New insight into the risk stratification and treatment priority of breast cancer: TMT scoring system

Abstract

Background:

Breast cancer continues to pose a significant global health burden, accounting for a substantial proportion of cancer-related mortality in women worldwide. Despite advancements in therapeutic strategies, overall survival rates remain suboptimal, underscoring the critical need for innovative prognostic tools to improve patient outcomes.

Methods:

In this study, we introduced a novel tumor mitochondria transfer (TMT) scoring system based on 17 core genes to assess mitochondrial dynamics and their potential impact on the tumor microenvironment. Utilizing comprehensive datasets from TCGA-BRCA and GEO databases, we investigated the association between TMT scores and immune-metabolic features at both multi-cell and single-cell resolutions. Single-cell RNA sequencing (scRNA-seq) profiles were used to delineate the biological consequences of mitochondrial transfer. Additionally, co-culture experiments were conducted to validate our findings.

Results:

Our analysis revealed that elevated TMT scores are strongly associated with adverse clinical outcomes in patients with breast cancer. Tumors with high TMT scores exhibited pronounced hypoxia, immune suppression, and metabolic reprogramming. Specifically, these tumors demonstrated impaired T-cell functionality and enhanced mitochondrial transfer to tumor cells, suggesting a pivotal role for mitochondrial dynamics in promoting immune evasion and metabolic adaptations that drive tumor progression.

Conclusions:

The TMT score may represent a novel prognostic biomarker in breast cancer, highlighting the intricate relationship between mitochondrial transfer and tumor pathophysiology. Our findings suggested that targeting mitochondrial dynamics could be a potential therapeutic avenue for exploration, which might enhance breast cancer management strategies. Further exploration of the mitochondrial mechanisms in cancer biology may pave the way for more precise and effective therapeutic interventions

Introduction

We utilized an AI-powered language model (DeepSeek) to assist in polishing and refining the language and readability of this manuscript and has accomplished the TITAN CHECKLIST^[1]. Breast cancer represents a significant global health challenge, posing a substantial threat to women’s health, and ranking as the second leading cause of cancer-related mortality among women worldwide. According to GLOBOCAN 2022, approximately 2.3 million people were diagnosed with breast cancer, accounting for 11.6% of all new cancer cases and 15.4% of all cancer-related deaths in women^[2,3]. Despite remarkable advancements in therapeutic modalities, including surgery, radiotherapy, hormone therapy, and targeted therapies, the mortality rate of breast cancer remains persistently high^[4,5]. To improve treatment management and enhance survival outcomes, there is an urgent need to develop reliable prognostic tools capable of predicting treatment responses and clinical outcomes in breast cancer patients.

HIGHLIGHTS

• The TMT score predicts aggressive breast cancer through immune and metabolic profiling.

• Tumor cells steal mitochondria from T cells, enhancing their own growth while impairing immunity.

• High TMT scores correlate with reduced T-cell efficacy due to mitochondrial depletion.

• Mitochondrial transfer exhibits dual roles: promoting or suppressing tumor growth based on its source.

• The TMT score enables better breast cancer prognosis and guides personalized therapeutic strategies.

T cells play pivotal roles in immune surveillance, maintaining homeostasis, and preventing disease progression. These cells are primarily categorized into CD4+ and CD8+ subsets, which mediate direct cytotoxicity, immunomodulatory functions, and long-term immune memory^[6,7]. Although tumor-infiltrating lymphocytes (TILs) are recruited to tumor sites to combat malignancies, their anti-tumor efficacy is often compromised within the tumor microenvironment (TME), a phenomenon referred to as T cell exhaustion^[8,9]. The functional status and abundance of T cells exhibit significant heterogeneity across different cancer types, which may critically influence the clinical outcomes and therapeutic responses.

