Access to this full-text is provided by Springer Nature.
Content available from Scientific Reports
This content is subject to copyright. Terms and conditions apply.
Exploring the tumor
microenvironment of breast cancer
to develop a prognostic model and
predict immunotherapy responses
Ye Tian1,3, Yong Yang1,3, Lei He2, Xiaocheng Yu1, Hu Zhou2 & Juan Wang2
Breast cancer is the most prevalent malignancy in women and exhibits signicant heterogeneity.
The tumor microenvironment (TME) plays a critical role in tumorigenesis, progression, and response
to therapy. However, its impact on the prognosis and immunotherapy responses is incompletely
understood. Using public databases, we conducted a comprehensive investigation of transcriptome
and single-cell sequencing data. After performing immune inltration analysis, we conducted
consensus clustering, weighted gene co-expression network analysis (WGCNA), Cox regression,
and least absolute shrinkage and selection operator (Lasso) regression to identify independent
prognostic genes in breast cancer. Subsequently, we developed a prognostic model for patients with
breast cancer. Tumor Immune Dysfunction and Exclusion (TIDE) values were used to assess patient’s
responsiveness to breast cancer. Based on single-cell RNA-sequencing data, we identied various
cell types through cluster analysis and investigated the expression of prognostic model genes in
each cell type. The drug sensitivity of targeted therapeutic agents for breast cancer treatment was
analyzed in dierent cell types. We identied 12 independent prognostic genes associated with
breast cancer and used these genes to construct a prognostic model. The prognostic model accurately
discriminated between patients classied as high- and low-risk, providing precise prognostic
predictions for individual patients. Additionally, our model exhibited a robust capacity to predict
the immunotherapeutic response in breast cancer patients. Our investigation revealed a notable
association between the proportion of endothelial cells (ECs) and patient prognosis in breast cancer. A
prognostic model for breast cancer was formulated that showed close associations between prognosis
and response to immunotherapy. For patients predicted by our model to not respond eectively to
immunotherapeutic agents, it may be considered to combine immunotherapeutic agents with targeted
therapeutic agents identied through our drug sensitivity analysis, which could potentially enhance
treatment ecacy.
Keywords Breast cancer, Single-cell sequencing, Bulk RNA-sequencing, Prognostic model, Immunotherapy
Breast cancer is the most common in women and the second leading cause of cancer mortality globally1. Its
morbidity and mortality rates exceed 20% and 10%, respectively2, signicantly impacting women’s lives and
healthcare systems3. Treatments for breast cancer include surgery, chemotherapy, radiotherapy, endocrine
therapy, and targeted therapy4; however, the high heterogeneity of breast cancer limits the eectiveness of
endocrine and targeted molecular therapies5. Recently, immunotherapy, particularly immune checkpoint
inhibitors (ICIs), has shown promise in cancer treatment6, though its ecacy in breast cancer remains challenged
owing to its notable heterogeneity7. erefore, there is an urgent need to develop enhanced prognostic tools and
biomarkers to accurately predict and eectively treat breast cancer.
e tumor microenvironment (TME) is widely recognized as playing a key role in cancer progression8 and
therapy resistance9. ICIs have proven eective in treating triple-negative breast cancer (TNBC)10 and their
role in HER2-positive breast cancer is of signicant interest to researchers. However, biomarkers or molecular
models capable of predicting immunotherapy response in patients are limited. Further exploration of how the
1Department of Thyroid and Breast Surgery, Wuhan No.1 Hospital, Tongji Medical College, Huazhong University of
Science and Technology, Wuhan, China. 2Department of Blood Transfusion, Tongji Hospital, Tongji Medical College,
Huazhong University of Science and Technology, Wuhan, China. 3Ye Tian and Yong Yang contributed equally to this
work and share rst authorship. email: juanwangsxk@tjh.tjmu.edu.cn
OPEN
Scientic Reports | (2025) 15:12569 1
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports
Content courtesy of Springer Nature, terms of use apply. Rights reserved
TME inuences immunotherapy response is needed. e Tumor Mutational Burden (TMB) stands as a potential
prognostic indicator, albeit the nexus between TMB and the prognosis of tumor patients remains a contested
point of discussion11.
Single-cell RNA sequencing (scRNA-seq) enables transcriptomic analysis at the cellular level, revealing
the diverse cell types while characterizing each cell’s transcription with specicity12,13. is technique oers
valuable insights into diverse cell states and the heterogeneity of cell populations, making it a valuable tool
for elucidating the characteristics of various cell types present in and surrounding breast cancer tumors.
is technique has identied diverse populations potentially associated with an unfavorable prognosis and
resistance to treatment14–16. Additionally, this method can also uncover the heterogeneity within the tumor
microenvironment, where such subpopulations might act as prospective targets for immunotherapy17. Numerous
studies have suggested that M2 macrophage inltration of the TME contributes to immunosuppression and
facilitates cancer progression, angiogenesis, invasion, and metastasis18, while angiogenesis in endothelial cells
(ECs) inuence progression and metastasis in cancer cells19.
In this study, we utilized public breast cancer data to assess immune cell inltration. Consensus clustering,
WGCNA, Cox regression, and Lasso regression analyses were applied to identify independent prognostic genes
for breast cancer, leading to the development of a prognostic model for patients with breast cancer. We also
evaluated the role of TMB in breast cancer prognosis and used single-cell sequencing analysis to identify high-
risk cell types. Next, we conducted further analyses on high-risk cell types. Further analyses were conducted
to assess variation in the drug sensitivity of dierent cell types providing novel insights into the prognosis and
treatment of breast cancer.
Materials and methods
Data sources
We downloaded e Cancer Genome Atlas (TCGA) breast cancer RNA sequencing data for gene expression
(FPKM and count values) (n = 1217) and somatic mutations (MuTect2 version, n= 986) from UCSC Xena20,
along with the corresponding prognosis information of breast cancer patients (n= 1260). Simultaneously, we
downloaded the GSE58812 and GSE23428 expression proling datasets from the GEO database (e Gene
Expression Omnibus)21 as validation sets. e GSE76275 dataset was downloaded and subsequently employed
for validation purposes.
We downloaded the breast cancer single-cell dataset GSE15839922 from the GEO database as the basis for
analysis. e species of the GSE158399 dataset was Homo sapiens, the sequencing platform of the GSE158399
was GPL20795 (HiSeq X Ten), and it contained three samples. Based on previous studies23,24, we selected an in
situ breast cancer sample (GSM4798908_B2019-1.expression_matrix_input.txt) (Supplementary Table S1) for
subsequent analyses.
Immune inltration analysis
e TME encompasses the immediate surroundings of tumor cells, including blood vessels, immune cells,
broblasts, diverse signaling molecules, and the extracellular matrix. e R package xCell25 was used to calculate
the enrichment scores of the immune and stromal cell types among the 64 cell types. e xCell analysis of
immune cell inltration was performed using the IOBR package26. e R package ggplot2 (version 3.4.2) was
used to generate box plots and stacked bar graphs to display the results. e Corrplot package (version 0.92) was
used to construct a correlation heat map.
Consensus clustering analysis
e consensus clustering27 algorithm identies potential clusters characterized by inherent heterogeneity,
ensuring stability across multiple runs28. Using the R package ConsensusClusterPlus (version 1.62.0)29, we
performed the clustering of the immune inltration results deconvoluted by xCell in TCGA-Breast cancer
expression data to better distinguish dierent subtypes of breast cancer samples. To verify the relationship
between the consensus-clustering subgroups and the overall survival (OS) rate, we used the survival package
(version 3.5.3) to analyze survival dierences of the consensus-clustering subgroups using the Kaplan-Meier
method. Subsequently, we examined dierences in the expression of the immune checkpoints PD1 (PDCD1)
and PDL1 (CD274) in the consensus-clustering subgroups.
For the TCGA-Breast Cancer expression data, the R package DESeq2 (version 1.38.3)30 was used for analysis
of dierentially expressed genes (DEGs) between consensus clustering subgroups. Genes with |logFC| ≥1 and P.
adj < 0.05 were considered to be DEGs. Volcano plots and heat maps were drawn to visualize the DEGs. We also
analyzed dierences in gene expression related to breast cancer between the consensus clustering subgroups.
Weighted gene co-expression network analysis
WGCNA evaluates the patterns of gene expression across multiple samples31. is method clusters genes
with analogous expression proles and investigates associations between gene modules and specic traits or
phenotypes. Genes exhibiting consistent expression trends across physiological processes or tissues may have
functional relationships and can be delineated as modules. Each gene module is denoted by a unique color.
