samples, providing essential information for further analysis and
   experimental design. During the cDNA synthesis phase, the choice to use
   2 μg of purified RNA in conjunction with oligo dT primers and the
   RevertAid First Strand cDNA Synthesis Kit was made to ensure the
   generated cDNA possesses high quality and representative expression.
   This not only aids in preserving the integrity of RNA but also enhances
   the reliability of subsequent PCR amplification. PCR amplification
   represents a pivotal stage in the experiment, where precise control of
   temperature and time allows for the selective amplification of specific
   gene segments. The optimized PCR conditions, such as the initial
   incubation at 95 °C, denaturation at 94 °C, annealing at 55 °C, and
   extension at 72 °C, not only increase the yield of PCR products but
   also ensure the specificity and stability of the amplification.
   Primers' specific information was listed as follows: FASN forward
   primer (5‐AAGGACCTGTCTAGGTTTGATGC‐3) and FASN reversed primer
   (5‐TGGCTTCATAGGTGACTTCCA‐3); GAPDH (GAPDH, an endogenous control)
   forward primer (5‐CTGGGCTACACTGAGCACC‐3) and GAPDH reversed primer
   (5‐AAGTGGTCGTTGAGGGCAATG‐3).

2.9. Cell counting Kit-8 (CCK-8)

   In this experiment, we utilized a concentration of 2 × 10^4 cells per
   milliliter, allocating them into a 96-well plate with sufficient
   experimental space, allowing for concurrent experiments or testing of
   cells' responses under different conditions. The 96-well plate provided
   a versatile platform for conducting multiple experiments
   simultaneously. The Cell Counting Kit-8 (CCK-8) employed in this study
   is manufactured by Dojindo, based in Kumamoto, Japan. This solution is
   widely used for assessing cell viability, relying on the measurement of
   cellular metabolic activity. By adding 10 μL of CCK-8 solution to each
   well at predefined time points, we were able to monitor changes in cell
   metabolic activity throughout the experimental process. After a 2-h
   incubation period, we employed a microplate reader to measure
   absorbance at 450 nm. This specific wavelength measurement was chosen
   to accurately assess the reaction products of CCK-8, reflecting the
   cellular viability. Recording absorbance values provided us with data,
   enabling the construction of a cell viability curve. The cell viability
   curve is a dynamic graphical representation illustrating the survival
   status of cells at different time points. Through this approach, we
   could observe variations in cell viability over time, providing crucial
   insights into how cells respond under diverse conditions.

2.10. Colony formation

   In the experiments conducted using six-well plates, an initial
   inoculation of 500 cells was performed in each well. Following cell
   adhesion, various treatments were applied. After a period of 10–14
   days, the counting of clones was assessed. The cleaning of clones was
   carried out using phosphate-buffered saline (PBS), followed by fixation
   with formaldehyde for 30 min. Crystalline purple was used for staining
   the six-well plates, and the plates were then set aside for an
   additional 20 min. At the outset of the experiments utilizing six-well
   plates, 500 cells were initially seeded into each well, marking the
   commencement of the study. After allowing the cells to adhere to the
   plate surface, different treatments were introduced to observe their
   effects. After a duration of 10–14 days, an evaluation of clone
   proliferation was conducted. To ensure precise analysis, a meticulous
   protocol was employed for clone processing. Thorough washing with PBS
   was carried out to remove any residual substances. Subsequently,
   fixation of clones was achieved by treating them with formaldehyde for
   a duration of 30 min. In the subsequent steps, staining of the six-well
   plates was performed using crystalline purple, providing a visual
   marker for the assessment of cell clones. This staining process was
   executed with precision, and the plates were left aside for an
   additional 20 min to ensure optimal color development.

2.11. Cell apoptosis evaluation

   Then, we examined the occurrence of cell apoptosis in two specific cell
   lines, namely RT4 and SW780. To initiate the experimental process,
   these cultured cells underwent treatment with 20 μmol/l proteinase K
   from Sigma, followed by fixation using 4 % paraformaldehyde. The
   evaluation of cell apoptosis was conducted through the terminal
   transferase UTP nick-end labeling (TUNEL) assay, using a kit provided
   by Roche and following the recommended protocol from the kit
   manufacturer. Notably, in the TUNEL assay, apoptotic cells were marked
   with red fluorescence, which serves as a distinctive feature of this
   experiment. Subsequently, visualization of these fluorescently labeled
   cells was performed using an inverted fluorescence microscope.
   Following visualization, detailed analysis of the obtained images was
   carried out using the Image J software developed by the National
   Institutes of Health (NIH). This comprehensive approach allows for a
   detailed exploration of the cellular response to experimental
   conditions in RT4 and SW780 cells, providing valuable insights into the
   apoptotic status of the cells.