Mitochondria are ubiquitous organelles in mammalian cells and plays a central role in critical cellular processes such as energy production, calcium signaling, biosynthesis of essential molecules, and regulation of cell death^[10–12]. These organelles possess their own 16.6 kb circular genome (16.3 kb in mice), known as mitochondrial DNA (mtDNA), which exhibits a high mutation rate^[13,14]. For decades, mitochondria have been believed to function exclusively in the cytoplasm of individual cells. However, emerging evidence has revealed extensive transfer of mitochondrial components between cells, challenging this traditional view^[15]. This intercellular transfer occurs primarily through tunneling nanotubes (TNTs), as well as via extracellular vesicles and tight junctions^[16,17]. Recent studies have highlighted the role of nanotube-mediated communication between cancer and immune cells, demonstrating unidirectional mitochondrial transfer from T cells to adjacent cancer cells^[18]. Given that mitochondria are essential for providing the energy and metabolites required for T cell activation, this phenomenon may represent a novel mechanism underlying T cell dysfunction in the TME^[19]. These findings underscore the complex interplay between mitochondrial dynamics and immune cell function and offer new insights into the mechanisms of tumor immune evasion

In this study, we utilized the tumor mitochondrial transfer (TMT) scoring system, which is based on a previously reported set of 17 genes (ABI1, ARF6, F11R, BAIAP2L1, MYADM, ACTR3, ARHGEF5, SRC, CAPZA2, GOLPH3, RAB10, ARF4, CAP1, PSEN1, YWHAG, YWHAZ, and HSPA4), to predict mitochondrial transfer from T cells to tumor cells^[20]. Tumor heterogeneity is a pervasive phenomenon, existing not only across different cancer types but also strikingly within a single type^[21,22]. A prime example is breast cancer, where subtypes such as triple-negative, HER2-positive, and Luminal show vast differences in their pathogenesis and clinical management strategies^[23,24]. Given this, rather than pan-cancer validation studies, we seek to apply this scoring framework to reveal the functional and mechanistic implications of mitochondrial metastasis in a specific cancer context. Specifically, by applying to the TCGA-BRCA cohort, we found that high TMT scores were associated with profound immunosuppression and metabolic reprogramming in the TME. Furthermore, we evaluated the impact of mitochondrial transfer on tumor and T cells at the single-cell level. Experimental validation has confirmed that mitochondrial transfer from T cells to tumor cells is enhanced under hypoxic conditions, contributing to T cell exhaustion. These findings highlight the critical role of mitochondrial transfer in shaping the immunosuppressive and hypoxic TME, ultimately promoting tumor progression.

Results

A high TMT score is correlated with poor prognosis of breast cancer

The TMT gene set, comprising 17 genes, has been previously reported to predict mitochondrial transfer from T cells to tumor cells^[20]. Using log2TPM-normalized mRNA expression profiles of these genes from the TCGA-BRCA database, we compared their expression levels between normal and tumor tissues. The results revealed that 16 of the 17 genes exhibited significantly different expression patterns in tumor tissues, with TMT scores being markedly higher in tumors compared to normal tissues (Figure 1A). To further explore the functional relationships among these genes, we constructed a protein-protein interaction (PPI) network and mapped their chromosomal locations (Figure 1B and C). Mutation analysis of the 17 TMT genes in TCGA-BRCA samples indicated that amplification was the most common mutation type, with no gene exceeding a mutation frequency of 17%, highlighting the genetic stability of these genes (Figure 1D). Univariate Cox regression analysis was performed for each of the 17 genes and the TMT scores (Figure 1E). The hazard ratio (HR) for the TMT score was significantly higher than that for the individual genes, demonstrating its robust prognostic performance. We also investigated the correlation between TMT scores and clinical characteristics of the patients. Notably, the PAM50 classification showed a significant association with TMT scores, where as other clinical features, such as age, TNM stage, or pathological stage, did not exhibit such correlations (Figure 1F). Specifically, among the PAM50 subtypes, HER2-enriched breast cancers exhibited the highest TMT scores, followed by Luminal B and Basal-like subtypes. In contrast, the Normal-like subtype demonstrated the lowest TMT scores. This pattern suggests that the TMT score may reflect molecular and metabolic heterogeneity captured by the PAM50 classification, particularly highlighting distinct mitochondrial metabolic activity in HER2-driven and more aggressive tumor subtypes. Both univariate and multivariate regression analyses confirmed a significant association between TMT scores and poor prognosis (Figure 1G and H). Further validation in an independent GEO dataset supports the prognostic value of the TMT score (Supplemental Digital Content Figure 1, available at: http://links.lww.com/JS9/G567), suggesting its potential as a promising biomarker in breast cancer. These results consistently indicate that the TMT score is not only a marker that distinguishes tumor cells from normal tissues in terms of the important biological phenomenon of the microenvironment “hijack” mitochondria, but also a potential prognostic indicator independent of traditional clinical parameters. Its level directly reflects the extent to which tumors utilize this mechanism, thereby participating in and driving invasive progression.