Using the R package WGCNA (version 1.72.1)32, we applied WGCNA to the DEGs of consensus clustering
subgroups from TCGA-Breast cancer data. To identify the network modules, WGCNA was employed with the
following parameters: soPower set to 3, mergeCutHeight at 0.25, and a minimum module size of 30. Aer
selecting the module of interest, all genes within the module were identied as expressed genes that were highly
correlated with the consensus clustering subgroups.
Scientic Reports | (2025) 15:12569 2
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Enrichment analysis
Gene Ontology (GO)33 enrichment analysis aims to elucidate and categorize the functions of genes. e Kyoto
Encyclopedia of Genes and Genomes (KEGG)34 is a widely used biological database that catalogues advanced
biological functions. For the expressed genes highly correlated with the consensus clustering subgroups,
enrichment analyses based on GO and KEGG were conducted using clusterProler (version 4.7.1.3)35 package.
e enrichment results were visualized using R packages GOplot (version 1.0.2)36 and enrichplot (version
1.18.4).
Screening prognostic-related genes and constructing the prognostic model
To identify the prognostic genes closely related to the consensus-clustering subgroups based on immune
inltration analysis, we conducted univariate and multivariate Cox regression analyses of the genes associated
with the consensus-clustering subgroups and selected independent prognostic genes. LASSO Cox regression
analysis was then applied for additional screening. Utilizing the cv.glmnet function from the glmnet package
(version 4.1.7), the optimal lambda value was determined through 10-fold cross-validation with a minimum
partial likelihood deviance37. Genes with non-zero coecients in the LASSO regression analysis were specically
chosen to construct the prognostic model. A multiple regression model was applied to calculate the risk score
using the independent prognostic genes selected by lasso-cox regression analysis. Risk scores were computed
using gene expression levels in combination with multivariate Cox regression coecients. e risk scores were
calculated as follows:
riskS
core =
∑
i
Coef f icient (hub genei)∗mRN A E xpression (hub genei
)
We utilized the median risk score as a cuto point to categorize the TCGA breast cancer samples into high-
and low-risk groups, based on the following considerations: First, the median risk score oers an objective
and balanced method for division, ensuring that the sample sizes of both groups are comparable and thereby
enhancing the stability of statistical analyses. Second, employing the median risk score as a cuto point is a
widely accepted approach in bioinformatics research. To assess the independent predictive ability of the risk
score for OS, KM survival analysis was conducted.
To further explore the relationship between high- and low-risk groups and immunotherapy, we analyzed
the immune checkpoint expression in the high- and low-risk groups. TIDE was used to predict the response
to immune checkpoint blockade38,39, with TIDE score computed for individual patients with tumors holding
promise as a novel biomarker, oering insights into the potential ecacy of immune checkpoint blockade.
TIDE scores were calculated using TCGA breast cancer data to ensure consistency in data analysis and facilitate
comparability across samples. e TIDE scores for each patient were obtained from the TIDE website ( h t t p s : / / t i
d e . d f c i . h a r v a r d . e d u / l o g i n / ) . Subsequently, the dierences in TIDE scores between the high- and low-risk groups
was analyzed. e R package ggExtra (v0.10.0) was used to draw scatter plots and t the correlation curves. Aer
completing the above analyses, we further validated the relationship between TIDE scores and risk scores using
GSE76275 as an external dataset.
Dierences in the TMB between the high- and low-risk groups were calculated. Subsequently, we employed
the surv_cutpoint function from the survminer package (version 0.4.9) in R to ascertain the optimal cuto point
serving as a threshold, dividing patients into high- and low-TMB groups, and then conducted KM survival
curve analysis between the high- and low-TMB groups. Additionally, we determined the mutational status of the
prognostic model genes using the maools package (version 2.14.0)40.
Single-cell clustering
To ensure data quality, we removed cells expressing fewer than 200 genes or more than 5000 genes, as well as cells
with mitochondrial gene expression exceeding 30%. Using the Seurat R package (version 4.3.0)41, we normalized
the single-cell data from GSE158399 (default parameters). We identied the top 2000 genes exhibiting the
greatest variability as highly variable genes (HVG), subsequently conducting principal components analysis
(PCA) on the HVG to reduce the data’s dimensionality. We selected the rst 30 principal components. e
nearest neighbors were identied among cells using the k-nearest neighbor (KNN) method (FindNeighbours
function), aer which cell clustering was performed using the Louvain algorithm. Unsupervised Uniform
Manifold Approximation and Projection (UMAP) was employed to visualize the cluster, followed by cluster
annotation. Single-cell RNA sequencing data has higher levels of technical noise and variability compared to
bulk RNA-seq data. erefore, in order to obtain more potential biological information and to achieve a more
comprehensive understanding of the functions of dierent cell types in subsequent analyses, we adopted a more
relaxed threshold in single-cell clustering analysis by referencing existing literature42. DEGs between cell types
were identied using GSE158399, the FindAllMarkers function of the R package Seurat (v4.33.0)41 with |logFC|
≥ 0.25 and P. adj < 0.05 as the criteria for DEGs selection.
To further investigate the relationship between cell type and prognostic model genes, we visualized
the expression of prognostic model genes in each cell type using bubble plots. Subsequently, we used the
AddModuleScore function of the R package Seurat (version 4.33.0) to calculate the scores of the model genes
in each cell type and to identify the cell types related to the model genes. en we utilized the R package
clusterProler(version 4.7.1.3) to conduct gene set enrichment analysis (GSEA)43on the DEGs in cell types
related to model genes and other cell types. GSEA was employed using the “h.all.v7.4. symbols.gmt” gene set
obtained from MSigDB44.
Scientic Reports | (2025) 15:12569 3
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Deconvolution analysis of the proportions of various cell types in TCGA breast cancer
samples
MuSiC45 is a method that employs cross-subject single-cell RNA sequencing to determine cell-type proportions
in bulk RNA sequencing data. Using breast cancer single-cell RNA sequencing data as a reference, we inferred
the cell type proportions in TCGA breast cancer samples using the R package MuSiC (version 1.0.0).
Based on the median proportion of high-risk cell type in each sample, we classied the samples into high-
and low-proportion high-risk cell groups. A KM survival curve analysis was conducted to analyze the survival
dierences between the high- and low-proportion high-risk cell groups. e proportion of high-risk cell type
might be correlated with the risk score; therefore, the ggExtra package (version 0.10.0) was used to create a
scatter plot between the proportion of high-risk cell type and risk scores and t the correlation curve.
Prediction of drug sensitivity for various cell types
To predict single-cell drug sensitivity, we used drug estimation from single-cell expression proles (DREEP)46.
Data on the sensitivity of the cancer cell lines to various drugs were obtained from the Cancer erapeutic
Response Portal (CTRP v2)47. e ecacy of each drug in each cell line was quantied using the Area Under
the Curve (AUC) of the dose-response curve. As the AUC mirrors a cell line’s in vitro response to varying
drug concentrations over a 72-hour timeframe, lower AUC values imply sensitivity, whereas higher values
suggest resistance to the drug. Consequently, genes exhibiting a positive correlation with the AUC might serve
as potential markers of resistance, whereas those that are negatively correlated may indicate sensitivity. We
established a list, in order of relevance, of biomarkers associated with drug sensitivity and resistance for each
drug. To predict single cell sensitivity to these drugs, we selected the top 250 expressed genes from each cell,
and used them as inputs for GSEA against the ordered list of biomarkers for every individual drug, as previously
described. erefore, a negative enrichment score (ES) signies that a cell highly expresses genes associated with
drug sensitivity, whereas a positive ES suggests that the cell expresses genes that confer drug resistance.
Statistical analysis
Data calculations and statistical analyses were performed using the R programming soware (version 4.2.3).
e Benjamini-Hochberg (BH) method was used for multiple test corrections to control the false discovery rate
(FDR). Specically, BH correction was implemented for dierential expression analysis using DESeq2, gene
ontology enrichment analysis via clusterProler, and Cox regression analysis for prognostic gene identication.
Due to the fact that single-cell RNA sequencing data and bulk RNA-seq data typically do not follow a normal
distribution, this study employed the Mann-Whitney U test to compare the two groups of data. e Wald test was
employed to identify DEGs. Survival analyses were performed using Kaplan-Meier curves, and the signicance
of the results was determined by the log-rank test using the survival package in R48. All statistical tests were two-
sided, with a P value < 0.05 as statistically signicant.
