ISEBC: Using A Novel Breast Cancer Tertiary Lymphoid Structures Signature To Build An Immune-favourable Status Evaluator For Breast Cancer

ISEBC: Using A Novel Breast Cancer Tertiary Lymphoid Structures Signature To Build An Immune-favourable Status Evaluator For Breast Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article ISEBC: Using A Novel Breast Cancer Tertiary Lymphoid Structures Signature To Build An Immune-favourable Status Evaluator For Breast Cancer Xiaokai Fan, Xuan Yu, Liang Chen This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5376285/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Tertiary lymphoid structures (TLSs), as a special type of immune-infiltrated region in tumor tissues, plays an important role in benefiting from immunotherapy or improving immune state of breast cancer patient. However, breast cancer-specific TLS signature gene sets are still lacking. Therefore, we extracted gene features with machine learning algorithm LightGBM, and differential expression genes with statistical test on multiple spatial transcriptome datasets, to finally obtain a novel breast cancer-specific TLS gene set (NBCTS). Compared with previous gene sets, it has stronger characterization ability of TLS in breast cancer samples. Since TLS have unique immune characteristics, we classified three different immune states using this gene set for breast cancer patients and get an immune state. To better facilitate evaluating this immune statuses of breast cancer patients or samples, we developed a user-friendly web tool (Immune State Evaluator for Breast Cancer, (ISEBC), www.omegene.tech:3838/ISEBC ) to make it more convenient for researchers and clinicians to use. Tertiary lymphoid structures (TLS) Breast cancer Immune state evaluation Biomarker research Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Breast cancer is a prevalent malignancy in females worldwide that remains difficult to completely cure in all subtypes of patients[ 1 ]. According to existing research, application of immunotherapy to breast cancer patients can effectively relieve or even cure the disease. The accurate evaluating the immune-favourable status of breast cancer is critical for assessing the disease state or choosing the suitable treatment way. However, effective open-source tools for evaluating the immune-favourable status of breast cancer patients are still lacking[ 2 – 3 ]. Some reported immune microenvironment biomarkers cannot be uniformly applied to assessments of the breast cancer immune microenvironment and evaluating criterion of these biomarkers still not be illustrate clearly. Meanwhile, timely and accurately evaluating the immune microenvironment of breast cancer patients is also important for diagnosis, disease progression monitoring and treatment selection. Tertiary lymphoid structures (TLSs) in tumor tissues are composed of infiltrated immune cells. Regularly, high infiltration of TLSs in tumor mircoenviroment often indicates better patient prognosis and a higher possibility of response to immunotherapy[ 4 – 5 , 8 ]. Currently, breast cancer still lacks cancer-specific TLS signature gene sets. Therefore, we used spatial transcriptome data of breast cancer patients and TLS regions in corresponding H&E slides annotated by pathologists. Together with feature extraction algorithms in machine learning, we identified a novel TLS signature gene set for breast cancer patient. This gene set has stronger characterization ability for breast cancer patient TLSs compared to previous gene sets and can assess the degree of TLS infiltration in breast cancer patients. Further investigation found that this gene set can also be used for evaluations of breast cancer patient immune status. Thus, we constructed an easy-to-use web tool for breast cancer immune status evaluation using this gene set to help researchers and physicians in this field timely analyze and assess the immune-favourable status of breast cancer patients or experimental samples(Fig. 1 A). Materials and methods 1. Data collection This study mainly used breast cancer spatial transcriptome data and bulk RNA-seq data. Breast cancer spatial transcriptomics data were primarily collected from the Gene Expression Omnibus(GEO) database, 10x Genomics official resources, and Zenodo database. Among them, the 4 patient H&E staining and spatial transcriptomics data of GSE176078(1160920F, CID44971, CID4535, CID4465, https://doi.org/10.5281/zenodo.4739739 ) and 3 patient H&E staining slices(G1, E1, F1, https://zenodo.org/records/4751624 ) respectively were obtained from Zendo database. The data from 10x Genomics official resources contained two consecutive slices from the same patient ( https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_1,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_2 ). An additional single slice was obtained from GEO dataset GSM6177599. In total, 10 slices were included in our used spatial transcriptome data. Bulk transcriptomic data originated from several prominent sources. The TCGA bulk breast cancer RNA data was mainly obtained from the UCSC Xena browser, comprising 1069 unique samples after deduplication and removal of adjacent normal samples. The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset was mainly obtained from cBioPortal. The GSE1456 and GSE0886 data were mainly obtained from the GEO database, including 318 and 178 patients respectively. The single cell RNA sequencing data was also obtained from the GEO database via accession GSE169246. 2. Data processing Spatial transcriptome data processing We used the Load10X_Spatial() function in Seurat to import count data in standard cellranger output format. For non-standard cellranger output data, scale.factor.json files were manually created and then read separately before using the Seurat CreateSeuratObject() function to create Seurat objects. After creating the Seurat objects, normalization was performed using SCTransform(), dimensionality reduction was conducted via RunPCA(). The first 30 PCAs were used for UMAP by RunUMAP() function. Neighbor detection utilized FindNeighbors(), and FindClusters() with a resolution of 0.7 assigned spots to clusters. SpatialFeaturePlot() was used for visualization of gene expression across slices. For the evaluation of spot expression intensity of our gene set, the 12 chemokine factor gene set and the 2020 Nature gene set, AUCell package was used for scoring, and then scores were normalized from 0–1 using standardization for comparison. Experienced pathologists annotated TLS regions. Single cell RNAseq data processing We read separately files and use scDblFinder to remove doublets/multiplets with default parameters, then merged these expression files and use Seurat CreateSeuratObject() to create Seurat object. Then with using harmony to remove batch effect that existed in the each samples. Finally we use AggregateExpression() function to sum counts in each sample and get the pseduo-bulk counts of each sample. Bulk RNA transcriptome data processing The TCGA breast cancer patient dataset was processed using the R programming language, initially filtering to remove duplicates, metastatic, and samples lacking survival data, retaining 1069 patients. GEO series GSE1456 and GSE0886 were matched by probe/gene ID, converting probes to genes with summation averaging for repeats. Processed matrices were linked to clinical information. The standardized expression matrix data in METABRIC was selected and matched with patient clinical information. 3. Identification of a novel breast cancer TLS signature gene set Machine learning algorithms, such as LightGBM, have demonstrated strong performance in the task of automated feature extraction from high-dimensional biological data. We leveraged LightGBM's capabilities to identify feature genes that can distinguish TLS regions or non-TLS regions.Additionally, we calculated differentially expressed genes between TLS and non-TLS areas primarily using the FindAllMarkers function within the Seurat package. Genes were considered differentially upregulated in TLS spots based on adjusted p-value thresholds and logFC value. To refine our novel breast cancer TLS signature gene set, we required candidate genes to satisfy two strict criteria. First, they must have been extracted as predictive TLS features by our LightGBM model feature extraction process. Second, they were required to exhibit robust highly express in TLS regions compared with non-TLS region. After applying these feature extraction process and differential expression analysis, we identified a novel 11-gene breast cancer signature of TLS. The multi-step biomarker discovery and optimization process thus yielded a rigorously-validated signature. To validate the characterization ability of our TLS signature relative to other gene sets, we performed receiver operating characteristic (ROC) curve analysis. ROC curves provide an objective assessment of a biomarker's ability to distinguish between two regions. The ROC curve plotting process began by calculating true positive rate (TPR) and false positive rate (FPR) values across a range of expression thresholds between 0–1 with 0.01 increments. TPR represents the proportion of true positives that are correctly identified, while FPR is the proportion of false positives. The area under the ROC curve (AUC) was also computed as a single measure of discriminatory power, with higher values indicating superior classification ability. The TRP, FPR and AUC are computed using the following formula: where TP is true positive, FN is false negative, TN is true negative, FP is false positive, \(\:{f}_{{\alpha\:}}\) is the function of ROC curve, \(\:{\text{t}}_{\text{i}}\) is threshold that belong to 0–1 4. Pathway expression scoring To score spot expression intensities for our novel gene set, the 12 chemokine gene set, and the Nature2020 set, we employed the AUCell R package. AUCell calculates gene expression scores for each spot based on provided gene sets. Standardization from 0–1 was then performed on the AUCell scores to normalize the quantitative ranges between 0 and 1. This preprocessing step ensured the scores were placed on comparable scales, facilitating direct comparisons between the performance of different gene sets. For pathway activity assessments of bulk RNA-seq data, we primarily utilized Gene Set Variation Analysis (GSVA). GSVA estimates pathway expression strength on a sample-by-sample manner. It has demonstrated superior ability over gene set enrichment analysis methods like single-sample GSEA (ssGSEA) to accurately gauge pathway enrichment levels from transcriptomic profiles. We opted to apply GSVA for this reason, to sensitively interrogate pathway signatures at the individual patient level throughout. 