Figure 1.

The prognostic significance of TMT scores in breast cancer. (A) Gene expression of the TMT genes and TMT score comparison between the normal and the cancer tissue; (B) PPI of the TMT protein; (C) Chromosome localization of the TMT genes; (D) Gene alterations of the TMT genes; (E0 COX regression of TMT score and TMT genes; (F) The relationship between TMT score and some clinical features; (G) Univariate and multivariate COX regression of TMT score and clinical features; H. Multivariate COX regression of TMT score and clinical features.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.001.

TMT score predicts different immune infiltration and metabolic reprogramming

To explore the biological significance of the mitochondrial transfer capacity, we stratified TCGA-BRCA samples into two subgroups based on the median TMT score as the threshold. Differential gene expression (DEG) analysis was performed using the “limma” package, with stringent thresholds (|log2FC| ≥ 0.8, P < 0.05). This analysis identified 161 significantly up-regulated genes and 103 down-regulated genes in TMT-high tumors compared to those in TMT-low tumors (Figure 2A). Pathway enrichment analysis using GO and KEGG was conducted separately for up-regulated and down-regulated genes. The up-regulated genes were predominantly enriched in pathways related to extracellular matrix (ECM) organization and intercellular communication, such as extracellular matrix organization and endoderm development (Figure 2B and D). The excessive deposition and hardening of the ECM not only provides a physical barrier to prevent immune cell infiltration, leading to immunosuppression, but also promotes the survival of cancer cells and chemotherapy resistance through signal transduction such as integrins^[25–27]. Active intercellular communication indicates that there is a complex “dialogue” between cancer cells and the surrounding fibroblasts and immune cells, jointly creating an ecological niche that supports tumor growth and metastasis^[28,29]. In contrast, down-regulated genes were primarily associated with innate and adaptive immunity, including activation and proliferation of myeloid cells and lymphocytes (Figure 2C and E). It is indicated that that this subtype or model of breast cancer adopts a strategy of actively suppressing local immune responses, that is, shaping an “immunosuppressive” or “cold” TME. This extensive down-regulation of immune-related gene expression may lead to defects in antigen presentation function, blocked T cell activation, and reduced immune cell infiltration, thereby providing a crucial “immune escape” mechanism for cancer cells, enabling them to evade immune surveillance and grow without limit^[30,31].

Figure 2.

Transcriptional landscape and functional enrichment of TMT high versus low group in breast cancer. (A) Differential gene analysis between TMT high and low group; (B and C) GO/KEGG enrichment analysis of up-regulated (B) and down-regulated (C) genes; (D and E) EMAP figure of GO/KEGG enrichment analysis of up-regulated (D) and down-regulated (E) genes.

TMT score is associated with immunosuppression

To investigate changes in immune characteristics, we conducted immunological analysis of the TCGA-BRCA dataset. The abundance of various immune cells within the tumor tissues was quantified, and differences in immune cell infiltration were compared between the TMT-high and TMT-low groups (Figure 3A). The results revealed that immune, stromal, and microenvironment scores were significantly reduced in the TMT-high group, indicating lower infiltration of TME cells. Correlation analysis further demonstrated a negative association between TMT scores and the infiltration of almost all immune cell types (Figure 3B)^[32]. Additionally, we observed that the expression of several key effector and regulatory molecules was significantly associated with TMT scores, highlighting distinct immune characteristics between the TMT-high and TMT-low groups (Figure 3C). Notably, co-stimulatory molecules, cytokines, and effector molecules were downregulated in the TMT-high group, contributing to an immunosuppressive TME. Subsequent analysis of pivotal immune processes confirmed a predominantly immunosuppressed TME in the TMT-high group (Figure 3D)^[33]. To evaluate the potential impact of mitochondrial transfer on immunotherapy responses, we applied the TIDE algorithm to compare TMT-high and TMT-low groups. The TMT-high group exhibited lower TIDE, exclusion, and MSI scores but higher dysfunction scores, suggesting poorer responses to immunotherapy (Figure 3E)^[34]. These findings suggest that the transfer of mitochondria from T cells to tumor cells may play a key role in tumor-mediated immunosuppression. It not only directly deprives T cells of the energy necessary for performing effector functions, leading to their exhaustion and dysfunction, but also may enhance the oxidative phosphorylation ability of tumor cells, thereby promoting their proliferation and survival. On the one hand, this metabolic theft mechanism weakens the strength and persistence of the anti-tumor immune response, and on the other hand, provides metabolic adaptability advantages for tumor cells. Eventually, they jointly lead to a decline in the reactivity of immunotherapy (such as checkpoint inhibitors), and the occurrence of clinical treatment resistance.