Results
Consensus clustering based on immune inltration
A owchart of the study is shown in Fig.1. Using the TCGA breast Cancer data, we calculated scores for 64 cell
types using the xCell algorithm to explore the degree of inltration of immune-related cell types. Subsequently,
the samples were clustered. e optimal number of subgroups was determined to be two: cluster 1 (C1) and
cluster 2 (C2) (Fig.2A). Figure2B shows strong clustering eects between the two clusters, where C1 consisted
of 972 samples and C2 included 245 samples (Supplementary Table S2). Disparities in immune cell inltration
levels between C1 and C2 were quantied (Fig.2C). Among the 38 types of immune cells, there were 8 types
that showed no signicant dierence in immune inltration levels between C1 and C2. ese 8 cell types were:
common myeloid progenitors (CMP), mast cells, Eosinophils, Class-switched memory B cells, central memory
CD4 + T cells (CD4 Tcm), Regulatory T cells (Tregs), natural killer cells (NK cells), and natural killer T cells
(NKTs). e remaining 30 types of immune cells exhibited signicant dierences in immune inltration levels
between C1 and C2. Common lymphoid progenitor cells (CLP), granulocyte-monocyte progenitors (GMP),
Basophils, 1 cells, and 2 cells had higher immune inltration levels in C1 than that in C2. Furthermore,
we studied the ratios of immune cell inltration between C1 and C2 and created a stacked bar graph (Fig.2D)
illustrating that the proportion of T cell subpopulations in C2 was lower than that in C1.
To further compare the prognostic dierences between C1 and C2, we performed a KM survival curve
analysis on C1 and C2 (Fig.3A). ere were 953 samples with survival information for C1 and 241 for C2. e
results showed that the prognosis of C2 was worse than that of C1. Next, we explored the dierences in the
expression of immune checkpoints PD1 (PDCD1) and PDL1 (CD274) between C1 and C2 (Fig.3B) and found
that the expression of PD1 (PDCD1) and PDL1 (CD274) in C2 was higher than that in C1. Cancer cells evade
immune surveillance when immune checkpoint molecules are overexpressed49. Subsequently, we studied the
Pearson’s correlation of the signicantly dierent immune cell types between C1 and C2 within the two clusters.
Figure3C illustrates the Pearson correlation among the immune cell types within C1. e results showed that
most immune cell types were positively correlated, especially CD8 + Tcm and CD4 + memory T cells (COR
= 0.724, P = 9.78E–159). e Pearson correlation among immune cell types within C2 is presented (Fig.3D), and
the results showed that most immune cell types also had a positive correlation, especially between the CD4 and
CD8 subpopulations, and the correlation was particularly high among B cell subpopulations, such as naïve B
cells and memory B cells (cor = 0.899, P = 1.95E–86). A small proportion of cell types were negatively correlated,
such as Megakaryocytes and 1 cells (cor = −0.575, P = 5.42E−23).
Scientic Reports | (2025) 15:12569 4
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
The results of weighted gene co-expression network analysis
To reveal the biological distinctions between C1 and C2 at the transcriptome level, we performed PCA using
consensus-clustering subgroups as labels. e results showed that there was a certain degree of dierence between
samples C1 and C2, and dierential expression analysis was performed (Fig.4A). We performed dierential gene
expression analysis between C1 and C2, and 5840 genes were identied as DEGs. 2516 up-regulated genes were
obtained, and 3324 down-regulated genes were identied. We generated a volcano plot of DEGs, which showed
Fig. 1. Flowchart of this study. TCGA-BRCA, the cancer genome atlas breast cancer; DEG, dierentially
expressed genes; WGCNA, Weighted gene co-expression network analysis.
Scientic Reports | (2025) 15:12569 5
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Fig. 2. Consensus clustering analysis of the TCGA breast cancer samples. (A) e CDF curve employed to
identify the optimal clustering method. (B) Cluster patients with breast cancer in the TCGA breast cancer set.
(C) Dierences of Immune cell composition in groups with C1 and C2. (D) Proportions of Immune cells in C1
and C2.
Scientic Reports | (2025) 15:12569 6
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
that the number of upregulated genes was lower than that of the downregulated genes (Fig.4B). A heatmap of
DEGs is also presented, indicating that the number of upregulated genes in C1 was slightly lower than that of the
downregulated genes (Fig.4C). We explored the dierential expression of genes closely related to breast cancer
between C1 and C2 (Fig.4D). ATM expression was lower in C1 than in C2, whereas BARD1, BRCA1, BRCA2,
CDH1, CHEK2, and RAD51D were expressed at higher levels in C1 than in C2.
We conducted WGCNA based on the DEGs between C1 and C2 to identify gene modules associated with
consensus clustering subgroups for immune inltration analysis. Using soPower 3, we constructed a scale-free
network (scale-free R2 = 0.9) (Fig.5A). Modules were generated and merged using a mergeCutHeight of 0.25
(Fig.5B). We obtained 8 gene modules (turquoise, brown, green, blue, red, black, yellow, gray). According to the
correlation analysis between the gene modules and traits, the turquoise module was associated with consensus-
clustering subgroups based on immune inltration (cor = 0.61, P = 1E-124) (Fig.5C). e scatter plot shows that
the correlation between module membership in the turquoise module and gene signicance for the cluster was
cor = 0.8, P < 1E-200 (Fig.5D). is observation suggests that genes with substantial relevance to a trait frequently
represent pivotal components within the modules that are intricately linked to that trait. Subsequently, GO and
KEGG enrichment analyses were conducted on a cohort of 1752 genes encompassing the turquoise module. e
ndings revealed that GO analysis was mainly enriched in functions such as ameboidal-type cell migration and
vascular process in circulatory system (Fig.5E). KEGG analysis was mainly enriched in pathways such as the
AMPK signaling pathway and cAMP signaling pathway (Fig.5F).
Fig. 3. Prognostic and immunological landscape of breast cancer identied through consensus clustering
analysis. (A) e prognosis between C1 and C2. (B) Expression dierences of PD1 (PDCD1) and PDL1
(CD274) in C1 and C2. (C) Investigation of interactions among immune Cells in C1. (D) Investigation of
interactions among immune Cells in C2.
Scientic Reports | (2025) 15:12569 7
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Construction and validation of a prognosis model
A prognostic model was constructed based on the identied module genes. Univariate Cox regression analysis
was performed to identify genes associated with prognosis, resulting in the identication of 925 genes. We
extended our study by conducting multivariate Cox regression analysis to identify independent prognostic
genes, obtaining 795 independent prognostic genes. Subsequently, we employed Lasso Cox regression analysis
to identify independent prognostic genes (Fig.6A, B), retaining 12 genes with non-zero coecients (EMP1,
RP11-100L22.1, ABCB5, XG, ADH4, RP5-1102E8.3, KB-1448 A5.1, LINC00890, NLGN1, LINC01082, RP1-66
C13.4, FAM71 A). Based on previous studies50, we constructed a prognostic model using these 12 genes. Using
the following formula, we calculated the risk scores:
risk score =0.034 ∗exp (EMP1)−0.214 ∗exp (RP11 −100L22.1)+0.156 ∗exp (ABCB5)
+0.205∗exp (XG)+0.007∗exp (ADH4)+0.483 ∗exp (RP5 −1102E8.3)
+0.103∗exp (KB −1448A5.1)+0.568 ∗exp (LINC00890)+0.093 ∗exp (NLGN1)
+0
.
515 ∗exp (LINC01082)+1
.
117 ∗exp (RP1 −66C13
.
4)+1
.
153 ∗exp (FAM71A).
We divided the samples into high- and low-risk groups using the median risk score as the threshold. We used
TCGA breast cancer dataset as a training set and conducted KM survival analysis (Fig.6C), nding that the
prognosis of the high-risk group was worse than that of the low-risk group. To further verify the stability
and accuracy of the prognostic model, we performed KM survival analysis on the datasets (GSE58812 and
GSE23428) as validation sets. e results obtained from the validation set were consistent with those observed
in the training set (Fig.6D, E).
Our study examined the dierences in immune checkpoint expression between the high- and low-risk
groups to determine whether immunotherapy was correlated with the high- and low-risk groups (Fig.6F). e
results showed that CD274, CD47, HAVCR2, IDO1, SIRPA, CTLA4 were signicantly highly expressed in the
Fig. 4. Dierentially expressed genes (DEGs) between C1 and C2. (A) PCA (3D) of C1 (red color) and C2
(blue color). (B) e volcano map obtained from the dierence analysis. (C) e heat map obtained from the
dierence analysis. (D) Dierential expression of breast cancer-related genes between C1 and C2.
Scientic Reports | (2025) 15:12569 8
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
high-risk group. TIDE scores were calculated for each sample, followed by a dierential analysis between the
high- and low-risk groups (Fig.6G). e TIDE score was signicantly higher in the high-risk group than in the
low-risk group (P< 0.0001). A higher TIDE predictive score signies an increased likelihood of immune escape,
suggesting a reduced potential benet of immunotherapy51. Our study examined the correlation between TIDE
Fig. 5. Weighted gene co-expression network analysis and enrichment analysis of module genes. (A–D)
Weighted gene co-expression network analysis. e turquoise module was linked to consensus clustering
subgroups, from which 1752 genes were identied. (E) GO enrichment analysis results of turquoise module
genes. (F) KEGG34 enrichment analysis results of turquoise module genes.