5. Identification of breast cancer immune Status To classify breast cancer patient immune subtypes, we utilized the gene expression profiles of the 11 genes comprising our novel breast cancer TLS signature within The Cancer Genome Atlas (TCGA) cohort. Unsupervised consensus clustering was performed using the ConsensusClusterPlus package in R to robustly group patient samples based on their gene expression signatures without predefining the number of clusters. This algorithm generates consensus matrices from subsampling to determine the stable cluster solution. Through consensus clustering of our novol TLS signature genes expression profile across TCGA samples, we reliably identified three major immune status within the patient population which intuitively depict distinct immune phenotypes present among breast cancer patients. In parallel, we applied UMAP dimensionality reduction to project the high-dimensional gene expression space into two dimensions for visualization. The ggplot2 package was primarily used to construct UMAP plots delineating the spatial relationships between patient samples belonging to different immune statuses . 6. Survival analysis We thoroughly investigated the impact of gene set expression levels on patient survival outcomes. The R packages survival and survminer were primarily leveraged to conduct these analyses and generate Kaplan-Meier plots. For GSE1456, GSE0886, and TCGA-BRCA cohorts, median expression of the gene sets was used as the cutoff point to segregate patients into high and low expression groups for comparison. Due to continuous clinical variable formatting in METABRIC, we applied an optimized threshold determination method to dichotomize patients based on signature scores. The ggsurvplot() function was mainly utilized to visualize and compare Kaplan-Meier survival plots between the stratified patient subgroups within each cohort. Additionally, pairwise log-rank tests of differences in survival distributions between immune subtypes were performed using pairwise_survdiff(). Results from these tests were also plotted using ggsurvplot() to depict divergent prognostic outcomes associated with characterized phenotypes. Through rigorous statistical evaluation and visualization of survival trends, we comprehensively demonstrated our gene signature's utility as a prognostic biomarker across independent patient populations. The effect of gene set expression strength on patient survival was mainly analyzed and plotted using the survival and survminer packages. For GSE1456, GSE0886 and TCGA-BRCA patients, median expression was used to divide patients into high and low expression groups. For METABRIC, an optimized method was mainly used to determine the division of high and low expression patients. The ggsurvplot() function was used to plot the results. Survival analysis between subtypes was mainly analyzed using the pairwise_survdiff() function, and the results were also plotted using the ggsurvplot() function. 7. Analysis of deconvolved immune cell infiltration in patients To characterize tumor-infiltrating immune cell compositions, we employed the R immunedeconv package, which utilizes validated reference target cell related genes to quantitate the proportions of various cell populations within the tumor microenvironment. This package intergrates 8 algorithms(quanTIseq、TIMER、CBIERSORT、MCP-Counter、Xcell、EPIC、ABIS、ConsensusTME、Estimate) Standardization of the immune cell scores was performed by following formula, \(\:{S}_{{c}_{i}}=\:\frac{{E}_{{c}_{i}}}{Var\left(\varvec{E}\right)}\) , which \(\:{E}_{{c}_{i}}\) is the estimated immune cell scores of cell type \(\:{c}_{i}\) , \(\:\varvec{E}\) is a vector of estimated immune cell scores of all cell types, \(\:{S}_{{c}_{i}}\) is the standardization score of cell type \(\:{c}_{i}\:\) .The ComplexHeatmap package was then leveraged to generate intuitive visualizations of the results. It produced heatmaps comparing the median relative infiltration levels of each immune cell subset across the three characterized immune Status . 8. BulkRNAseq differential expression analysis and biological process enrichment analysis To gain insight into molecular determinants of the second immune state, we calculated differentially expressed genes between it and the other statuses using DESeq2. DESeq2 is a widely used R package that models count data to determine differential expression while controlling for batch effects and considering biological variability. It robustly identifies changes in gene expression levels across different conditions. We then leveraged clusterProfiler, an enrichment analysis tool, to characterize the biological roles of top upregulated genes within the second state. Gene Ontology (GO) term overrepresentation analysis highlighted functional categories most significantly associated with these upregulated genes. And also we removed redundant processes within the enriched GO terms to condense the biological themes driving distinct immune responsiveness in the favorable second immune state. 9. Construction and validation of breast cancer immune status prediction model We leveraged the TCGA breast cancer RNA-seq dataset to train and evaluate an immune statues classification model. The top 100 most variably expressed genes across patients were selected as predictive input features. These genes underwent normalization to scale expression values from 0 to 1, the following formula: $$\:{n}_{i}\:=\:\frac{{x}_{i}-min\left(\varvec{x}\right)}{max\left(\varvec{x}\right)\:-\:min\left(\varvec{x}\right)}$$ where \(\:{x}_{i}\:\) is the normalized value of gene \(\:i\) , the vector \(\:\varvec{x}\) is these top 100 most variably expressed gene expression value of different patients, \(\:{n}_{i\:}\) is the normalized value ranging from 0 to 1. The preprocessed dataset was then randomly partitioned using 70% for model training and the remaining 30% held-out for testing. This train-test split procedure was repetitively conducted 3 times with different random seeds to validate model performance. For each split, a LightGBM classifier was trained on the 70% training subset. LightGBM aims to learn a function \(\:{f}_{\phi\:}(\varvec{n};\theta\:)\) that maps gene expression profiles to immune statues, where \(\:\theta\:\) represents the model parameter set. Hyperparameter optimization via grid search was performed to identify \(\:\theta\:\) values minimizing a loss function: \(\:L\:={f}_{\phi\:}(\varvec{n};{\theta\:}_{i})-\varvec{c}\:\) \(\:\varvec{c}\) is the classes of immune statues ,with the aim of finding an optimal \(\:\theta\:\) corresponding to lower \(\:L\) . The fine-tune process is to find a suitable parameters of \(\:{\theta\:}_{i}\) from \(\:\{{\theta\:}_{1},{\theta\:}_{2}...{\theta\:}_{L}\}\) . Finally the trained lightGBM model is the \(\:{f}_{\phi\:}(\varvec{n};{\theta\:}_{i})\) . Prediction performance was subsequently assessed using the separate 30% test subset. The scikit-learn metrics module was primarily used to analyze model performance. Confusion matrices generated by confusion_matrix() visualized predicted versus true class assignments. AUC scores computed by auc() provided a summary metric of discriminatory power between immune status. This rigorous validation workflow objectively evaluated our machine learning approach for immune status classification of new patient samples. 10. Construction of a breast cancer immune status evaluation web tool We used R shiny as the main framework to build an interactive web application. For queries containing few samples(1-400), it primarily computes Spearman's correlation coefficients between the top 100 variant genes from our training dataset and the overlapping genes in the user-uploaded expression matrix. For Larger sample sets(> 400) first undergo normalization of the top variant genes to the 0–1 range. The scaled expression profiles are then fed into our pre-trained LightGBM classification model for immune status prediction. The ggplot2 package was mainly leveraged to generate publication-quality data displays for intuitive interpretation. Users can conveniently query breast cancer patients or samples expression data without requirement of specialized computational skills. This optimized workflow empowers immune status classification to advance research and guide clinical decisions. Results 1. A novel breast cancer-specific tertiary lymphoid structures signature Spatial transcriptomics, as an impactful biotechnology in 2022, characterizes gene expression patterns across different locations and regions within a biological sample. By resolving mRNA transcripts into spatially encoded spots, it allows researchers to study gene expression and functions of specific genes in spatial locations from a new perspective, something not possible with conventional single cell RNA sequencing or bulk RNA sequencing. Therefore we use spatial transcriptomics data to study the TLSs. TLSs are immune cell aggregates that form within tumors and play an important role in the anti-cancer response. To define an breast cancer-specific signature of TLS, we collected multi-center breast cancer patient spatial transcriptome datasets from different sources (Supplementary table 1), and used the feature extraction algorithm lightGBM to identify TLS feature genes. Then, we calculated specific highly expressed genes in TLS regions using wilcox’test, to intersect these highly expressed genes in TLS regions and extracted gene features, we finally established a novel breast cancer tertiary lymphoid structure signature gene set (NBCTS). This gene set include 11 genes, which are CD52, CD48, CXCL9, CXCL13, TRBC2, CCL19, MS4A1, LCP1, CSN1S1, TBC1D10C, RAC2. CXCL13 is a well-defined marker gene of TLS regions based on published studies. CCL19 participate in recruiting the B cells and T cells which are cornerstone of the TLS formation in the initial stage. Compared to established gene sets like 12 classic chemokine factor genes[ 6 ] and a gene set published in Nature in 2020 by Rita Carita[ 7 ], our new breast cancer TLS gene set exhibited larger differences in expression between inner TLS regions versus outside, as well as stronger signal detection within testing set slices (E1, G1, F1, GSM6177599)(Fig. 1 B). Additionally, receiver operating characteristic (ROC) analysis indicated our gene set outperformed others in predicting TLS infiltration strength in breast cancer patients (Fig. 1 C). Kaplan-Meier survival analysis of the NBCTS across independent transcriptomic cohorts also illustrated its utility for characterizing patient prognosis. Overall, by leveraging spatial transcriptomics datasets, we defined a robust signature that can be used in the prediction or scoring the infiltration level of TLS and exists translational potential. 