2.12. Wound healing assay

   In this experimental procedure, RT4 and SW780 cells in the logarithmic
   phase were meticulously prepared by seeding them in six‐well plates,
   followed by incubation at 37 °C until the cell monolayer achieved an
   approximate 90 % confluence. The subsequent step involved the
   initiation of a wound healing assay, wherein uniform and straight
   scratches were carefully introduced into the cell monolayers using
   pipette tips. To ensure the removal of any cellular debris, the
   scratched cell monolayers underwent a thorough wash with
   phosphate‐buffered saline (PBS). Subsequently, a crucial post-treatment
   phase unfolded, involving the incubation of cells with 2 mL complete
   culture medium for an additional 24 h. The dynamic process of wound
   healing was diligently observed and documented through the acquisition
   of images at both the initiation of the assay (0 h) and after the 24-h
   incubation period. To quantify and analyze the rate of wound healing,
   ImageJ software, a powerful tool developed by the National Institutes
   of Health (NIH), was employed. This comprehensive approach not only
   facilitated the visualization of the wound healing process but also
   enabled a quantitative assessment, providing valuable insights into the
   migratory and regenerative capabilities of RT4 and SW780 cells under
   the experimental conditions.

2.13. Cell invasion assays

   In preparation for invasion assays, Matrigel sourced from BD
   Biosciences, CA, underwent a meticulous dilution process to achieve a
   concentration of 1 mg/mL, utilizing DMEM without FBS medium. This
   appropriately diluted Matrigel was promptly applied to the upper
   chamber of the trans-well system. Following this application, a period
   of hydration lasting 3–4 h ensued. Subsequently, 1 × 10^5 cells in
   200 μl of DMEM without FBS medium were carefully introduced into the
   upper chamber. Simultaneously, the lower chamber received an addition
   of 800 μl DMEM supplemented with 10 % FBS per well, creating a
   conducive environment for cell migration. The assembled trans-well
   system was then subjected to a 24-h incubation period at 37 °C and 5 %
   CO[2]. Post-incubation, cells that had traversed to the bottom surface
   of the membrane were meticulously fixed with 100 % methanol for a
   duration of 30 min. Subsequent to fixation, a staining step ensued,
   involving the application of 0.1 % crystal violet for 10 min. Following
   staining, a thorough wash with water was performed three times to
   remove excess stain. The final step involved the counting and imaging
   of invading cells, a process carried out in at least 10 random fields
   under a light microscope. It is noteworthy that this comprehensive
   invasion assay protocol was executed three times independently,
   ensuring the reliability and consistency of the obtained results.

3. Western blotting

   In this experiment, transfected BC cells (BC cells) were lysed, and
   total protein was extracted using RIPA buffer, followed by
   quantification using a BCA kit. Subsequently, each protein sample
   (30 μg) underwent denaturation in sample buffer at 100 °C for 5 min and
   was subjected to 10 % sodium dodecyl sulfate-polyacrylamide gel
   electrophoresis for 50 min. The gel was then transferred onto a
   nitrocellulose membrane at a constant current of 200 mA for 60 min.
   After blocking the nitrocellulose membrane with 5 % skim milk at room
   temperature for 2 h, it was incubated overnight at 4 °C with a properly
   diluted primary protein antibody. The following day, the membrane was
   washed three times with phosphate-buffered saline with Tween 20 (PBST)
   solution, incubated with a secondary antibody at room temperature for
   45 min, and washed again with PBST. Finally, the membrane was developed
   using an ECL solution to capture the target band. This procedure allows
   researchers to analyze protein expression levels and detect the
   presence of specific proteins in the given cell samples.

4. Statistical analysis

   Graphpad Prism 8.0 and R version 3.5.2 were employed as key tools in
   this study for the generation of charts and comprehensive data
   analysis. The research methodology involved a meticulous approach,
   ensuring that results were derived from a minimum of three independent
   experiments. To ensure clarity and accuracy, experimental data were
   concisely presented as means with their respective standard deviations
   (x ± s). Statistical analysis was utilized to scrutinize the nuances
   within the data. The widely adopted Student's t-test was used for
   comparisons between two groups. In scenarios involving more than two
   groups, a choice was made between Student's t-test or ANOVA, depending
   on the context, with SPSS serving as the analytical platform. Through
   Kaplan-Meier analysis, we explored the potential association between
   FASN expression and survival outcomes in BC patients. An important
   consideration throughout the entire study was the adoption of a
   significance threshold of P < 0.05, designating findings below this
   predetermined level as statistically significant.