Figure 3.

Immune profiling of breast cancer subgroups Based on TMT score. (A) Differences of immune cells infiltration between TMT high and low group; (B) Heatmap of correlation of TMT and immune cells; (C) Heatmap of checkpoint expression; (D) Differences of immune-process between TMT high and low group; (E) TIDE scores between TMT high and low group.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.001.

Tumor with high TMT score is characterized with proliferative capability

To explore the metabolic characteristics of TCGA-BRCA samples stratified by TMT scores, we evaluated the expression of 84 metabolic pathway gene lists from the KEGG database. Our analysis revealed that the majority of metabolic pathways were significantly altered by the TMT status (Figure 4A). Specifically, key metabolic pathways, including the TCA cycle (Spearman r = 0.56, P < 0.0001), and synthesis pathways related to extracellular matrix (ECM) biogenesis, such as N-glycan biosynthesis (Spearman r = 0.56, P < 0.0001), exhibited strong associations (|Spearman r| > 0.5) with TMT scores (Figure 4B).To validate these metabolic features, we introduced an independent metabolomic dataset and confirmed significant correlations between TMT scores and the most associated metabolites (Figure 4C). Principal Component Analysis (PCA) demonstrated that TMT classifications closely corresponded to distinct metabolic profiles in tumors, and dendrogram plots further confirmed the heterogeneity (Figure 4D and E). Differential and enrichment analyses identified pathways linked to ECM synthesis and cell growth, particularly in TMT-high tumors (Figure 4F and G). These findings systematically confirmed that TMT classification can effectively identify a class of tumor subtypes with unique metabolic reprogramming characteristics. This type of tumor exhibits a significant synergistic activation of extracellular matrix (ECM) remodeling and proliferation-related signaling pathways, suggesting that it is in a highly invasive and rapidly proliferating malignant state. The reconstruction of ECM not only provides a structural basis for the migration and distant metastasis of tumor cells, but may also drive proliferation and survival signals through integrin signals and mechanical transduction pathways^[26,28]. Meanwhile, the continuous activation of proliferation signals implies that this type of tumor may be highly dependent on the cell cycle and anabolic processes, thereby providing potential targets for novel therapeutic strategies targeting the metabolic-ECM-proliferation axis.

Figure 4.

Comparative metabolomic profiling reveals distinct metabolic landscapes between TMT high and low group. (A) Heatmap of KEGG metabolic pathway; (B) Correlation dot plot of the pathway that |Spearman r| > 0.5; (C) Heatmap of metabolomics concerning TMT sores; (D) PCA plot of metabolomics between two groups; (E) Tree plot of metabolomics between two groups; (F) Volcano map of different metabolites between two groups; (G) Enrichment score of the differential metabolites.*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.001.

TMT score is associated with T cell depletion and oxidative respiration enhancement