Scientic Reports | (2025) 15:12569 9
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Fig. 6. Construction and validation of the prognostic model. (A, B) Determining the number of factors
using the LASSO algorithm. (C) KM survival analysis for high- and low-risk groups in training set. (D, E)
KM survival analysis for high- and low-risk groups in validation sets. (F) Dierential expression of immune
checkpoints between high- and low-risk groups. (G) e dierence in TIDE scores between high and low-
risk groups. (H) Scatter plot of the correlation between TIDE scores and risk scores. LASSO: Least Absolute
Shrinkage and Selection Operator; TIDE: Tumor Immune Dysfunction and Exclusion.
Scientic Reports | (2025) 15:12569 10
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
scores and risk scores (Fig.6H) and revealed a positive association, where higher risk scores corresponded to
increased TIDE scores (P < 0.001). Next, we used GSE76275 as an external dataset for validation. e results
indicated that PDCD1 expression was signicantly elevated in the high-risk group (Fig. S1A). e TIDE score
was higher in the high-risk group compared to the low-risk group; however, no signicant dierence was found
between the two groups (Fig. S1B). A higher TIDE score was linked to an increased risk score, but this correlation
was not statistically signicant (Fig. S1C).
Dierential analysis and enrichment analysis between high- and low-risk groups
We further analyzed the DEGs between the high- and low-risk groups. We identied 2064 DEGs with upregulated
and 414 downregulated DEGs (Fig.7A, B, Supplementary Table S3). GO and KEGG enrichment analyses were
conducted for the DEGs (Supplementary Table S4). GO enrichment analysis indicated that up-regulated DEGs
were primarily associated in “response to peptide hormone” (GO:0043434) and “vascular process in circulatory
system” (GO:0003018) categories (Fig.7C). In contrast, down-regulated DEGs enriched in “signal release”
(GO:0023061) and “regulation of neuronal synaptic plasticity” (GO:0048168) categories (Fig.7D). Additionally,
KEGG pathway enrichment analysis revealed that the up-regulated DEGs were enriched in pathways related to
“Viral protein interaction with cytokine and cytokine receptor” and “Tyrosine metabolism” (Fig.7E), while the
down-regulated DEGs were found to be enriched in pathways such as “Neuroactive ligand-receptor interaction”
and “Metabolism of xenobiotics by cytochrome P450” (Fig.7F).
Mutation analysis of prognostic model genes
We investigated the mutation status of 12 prognostic model genes (EMP1, RP11-100L22.1, ABCB5, XG, ADH4,
RP5-1102E8.3, KB-1448 A5.1, LINC00890, NLGN1, LINC01082, RP1-66 C13.4, FAM71 A) in high- and low-risk
groups. e principal type of mutation identied among the 12 prognostic model genes within the high-risk
group was a missense mutation (Fig.8A). Missense mutation was the predominant mutation type in the 12
prognostic model genes in the low-risk group (Fig.8B). We calculated the dierences in TMB between the
high- and low-risk groups. Following this, we employed the optimal cuto point as a threshold to categorize
the TCGA-Breast cancer samples into high- and low-TMB groups. In our KM survival analysis, patients in the
high TMB group had a worse prognosis than those in the low TMB group (Fig.8C). We integrated the high- and
low-TMB groups with the high- and low-risk groups to delve deeper into Kaplan-Meier survival analysis. e
ndings revealed that the group characterized by low-TMB and low-risk exhibited the most favorable prognosis
(Fig.8D).
Single-cell clustering and single-cell annotation
To elucidate the cellular heterogeneity and unravel the biological transformations associated with breast cancer,
we conducted an in-depth analysis of single-cell data obtained from breast cancer samples. Following quality
control and ltering, we obtained single-cell transcriptome data from 11,213 cells (Fig.9A). Seven cell types were
clustered and identied using the Louvain algorithm and visualized using UMAP (Fig.9B). e marker genes for
the cancer cells were KRT19, KRT8, KRT18, ERBB2, ESR1; e marker genes for the cancer stem cells included
KRT19, KRT8, KRT18, MKI67; e marker genes for the Macrophages (Macro) were MSR1, CD163; e marker
genes for the Plasma Cells (Plasma) were MZB1, IGHG3, JCHAIN; e marker genes for the Endothelial Cells
(EC) were VWF, PECAM1; e marker genes for the Pericytes were MCAM, MYH11, SUSD2, TAGLN, ACTA2;
e marker genes for the Myobroblasts (Myob) were DCN, LUM, COL1 A2 (Fig.9C)22. Subsequently, we
identied the DEGs among the cell types and visualized them in a heatmap (Fig.9D).
Expression of prognostic model genes in cell types of breast cancer
We examined the expression of the 12 prognostic model genes in various cell types. Among the seven cell
types investigated, only four genes were expressed, with EMP1 exhibiting the highest level of expression in ECs
(Fig.10A). We calculated the scores of prognostic model genes for each cell type. e results indicated that
the EC type achieved the highest score, indicating a strong association with prognosis and high-risk cell type
(Fig.10B). We present scatter plots of the expression of marker genes (VWF, PECAM1) in the dierent cell
types (Fig.10C, D). Subsequently, GSEA analysis was conducted on the DEGs identied in EC and other cell
types, and the analysis revealed that the DEGs were signicantly enriched in hallmark epithelial mesenchymal
transition and hallmark estrogen response late (Fig.10E).
The results of deconvolution analysis of the proportions of various cell types in TCGA breast
cancer samples
Using breast cancer single-cell RNA sequencing data as a reference, the proportion of cell types within the breast
cancer samples were determined. e results indicated that EC had the highest proportion among all samples
(Fig.11A). As a high-risk cell type, EC was used to categorize samples into high-proportion and low-proportion
EC groups based on the median EC proportion. KM survival curve analysis was subsequently conducted to
evaluate the dierences in survival rates between groups with high- and low-proportion of EC (Fig.11B), nding
that the high-proportion EC group had worse prognosis than the low-proportion EC group. Additionally, we
discovered a positive correlation between the proportion of EC and risk score (Fig.11C). By combining high-
and low-proportion EC groups with high- and low-risk groups, we conducted a deeper KM survival analysis,
revealing that a high-proportion of EC and high-risk group had the worst prognoses (Fig.11D).
Drug sensitivity of various cell types
Intratumoral heterogeneity is associated with aggressive tumor behavior, therapy resistance, and unfavorable
patient outcomes52. Identifying drugs that are sensitive to various cell types within tumors plays a vital role in
Scientic Reports | (2025) 15:12569 11
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Fig. 7. Enrichment analysis of dierentially expressed genes (DEGs) between high- and low-risk groups.
(A) e volcano map obtained from the dierence analysis. (B) e heat map obtained from the dierence
analysis. (C, D) GO enrichment analysis results of DEGs. (E) KEGG enrichment analysis results of up-
regulated DEGs. (F) KEGG enrichment analysis results of down-regulated DEGs.
Scientic Reports | (2025) 15:12569 12
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
cancer treatment. erefore, we identied the dierences in sensitivity of various cell types to commonly used
targeted therapeutic agents for breast cancer treatment (Fig.12A–F). Our ndings revealed that, compared to
other cell types, cancer stem cells exhibited higher sensitivity to Lapatinib, Neratinib, and Olaparib. Compared
with other cell types, our results indicated lower sensitivity to PRIMA-1-Met, temsirolimus, and Tipifarnib-p2
in EC.
Discussion
Breast cancer is a heterogeneous and intricate disease with a pronounced incidence and high mortality among
females globally53. e global incidence of breast cancer is 24.2%, with a notably high prevalence among women
in developed regions54. Immunotherapy can eciently mobilize autoimmune functions, reinstate cellular
immune responses, and exert an immune-mediated eradication eect on tumors. Following the success of
immunotherapy, breast cancer, which was formerly perceived as having low immunogenicity, has progressed
to the realm of immunotherapy. Personalized immunotherapy is currently receiving increasing attention and
is being developed in the eld of cancer therapy. Previous studies have shown that the TME in breast cancer
encompasses immune inltrates that play signicant roles in oncogenesis, disease progression, and response
to immunotherapy55. To date, predictors of immunotherapy response in breast cancer include factors such as
PD-L1 status, TMB, and TILs; however, none of these variables have accumulated sucient evidence to serve as
stratication factors56. erefore, there is an urgent need to identify novel biomarkers or molecular models to
predict immunotherapy responses in patients with breast cancer. As new studies emerge, additional risk factors
and prognostic markers are constantly being discovered, highlighting the potential drawbacks and limitations
of the existing breast cancer prognosis models. Furthermore, as precision medicine rapidly evolves, there is
an increasing need for more accurate prognostic models to guide clinical decisions and create personalized
healthcare plans for patients.