5. Results

5.1. Identification of the DE-ARGs in BC specimens and functional enrichment
analysis

   To systematically assess the genetic landscape associated with BC
   progression, we performed an in-depth analysis of the [64]GSE13507
   datasets employing the Student's t-test. Our investigation unveiled 300
   DE-ARGs in BC samples, delineating 149 genes experiencing
   downregulation and 151 genes undergoing upregulation, as depicted in
   [65]Fig. 1A. Then, we performed DO analysis to explore the possible
   association between 300 DE-ARGs and various diseases. As shown in
   [66]Fig. 1B, we found that 300 DE-ARGs were mainly enriched in
   non-small cell lung carcinoma, breast carcinoma, urinary system cancer,
   musculoskeletal system cancer, female reproductive organ cancer and
   male reproductive organ cancer. The results of GO analysis revealed
   that 300 DE-ARGs were mainly related to regulation of protein
   serine/threonine kinase activity, positive regulation of cell adhesion,
   cell-substrate adhesion, epithelial cell proliferation, focal adhesion,
   cell−substrate junction, cell leading edge, cell−cell junction, protein
   serine/threonine kinase activity, DNA-binding transcription factor
   binding, cadherin binding ([67]Fig. 1C). The results of KEGG analysis
   indicated that 300 DE-ARGs were mainly related to Pathways in cancer,
   Focal adhesion, Regulation of actin cytoskeleton, Small cell lung
   cancer and Prostate cancer ([68]Fig. 1D). Our findings provide a
   comprehensive insight into the altered genetic landscape during BC
   progression, pinpointing potential molecular markers and pathways that
   could be instrumental in advancing our understanding and potential
   therapeutic strategies for BC.

Fig. 1.

   [69]Fig. 1
   [70]Open in a new tab

   Identification of the DE-ARGs in BC specimens and functional enrichment
   analysis. (A) The analysis of [71]GSE13507 datasets (including 67
   normal samples and 165 tumor samples) using Student's t-test revealed
   300 DE-ARGs in BC samples, including 149 downregulated genes and 151
   upregulated genes. (B) Disease Ontology (DO) Analysis of DE-ARGs. The
   distribution of 300 DE-ARGs across various diseases, as shown in the DO
   analysis. (C) Gene Ontology (GO) Analysis of DE-ARGs. Functional
   enrichment analysis depicting the biological processes associated with
   the 300 DE-ARGs. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG)
   Analysis of DE-ARGs. Pathway enrichment analysis indicating the main
   pathways associated with the 300 DE-ARGs.

5.2. Identification and validation of diagnostic feature biomarkers

   In our quest to discern variations between BC and normal specimens, we
   sought to evaluate the diagnostic potential of Differentially Expressed
   Annotated Genes (DE-ARGs). Employing two distinct machine learning
   algorithms, namely LASSO and SVM-RFE, on the [72]GSE13507 dataset, we
   aimed to identify significant DE-ARGs capable of distinguishing BC from
   normal specimens. The LASSO logistic regression algorithm,
   incorporating penalty parameter tuning through 10-fold
   cross-validation, revealed 28 BC-related features (depicted in [73]Fig.
   2A and B). The expression patterns of these 28 novel diagnostic genes
   are illustrated in [74]Fig. 2C. Subsequently, the SVM-RFE algorithm was
   applied to filter DE-ARGs and identify an optimal combination of
   feature genes, resulting in the selection of 13 genes ([75]Fig. 3A and
   B), whose expression patterns are visualized in [76]Fig. 3C. By
   intersecting the marker genes obtained from the LASSO and SVM-RFE
   models, we identified 9 marker genes (PTGS2, CHEK2, BIRC3, PRKCQ, NRAS,
   CDH3, LRP1, FASN, and SIRT6) for further analysis (depicted in [77]Fig.
   4A). Utilizing the R package glm, we constructed a logistic regression
   model based on these 9 marker genes. ROC curves demonstrated that the
   logistic regression model effectively differentiated between normal and
   BC samples, achieving an AUC of 0.948 ([78]Fig. 4B). Additionally, we
   assessed the diagnostic capability of individual genes through ROC
   curves, all of which exhibited AUC values exceeding 0.65. The
   robustness of the new model was further validated in the [79]GSE3167
   datasets ([80]Fig. 4C). Expression patterns of the 9 marker genes in BC
   specimens from both [81]GSE13507 ([82]Fig. 4D) and [83]GSE3167
   ([84]Fig. 4E) datasets were also elucidated. In summary, our findings
   suggest that the logistic regression model, incorporating the 9
   identified marker genes, provides superior accuracy and specificity in
   distinguishing BC samples from normal samples compared to individual
   marker genes.