To explore the impact of mitochondrial transfer on tumor cells and T cells, we investigated the TMT score at the single-cell level^[35]. After quality control procedures, a total of 281 302 cells were analyzed, and 19 subpopulations were identified through dimensionality reduction clustering. Based on classic annotation markers, we classified eight main subpopulations: epithelial cells, T cells, B cells, plasma cells, myeloid cells, endothelial cells, mesenchymal cells, and cycling cells (Figure 5A–D). We extracted malignant epithelial cells and performed pseudo-bulk cell analysis to assess TMT expression levels. The results revealed that TMT scores were negatively correlated with T cell infiltration, whereas the proportion of T cell subtypes remained unchanged across TMT scores (Figures 5E and F). Differential gene expression (DEG) analysis indicated that tumor cells with higher TMT scores exhibited enhanced proliferation capabilities (Figure 5G and H). To investigate metabolic changes in tumor cells receiving mitochondria from T cells, we analyzed the metabolic status of tumor cells stratified by TMT scores. The data showed that energy metabolism pathways, such as oxidative phosphorylation and TCA cycle, were significantly upregulated in the TMT-high group (Figure 5I). Next, we explored the effects of mitochondrial loss on T cells by comparing T cells from samples with the highest and lowest TMT scores. DEG analysis revealed that SCGB2A2 and HSPA1B were significantly upregulated, whereas CCL5 and CCR4 were downregulated, suggesting suppression of immune responses and activation of cellular stress-related pathways (Figure 6A). Pathway enrichment analysis further demonstrated that upregulated genes were associated with ER stress, whereas downregulated genes were linked to T cell activation (Figure 6B and C). These findings indicate that T cells experiencing mitochondrial loss undergo severe energy dysregulation, leading to impaired protein folding, ER stress, and ultimately T cell depletion and exhaustion.

Figure 5.

The immune-metabolic landscape of TMT-high deciphered by single-cell transcriptomics. (A and B) UMAP view of Pre (A) and Post (B) of cell subpopulation identification; (C and D) Marker gene expression for each cell type; (E) Bar plot of cell composition distribution in each sample; (F) Bar plot of cell composition distribution of T cells in each sample; (G) Volcano plot of DEG between TMT high and low cells; (H) GO enrichment analysis of up-regulated genes; (I) Correlation dot plot of the pathway that |Spearman r| > 0.5.

Figure 6.

Transcriptional landscape and functional enrichment of TMT high versus low group in T cells. (A) Volcano plot of DEG between TMT high and low T cells; (B) GO enrichment analysis of up-regulated genes; (C) GO enrichment analysis of down-regulated genes.

Tumor cells can gain proliferation advantage by hijacking T cell mitochondria

To validate the above findings, T cells were isolated from donor peripheral blood, cultured, and expanded in vitro for 7 days. Mitochondria were labeled with red fluorescence, and T cells were co-cultured with tumor cells under both normoxic and hypoxic conditions in either direct contact or non-contact settings (Figure 7A). After 48 hours of co-culture, the fluorescence intensity in the tumor cells was measured under identical exposure conditions. These results demonstrated that direct contact, rather than extracellular vesicle-mediated transfer, significantly increased the acquisition of T cell mitochondria by tumor cells. Hypoxia further enhanced mitochondrial transport (P < 0.05; Figure 7B and C). Additionally, the expression levels of TMT-related genes were significantly upregulated under these conditions (Figure 7D and E). In the HCC1954 cell line, a HER2-positive breast cancer model, contact-based co-culture led to a marked upregulation in the relative expression of nearly all genes examined. In contrast, in MDA-MB-231, a triple-negative breast cancer (TNBC) model, significant elevation was observed only in a limited set of genes such as ARHGEF5, F11R, and HSPA4. This subtype-specific variation in gene expression appears to align with our earlier finding that PAM50-based breast cancer subtypes significantly influence the efficiency of mitochondrial transfer. To assess the impact of mitochondrial loss on T cells, we measured the expression of XBP1, a marker of endoplasmic reticulum (ER) stress, in T cells across the four experimental groups. The results showed that mitochondrial loss in T cells induced ER stress, and hypoxia exacerbated this effect (Figure 7F). To investigate the functional consequences of tumor cells acquiring mitochondria, we extracted mitochondria from T cells and co-cultured them with tumor cells for 48 hours. CCK-8 and Transwell assays revealed that tumor cells receiving mitochondria exhibited significantly enhanced proliferation and migration capabilities (Figure 7G-I). These experimental findings confirmed that direct contact between tumor cells and T cells, particularly under hypoxic conditions, facilitates mitochondrial transfer, leading to enhanced tumor cell proliferation and migration. Concurrently, mitochondrial loss in T cells induces ER stress, which contributes to T cell dysfunction.

Figure 7.