Based on the immune inltration analysis, we performed a consensus clustering analysis, resulting in the
classication of the samples into two subgroups, denoted as C1 and C2. e ndings from the KM survival
analysis demonstrated that patients in C2 exhibited a markedly decreased OS rate compared to those in C1.
e expression levels of PD1 (PDCD1) and PDL1 (CD274) in C2 were higher than those in C1. e PD-1/
PD-L1 immune checkpoint serves a crucial function in facilitating tumor immune evasion and contributes to
the development of the tumor microenvironment57. Consequently, the PD-1/PD-L1 blockade has emerged as
a landmark in immunotherapy. Considering the signicant biological dierences between C1 and C2, further
studies are warranted. We performed WGCNA using the DEGs identied between C1 and C2. Additionally,
Fig. 8. Gene mutation analysis. (A) e independent prognostic gene variants in high-risk group. (B) e
independent prognostic gene variants in low-risk group. (C, D) Correlation analysis between tumor mutation
burden and prognosis.
Scientic Reports | (2025) 15:12569 13
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
we conducted enrichment analyses of 1752 genes within the turquoise module. e ndings indicated that
GO analysis revealed enrichment primarily in functions related to ameboidal-type cell migration and vascular
process in the circulatory system; KEGG analysis demonstrated enrichment primarily in pathways such as
AMPK signaling pathway and cAMP signaling pathway. Despite the association between AMPK and tumor-
suppressive functions, mounting evidence suggests that AMPK activation may exhibit protumorigenic eects in
a context-dependent manner58,59. Numerous studies have shown that cAMP can enhance the survival of tumor
cells through diverse mechanisms60. Further investigation is required to examine the underlying mechanisms.
Cox and Lasso regressions were utilized to develop a novel prognostic model based on the module genes.
We have collectively screened out 12 independent prognostic genes, namely EMP1, RP11-100L22.1, ABCB5, XG,
ADH4, RP5-1102E8.3, KB-1448 A5.1, LINC00890, NLGN1, LINC01082, RP1-66 C13.4, and FAM71 A. EMP1 has
been identied as a biomarker linked to getinib resistance in lung cancer and is associated with prednisolone
resistance in patients with ALL61. ABCB5 enhances the eux of drug molecules from cancer cells, thereby
promoting resistance62. Based on a previous study, the levels of ADH4 mRNA and protein expression were
notably decreased in hepatocellular carcinoma tissues and demonstrated a signicant correlation with survival
outcomes63. A previous study showed the upregulation of NLGN1 in colorectal cancer, suggesting that increased
NLGN1 expression could potentially function as an independent predictor of adverse outcomes in individuals
with colorectal cancer64. LINC01082 may suppress the incidence and progression of colon cancer by inhibiting
cell proliferation, migration, and invasion65. ere have been a few studies related to other independent
prognostic genes, and these genes are good candidates for further research. We constructed a novel prognostic
model utilizing 12 independent prognostic genes that demonstrated good performance in both training and
validation sets. Next, we explored the correlation between the high- and low-risk groups and immunotherapy.
TIDE algorithms are commonly used to predict responses to cancer immunotherapy, where higher TIDE
scores indicate worse outcomes from immunotherapy39. In TCGA breast cancer data analysis, the TIDE scores
were notably elevated in the high-risk group, suggesting a reduced ecacy of immunotherapy in breast cancer
patients classied in the high-risk group. Furthermore, we used GSE76275 as an external dataset for further
validation. e GSE76275 dataset originates from a study conducted by Burstein et al.66 that focused on patients
with TNBC. e GSE76275 dataset comprises 265 breast cancer samples, of which 198 are TNBC tumors, while
Fig. 9. Annotation of cell types using single-cell sequencing data and identication of DEGs. (A) Quality
control chart of GSE158399 single-cell data set. (B) Clustering of seven cell types using the Louvain algorithm
and visualization with UMAP. (C) e expression of marker genes in various cell types. (D) e expression of
DEGs in various cell types. UMAP: Uniform Manifold Approximation and Projection.
Scientic Reports | (2025) 15:12569 14
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
the remaining 67 samples are not triple-negative tumors. e results revealed a trend consistent with the TCGA
breast cancer dataset, but there was no statistical signicance. e reasons for these results are presented below.
First, e TCGA breast cancer dataset has a larger sample size, which enhanced the statistical power to detect
signicant dierences. In contrast, the GSE76275 dataset has a small sample size, which may lead to a lack of
statistical signicance even when the trends were similar. Second, the majority of samples in the GSE76275
dataset are TNBC. TNBC exhibits signicant dierences in biological characteristics and TME compared to
other types of breast cancer, which may inuence the TIDE score. erefore, we need more external datasets and
larger samples for further validation in the future. Our model could eectively predict the prognosis of patients
with breast cancer and their responses to immunotherapy.
We performed enrichment analyses of the DEGs between the high- and low-risk groups. e GO enrichment
analysis revealed that the up-regulated DEGs were predominantly enriched in the “response to peptide hormone”
(GO:0043434) and “vascular process in circulatory system” (GO:0003018) categories. KEGG enrichment
analysis revealed that the upregulated DEGs were signicantly enriched in immune-related and metabolic
pathways. Immune-related pathways were the most enriched pathways in the genes of tumor tissues, suggesting
their potential inuence on immune activity67. Tyrosine metabolism has surfaced as a pivotal dysfunctional
pathway in aggressive cancer cell lines, with documented interactions between cancer-related pathways and
tyrosine metabolism68,69. Considering that the TMB is a prognostic indicator of numerous cancer types, its
potential role in determining the prognostic status of breast cancer has been examined70. Breast cancer samples
from TCGA were classied into high- and low-TMB groups. Poorer prognosis was observed in the high-TMB
group. We combined the high- and low-TMB groups with the high- and low-risk groups to conduct KM survival
analysis. e outcomes indicated that the low-TMB and low-risk group exhibited the most favorable prognoses.
erefore, these two classication systems can enhance the predictive utility as prognostic signatures.
Fig. 10. High-risk cell type in single-cell sequencing data. (A) Expression levels of independent prognostic
genes in various cell types of breast cancer. (B) Scores of independent prognostic genes across dierent cell
types. (C, D) e expression of marker genes of high-risk cell types in various cell types. (E) e GSEA
enrichment analysis results of DEGs between high-risk cell type and other cell types.
Scientic Reports | (2025) 15:12569 15
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
scRNA-seq data were used to enhance the comprehensive understanding of gene expression among a variety
of cell types. Clustering was conducted utilizing the Louvain algorithm, subsequently applying UMAP for data
visualization. Seven distinct cell types were identied: cancer cells, cancer stem cells, macrophages, plasma cells,
endothelial cells, pericytes, and myobroblasts. A previous study suggested that B and plasma cells play pivotal
roles in the ecacy of PD-L1 blockade in non-small cell lung cancer71. Angiogenesis and tumor progression
require communication between the tumor and endothelial cells72. Endothelial cells have the capability to
secrete a variety of cytokines that hold signicant roles in driving tumor progression and metastasis73. Pericytes
interact multifariously with various components of the TME, such as promoting cancer cell proliferation and
drug resistance via paracrine actions and inducing M2 macrophage polarization74. Myobroblasts play an active
role in fostering the growth and spread of neoplastic epithelial cells75. e expression of 12 genes related to
prognosis was investigated in the seven cell types. We found that the EC type was closely related to breast cancer
prognosis, representing a high-risk cell type. GSEA was performed on DEGs identied in EC and other cell
types. e outcome revealed a signicant enrichment of hallmark epithelial mesenchymal transition among the
DEGs. Epithelial-mesenchymal transition plays a crucial role in cancer progression and metastasis of cancer76.
In this study, we used the MuSiC method to perform deconvolution analysis on TCGA breast cancer data
and observed a high proportion of ECs. Possible reasons for this include: there may be more vascular-related
components present in TCGA breast cancer samples; ECs in the TME may play an important role in tumor
progression77,78. erefore, this result may reect, to some extent, the characteristics of the TME specic to
breast cancer. It is noteworthy that the two panels in Fig.11A exhibit similar proportions of cell types. is may
be due to the fact that both the high- and low-risk groups were derived from the TCGA breast cancer samples,
sharing similar pathological states. In addition, we divided TCGA breast cancer samples into high-proportion
EC group and low-proportion EC group. e high-proportion EC group had worse prognosis than the low-
proportion EC group. We integrated the high- and low-proportion EC groups with high- and low-risk groups
for a detailed exploration of the KM survival analysis, which revealed that the high-proportion EC and high-
risk group exhibited the poorest prognosis. ese two classication systems complement each other in terms of
prognostic signatures.