Fig. 2.

   [85]Fig. 2
   [86]Open in a new tab

   Identification of BC-Related biomarkers Using LASSO Algorithm. (A and
   B)Application of the LASSO logistic regression algorithm on the
   [87]GSE13507 dataset, with penalty parameter tuning through 10-fold
   cross-validation, resulted in the selection of 28 BC-related features.
   (C) Expression Patterns of 28 Novel Diagnostic Genes. Visualization of
   the expression patterns of the 28 identified BC-related features,
   providing insights into their potential diagnostic significance.

Fig. 3.

   [88]Fig. 3
   [89]Open in a new tab

   Selection of Optimal Feature Genes Using SVM-RFE Algorithm. (A and B)
   The SVM-RFE algorithm was applied to filter DE-ARGs, identifying an
   optimal combination of 13 feature genes. (C)Expression Patterns of 13
   Optimal Feature Genes. Visual representation of the expression patterns
   of the 13 optimal feature genes, offering insights into their potential
   diagnostic utility.

Fig. 4.

   [90]Fig. 4
   [91]Open in a new tab

   Identification of 9 Marker Genes and the development of a novel
   diagnostic model. (A) Intersection of marker genes obtained from LASSO
   and SVM-RFE models, resulting in the selection of 9 marker genes
   (PTGS2, CHEK2, BIRC3, PRKCQ, NRAS, CDH3, LRP1, FASN, and SIRT6) for
   further analysis. (B) Evaluation of Logistic Regression Model.
   Construction of a logistic regression model based on the 9 identified
   marker genes, demonstrating its effectiveness in differentiating
   between normal and BC samples with an AUC of 0.948 in ROC curves. (C)
   Validation in [92]GSE3167 Datasets. Assessment of the diagnostic
   capability of the logistic regression model in [93]GSE3167 datasets,
   confirming its robustness. (D and E)Expression patterns of 9 marker
   genes in BC Specimens. Visualization of the expression patterns of the
   9 marker genes in BC specimens from both [94]GSE13507 and [95]GSE3167
   datasets.

5.3. The expression and prognostic value of 9 marker genes in BC based on
TCGA datasets

   Subsequently, we scrutinized the expression profiles of the 9
   identified marker genes in BC using TCGA datasets. The findings
   revealed aberrant levels in eight marker genes within BC specimens,
   namely PTGS2, BIRC3, PRKCQ, NRAS, CDH3, LRP1, FASN, and SIRT6 (depicted
   in [96]Fig. 5A). Subsequent survival analysis unveiled that LRP1, FASN,
   and SIRT6 were significantly associated with the clinical outcomes of
   BC patients ([97]Fig. 5B). Additionally, a pan-cancer analysis
   demonstrated dysregulated expression patterns of SIRT6 ([98]Fig. 5C),
   LRP1 ([99]Fig. 5D), and FASN ([100]Fig. 5E) across multiple tumor
   types. The dysregulated expression observed in various cancers
   positions SIRT6, LRP1, and FASN as potential candidates for cancer
   biomarkers. This implies a potential association between the expression
   levels of these genes and cancer development, clinical progression, or
   treatment response. The consistent dysregulation observed across
   diverse cancer types suggests shared biological features or pathways,
   offering crucial insights for understanding common mechanisms in cancer
   development and facilitating the design of cross-cancer treatment
   strategies.

Fig. 5.

   [101]Fig. 5
   [102]Open in a new tab

   Identification of the critical survival-related marker genes in BC
   based on TCGA datasets. (A) Expression Profiles of 9 Marker Genes in BC
   Specimens based on TCGA datasets (including 19 normal samples and 412
   tumor samples). (B) Prognostic Value of LRP1, FASN, and SIRT6 in BC
   Patients. Survival analysis demonstrated a significant association
   between LRP1, FASN, and SIRT6 expression levels and the clinical
   outcomes of BC patients. (C–E) Pan-Cancer Analysis of SIRT6, LRP1, and
   FASN. Examination of dysregulated expression patterns of SIRT6, LRP1,
   and FASN across various tumor types. *p < 0.05, **p < 0.01,
   ***p < 0.001.