2. Application of NBCTS in evaluating patient tumor immune status Since TLS within tumors play a critical role in recruiting and activating anti-cancer immune cells, thereby enhancing patient survival outcomes,we therefore explore whether leveraging our novel breast cancer-specific TLS signature gene set can comprehensively evaluate the degree and composition of immune cell infiltration across patient tumor tissues. If the envisage can be realized, this would give a new way to assess tumor immune status of each patient. Through unsupervised consensus clustering analysis of NBCTS gene expression profiles from The Cancer Genome Atlas (TCGA) breast cancer cohort, we found three different immune statuses within the patient population (Fig. 2 A). Further analyses showed the patients of second immune status show significantly better prognosis and survival compared to the first and third status. Patients assigned to the second status also exhibited higher comprehensive scores for our novel TLS signature gene set versus the other statuses (Fig. 2 B) . To gain deeper insight, we employed multiple computational cell deconvolution algorithms like quanTIseq, TIMER, CBIERSORT, MCP-Counter, Xcell, EPIC, ABIS, ConsensusTME to estimate the relative proportions of different immune cell populations infiltrating tumors across immune statuses. Results demonstrated the second immune status had markedly elevated levels of both effector memory and naive CD8 + T cells relative to the other groups. Given the importance of B cells in TLS formation, we observed B cell abundances were also significantly greater in patients within the second immune status versus the other two immune status, which is consistently with NBCTS scores of these patients(Fig. 2 C, Supplementary Fig. 1A). The antigen processing and presentation pathway also found significantly up-regulated in the second immune status (Fig. 2 D,E). By doing differential expression genes analyses compare the second immune status with the remaining patients, we found that some genes are closely related to enhancing patient anti-tumor immune ability, such as CXCL13 and IFNG, were highly expressed in the second status (Fig. 2 F). Gene Ontology enrichment analysis of highly expressed genes in the second status found that they were mainly enriched in biological processes related to regulating immune antibody production and lymphocyte-mediated immune response regulation (Fig. 2 G). We also use a single cell dataset (Supplementary table 2) of combining immuntherapy and chemotherpy cohort to make a pseudo-bulk expression matrix, with GSVA algorithm we calcualte the enrichment score of our NBCTS, 12 classic chemokine factor genes and a gene set published in Nature in 2020 by Rita Carita. Compare with these gene sets and marker genes (CD8A, CXCL13) which are widely proven as favourable markers for immunetherapy, our NBCTS shows significantly different between partial response patients and stable disease patients (Fig. 2 H). By synthesizing these results, we can reasonably conclude that the second immune status represents an immune-favorable microenvironment for breast cancer patients undergoing immunotherapy. Additionally, our analysis indicated that the second immune status correlated with a TLSs-dominant profile. Indeed, through our whole immune status classification process, the second immune status emerged as the most important immune-favorable microenvironment profile identified. 3. Construction of a breast cancer immune status classifier model based on the NBCTS Currently, some gene expression-based signatures have been proposed for evaluating the immune infiltration status of breast cancer patients. However, they generally lack standardized implementation protocols that lead to some difficulties in practical applications. This problem can largely be attributed to two main factors: the high biological heterogeneity between patients' tumor immune microenvironments, and the absence of well-defined thresholds for classifying gene expression levels as high or low across different patients and studies. As a result, the utility and reproducibility of previous gene expression signatures for classify tumor’s immune statues remains difficulty. To develop a more broadly applicable and transferable immune statuses evaluating solution, we constructed a three-class immune status classifier using gene expression data from TCGA breast cancer cohort and actually we mainly want to build a model that can successfully classify the second immune-favorable status. Patients were annotated with the matching immune status based on previous classifier results, which categorized samples according to the expression pattern of our novel TLS signature gene set. To optimize characterization and standardization of the training data, we first calculated gene variance and identified the top 100 most variably expressed genes as training features. The expression values of these top 100 genes were then normalized from 0 to 1 across all patients to pre-process the data prior to classifier training. The classifier performance was rigorously evaluated using multiple testing strategies. By initializing different random seeds and conducting repeated train-test splits of the TCGA data, the out-of-sample classification predictions from the test sets achieved excellent accuracy under all random seeds. Specifically, the area under the receiver operating characteristic curve exceeded 0.99 for each immune status class and resampling scenario, indicating exceptional discriminatory power (Fig. 3 A,B). To assess generalizability beyond the TCGA cohort, we next applied our pre-trained classifier to an independent breast cancer gene expression dataset from the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC). Reassuringly, the predicted immune status distributions from this external validation set closely recapitulated the cell infiltration profiles (Supplementary Fig. 1B) and survival trends of the reference TCGA subtypes(Supplementary Fig. 1A). This consistency provides strong evidence for the classifier's robust predictive performance across studies. 4. A web tool for breast cancer patient immune status evaluation In order to better apply this immune status prediction model and make the tool more accessible and user-friendly for various users, we constructed a web application called ISEBC (Immune State Evaluation tool for Breast Cancer, www.omegene.tech:3838/ISEBC ) for assessing the immune status of breast cancer patients (Fig. 4 A). The ISEBC web application framework was built using the R shiny platform and provides two analysis options depending on sample size. For evaluations involving 1-400 samples, Spearman correlation analysis is performed due to the relatively small number of patient conditions represented. It calculates the similarity of highly variable gene expression between the uploaded data and reference TCGA breast cancer patients for each immune status. The overall highest similarity to a given reference immune status indicates a higher probability that the sample(s) belong to that status, and this result is the of web analysis output (Fig. 4 B). When evaluating larger sample sets containing more than 400 patients, users can opt for the "large sample assessment" mode. In this scenario, ISEBC will predict the immune status of each individual patient by applying the pre-trained LightGBM machine learning model. This model was trained on gene expression profiles from TCGA patients annotated with immune statuses using the top 100 most variably expressed genes. Its predictions leverage the broader expression profile characterization captured during training from a larger and more diverse set of reference samples (Fig. 4 C). Overall, ISEBC provides a user-friendly web interface for evaluating the immune status of breast cancer patients or samples using either a correlation-based or machine learning-based approach, depending on the number of samples analyzed. This tool will help facilitate practical application of evaluating the immune status for breast cancer. Discussion Compared to other tumor types, the immunogenicity of breast cancer is still lower than other cancer types, especially highly immunogenic cancer types such as melanoma and lung cancer[ 2 ]. Accurately characterizing the tumor immune status is critically important for tailoring optimal treatment strategies to individual patients.The presence of TLSs within tumors play an important role in facilitating the recruitment of cytotoxic CD8 + T cells and immune-activating B cells, which help eliminate cancer cells. Numerous previous studies have demonstrated robust correlations between increased TLS infiltration and more favorable patient outcomes, including improved overall survival and prognosis[ 3 ]. The formation of TLS within the tumor microenvironment is believed to enhance the adaptive anti-tumor immune response by supporting interactions between lymphocytes and antigen-presenting cells that are needed to activate and boost anti-tumor immunity[ 9 ]. Through analyzing spatial transcriptomic datasets, we identified a novel breast cancer-specific TLS gene signature (NBCTS). Compared with other TLS gene sets, this novel TLS gene set identified from breast cancer spatial transcriptome datasets in our study has better performance in predicting TLS infiltration. Some previous gene sets or methods based on the distance between CD8 + T cells and tumor cells or based on the infiltration of immune cells also can cluster patients into different subclusters, but the usage protocol was not clear and researchers hardly to use these gene sets or methods. For the widely used, We not only training a immune state classification model, but also construct a web tool and the ability to evaluate patient immune states in different sample amount modes, it allows users to more conveniently analyze the immune state of samples. For validation using independent cohorts, we applied a 0–1 normalization transform preserving gene rank orders. This standardized preprocessing approach may enhance future evaluations of patient immune statuses through more robust cross-study comparisons and integrated analyzes. Training machine learning models on high-quality reference datasets is also paramount for developing predictive models with translational potential. Based on immune states defined by our evaluator, the second phenotype showed most prominent immune traits while the first showed weakest. The second immune state also show significant B cell-related features and obviously exhibit TLS-dominant phenotype. Comparing evaluation of CXCL13 + CD8 + T cell infiltration level as originally reported, patients with high expression of the NBCTS gene set showed significantly greater benefit from immunotherapy than patients with low expression of the NBCTS gene set. This indicated that our NBCTS gene set have potential ability for evaluating the immune therapy response rate. However, due to the small number of breast cancer immunotherapy treatment open-source data currently