Fig. 11. Deconvolution analysis to assess the composition of dierent cell types in TCGA breast cancer
samples. (A) e proportions of each cell type in all samples of the high- and low-risk groups. (B) e eect of
the proportion of EC in breast cancer samples on prognosis. (C) e correlation between proportion of EC and
risk score. (D) Combining high- and low-proportion EC groups with high and low-risk groups to conduct the
KM survival analysis.
Scientic Reports | (2025) 15:12569 16
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
ICIs have demonstrated ecacy in treating various advanced solid tumors, leading to their rapid adoption and
advancement in the eld of antitumor drug research79. Recent studies have demonstrated that certain subgroups
of metastatic breast cancers exhibit positive responses to anti-PD-1/PD-L1 agents80. However, patients with
liver or brain metastases exhibit a lower response rate to anti-PD-1/PD-L1 agents than those with metastases to
other locations81,82. erefore, the development of novel therapeutic approaches is imperative. We determined
variations in sensitivity to targeted therapeutic agents commonly used in breast cancer treatment among diverse
cell types. e combination of ICIs with DNA damage repair inhibitors, such as PARP inhibitors, has emerged
as a viable strategy for patients with breast cancer harboring BRCA mutations83. In the TNBC tumor model,
niraparib activated interferon signaling pathways, thereby enhancing the antitumor ecacy of the anti-PD-1
antibody84. erefore, the developed prognostic model can be used to assess the ecacy of immunotherapeutic
agents in patients with breast cancer. For patients who may not respond eciently to immunotherapy agents,
Fig. 12. Drug sensitivity analysis of targeted therapy agents for breast cancer in dierent cell types. (A)
Lapatinib. (B) Neratinib. (C) Olaparib. (D) PRIMA-1-Met. (E) Temsirolimus. (F) Tipifarnib-p2.
Scientic Reports | (2025) 15:12569 17
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
a combination of immunotherapy agents with targeted therapy agents identied through our drug sensitivity
analysis could improve ecacy.
Although the ndings of the present study are valuable, there are certain limitations. Firstly, while our
prognostic model was rigorously veried across diverse datasets, further corroboration using real-world data
and more external datasets is required. Additionally, thorough functional experiments are imperative to clarify
the specic mechanisms associated with the implicated genes.
Conclusion
In conclusion, our study examined the correlation between breast cancer prognosis and dynamic alterations
in TME. A prognostic model was developed to predict survival and response to immunotherapy using the 12
independent prognostic genes identied in this study. ese independent prognostic genes could serve as guides
for precision breast cancer therapy. Our study found that the proportion of ECs was associated with prognosis
in patients with breast cancer. ere are new opportunities for developing targeted therapies by focusing on
EC metabolism. Based on our prognostic model and drug sensitivity analysis, immunotherapy combined with
targeted therapy might be considered to enhance ecacy in patients with breast cancer who might have poor
responses to immunotherapy. is study deepens the understanding of breast cancer and lays the foundation for
individualized treatment strategies.
Data availability
e corresponding author will make the raw data supporting the conclusions of this article available without any
unnecessary restrictions.
Received: 31 October 2024; Accepted: 7 April 2025
References
1. Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2019. CA Cancer J. Clin. 69, 7–34. https://doi.org/10.3322/caac.21551
(2019).
2. Chen, W. et al. Cancer statistics in China, 2015. CA Cancer J. Clin. 66, 115–132. https://doi.org/10.3322/caac.21338 (2016).
3. Harbeck, N. et al. Breast cancer. Nat. Rev. Dis. Prim. 5, 66. https://doi.org/10.1038/s41572-019-0111-2 (2019).
4. Gradishar, W. J. et al. Breast cancer version 2.2015. J. Natl. Compr. Cancer Netw. 13, 448–475. h t t p s : / / d o i . o r g / 1 0 . 6 0 0 4 / j n c c n . 2 0 1 5 . 0
0 6 0 (2015).
5. Slamon, D. J. et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses
HER2. N Engl. J. Med. 344, 783–792. https://doi.org/10.1056/NEJM200103153441101 (2001).
6. Wang, X., Wang, F., Zhong, M., Yarden, Y. & Fu, L. e biomarkers of hyperprogressive disease in PD-1/PD-L1 blockage therapy.
Mol. Cancer 19, 81. https://doi.org/10.1186/s12943-020-01200-x (2020).
7. Marei, H. E., Hasan, A., Pozzoli, G. & Cenciarelli, C. Cancer immunotherapy with immune checkpoint inhibitors (ICIs): Potential,
mechanisms of resistance, and strategies for reinvigorating T cell responsiveness when resistance is acquired. Cancer Cell. Int. 23,
64. https://doi.org/10.1186/s12935-023-02902-0 (2023).
8. Lv, B. et al. Immunotherapy: Reshape the tumor immune microenvironment. Front. Immunol. 13, 844142. h t t p s : / / d o i . o r g / 1 0 . 3 3 8 9
/ m m u . 2 0 2 2 . 8 4 4 1 4 2 (2022).
9. Dong, M. et al. Perfusion air culture of precision-cut tumor slices: An ex vivo system to evaluate individual drug response under
controlled culture conditions. Cells https://doi.org/10.3390/cells12050807 (2023).
10. Villacampa, G. et al. Addition of immune checkpoint inhibitors to chemotherapy versus chemotherapy alone in rst-line metastatic
triple-negative breast cancer: A systematic review and meta-analysis. Cancer Treat. Rev. 104, 102352. h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . c t r v .
2 0 2 2 . 1 0 2 3 5 2 (2022).
11. Wang, J., Song, J., Liu, Z., Zhang, T. & Liu, Y. High tumor mutation burden indicates better prognosis in colorectal cancer patients
with KRAS mutations. Front. Oncol. 12, 1015308. https://doi.org/10.3389/fonc.2022.1015308 (2022).
12. Tang, F. et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nat. Methods 6, 377–382. h t t p s : / / d o i . o r g / 1 0 . 1 0 3 8 / n m e t h . 1
3 1 5 (2009).
13. Olsen, T. K. & Baryawno, N. Introduction to single-cell RNA sequencing. Curr. Protoc. Mol. Biol. 122, e57. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 2
/ c p m b . 5 7 (2018).
14. Bassez, A. et al. A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer. Nat. Med. 27,
820–832. https://doi.org/10.1038/s41591-021-01323-8 (2021).
15. Gong, Z., Li, Q., Li, J., Xie, J. & Wang, W. A novel signature based on autophagy-related LncRNA for prognostic prediction and
candidate drugs for lung adenocarcinoma. Transl. Cancer Res. 11, 14–28. https://doi.org/10.21037/tcr-21-1554 (2022).
16. Xu, R. et al. Single-cell sequencing analysis based on public databases for constructing a metastasis-related prognostic model for
gastric cancer. Appl. Bionics Biomech. 2022, 7061263. https://doi.org/10.1155/2022/7061263 (2022).
17. Pei, S. et al. Integrating single-cell RNA-seq and bulk RNA-seq to construct prognostic signatures to explore the role of glutamine
metabolism in breast cancer. Front. Endocrinol. (Lausanne) 14, 1135297. https://doi.org/10.3389/fendo.2023.1135297 (2023).
18. Haque, A. et al. CD206(+) tumor-associated macrophages promote proliferation and invasion in oral squamous cell carcinoma via
EGF production. Sci. Rep. 9, 14611. https://doi.org/10.1038/s41598-019-51149-1 (2019).
19. Maishi, N. & Hida, K. Tumor endothelial cells accelerate tumor metastasis. Cancer Sci. 108, 1921–1926. h t t p s : / / d o i . o r g / 1 0 . 1 1 1 1 / c
a s . 1 3 3 3 6 (2017).
20. Goldman, M. J. et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat. Biotechnol. 38, 675–678.
https://doi.org/10.1038/s41587-020-0546-8 (2020).
21. Barrett, T. et al. NCBI GEO: Archive for functional genomics data sets–update. Nucleic Acids Res. 41, D991–995. h t t p s : / / d o i . o r g / 1
0 . 1 0 9 3 / n a r / g k s 1 1 9 3 (2013).
22. Xu, K. et al. Integrative analyses of scRNA-seq and scATAC-seq reveal CXCL14 as a key regulator of lymph node metastasis in
breast cancer. Hum. Mol. Genet. 30, 370–380. https://doi.org/10.1093/hmg/ddab042 (2021).
23. Strand, S. H. et al. Molecular classication and biomarkers of clinical outcome in breast ductal carcinoma in situ: Analysis of
TBCRC 038 and RAHBT cohorts. Cancer Cell 40, 1521–1536. https://doi.org/10.1016/j.ccell.2022.10.021 (2022).