5.4. The prognostic value of FASN expression in BC patients

   Given the limited reports on the role of FASN expression in BC, our
   attention was directed towards a more in-depth exploration of FASN. To
   ascertain the potential of FASN as a novel biomarker for BC patients,
   we conducted comprehensive univariate and multivariate analyses using
   Cox's proportional hazard model. The univariate analysis revealed
   significant associations between overall survival of BC patients and
   variables such as age, clinical stage, and FASN expression (depicted in
   [103]Fig. 6A). Further strengthening this association, the multivariate
   Cox regression analyses affirmed that FASN expression stands as an
   independent prognostic indicator for overall survival in BC patients
   ([104]Fig. 6B). Considering the diverse clinical factors influencing BC
   patients, such as age, clinical stage, and FASN expression, and
   recognizing the need for a practical tool, we developed a nomogram
   model. This model aims to assist clinicians by providing a more
   comprehensive and individualized assessment of the survival prospects
   for patients with BC (illustrated in [105]Fig. 6C and D). The nomogram
   integrates multiple factors to create a visual and user-friendly tool
   that can aid in predicting the overall survival of BC patients, thereby
   contributing to more informed clinical decision-making.

Fig. 6.

   [106]Fig. 6
   [107]Open in a new tab

   The prognostic values of FASN expressions in BC patients. (A)
   Univariate Analysis of Prognostic Factors in BC Patients. Univariate
   analysis using Cox's proportional hazard model illustrating the
   associations between overall survival in BC patients and key variables.
   (B) Multivariate Cox Regression Analysis of FASN Expression.
   Multivariate Cox regression analyses confirming the independent
   prognostic value of FASN expression for overall survival in BC
   patients. (C and D) Nomogram model for survival prediction in BC
   Patients. Development of a nomogram model integrating age, clinical
   stage, and FASN expression as factors to provide a comprehensive and
   individualized assessment of overall survival in BC patients.
   *p < 0.05, **p < 0.01, ***p < 0.001.

5.5. The association between FASN expression and clinical factors of BC
patients

   Considering the prognostic significance of FASN in BC patients, our
   objective was to unravel the specific correlations between FASN
   expression levels and key clinical parameters in BC. We specifically
   focused on variables such as age, grade, clinical stage, T stage, M
   stage, and N stage. Notably, we observed that low expression of FASN
   was notably associated with age and gender, although no significant
   associations were found with other clinical factors (depicted in
   [108]Fig. 7A). To provide a visual representation of the variations in
   the distribution of clinical variables among individuals with high or
   low FASN expression, we constructed a heatmap illustrating these
   differences ([109]Fig. 7B). This heatmap serves as a comprehensive tool
   to highlight the nuanced relationships between FASN expression levels
   and various clinical parameters, shedding light on potential
   associations that may contribute to a more thorough understanding of
   the role of FASN in BC.

Fig. 7.

   [110]Fig. 7
   [111]Open in a new tab

   Association between FASN expression and clinical parameters in BC. (A)
   Exploration of the correlations between FASN expression levels and key
   clinical factors, including age, gender, stage and grade. (B) Heatmap
   of Clinical Variables and FASN Expression. Construction of a heatmap
   visualizing the distribution of clinical variables among individuals
   with high or low FASN expression. *p < 0.05, **p < 0.01, ***p < 0.001.

5.6. The biological functions of FASN in BC

   Subsequently, we identified a diverse set of DEGs from individuals
   characterized by either elevated or diminished FASN levels. A total of
   1421 DEGs were discerned through our screening process (depicted in
   [112]Fig. 8A). Then, we performed DO analysis and found that 1421 DEGs
   were mainly enriched in non-small cell lung carcinoma, breast
   carcinoma, urinary system cancer, musculoskeletal system cancer, female
   reproductive organ cancer and male reproductive organ cancer ([113]Fig.
   8B). The results of GO analysis revealed that 1421 DEGs were mainly
   related to phagocytosis, humoral immune response, lymphocyte mediated
   immunity, external side of plasma membrane, immunoglobulin complex,
   collagen-containing extracellular matrix, antigen binding, receptor
   ligand activity and signaling receptor activator activity ([114]Fig.
   8C). The results of KEGG analysis revealed that 1421 DEGs were mainly
   related to Cytokine-cytokine receptor interaction, Neuroactive
   ligand-receptor interaction, Hematopoietic cell lineage and Osteoclast
   differentiation ([115]Fig. 8D).

Fig. 8.