available, this conclusion may require larger sample size datasets in the future to validate. Overall, the novel breast cancer TLS signature gene set identified in our study has better characterization ability than previous gene sets and can well classify breast cancer patients into different immune states. And we also find a survival and immune favorable statues of breast cancer patients. The developed web tool empowers widespread interrogation of tumor immune statues to deepen understanding and guide personalized medicine decisions. Abbreviations TLS:Tertiary lymphoid structure; NBCTS: novel breast cancer tertiary lymphoid structure signature gene set; OS: overall survival; TCGA: The Cancer Genome Atlas; METABRIC: Molecular Taxonomy of Breast Cancer International Consortium Declarations Acknowledgments We sincerely thank Alexander Swarbrick (Garvan Institute of Medical Research) for sharing their single-cell RNA sequencing data in the GEO database and spatial transcriptome data in the Zenodo database. Authors’ contributions XK.F., L.C. conceived and designed the study and drafted the manuscript. XK.F., X.Y performed the analysis of the data. The author(s) read and approved the final manuscript. Code availability The code will be uploaded to github after paper published. Funding Not applicable. Availability of data and materials Our re-analysis single-cell RNA data were deposited at the National Center for Biotechnology Information (NCBI) GEO and are accessible through the accession number GSE169246. The spatial transcriptome data of our used were deposited at the Zenodo database and 10x offical resource, the website is (https://doi.org/10.5281/zenodo.4739739, https://zenodo.org/records/4751624,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_1,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_2). Ethics approval Not applicable. consent to participate Not applicable. Competing interests The authors declare no potential conflicts of interest. References Voduc KD et al (2010) Breast cancer subtypes and the risk of local and regional relapse. J Clin Oncol 28(10):1684–1691 Barzaman K et al (2020) Breast cancer: Biology, biomarkers, and treatments. Int Immunopharmacol 84:106535 Colbeck EJ et al (2017) Tertiary Lymphoid Structures in Cancer: Drivers of Antitumor Immunity, Immunosuppression, or Bystander Sentinels in Disease? Front Immunol 8:1830 Dieu-Nosjean MC et al (2016) Tertiary lymphoid structures, drivers of the anti-tumor responses in human cancers. Immunol Rev 271(1):260–275 Seow DYB et al (2020) Tertiary lymphoid structures, and associated plasma cells play an important role in the biology of triple-negative breast cancers. Breast Cancer Res Treat 180(2):369–377 Prabhakaran S et al (2017) Evaluation of invasive breast cancer samples using a 12-chemokine gene expression score: correlation with clinical outcomes. Breast Cancer Res 19(1):71 Cabrita R, Lauss M, Sanna A et al (2020) Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature 577:561–565 Schumacher TN, Thommen DS (2022) Tertiary lymphoid structures in cancer. Science 375(6576):eabf9419 Munoz-Erazo L et al (2020) Tertiary lymphoid structures in cancer - considerations for patient prognosis. Cell Mol Immunol 17(6):570–575 Additional Declarations No competing interests reported. Supplementary Files Supplementarytables.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5376285","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":377263653,"identity":"01db3096-5390-41ab-bb15-1fa4a8dc0bed","order_by":0,"name":"Xiaokai Fan","email":"","orcid":"","institution":"Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Xiaokai","middleName":"","lastName":"Fan","suffix":""},{"id":377263654,"identity":"83e707a3-c811-478a-9219-121e41a00f51","order_by":1,"name":"Xuan Yu","email":"","orcid":"","institution":"Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Xuan","middleName":"","lastName":"Yu","suffix":""},{"id":377263655,"identity":"7873cfad-950b-4156-86bc-43a73155101f","order_by":2,"name":"Liang Chen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5klEQVRIiWNgGAWjYBACPmYIncDAwHwARBIGbAgtbAlA0oAILQxwLTwg5cRoYWd+9vALg10ev3TP5w8Pav4w8LcfYPxcgNdhbObGMgzJxZJzzm6TSDhmwCBxJoFZegZ+v5hJSzAcSNxwI3cbQwIb0GE3gII8eLWwfwNr2X8j5/GHhH8GDPKEtfCYSX4A2SKRwyCR2GbAYECEljJpBoPkxBk30swkEvuMeQzPJDZL49PCz398m+SPCrvE/hnJjz/++CYnJ3f88MHP+LSAADMPUmwAFTM2ENAAVPKDoJJRMApGwSgY0QAAbmA/3xiUUqMAAAAASUVORK5CYII=","orcid":"","institution":"Chinese Academy of Sciences","correspondingAuthor":true,"prefix":"","firstName":"Liang","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2024-11-02 04:08:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5376285/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5376285/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":70922009,"identity":"af9eceba-871a-4ba9-b59d-f4a64f268025","added_by":"auto","created_at":"2024-12-09 08:44:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":605917,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOverview of ISEBC and the performance of NBCTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) The main research content and specific workflow of this study. We collected breast cancer spatial transcriptome data and used machine learning algorithms and differential gene analysis to obtain a novel breast cancer TLS-specific signature gene set. We used this gene set to assess breast cancer patient immune active subtypes and trained a LightGBM multi-classification model to finally build the ISEPBC web analysis tool to facilitate user use.\u003c/p\u003e\n\u003cp\u003e(B) Three genes were scored in the TLS regions and outside regions of slices left as the test set. The average scores inside and outside regions were calculated after normalization, and the difference was used as the evaluation performance index score. The boxplot shows the D-values of each gene set. A spatial transcriptomics slice was also selected to show the scoring of each gene set, which can clearly see the better characterization ability of NBCTS.\u003c/p\u003e\n\u003cp\u003e(C) The ROC curve shows the ability of each gene set to classify TLS regions. The classification ability is characterized by the area under the curve.\u003c/p\u003e\n\u003cp\u003e(D) Survival analysis was performed on four groups of breast cancer patients from different sources of bulk RNA-seq data. The statistical method uses log-rank test, and p \u0026lt; 0.05 was considered to indicate a survival difference between the high and low expression groups.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/b395282b3c7f923a2d7c4e97.png"},{"id":70922008,"identity":"daf0e8a1-17ce-4755-9d52-75efe05f0995","added_by":"auto","created_at":"2024-12-09 08:44:15","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":980940,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe NBCTS used for immune statues evaluation and corresponding characteristic of immune statue-II\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Consensus clustering result of TCGA breast cancer patients, with each square representing an immune status class.\u003c/p\u003e\n\u003cp\u003e(B) Survival analysis results of different immune statuses in TCGA breast cancer patients.\u003c/p\u003e\n\u003cp\u003e(C) Score of novel breast cancer TLS gene set and median heatmap of gene expression of constituent gene sets, various infiltrating cell scores of different immune statuses in TCGA breast cancer patients.\u003c/p\u003e\n\u003cp\u003e(D) Scatter plot showing umap dimension reduction results of TCGA breast cancer patients, different immune statuses characterized by different colors.\u003c/p\u003e\n\u003cp\u003e(E) Scatter plot showing GSVA calculated Antigen processing and presentation gene set scores.\u003c/p\u003e\n\u003cp\u003e(F) Volcano plot showing differentially expressed genes between the second class and the other two immune status subtypes, with the standard of absolute logFC greater than or equal to 1.\u003c/p\u003e\n\u003cp\u003e(G) Bar chart showing the top 10 biological processes enriched by the high expression genes of the second subtype.\u003c/p\u003e\n\u003cp\u003e(H) The expression levels of CD8A and CXCL13, as well as the enrichment scores of 12 chemokines from the Nature2020_RitaCabrita signature and our novel TLS gene signature, are shown. The Wilcoxon signed-rank test was used to compare groups, with p-values less than 0.05 considered statistically significant.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/111a6300528f984f9c8b7a22.png"},{"id":70922695,"identity":"3a825249-d48f-4d76-9e26-89ec78eae5df","added_by":"auto","created_at":"2024-12-09 08:52:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":445980,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe performance of immune statues evaluator\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Prediction effect and accuracy of the LightGBM breast cancer patient immune subtype classifier on the training set and test set.\u003c/p\u003e\n\u003cp\u003e(B) The ROC (Receiver operating characteristic) curve and AUC (Area under curve) value shows the performance of model classification.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/a00d05b2c67f13c50f5f1564.png"},{"id":70922696,"identity":"2a69f707-9ad5-4210-a2ec-bc4f88e3a6f9","added_by":"auto","created_at":"2024-12-09 08:52:15","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":340889,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe main function modules of ISEBC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) The main construction content and functions of the ISEBC breast cancer patient immune status web evaluation tool. Users can select the appropriate analysis module according to the number of samples in the uploaded data.\u003c/p\u003e\n\u003cp\u003e(B) The results provide a prediction of the minimum sample size. This includes: 1) A table presenting the immune status profile that is most likely based on varying sample sizes. 2) A boxplot showing the correlation between each sample and reference TCGA samples using high variable genes, with p-values determined by Wilcoxon rank-sum test to assess significance (p \u0026lt; 0.05).\u003c/p\u003e\n\u003cp\u003e(C) When the sample size is set to 'Large-scale sample collector', a pre-trained LightGBM model is used to predict the immune status of each sample. The table presents these predicted immune status results. A bar plot additionally displays the number of samples classified into each predicted immune status category.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/f3c301f7ff138b3fd671ed10.png"},{"id":70924994,"identity":"29b81e1c-4f06-46b8-8a77-b642c05c618c","added_by":"auto","created_at":"2024-12-09 09:08:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2942037,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/5c946565-d353-4c38-9568-43d1abcdf4ab.pdf"},{"id":70922011,"identity":"8f772cce-733c-4e41-bd23-1aa278462a45","added_by":"auto","created_at":"2024-12-09 08:44:15","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4457009,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytables.docx","url":"https://assets-eu.researchsquare.com/files/rs-5376285/v1/2295cf6ac54720c2cb324f53.