Function analysis of tumor cells after mitochondria transfer. (A) Experimental workflow; (B and C) Fluorescence intensity of MDA-MD-231 (B) and HCC1954 (C) after co-culture; (D and E) MDA-MD-231 (D) and HCC1954 (E) expression levels of TMT genes; (F) Western Blot of the T cells after co-culture; (G and H) Cell proliferation experiments of MDA-MD-231 (G) and HCC1954 (H); (I) Transwell experiments for MDA-MD-231 and HCC1954.

Discussion

Intercellular mitochondrial transfer is a widespread phenomenon observed across various cell types under both physiological and pathological conditions^[17,36]. This process significantly alters the energy metabolism of recipient cells, notably enhancing oxidative phosphorylation (OXPHOS) and ATP production, which is consistent with our finding that the TMT score correlates with the tricarboxylic acid (TCA) cycle^[37]. Zhang et al further enriched this understanding by developing the MERCI algorithm by integrating single-cell sequencing data, tracking mitochondrial mutations in T cells and identifies tumor cells that obtain mitochondria, thereby constructing a characteristic gene set of TMT score^[20]. Based on this, our study employed the TMT scoring system as a tool to investigate how mitochondrial metastasis from T cells to tumor cells functionally affects the progression of breast cancer. In this study, our findings suggest that the TMT score may serve as a potential biomarker for aggressive features in breast cancer, particularly in relation to its immune and metabolic profiles. The introduction of the TMT score provides a new perspective on the complex pathophysiology of breast cancer, illuminating how mitochondrial dynamics can influence both the immune landscape and metabolic reprogramming within the TME. Specifically, our findings indicate that the transfer behavior of mitochondria to tumor cells, on the one hand, endows tumor cells with growth advantages by reshaping their metabolic adaptability, and on the other hand, profoundly weakens the function of immune cells by inducing an immunosuppressive state, jointly driving the malignant evolution of tumors. This links metabolic reprogramming with immunosuppressive mechanisms through the dynamic process of mitochondrial metastasis, providing a new perspective for systematically explaining the mechanism of breast cancer progression. TNTs are recognized as primary intercellular conduits for mitochondrial transfer, and their formation is notably augmented by hypoxic conditions in various cancers^[38]. Consequently, the increased formation of TNTs under hypoxic conditions may explain the elevated mitochondrial transfer observed in hypoxic cells in our experimental models.

It is well-established that cancer, as a systemic disease, exhibits malignant characteristics not only through the ability of tumor cells to evade growth inhibition signals, reprogram energy metabolism, and escape immune surveillance, but also through a dynamic ecosystem comprising tumor cells, stromal cells, and immune cells, collectively termed the TME^[39,40]. Tumor cells engage in molecular exchange and signal cascade transmission with components of the TME through specialized mechanisms, such as mitochondrial transfer. This process directly drives the reconstruction of respiratory function, accelerates tumor invasiveness, and influences multiple therapeutic outcomes^[36,41]. The TMT score has emerged as a reliable indicator associated with aggressive tumor phenotypes characterized by metabolic reprogramming. The correlation between high TMT scores and increased tumor aggressiveness underscores the pivotal role of mitochondrial transfer in modulating immune responses and metabolic profiles in cancer cells. Supporting this, Tanmoy et al demonstrated that tumor cells co-cultured with T cells exhibited significantly higher proliferation rates than those cultured alone^[18]. Conversely, another study revealed that tumor cells receiving normal mitochondria experienced suppression of proliferation^[30], suggesting that the source of mitochondria critically influences recipient cell behavior^[42].