24. Onkar, S. et al. Immune landscape in invasive ductal and lobular breast cancer reveals a divergent macrophage-driven
microenvironment. Nat. Cancer 4, 516–534. https://doi.org/10.1038/s43018-023-00527-w (2023).
25. Aran, D., Hu, Z. & Butte, A. J. xCell: Digitally portraying the tissue cellular heterogeneity landscape. Genome Biol. 18, 220. h t t p s : /
/ d o i . o r g / 1 0 . 1 1 8 6 / s 1 3 0 5 9 - 0 1 7 - 1 3 4 9 - 1 (2017).
Scientic Reports | (2025) 15:12569 18
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
26. Zeng, D. et al. Multi-omics immuno-oncology biological research to decode tumor microenvironment and signatures. Front.
Immunol. 12, 687975. https://doi.org/10.3389/mmu.2021.687975 (2021).
27. Lock, E. F. & Dunson, D. B. Bayesian consensus clustering. Bioinformatics 29, 2610–2616. h t t p s : / / d o i . o r g / 1 0 . 1 0 9 3 / b i o i n f o r m a t i c s /
b t t 4 2 5 (2013).
28. Feng, S. et al. Immunogenic cell death related risk model to delineate ferroptosis pathway and predict immunotherapy response of
patients with GBM. Front. Immunol. 13, 992855. https://doi.org/10.3389/mmu.2022.992855 (2022).
29. Wilkerson, M. D. & Hayes, D. N. ConsensusClusterPlus: A class discovery tool with condence assessments and item tracking.
Bioinformatics 26, 1572–1573. https://doi.org/10.1093/bioinformatics/btq170 (2010).
30. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome
Biol. 15, 550. https://doi.org/10.1186/s13059-014-0550-8 (2014).
31. Liu, R., Hu, Y., Liu, T. & Wang, Y. Proles of immune cell inltration and immune-related genes in the tumor microenvironment
of osteosarcoma cancer. BMC Cancer 21, 1345. https://doi.org/10.1186/s12885-021-09042-6 (2021).
32. Langfelder, P. & Horvath, S. WGCNA: An R package for weighted correlation network analysis. BMC Bioinform. 9, 559. h t t p s : / / d o
i . o r g / 1 0 . 1 1 8 6 / 1 4 7 1 - 2 1 0 5 - 9 - 5 5 9 (2008).
33. e Gene Ontology Consortium. e gene ontology resource: 20 years and still going strong. Nucleic Acids Res. 47, D330–D338.
https://doi.org/10.1093/nar/gky1055 (2019).
34. Kanehisa, M. & Goto, S. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27–30. h t t p s : / / d o i . o r g / 1 0 . 1 0 9 3 /
n a r / 2 8 . 1 . 2 7 (2000).
35. Yu, G., Wang, L. G., Han, Y. & He, Q. Y. ClusterProler: An R package for comparing biological themes among gene clusters.
OMICS 16, 284–287. https://doi.org/10.1089/omi.2011.0118 (2012).
36. Walter, W., Sanchez-Cabo, F. & Ricote, M. GOplot: An R package for visually combining expression data with functional analysis.
Bioinformatics 31, 2912–2914. https://doi.org/10.1093/bioinformatics/btv300 (2015).
37. Lo Russo, G. et al. PEOPLE (NTC03447678), a phase II trial to test pembrolizumab as rst-line treatment in patients with advanced
NSCLC with PD-L1 < 50%: A multiomics analysis. J. Immunother. Cancer. https://doi.org/10.1136/jitc-2023-006833 (2023).
38. Xiang, L. et al. A potential biomarker of combination of tumor mutation burden and copy number alteration for ecacy of
immunotherapy in KRAS-mutant advanced lung adenocarcinoma. Front. Oncol. 10, 559896. h t t p s : / / d o i . o r g / 1 0 . 3 3 8 9 / f o n c . 2 0 2 0 . 5 5
9 8 9 6 (2020).
39. Jiang, P. et al. Signatures of T cell dysfunction and exclusion predict cancer immunotherapy response. Nat. Med. 24, 1550–1558.
https://doi.org/10.1038/s41591-018-0136-1 (2018).
40. Mayakonda, A., Lin, D. C., Assenov, Y., Plass, C. & Koeer, H. P. Maools: Ecient and comprehensive analysis of somatic variants
in cancer. Genome Res. 28, 1747–1756. https://doi.org/10.1101/gr.239244.118 (2018).
41. Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587. https://doi.org/10.1016/j.cell.2021.04.048
(2021).
42. Xu, J. et al. Delving into the heterogeneity of dierent breast cancer subtypes and the prognostic models utilizing scRNA-Seq and
bulk RNA-Seq. Int. J. Mol. Sci. 23. https://doi.org/10.3390/ijms23179936 (2022).
43. Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression
proles. Proc. Natl. Acad. Sci. U. S. A. 102, 15545–15550. https://doi.org/10.1073/pnas.0506580102 (2005).
44. Liberzon, A. et al. e molecular signatures database (MSigDB) hallmark gene set collection. Cell. Syst. 1, 417–425. h t t p s : / / d o i . o r g
/ 1 0 . 1 0 1 6 / j . c e l s . 2 0 1 5 . 1 2 . 0 0 4 (2015).
45. Wang, X., Park, J., Susztak, K., Zhang, N. R. & Li, M. Bulk tissue cell type deconvolution with multi-subject single-cell expression
reference. Nat. Commun. 10, 380. https://doi.org/10.1038/s41467-018-08023-x (2019).
46. Gambardella, G. et al. A single-cell analysis of breast cancer cell lines to study tumour heterogeneity and drug response. Nat.
Commun. 13, 1714. https://doi.org/10.1038/s41467-022-29358-6 (2022).
47. Rees, M. G. et al. Correlating chemical sensitivity and basal gene expression reveals mechanism of action. Nat. Chem. Biol. 12,
109–116. https://doi.org/10.1038/nchembio.1986 (2016).
48. Durisova, M. & Dedik, L. SURVIVAL–an integrated soware package for survival curve estimation and statistical comparison of
survival rates of two groups of patients or experimental animals. Methods Find. Exp. Clin. Pharmacol. 15, 535–540 (1993).
49. Ohaegbulam, K. C., Assal, A., Lazar-Molnar, E., Yao, Y. & Zang, X. Human cancer immunotherapy with antibodies to the PD-1 and
PD-L1 pathway. Trends Mol. Med. 21, 24–33. https://doi.org/10.1016/j.molmed.2014.10.009 (2015).
50. Chen, Y., Li, Z. Y., Zhou, G. Q. & Sun, Y. An immune-related gene prognostic index for head and neck squamous cell carcinoma.
Clin. Cancer Res. 27, 330–341. https://doi.org/10.1158/1078-0432.CCR-20-2166 (2021).
51. Lu, J. et al. A novel prognostic model based on single-cell RNA sequencing data for hepatocellular carcinoma. Cancer Cell. Int. 22,
38. https://doi.org/10.1186/s12935-022-02469-2 (2022).
52. Davies, A. E. et al. Systems-level properties of EGFR-RAS-ERK signaling amplify local signals to generate dynamic gene expression
heterogeneity. Cell Syst. 11, 161–175. https://doi.org/10.1016/j.cels.2020.07.004 (2020).
53. Wang, X. et al. e circACTN4 interacts with FUBP1 to promote tumorigenesis and progression of breast cancer by regulating the
expression of proto-oncogene MYC. Mol. Cancer 20, 91. https://doi.org/10.1186/s12943-021-01383-x (2021).
54. Ferlay, J. et al. Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods. Int. J. Cancer 144,
1941–1953. https://doi.org/10.1002/ijc.31937 (2019).
55. Omilian, A. R. et al. Multiplexed digital spatial proling of invasive breast tumors from black and white women. Mol. Oncol. 16,
54–68. https://doi.org/10.1002/1878-0261.13017 (2022).
56. Rizzo, A. & Ricci, A. D. Biomarkers for breast cancer immunotherapy: PD-L1, TILs, and beyond. Expert Opin. Investig. Drugs 31,
549–555. https://doi.org/10.1080/13543784.2022.2008354 (2022).
57. Wang, Q. et al. e roles of MicroRNAs in regulating the expression of PD-1/PD-L1 immune checkpoint. Int. J. Mol. Sci. 18 h t t p s :
/ / d o i . o r g / 1 0 . 3 3 9 0 / i j m s 1 8 1 2 2 5 4 0 (2017).
58. Hardie, D. G. Molecular pathways: is AMPK a friend or a foe in cancer? Clin. Cancer Res. 21, 3836–3840. h t t p s : / / d o i . o r g / 1 0 . 1 1 5 8 /
1 0 7 8 - 0 4 3 2 . C C R - 1 4 - 3 3 0 0 (2015).