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"ISEBC: Using A Novel Breast Cancer Tertiary Lymphoid Structures Signature To Build An Immune-favourable Status Evaluator For Breast Cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eBreast cancer is a prevalent malignancy in females worldwide that remains difficult to completely cure in all subtypes of patients[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. According to existing research, application of immunotherapy to breast cancer patients can effectively relieve or even cure the disease. The accurate evaluating the immune-favourable status of breast cancer is critical for assessing the disease state or choosing the suitable treatment way. However, effective open-source tools for evaluating the immune-favourable status of breast cancer patients are still lacking[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Some reported immune microenvironment biomarkers cannot be uniformly applied to assessments of the breast cancer immune microenvironment and evaluating criterion of these biomarkers still not be illustrate clearly. Meanwhile, timely and accurately evaluating the immune microenvironment of breast cancer patients is also important for diagnosis, disease progression monitoring and treatment selection. Tertiary lymphoid structures (TLSs) in tumor tissues are composed of infiltrated immune cells. Regularly, high infiltration of TLSs in tumor mircoenviroment often indicates better patient prognosis and a higher possibility of response to immunotherapy[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Currently, breast cancer still lacks cancer-specific TLS signature gene sets. Therefore, we used spatial transcriptome data of breast cancer patients and TLS regions in corresponding H\u0026amp;E slides annotated by pathologists. Together with feature extraction algorithms in machine learning, we identified a novel TLS signature gene set for breast cancer patient. This gene set has stronger characterization ability for breast cancer patient TLSs compared to previous gene sets and can assess the degree of TLS infiltration in breast cancer patients. Further investigation found that this gene set can also be used for evaluations of breast cancer patient immune status. Thus, we constructed an easy-to-use web tool for breast cancer immune status evaluation using this gene set to help researchers and physicians in this field timely analyze and assess the immune-favourable status of breast cancer patients or experimental samples(Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e1. Data collection\u003c/h2\u003e\n \u003cp\u003eThis study mainly used breast cancer spatial transcriptome data and bulk RNA-seq data. Breast cancer spatial transcriptomics data were primarily collected from the Gene Expression Omnibus(GEO) database, 10x Genomics official resources, and Zenodo database. Among them, the 4 patient H\u0026amp;E staining and spatial transcriptomics data of GSE176078(1160920F, CID44971, CID4535, CID4465, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5281/zenodo.4739739\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"Underline\"\u003e)\u003c/span\u003e and 3 patient H\u0026amp;E staining slices(G1, E1, F1, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://zenodo.org/records/4751624\u003c/span\u003e\u003c/span\u003e) respectively were obtained from Zendo database. The data from 10x Genomics official resources contained two consecutive slices from the same patient (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_1,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_2\u003c/span\u003e\u003c/span\u003e). An additional single slice was obtained from GEO dataset GSM6177599. In total, 10 slices were included in our used spatial transcriptome data. Bulk transcriptomic data originated from several prominent sources. The TCGA bulk breast cancer RNA data was mainly obtained from the UCSC Xena browser, comprising 1069 unique samples after deduplication and removal of adjacent normal samples. The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) dataset was mainly obtained from cBioPortal. The GSE1456 and GSE0886 data were mainly obtained from the GEO database, including 318 and 178 patients respectively. The single cell RNA sequencing data was also obtained from the GEO database via accession GSE169246.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003e2. Data processing\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial transcriptome data processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe used the Load10X_Spatial() function in Seurat to import count data in standard cellranger output format. For non-standard cellranger output data, scale.factor.json files were manually created and then read separately before using the Seurat CreateSeuratObject() function to create Seurat objects. After creating the Seurat objects, normalization was performed using SCTransform(), dimensionality reduction was conducted via RunPCA(). The first 30 PCAs were used for UMAP by RunUMAP() function. Neighbor detection utilized FindNeighbors(), and FindClusters() with a resolution of 0.7 assigned spots to clusters.\u003c/p\u003e\n\u003cp\u003eSpatialFeaturePlot() was used for visualization of gene expression across slices.\u003c/p\u003e\n\u003cp\u003eFor the evaluation of spot expression intensity of our gene set, the 12 chemokine factor gene set and the 2020 Nature gene set, AUCell package was used for scoring, and then scores were normalized from 0\u0026ndash;1 using standardization for comparison. Experienced pathologists annotated TLS regions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle cell RNAseq data processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe read separately files and use scDblFinder to remove doublets/multiplets with default parameters, then merged these expression files and use Seurat CreateSeuratObject() to create Seurat object. Then with using harmony to remove batch effect that existed in the each samples. Finally we use AggregateExpression() function to sum counts in each sample and get the pseduo-bulk counts of each sample.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBulk RNA transcriptome data processing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe TCGA breast cancer patient dataset was processed using the R programming language, initially filtering to remove duplicates, metastatic, and samples lacking survival data, retaining 1069 patients.\u003c/p\u003e\n\u003cp\u003eGEO series GSE1456 and GSE0886 were matched by probe/gene ID, converting probes to genes with summation averaging for repeats. Processed matrices were linked to clinical information. The standardized expression matrix data in METABRIC was selected and matched with patient clinical information.\u003c/p\u003e\n\u003ch3\u003e3. Identification of a novel breast cancer TLS signature gene set\u003c/h3\u003e\n\u003cp\u003eMachine learning algorithms, such as LightGBM, have demonstrated strong performance in the task of automated feature extraction from high-dimensional biological data. We leveraged LightGBM\u0026apos;s capabilities to identify feature genes that can distinguish TLS regions or non-TLS regions.Additionally, we calculated differentially expressed genes between TLS and non-TLS areas primarily using the FindAllMarkers function within the Seurat package. Genes were considered differentially upregulated in TLS spots based on adjusted p-value thresholds and logFC value. To refine our novel breast cancer TLS signature gene set, we required candidate genes to satisfy two strict criteria. First, they must have been extracted as predictive TLS features by our LightGBM model feature extraction process. Second, they were required to exhibit robust highly express in TLS regions compared with non-TLS region. After applying these feature extraction process and differential expression analysis, we identified a novel 11-gene breast cancer signature of TLS. The multi-step biomarker discovery and optimization process thus yielded a rigorously-validated signature. To validate the characterization ability of our TLS signature relative to other gene sets, we performed receiver operating characteristic (ROC) curve analysis. ROC curves provide an objective assessment of a biomarker\u0026apos;s ability to distinguish between two regions. The ROC curve plotting process began by calculating true positive rate (TPR) and false positive rate (FPR) values across a range of expression thresholds between 0\u0026ndash;1 with 0.01 increments. TPR represents the proportion of true positives that are correctly identified, while FPR is the proportion of false positives. The area under the ROC curve (AUC) was also computed as a single measure of discriminatory power, with higher values indicating superior classification ability. The TRP, FPR and AUC are computed using the following formula:\u003c/p\u003e\n\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n \u003cdiv id=\"FileID_Equa\" class=\"mathdisplay\"\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003ewhere TP is true positive, FN is false negative, TN is true negative, FP is false positive, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{{\\alpha\\:}}\\)\u003c/span\u003e\u003c/span\u003e is the function of ROC curve,\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\text{t}}_{\\text{i}}\\)\u003c/span\u003e\u003c/span\u003e is threshold that belong to 0\u0026ndash;1\u003c/p\u003e\n\u003ch3\u003e4. Pathway expression scoring\u003c/h3\u003e\n\u003cp\u003eTo score spot expression intensities for our novel gene set, the 12 chemokine gene set, and the Nature2020 set, we employed the AUCell R package. AUCell calculates gene expression scores for each spot based on provided gene sets.\u003c/p\u003e\n\u003cp\u003eStandardization from 0\u0026ndash;1 was then performed on the AUCell scores to normalize the quantitative ranges between 0 and 1. This preprocessing step ensured the scores were placed on comparable scales, facilitating direct comparisons between the performance of different gene sets.\u003c/p\u003e\n\u003cp\u003eFor pathway activity assessments of bulk RNA-seq data, we primarily utilized Gene Set Variation Analysis (GSVA). GSVA estimates pathway expression strength on a sample-by-sample manner. It has demonstrated superior ability over gene set enrichment analysis methods like single-sample GSEA (ssGSEA) to accurately gauge pathway enrichment levels from transcriptomic profiles. We opted to apply GSVA for this reason, to sensitively interrogate pathway signatures at the individual patient level throughout.