   [116]Fig. 8
   [117]Open in a new tab

   The Biological Functions of FASN in BC using Functional enrichment
   analysis. (A) Identification of DEGs Associated with FASN Levels. (B)
   DO Analysis of 1421 DEGs. (C) GO Analysis of 1421 DEGs. (D) KEGG
   Analysis of 1421 DEGs.

5.7. Knockdown of FASN suppressed BC progression via Wnt/β‐catenin pathway

   The exploration of FASN's potential role in BC cells commenced with an
   examination of its expression across various BC cell lines. Remarkably,
   both FASN mRNA and protein levels exhibited distinct elevation in five
   different BC cell lines, as depicted in [118]Fig. 9A and Supplementary
   S1A. Given that the expression of FASN exhibited a higher level in RT4
   and SW780 cells, we chose them as further experiments. To investigate
   the functional implications, we employed shRNA plasmids to effectively
   knock down FASN expression in RT4 and SW780 cells, achieving an
   efficiency exceeding 60 %, as demonstrated in [119]Fig. 9B and
   Supplementary S1B. Subsequent assessments through CCK-8 assays and
   clonogenic assays unequivocally revealed that the knockdown of FASN
   significantly impeded the proliferation of RT4 and SW780 cells
   ([120]Fig. 9C and D). Furthermore, TUNEL assays demonstrated a notable
   increase in apoptosis upon FASN knockdown in both RT4 and SW780 cells
   ([121]Fig. 9E). Expanding our inquiry, we explored the impact of FASN
   inhibition on the proliferation and metastasis of BC cells. Results
   from the wound healing experiment highlighted a substantial reduction
   in the migratory capacity of BC cells following FASN knockdown
   ([122]Fig. 10A). Further insights were gained through transwell assays,
   which indicated a diminished invasive ability of RT4 and SW780 cells
   after FASN knockdown ([123]Fig. 10B). Additionally, we scrutinized the
   expressions of key proteins associated with the Wnt/β‐catenin/EMT
   pathway, including N‐cadherin, E‐cadherin, MMP‐9, and β‐catenin. Our
   findings revealed that the silencing of FASN led to a suppression of
   N‐cadherin, β‐catenin, and MMP‐9 expressions, while concurrently
   promoting the expression of E‐cadherin ([124]Fig. 10C and Supplementary
   S1C).

Fig. 9.

   [125]Fig. 9
   [126]Open in a new tab

   Knockdown of FASN distinctly suppressed the proliferation of BC cells.
   (A) Elevated Expression of FASN in BC Cell Lines by RT-PCR and Western
   blot. The original Western blot was shown in Supplementary S1A. (B)
   Efficient Knockdown of FASN in RT4 and SW780 Cells. Implementation of
   shRNA plasmids to achieve a robust knockdown of FASN expression in RT4
   and SW780 cells. The original Western blot was shown in Supplementary
   S1B. (C) CCK-8 assays. (D)Clonogenic assays. (E) TUNEL assays. Scale
   bars: 20μm. *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 10.

   [127]Fig. 10
   [128]Open in a new tab

   Knockdown of FASN distinctly suppressed the metastasis of BC cells via
   Wnt/β‐catenin/EMT Pathway. (A) Wound healing experiment highlighting a
   substantial reduction in the migratory capacity of BC cells following
   FASN knockdown. Scale bars: 200 μm. (B) Diminished Invasive Ability
   After FASN Knockdown by transwell assays. Scale bars: 100 μm. (C)
   Modulation of Wnt/β‐catenin/EMT Pathway proteins by FASN silencing
   using Western blot. The original Western blot was shown in
   Supplementary S1C. *p < 0.05, **p < 0.01, ***p < 0.001.

6. Discussion

   Clinical tumor diagnostic markers for BC include Bladder Tumor Antigen
   (BTA), urine cytology (UroVysion), Urothelial Cancer-Associated Antigen
   1 (UCA1), Cytokeratin-19 Fragment (CYFRA 21-1), and Bladder Cancer MCM6
   marker [[129]22,[130]23]. These markers play diverse roles in the
   diagnosis, staging, and prognostic assessment of BC. While they exhibit
   variations in specificity and sensitivity, each has its own advantages
   and limitations [[131]6,[132]24]. For instance, BTA and UroVysion,
   though non-invasive, demonstrate relatively lower specificity, whereas
   UCA1 and MCM6 markers show potential advantages in early diagnosis
   [[133]25,[134]26]. Overall, the comprehensive use of these markers
   contributes to improved accuracy and reliability in diagnosing BC,
   providing valuable information for patient treatment. Despite certain
   challenges, these markers play a crucial role in facilitating early
   detection and treatment of BC. Enhancing the sensitivity of biomarkers
   is imperative for advancing the diagnostic capabilities in BC patients.