59. Kfoury, A. et al. AMPK promotes survival of c-Myc-positive melanoma cells by suppressing oxidative stress. EMBO J. 37 h t t p s : / / d
o i . o r g / 1 0 . 1 5 2 5 2 / e m b j . 2 0 1 7 9 7 6 7 3 (2018).
60. Wang, J., Tang, R., Lv, M., Zhang, J. & Shen, B. Selective unresponsiveness to the Inhibition of p38 MAPK activation by cAMP helps
L929 broblastoma cells escape TNF-alpha-induced cell death. Mol. Cancer 9, 6. https://doi.org/10.1186/1476-4598-9-6 (2010).
61. Wang, Y. W., Cheng, H. L., Ding, Y. R., Chou, L. H. & Chow, N. H. EMP1, EMP 2, and EMP3 as novel therapeutic targets in human
cancer. Biochim. Biophys. Acta Rev. Cancer 1868, 199–211. https://doi.org/10.1016/j.bbcan.2017.04.004 (2017).
62. Luo, Y. et al. Side population cells from human melanoma tumors reveal diverse mechanisms for chemoresistance. J. Invest.
Dermatol. 132, 2440–2450. https://doi.org/10.1038/jid.2012.161 (2012).
63. Wei, R. R. et al. Identication of ADH4 as a novel and potential prognostic marker in hepatocellular carcinoma. Med. Oncol. 29,
2737–2743. https://doi.org/10.1007/s12032-011-0126-3 (2012).
64. Yu, Q. et al. Upregulated NLGN1 predicts poor survival in colorectal cancer. BMC Cancer 21, 884. h t t p s : / / d o i . o r g / 1 0 . 1 1 8 6 / s 1 2 8 8
5 - 0 2 1 - 0 8 6 2 1 - x (2021).
65. Xiong, W. et al. Overexpression LINC01082 suppresses the proliferation, migration and invasion of colon cancer. Mol. Cell.
Biochem. 462, 33–40. https://doi.org/10.1007/s11010-019-03607-7 (2019).
Scientic Reports | (2025) 15:12569 19
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
66. Burstein, M. D. et al. Comprehensive genomic analysis identies novel subtypes and targets of triple-negative breast cancer. Clin.
Cancer Res. 21, 1688–1698. https://doi.org/10.1158/1078-0432.CCR-14-0432 (2015).
67. Zhong, Z. et al. Identication of specic cervical cancer subtypes and prognostic gene sets in tumor and nontumor tissues based
on GSVA analysis. J. Oncol. 2022, 6951885. https://doi.org/10.1155/2022/6951885 (2022).
68. Chen, W. Y. et al. Analyzing the dierentially expressed genes and pathway cross-talk in aggressive breast cancer. J. Obstet.
Gynaecol. Res. 41, 132–140. https://doi.org/10.1111/jog.12495 (2015).
69. Bendinelli, B. et al. Prediagnostic circulating metabolites in female breast cancer cases with low and high mammographic breast
density. Sci. Rep. 11, 13025. https://doi.org/10.1038/s41598-021-92508-1 (2021).
70. Yin, W. et al. Development and validation of a tumor mutation burden-related immune prognostic model for lower-grade glioma.
Front. Oncol. 10, 1409. https://doi.org/10.3389/fonc.2020.01409 (2020).
71. Patil, N. S. et al. Intratumoral plasma cells predict outcomes to PD-L1 blockade in non-small cell lung cancer. Cancer Cell 40,
289–300. https://doi.org/10.1016/j.ccell.2022.02.002 (2022).
72. Liang, W. et al. Vasorin stimulates malignant progression and angiogenesis in glioma. Cancer Sci. 110, 2558–2572. h t t p s : / / d o i . o r g
/ 1 0 . 1 1 1 1 / c a s . 1 4 1 0 3 (2019).
73. Buess, M., Rajski, M., Vogel-Durrer, B. M., Herrmann, R. & Rochlitz, C. Tumor-endothelial interaction links the CD44(+)/CD24(-
) phenotype with poor prognosis in early-stage breast cancer. Neoplasia 11, 987–1002. https://doi.org/10.1593/neo.09670 (2009).
74. Jiang, Z. et al. Pericytes in the tumor microenvironment. Cancer Lett. 556, 216074. https://doi.org/10.1016/j.canlet.2023.216074
(2023).
75. Al-Rakan, M. A., Hendrayani, S. F. & Aboussekhra, A. CHEK2 represses breast stromal broblasts and their paracrine tumor-
promoting eects through suppressing SDF-1 and IL-6. BMC Cancer 16, 575. https://doi.org/10.1186/s12885-016-2614-5 (2016).
76. Zhang, L. & Min, W. Bioorthogonal chemical imaging of metabolic changes during epithelial-mesenchymal transition of cancer
cells by stimulated Raman scattering microscopy. J. Biomed. Opt. 22, 1–7. https://doi.org/10.1117/1.JBO.22.10.106010 (2017).
77. Hansen, S. et al. Vascular grading of angiogenesis: Prognostic signicance in breast cancer. Br. J. Cancer 82, 339–347. h t t p s : / / d o i . o
r g / 1 0 . 1 0 5 4 / b j o c . 1 9 9 9 . 0 9 2 4 (2000).
78. Harris, M. A. et al. Towards targeting the breast cancer immune microenvironment. Nat. Rev. Cancer 24, 554–577. h t t p s : / / d o i . o r g
/ 1 0 . 1 0 3 8 / s 4 1 5 6 8 - 0 2 4 - 0 0 7 1 4 - 6 (2024).
79. Jiang, T., Chen, X., Ren, X., Yang, J. M. & Cheng, Y. Emerging role of autophagy in anti-tumor immunity: Implications for the
modulation of immunotherapy resistance. Drug Resist. Updates 56, 100752. https://doi.org/10.1016/j.drup.2021.100752 (2021).
80. Vonderheide, R. H., Domchek, S. M. & Clark, A. S. Immunotherapy for breast cancer: What are we missing? Clin. Cancer Res. 23,
2640–2646. https://doi.org/10.1158/1078-0432.CCR-16-2569 (2017).
81. Schmid, P. et al. Atezolizumab and nab-paclitaxel in advanced triple-negative breast cancer. N. Engl. J. Med. 379, 2108–2121.
https://doi.org/10.1056/NEJMoa1809615 (2018).
82. Zou, Y. et al. e Single-Cell landscape of intratumoral heterogeneity and the immunosuppressive microenvironment in liver and
brain metastases of breast cancer. Adv. Sci. 10, e2203699. https://doi.org/10.1002/advs.202203699 (2023).
83. Zhu, L., Liu, J., Chen, J. & Zhou, Q. e developing landscape of combinatorial therapies of immune checkpoint blockade with
DNA damage repair inhibitors for the treatment of breast and ovarian cancers. J. Hematol. Oncol. 14, 206. h t t p s : / / d o i . o r g / 1 0 . 1 1 8 6
/ s 1 3 0 4 5 - 0 2 1 - 0 1 2 1 8 - 8 (2021).
84. Zhu, S. et al. Recent advances in targeted strategies for triple-negative breast cancer. J. Hematol. Oncol. 16, 100. h t t p s : / / d o i . o r g / 1 0 .
1 1 8 6 / s 1 3 0 4 5 - 0 2 3 - 0 1 4 9 7 - 3 (2023).
Acknowledgements
We appreciate the valuable data provided by the TCGA and GEO databases.
Author contributions
YT, YY, and JW designed and executed the study, conducted data analysis, and authored the manuscript. LH and
XCY assisted in data analysis and visualization. HZ revised the manuscript and participated in the discussion.
All authors reviewed the manuscript.
Funding
is study was carried out without funding.
Declarations
Competing interests
e authors declare no competing interests.
Ethics statement
is study, which solely utilized data obtained from public databases and did not involve any human
participants, did not necessitate ethical approval.
Additional information
Supplementary Information e online version contains supplementary material available at h t t p s : / / d o i . o r g / 1
0 . 1 0 3 8 / s 4 1 5 9 8 - 0 2 5 - 9 7 7 8 4 - 9 .
Correspondence and requests for materials should be addressed to J.W.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Scientic Reports | (2025) 15:12569 20
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
Open Access is article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives
4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in
any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide
a link to the Creative Commons licence, and indicate if you modied the licensed material. You do not have
permission under this licence to share adapted material derived from this article or parts of it. e images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence
and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to
obtain permission directly from the copyright holder. To view a copy of this licence, visit h t t p : / / c r e a t i v e c o m m o
n s . o r g / l i c e n s e s / b y - n c - n d / 4 . 0 / .
© e Author(s) 2025
Scientic Reports | (2025) 15:12569 21
| https://doi.org/10.1038/s41598-025-97784-9
www.nature.com/scientificreports/
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