\u003c/p\u003e\n\u003ch3\u003e5. Identification of breast cancer immune Status\u003c/h3\u003e\n\u003cp\u003eTo classify breast cancer patient immune subtypes, we utilized the gene expression profiles of the 11 genes comprising our novel breast cancer TLS signature within The Cancer Genome Atlas (TCGA) cohort. Unsupervised consensus clustering was performed using the ConsensusClusterPlus package in R to robustly group patient samples based on their gene expression signatures without predefining the number of clusters. This algorithm generates consensus matrices from subsampling to determine the stable cluster solution.\u003c/p\u003e\n\u003cp\u003eThrough consensus clustering of our novol TLS signature genes expression profile across TCGA samples, we reliably identified three major immune status within the patient population which intuitively depict distinct immune phenotypes present among breast cancer patients.\u003c/p\u003e\n\u003cp\u003eIn parallel, we applied UMAP dimensionality reduction to project the high-dimensional gene expression space into two dimensions for visualization. The ggplot2 package was primarily used to construct UMAP plots delineating the spatial relationships between patient samples belonging to different immune statuses .\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e6. Survival analysis\u003c/h2\u003e\n \u003cp\u003eWe thoroughly investigated the impact of gene set expression levels on patient survival outcomes. The R packages survival and survminer were primarily leveraged to conduct these analyses and generate Kaplan-Meier plots.\u003c/p\u003e\n \u003cp\u003eFor GSE1456, GSE0886, and TCGA-BRCA cohorts, median expression of the gene sets was used as the cutoff point to segregate patients into high and low expression groups for comparison.\u003c/p\u003e\n \u003cp\u003eDue to continuous clinical variable formatting in METABRIC, we applied an optimized threshold determination method to dichotomize patients based on signature scores. The ggsurvplot() function was mainly utilized to visualize and compare Kaplan-Meier survival plots between the stratified patient subgroups within each cohort.\u003c/p\u003e\n \u003cp\u003eAdditionally, pairwise log-rank tests of differences in survival distributions between immune subtypes were performed using pairwise_survdiff(). Results from these tests were also plotted using ggsurvplot() to depict divergent prognostic outcomes associated with characterized phenotypes.\u003c/p\u003e\n \u003cp\u003eThrough rigorous statistical evaluation and visualization of survival trends, we comprehensively demonstrated our gene signature\u0026apos;s utility as a prognostic biomarker across independent patient populations.\u003c/p\u003e\n \u003cp\u003eThe effect of gene set expression strength on patient survival was mainly analyzed and plotted using the survival and survminer packages. For GSE1456, GSE0886 and TCGA-BRCA patients, median expression was used to divide patients into high and low expression groups. For METABRIC, an optimized method was mainly used to determine the division of high and low expression patients. The ggsurvplot() function was used to plot the results. Survival analysis between subtypes was mainly analyzed using the pairwise_survdiff() function, and the results were also plotted using the ggsurvplot() function.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003e7. Analysis of deconvolved immune cell infiltration in patients\u003c/h3\u003e\n\u003cp\u003eTo characterize tumor-infiltrating immune cell compositions, we employed the R immunedeconv package, which utilizes validated reference target cell related genes to quantitate the proportions of various cell populations within the tumor microenvironment. This package intergrates 8 algorithms(quanTIseq、TIMER、CBIERSORT、MCP-Counter、Xcell、EPIC、ABIS、ConsensusTME、Estimate)\u003c/p\u003e\n\u003cp\u003eStandardization of the immune cell scores was performed by following formula, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{S}_{{c}_{i}}=\\:\\frac{{E}_{{c}_{i}}}{Var\\left(\\varvec{E}\\right)}\\)\u003c/span\u003e\u003c/span\u003e, which \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{{c}_{i}}\\)\u003c/span\u003e\u003c/span\u003e is the estimated immune cell scores of cell type \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{c}_{i}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{E}\\)\u003c/span\u003e\u003c/span\u003e is a vector of estimated immune cell scores of all cell types, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{S}_{{c}_{i}}\\)\u003c/span\u003e\u003c/span\u003e is the standardization score of cell type \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{c}_{i}\\:\\)\u003c/span\u003e\u003c/span\u003e.The ComplexHeatmap package was then leveraged to generate intuitive visualizations of the results. It produced heatmaps comparing the median relative infiltration levels of each immune cell subset across the three characterized immune Status .\u003c/p\u003e\n\u003ch3\u003e8. BulkRNAseq differential expression analysis and biological process enrichment analysis\u003c/h3\u003e\n\u003cp\u003eTo gain insight into molecular determinants of the second immune state, we calculated differentially expressed genes between it and the other statuses using DESeq2. DESeq2 is a widely used R package that models count data to determine differential expression while controlling for batch effects and considering biological variability. It robustly identifies changes in gene expression levels across different conditions.\u003c/p\u003e\n\u003cp\u003eWe then leveraged clusterProfiler, an enrichment analysis tool, to characterize the biological roles of top upregulated genes within the second state. Gene Ontology (GO) term overrepresentation analysis highlighted functional categories most significantly associated with these upregulated genes. And also we removed redundant processes within the enriched GO terms to condense the biological themes driving distinct immune responsiveness in the favorable second immune state.\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e9. Construction and validation of breast cancer immune status prediction model\u003c/h2\u003e\n \u003cp\u003eWe leveraged the TCGA breast cancer RNA-seq dataset to train and evaluate an immune statues classification model. The top 100 most variably expressed genes across patients were selected as predictive input features. These genes underwent normalization to scale expression values from 0 to 1, the following formula:\u003c/p\u003e\n \u003cdiv id=\"Equb\" class=\"Equation\"\u003e\n \u003cdiv id=\"FileID_Equb\" class=\"mathdisplay\"\u003e$$\\:{n}_{i}\\:=\\:\\frac{{x}_{i}-min\\left(\\varvec{x}\\right)}{max\\left(\\varvec{x}\\right)\\:-\\:min\\left(\\varvec{x}\\right)}$$\u003c/div\u003e\u003c/div\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{i}\\:\\)\u003c/span\u003e\u003c/span\u003eis the normalized value of gene\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e, the vector \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{x}\\)\u003c/span\u003e\u003c/span\u003e is these top 100 most variably expressed gene expression value of different patients, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{n}_{i\\:}\\)\u003c/span\u003e\u003c/span\u003e is the normalized value ranging from 0 to 1. The preprocessed dataset was then randomly partitioned using 70% for model training and the remaining 30% held-out for testing. This train-test split procedure was repetitively conducted 3 times with different random seeds to validate model performance. For each split, a LightGBM classifier was trained on the 70% training subset. LightGBM aims to learn a function \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{\\phi\\:}(\\varvec{n};\\theta\\:)\\)\u003c/span\u003e\u003c/span\u003e that maps gene expression profiles to immune statues, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\theta\\:\\)\u003c/span\u003e\u003c/span\u003e represents the model parameter set. Hyperparameter optimization via grid search was performed to identify \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\theta\\:\\)\u003c/span\u003e\u003c/span\u003e values minimizing a loss function:\u003c/p\u003e\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:L\\:={f}_{\\phi\\:}(\\varvec{n};{\\theta\\:}_{i})-\\varvec{c}\\:\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u0026nbsp;\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{c}\\)\u003c/span\u003e\u0026nbsp;\u003c/span\u003e is the classes of immune statues ,with the aim of finding an optimal \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\theta\\:\\)\u003c/span\u003e\u003c/span\u003e corresponding to lower \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:L\\)\u003c/span\u003e\u003c/span\u003e. The fine-tune process is to find a suitable parameters of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\theta\\:}_{i}\\)\u003c/span\u003e\u003c/span\u003e from \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\{{\\theta\\:}_{1},{\\theta\\:}_{2}...{\\theta\\:}_{L}\\}\\)\u003c/span\u003e\u003c/span\u003e. Finally the trained lightGBM model is the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{f}_{\\phi\\:}(\\varvec{n};{\\theta\\:}_{i})\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003cp\u003ePrediction performance was subsequently assessed using the separate 30% test subset. The scikit-learn metrics module was primarily used to analyze model performance. Confusion matrices generated by confusion_matrix() visualized predicted versus true class assignments. AUC scores computed by auc() provided a summary metric of discriminatory power between immune status. This rigorous validation workflow objectively evaluated our machine learning approach for immune status classification of new patient samples.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e10. Construction of a breast cancer immune status evaluation web tool\u003c/h2\u003e\u003cp\u003eWe used R shiny as the main framework to build an interactive web application. For queries containing few samples(1-400), it primarily computes Spearman\u0026apos;s correlation coefficients between the top 100 variant genes from our training dataset and the overlapping genes in the user-uploaded expression matrix. For Larger sample sets(\u0026gt;\u0026thinsp;400) first undergo normalization of the top variant genes to the 0\u0026ndash;1 range. The scaled expression profiles are then fed into our pre-trained LightGBM classification model for immune status prediction. The ggplot2 package was mainly leveraged to generate publication-quality data displays for intuitive interpretation. Users can conveniently query breast cancer patients or samples expression data without requirement of specialized computational skills. This optimized workflow empowers immune status classification to advance research and guide clinical decisions.