   Anoikis, translated as “programmed cell death induced by loss of
   extracellular adhesion,” refers to a form of programmed cell death
   triggered when cells lose normal adhesion to surrounding cells or the
   extracellular matrix [[135][27], [136][28], [137][29]]. Cells rely on
   adhesion to neighboring cells or the matrix to maintain their normal
   function and survival. When cells lose this adhesion, they may undergo
   Anoikis, preventing unattached cells from entering other sites and
   forming aberrant growth or metastatic tumors [[138]30,[139]31]. Under
   normal circumstances, Anoikis serves as a protective mechanism,
   contributing to the maintenance of tissue structure stability and
   preventing the uncontrolled growth and spread of abnormal cells
   [[140]32,[141]33]. However, in tumors, the Anoikis mechanism is often
   disrupted, enabling cancer cells to grow and spread to other locations
   within the body. The resistance of tumor cells to Anoikis is a crucial
   factor in tumor metastasis. During the development of cancer, some
   tumor cells may overcome the restrictions of Anoikis, escape from the
   primary tumor site, enter the bloodstream or lymphatic system, and then
   settle in other tissues or organs, forming distant metastatic tumors
   [[142]34,[143]35]. This increases the invasiveness and metastatic
   potential of tumors, making treatment more challenging. Anoikis have
   been studied in many cancers, such as pancreatic cancer, gastric
   cancer, colorectal cancer and [[144][36], [145][37], [146][38],
   [147][39]]. However, the specific function of Anoikis-related genes in
   BC was rarely reported. In this study, we firstly identified 300
   DE-ARGs in BC specimens through a comprehensive analysis of
   [148]GSE13507 datasets. These genes exhibited significant associations
   with various diseases, particularly non-small cell lung carcinoma,
   breast carcinoma, and urinary system cancer. GO analysis revealed their
   involvement in key cellular processes, while KEGG analysis highlighted
   their association with cancer-related pathways. Furthermore, the
   research successfully developed a diagnostic model based on 9
   identified marker genes, demonstrating superior accuracy and
   specificity in distinguishing BC from normal samples. This model offers
   potential insights into the genetic landscape of BC and presents a
   valuable tool for improved diagnostic precision in clinical settings.

   Then, we performed survival analysis using nine diagnostic genes based
   on TCGA datasets and only found that SIRT6, LRP1, and FASN were
   associated with clinical prognosis of BC patients. Given that the
   function of FASN in BC was rarely reported. Then, we focused on FASN.
   Fatty Acid Synthase (FASN) is a multifunctional enzyme involved in de
   novo fatty acid synthesis [[149]40]. It plays a crucial role in
   cellular lipid metabolism by catalyzing the synthesis of long-chain
   fatty acids from acetyl-CoA and malonyl-CoA [[150]41,[151]42]. FASN is
   a large, homodimeric protein with multiple functional domains, making
   it a key player in lipid biosynthesis and energy storage. FASN has
   garnered significant attention in cancer research due to its
   association with tumorigenesis and cancer progression
   [[152]43,[153]44]. Several studies have implicated FASN in various
   cancers, highlighting its role in providing cancer cells with the
   necessary fatty acids for membrane synthesis, energy storage, and
   signaling pathways. The upregulation of FASN is commonly observed in
   cancer cells, contributing to their uncontrolled growth and survival
   [[154][45], [155][46], [156][47]]. In this study, we confirmed FASN
   expression as an independent prognostic indicator for overall survival
   in BC patients. Moreover, we performed functional experiments and
   confirmed that FASN was highly expressed in BC cells and its knockdown
   distinctly suppressed the proliferation and metastasis of BC cells. Our
   findings suggested FASN as a novel tumor promotor in BC. Therefore,
   therapeutic strategies targeting FASN may help inhibit the development
   and metastasis of BC, offering a new direction for the treatment of BC.