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e1. A novel breast cancer-specific tertiary lymphoid structures signature\u003c/h2\u003e \u003cp\u003eSpatial transcriptomics, as an impactful biotechnology in 2022, characterizes gene expression patterns across different locations and regions within a biological sample. By resolving mRNA transcripts into spatially encoded spots, it allows researchers to study gene expression and functions of specific genes in spatial locations from a new perspective, something not possible with conventional single cell RNA sequencing or bulk RNA sequencing. Therefore we use spatial transcriptomics data to study the TLSs. TLSs are immune cell aggregates that form within tumors and play an important role in the anti-cancer response. To define an breast cancer-specific signature of TLS, we collected multi-center breast cancer patient spatial transcriptome datasets from different sources (Supplementary table 1), and used the feature extraction algorithm lightGBM to identify TLS feature genes. Then, we calculated specific highly expressed genes in TLS regions using wilcox\u0026rsquo;test, to intersect these highly expressed genes in TLS regions and extracted gene features, we finally established a novel breast cancer tertiary lymphoid structure signature gene set (NBCTS). This gene set include 11 genes, which are CD52, CD48, CXCL9, CXCL13, TRBC2, CCL19, MS4A1, LCP1, CSN1S1, TBC1D10C, RAC2. CXCL13 is a well-defined marker gene of TLS regions based on published studies. CCL19 participate in recruiting the B cells and T cells which are cornerstone of the TLS formation in the initial stage. Compared to established gene sets like 12 classic chemokine factor genes[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] and a gene set published in Nature in 2020 by Rita Carita[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], our new breast cancer TLS gene set exhibited larger differences in expression between inner TLS regions versus outside, as well as stronger signal detection within testing set slices (E1, G1, F1, GSM6177599)(Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Additionally, receiver operating characteristic (ROC) analysis indicated our gene set outperformed others in predicting TLS infiltration strength in breast cancer patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). Kaplan-Meier survival analysis of the NBCTS across independent transcriptomic cohorts also illustrated its utility for characterizing patient prognosis. Overall, by leveraging spatial transcriptomics datasets, we defined a robust signature that can be used in the prediction or scoring the infiltration level of TLS and exists translational potential.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e2. Application of NBCTS in evaluating patient tumor immune status\u003c/h2\u003e \u003cp\u003eSince TLS within tumors play a critical role in recruiting and activating anti-cancer immune cells, thereby enhancing patient survival outcomes,we therefore explore whether leveraging our novel breast cancer-specific TLS signature gene set can comprehensively evaluate the degree and composition of immune cell infiltration across patient tumor tissues. If the envisage can be realized, this would give a new way to assess tumor immune status of each patient.\u003c/p\u003e \u003cp\u003eThrough unsupervised consensus clustering analysis of NBCTS gene expression profiles from The Cancer Genome Atlas (TCGA) breast cancer cohort, we found three different immune statuses within the patient population (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Further analyses showed the patients of second immune status show significantly better prognosis and survival compared to the first and third status. Patients assigned to the second status also exhibited higher comprehensive scores for our novel TLS signature gene set versus the other statuses (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB) .\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo gain deeper insight, we employed multiple computational cell deconvolution algorithms like quanTIseq, TIMER, CBIERSORT, MCP-Counter, Xcell, EPIC, ABIS, ConsensusTME to estimate the relative proportions of different immune cell populations infiltrating tumors across immune statuses. Results demonstrated the second immune status had markedly elevated levels of both effector memory and naive CD8\u0026thinsp;+\u0026thinsp;T cells relative to the other groups. Given the importance of B cells in TLS formation, we observed B cell abundances were also significantly greater in patients within the second immune status versus the other two immune status, which is consistently with NBCTS scores of these patients(Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC, Supplementary Fig.\u0026nbsp;1A). The antigen processing and presentation pathway also found significantly up-regulated in the second immune status (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD,E).\u003c/p\u003e \u003cp\u003eBy doing differential expression genes analyses compare the second immune status with the remaining patients, we found that some genes are closely related to enhancing patient anti-tumor immune ability, such as CXCL13 and IFNG, were highly expressed in the second status (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eF). Gene Ontology enrichment analysis of highly expressed genes in the second status found that they were mainly enriched in biological processes related to regulating immune antibody production and lymphocyte-mediated immune response regulation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eG). We also use a single cell dataset (Supplementary table 2) of combining immuntherapy and chemotherpy cohort to make a pseudo-bulk expression matrix, with GSVA algorithm we calcualte the enrichment score of our NBCTS, 12 classic chemokine factor genes and a gene set published in Nature in 2020 by Rita Carita. Compare with these gene sets and marker genes (CD8A, CXCL13) which are widely proven as favourable markers for immunetherapy, our NBCTS shows significantly different between partial response patients and stable disease patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eH). By synthesizing these results, we can reasonably conclude that the second immune status represents an immune-favorable microenvironment for breast cancer patients undergoing immunotherapy. Additionally, our analysis indicated that the second immune status correlated with a TLSs-dominant profile. Indeed, through our whole immune status classification process, the second immune status emerged as the most important immune-favorable microenvironment profile identified.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3. Construction of a breast cancer immune status classifier model based on the NBCTS\u003c/h2\u003e \u003cp\u003eCurrently, some gene expression-based signatures have been proposed for evaluating the immune infiltration status of breast cancer patients. However, they generally lack standardized implementation protocols that lead to some difficulties in practical applications. This problem can largely be attributed to two main factors: the high biological heterogeneity between patients' tumor immune microenvironments, and the absence of well-defined thresholds for classifying gene expression levels as high or low across different patients and studies. As a result, the utility and reproducibility of previous gene expression signatures for classify tumor\u0026rsquo;s immune statues remains difficulty.\u003c/p\u003e \u003cp\u003eTo develop a more broadly applicable and transferable immune statuses evaluating solution, we constructed a three-class immune status classifier using gene expression data from TCGA breast cancer cohort and actually we mainly want to build a model that can successfully classify the second immune-favorable status. Patients were annotated with the matching immune status based on previous classifier results, which categorized samples according to the expression pattern of our novel TLS signature gene set. To optimize characterization and standardization of the training data, we first calculated gene variance and identified the top 100 most variably expressed genes as training features. The expression values of these top 100 genes were then normalized from 0 to 1 across all patients to pre-process the data prior to classifier training.\u003c/p\u003e \u003cp\u003eThe classifier performance was rigorously evaluated using multiple testing strategies. By initializing different random seeds and conducting repeated train-test splits of the TCGA data, the out-of-sample classification predictions from the test sets achieved excellent accuracy under all random seeds. Specifically, the area under the receiver operating characteristic curve exceeded 0.99 for each immune status class and resampling scenario, indicating exceptional discriminatory power (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA,B).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo assess generalizability beyond the TCGA cohort, we next applied our pre-trained classifier to an independent breast cancer gene expression dataset from the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC). Reassuringly, the predicted immune status distributions from this external validation set closely recapitulated the cell infiltration profiles (Supplementary Fig.\u0026nbsp;1B) and survival trends of the reference TCGA subtypes(Supplementary Fig.\u0026nbsp;1A). This consistency provides strong evidence for the classifier's robust predictive performance across studies.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4. A web tool for breast cancer patient immune status evaluation\u003c/h2\u003e \u003cp\u003eIn order to better apply this immune status prediction model and make the tool more accessible and user-friendly for various users, we constructed a web application called ISEBC (Immune State Evaluation tool for Breast Cancer, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"http://www.omegene.tech:3838/ISEBC\" target=\"_blank\"\u003ewww.omegene.tech:3838/ISEBC\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.omegene.tech:3838/ISEBC\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) for assessing the immune status of breast cancer patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe ISEBC web application framework was built using the R shiny platform and provides two analysis options depending on sample size. For evaluations involving 1-400 samples, Spearman correlation analysis is performed due to the relatively small number of patient conditions represented. It calculates the similarity of highly variable gene expression between the uploaded data and reference TCGA breast cancer patients for each immune status. The overall highest similarity to a given reference immune status indicates a higher probability that the sample(s) belong to that status, and this result is the of web analysis output (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003eWhen evaluating larger sample sets containing more than 400 patients, users can opt for the \"large sample assessment\" mode. In this scenario, ISEBC will predict the immune status of each individual patient by applying the pre-trained LightGBM machine learning model. This model was trained on gene expression profiles from TCGA patients annotated with immune statuses using the top 100 most variably expressed genes. Its predictions leverage the broader expression profile characterization captured during training from a larger and more diverse set of reference samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC).\u003c/p\u003e \u003cp\u003eOverall, ISEBC provides a user-friendly web interface for evaluating the immune status of breast cancer patients or samples using either a correlation-based or machine learning-based approach, depending on the number of samples analyzed. This tool will help facilitate practical application of evaluating the immune status for breast cancer.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eCompared to other tumor types, the immunogenicity of breast cancer is still lower than other cancer types, especially highly immunogenic cancer types such as melanoma and lung cancer[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Accurately characterizing the tumor immune status is critically important for tailoring optimal treatment strategies to individual patients.The presence of TLSs within tumors play an important role in facilitating the recruitment of cytotoxic CD8\u0026thinsp;+\u0026thinsp;T cells and immune-activating B cells, which help eliminate cancer cells. Numerous previous studies have demonstrated robust correlations between increased TLS infiltration and more favorable patient outcomes, including improved overall survival and prognosis[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The formation of TLS within the tumor microenvironment is believed to enhance the adaptive anti-tumor immune response by supporting interactions between lymphocytes and antigen-presenting cells that are needed to activate and boost anti-tumor immunity[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Through analyzing spatial transcriptomic datasets, we identified a novel breast cancer-specific TLS gene signature (NBCTS). Compared with other TLS gene sets, this novel TLS gene set identified from breast cancer spatial transcriptome datasets in our study has better performance in predicting TLS infiltration. Some previous gene sets or methods based on the distance between CD8\u0026thinsp;+\u0026thinsp;T cells and tumor cells or based on the infiltration of immune cells also can cluster patients into different subclusters, but the usage protocol was not clear and researchers hardly to use these gene sets or methods. For the widely used, We not only training a immune state classification model, but also construct a web tool and the ability to evaluate patient immune states in different sample amount modes, it allows users to more conveniently analyze the immune state of samples.\u003c/p\u003e \u003cp\u003eFor validation using independent cohorts, we applied a 0\u0026ndash;1 normalization transform preserving gene rank orders. This standardized preprocessing approach may enhance future evaluations of patient immune statuses through more robust cross-study comparisons and integrated analyzes. Training machine learning models on high-quality reference datasets is also paramount for developing predictive models with translational potential.\u003c/p\u003e \u003cp\u003eBased on immune states defined by our evaluator, the second phenotype showed most prominent immune traits while the first showed weakest. The second immune state also show significant B cell-related features and obviously exhibit TLS-dominant phenotype. Comparing evaluation of CXCL13\u0026thinsp;+\u0026thinsp;CD8\u0026thinsp;+\u0026thinsp;T cell infiltration level as originally reported, patients with high expression of the NBCTS gene set showed significantly greater benefit from immunotherapy than patients with low expression of the NBCTS gene set. This indicated that our NBCTS gene set have potential ability for evaluating the immune therapy response rate. However, due to the small number of breast cancer immunotherapy treatment open-source data currently available, this conclusion may require larger sample size datasets in the future to validate.\u003c/p\u003e \u003cp\u003eOverall, the novel breast cancer TLS signature gene set identified in our study has better characterization ability than previous gene sets and can well classify breast cancer patients into different immune states. And we also find a survival and immune favorable statues of breast cancer patients. The developed web tool empowers widespread interrogation of tumor immune statues to deepen understanding and guide personalized medicine decisions.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eTLS:Tertiary lymphoid structure; NBCTS: novel breast cancer tertiary lymphoid structure signature gene set; OS: overall survival; TCGA: The Cancer Genome Atlas; METABRIC: Molecular Taxonomy of Breast Cancer International Consortium\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe sincerely thank Alexander Swarbrick (Garvan Institute of Medical Research) for sharing their single-cell RNA sequencing data in the GEO database and spatial transcriptome data in the Zenodo database.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eXK.F., L.C. conceived and designed the study and drafted the manuscript. XK.F., X.Y performed the analysis of the data. The author(s) read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe code will be uploaded to github after paper published.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur re-analysis single-cell RNA data were deposited at the National Center for Biotechnology Information (NCBI) GEO and are accessible through the accession number GSE169246. The spatial transcriptome data of our used were deposited at the Zenodo database and 10x offical resource, the website is (https://doi.org/10.5281/zenodo.4739739, https://zenodo.org/records/4751624,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_1,https://support.10xgenomics.com/spatial-gene-expression/datasets/1.1.0/V1_Breast_Cancer_Block_A_Section_2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003econsent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no potential conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eVoduc KD et al (2010) Breast cancer subtypes and the risk of local and regional relapse. J Clin Oncol 28(10):1684\u0026ndash;1691\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarzaman K et al (2020) Breast cancer: Biology, biomarkers, and treatments. Int Immunopharmacol 84:106535\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColbeck EJ et al (2017) Tertiary Lymphoid Structures in Cancer: Drivers of Antitumor Immunity, Immunosuppression, or Bystander Sentinels in Disease? Front Immunol 8:1830\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDieu-Nosjean MC et al (2016) Tertiary lymphoid structures, drivers of the anti-tumor responses in human cancers. Immunol Rev 271(1):260\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeow DYB et al (2020) Tertiary lymphoid structures, and associated plasma cells play an important role in the biology of triple-negative breast cancers. Breast Cancer Res Treat 180(2):369\u0026ndash;377\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrabhakaran S et al (2017) Evaluation of invasive breast cancer samples using a 12-chemokine gene expression score: correlation with clinical outcomes. Breast Cancer Res 19(1):71\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCabrita R, Lauss M, Sanna A et al (2020) Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature 577:561\u0026ndash;565\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchumacher TN, Thommen DS (2022) Tertiary lymphoid structures in cancer. Science 375(6576):eabf9419\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMunoz-Erazo L et al (2020) Tertiary lymphoid structures in cancer - considerations for patient prognosis. Cell Mol Immunol 17(6):570\u0026ndash;575\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Tertiary lymphoid structures (TLS), Breast cancer, Immune state evaluation, Biomarker research","lastPublishedDoi":"10.21203/rs.3.rs-5376285/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5376285/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTertiary lymphoid structures (TLSs), as a special type of immune-infiltrated region in tumor tissues, plays an important role in benefiting from immunotherapy or improving immune state of breast cancer patient. However, breast cancer-specific TLS signature gene sets are still lacking. Therefore, we extracted gene features with machine learning algorithm LightGBM, and differential expression genes with statistical test on multiple spatial transcriptome datasets, to finally obtain a novel breast cancer-specific TLS gene set (NBCTS). Compared with previous gene sets, it has stronger characterization ability of TLS in breast cancer samples. Since TLS have unique immune characteristics, we classified three different immune states using this gene set for breast cancer patients and get an immune state. To better facilitate evaluating this immune statuses of breast cancer patients or samples, we developed a user-friendly web tool (Immune State Evaluator for Breast Cancer, (ISEBC),\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"http://www.omegene.tech:3838/ISEBC\" target=\"_blank\"\u003ewww.omegene.tech:3838/ISEBC\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.omegene.tech:3838/ISEBC\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) to make it more convenient for researchers and clinicians to use.\u003c/p\u003e","manuscriptTitle":"ISEBC: Using A Novel Breast Cancer Tertiary Lymphoid Structures Signature To Build An Immune-favourable Status Evaluator For Breast Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-09 08:44:11","doi":"10.21203/rs.3.rs-5376285/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0ca027bd-506a-4946-80e7-73cde1e33c31","owner":[],"postedDate":"December 9th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-12-09T08:44:13+00:00","versionOfRecord":[],"versionCreatedAt":"2024-12-09 08:44:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5376285","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5376285","identity":"rs-5376285","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}