   Wnt/β-catenin is a signaling pathway that plays a crucial role in
   various cellular processes, including embryonic development, tissue
   homeostasis, and regulation of gene expression [[157][48], [158][49],
   [159][50]]. The pathway is initiated by Wnt ligands, which bind to cell
   surface receptors, leading to the activation of intracellular signaling
   cascades. In the absence of Wnt signaling, a “destruction complex” is
   formed, which targets β-catenin for degradation [[160]51]. When Wnt
   ligands bind to their receptors, the destruction complex is inhibited,
   allowing β-catenin to accumulate in the cytoplasm. Subsequently,
   β-catenin translocates into the nucleus, where it interacts with
   transcription factors of the T-cell factor/lymphoid enhancer factor
   (TCF/LEF) family [[161]52,[162]53]. This interaction leads to the
   activation of target genes involved in cell proliferation, survival,
   and differentiation. Aberrant activation of the Wnt/β-catenin pathway
   is associated with various diseases, including cancer
   [[163]54,[164]55]. In cancer, mutations or dysregulation of components
   in the pathway can result in the constitutive activation of Wnt
   signaling, promoting tumor growth and metastasis. The Wnt/β-catenin
   pathway is a key player in cancer stem cell maintenance,
   epithelial-mesenchymal transition (EMT), and resistance to apoptosis.
   It has been confirmed that the aberrant activation of the Wnt/β-catenin
   pathway in BC may promote tumor development and progression through
   various mechanisms [[165][56], [166][57], [167][58]]. In this study, we
   performed Western blot and found that the silencing of FASN led to a
   suppression of N‐cadherin, β‐catenin, and MMP‐9 expressions, while
   concurrently promoting the expression of E‐cadherin, suggesting that
   FASN may promote the proliferation and metastasis via modulating
   Wnt/β-catenin pathway.

   This article has a number of limitations. Firstly, the study heavily
   relies on the analysis of the [168]GSE13507 and [169]GSE3167 datasets.
   The representativeness of these datasets may be limited, and results
   might not be universally applicable to all BC cases. Secondly, while
   the study explores the functional role of FASN in BC cell lines, the
   findings primarily focus on cell proliferation, apoptosis, migration,
   and invasion. Further investigations into the underlying molecular
   mechanisms and in vivo studies are necessary for a comprehensive
   understanding of FASN's role in BC progression.

7. Conclusion

   We identified 300 differentially expressed annotated genes (DE-ARGs)
   closely associated with BC progression. These genes are predominantly
   enriched across various cancers, participating in the regulation of
   protein kinase activity, cell adhesion, and other biological processes.
   Utilizing machine learning algorithms, we identified 28 BC-related
   features and constructed a logistic regression model based on 9 marker
   genes, demonstrating superior accuracy in distinguishing BC from normal
   samples. Validation in TCGA datasets revealed significant correlations
   between LRP1, FASN, SIRT6, and clinical outcomes in BC patients,
   suggesting their potential as cancer biomarkers. FASN, identified as an
   independent prognostic indicator, regulates BC cell proliferation and
   metastasis through the Wnt/β-catenin pathway. These findings provide
   crucial insights for a deeper understanding of BC's molecular
   mechanisms and future therapeutic strategies.

Ethics approval and consent to participate

   Not applicable.

Consent for publication

   Not applicable.

Data availability statement

   The datasets analyzed during the current study are available in The
   Cancer Genome Atlas database (TCGA,
   [170]https://tcga-data.nci.nih.gov/tcga/) and the Gene Expression
   Omnibus database (GEO, [171]https://www.ncbi.nlm.nih.gov/geo/)
   [[172]GSE13507 and [173]GSE3167 datasets]. The datasets generated
   during and/or analyzed during the current study are available from the
   corresponding author upon reasonable request.

Fundings

   This research did not receive specific grants or fundings.

CRediT authorship contribution statement

   Ruoyu Peng: Writing – review & editing, Writing – original draft,
   Visualization, Software, Resources, Project administration,
   Investigation, Formal analysis, Data curation, Conceptualization.
   Xiaohan Ma: Writing – review & editing, Writing – original draft,
   Visualization, Software, Resources, Methodology, Investigation, Data
   curation, Conceptualization. Zhiyun Jiang: Writing – review & editing,
   Writing – original draft, Visualization, Resources, Project
   administration, Methodology, Investigation, Data curation,
   Conceptualization. Yu Duan: Writing – review & editing, Writing –
   original draft, Visualization, Resources, Project administration,
   Methodology, Investigation, Formal analysis, Data curation,
   Conceptualization. Shaogang Lv: Writing – review & editing, Writing –
   original draft, Visualization, Resources, Project administration,
   Methodology, Data curation, Conceptualization. Wei Jing: Writing –
   review & editing, Writing – original draft, Visualization, Supervision,
   Project administration, Methodology, Investigation, Formal analysis,
   Data curation, Conceptualization.

Declaration of competing interest

   The authors declare that they have no known competing financial
   interests or personal relationships that could have appeared to
   influence the work reported in this paper.

Acknowledgements