T cells, the primary mediators of immune surveillance within the TME, have long been a focal point of immunotherapy research^[43–45]. The role of T cells in the breast cancer microenvironment is dual and complex^[46,47]. On one hand, T-cell-mediated cellular immunity is the core defense mechanism of the body against breast cancer. Cytotoxic T cells can directly induce tumor cell apoptosis by releasing effector molecules such as perforin and granzymes, while T helper 1 (Th1) cells further amplify this antitumor immune response through cytokine secretion^[48–50]. On the other hand, under persistent antigen stimulation in the TME, T cells are prone to functional exhaustion, characterized by upregulated expression of inhibitory receptors such as PD-1, TIM-3, and LAG-3, thereby losing their ability to effectively kill tumor cells^[51–53]. Several strategies have been proposed to reverse T cell exhaustion and enhance cytotoxicity, including immune checkpoint inhibitors, adoptive cell transfer, and CAR-T therapy^[51,54]. In our study, we observed that mitochondrial reallocation not only compromises the anti-tumor immune response but also facilitates tumor adaptation and survival under metabolically restrictive conditions, which is consistent with previous reports^[55,56]. Moreover, recent findings further indicate that cancer cells can transfer mitochondria to T cells, thereby disrupting their respiratory and anti-tumor capabilities^[57]. Similarly, tumor cells can also promote tumor proliferation and impair anti-tumor immunity by transmitting mitochondria to tumor-associated fibroblasts^[58]. These provides novel insights into the role of mitochondrial transfer in tumor immunity.

These findings suggest a novel metabolic immune regulatory mechanism: tumor cells may acquire exogenous energy metabolic capacity by actively taking up mitochondria from T cells and even other TME components (such as cancer-associated fibroblasts or myeloid cells), driving themselves to undergo intense metabolic reprogramming and enhancing their proliferation and survival advantages. Meanwhile, this mitochondrial “theft” behavior may lead to mitochondrial function impairment and energy depletion in donor immune cells (especially T cells), thereby causing cell dysfunction or apoptosis and damaging the anti-tumor immune response. This bidirectional metabolic interference, together with the observed down-regulation of lymphocyte activation markers, constitutes a synergistic immunosuppressive network, which not only weakens the function of effector immune cells but also actively reshapes the metabolic microenvironment of tumors, thereby providing an efficient path for tumor immune evasion and ultimately promoting tumor progression and deterioration.

Specifically, our findings demonstrate that high TMT scores are associated with diminished T-cell efficacy due to mitochondrial depletion, concurrent with enhanced metabolic capacity in tumor cells. The incorporation of the TMT score into clinical practice holds transformative potential for breast cancer management by enabling more accurate prognostic evaluations and facilitating personalized therapeutic strategies. Potential interventions may include targeting mitochondrial transfer mechanisms to restore T-cell function or combining approaches with conventional therapies to amplify their efficacy. It was proposed that targeting TNT may inhibit mitochondrial transfer between tumor cells and T cells, thus enhancing the efficacy of immunotherapy^[18]. Moreover, transplanting mitochondria from normal bone marrow stromal cells to T cells can enhance the metabolic adaptability and anti-tumor efficacy of T cells^[59]. Our findings suggest a noteworthy possibility: tumors with elevated TMT scores may exhibit a specific dependency on mitochondrial metabolic pathways. This scoring system could potentially aid in future efforts to identify patient subpopulations that may respond to “targeted mitochondrial transfer” or inhibitors of related metabolic pathways. These provides new insights for treatments targeting mitochondrial metabolism. However, it should be emphasized that although this study has revealed the potential utility of the TMT score in prognostic stratification and evaluation of immune status, its direct clinical translatability remains uncertain. The eventual prospects for clinical application will hinge on rigorous validation in prospective, large-scale independent cohorts, as well as integration with established patient stratification and predictive tools – such as TNM staging and IHC staining – to develop more discriminative composite models. Such an approach could enable more accurate identification of high-risk patients and inform the design of personalized follow-up and management strategies.

The phenomenon of horizontal mitochondrial transfer between cells remains underexplored, and our research provides only indirect evidence. The MitoTracker has inevitable limitations, such as dye leakage along with mitochondria and the inability to achieve long-term and stable tracking in living cells. Therefore, in the subsequent research, we will adopt two more stable tracing strategies: tracking the mtDNA mutations specific to donor cells, or using engineered donor cells that stably express mitochondrial-targeted fluorescent proteins (such as Mito-GFP)^[57,59,60]. These methods can accurately distinguish and specifically label mitochondria derived from donors, thereby enabling long-term and non-invasive fate tracking of individual exogenous mitochondria within a single recipient cell in the living TME. Another notable limitation of this study was the absence of in vivo experiments to confirm that the identified genes directly mediate mitochondrial transfer within the TME. Although our approach of “feeding” isolated mitochondria demonstrated that exogenous mitochondria are sufficient to enhance tumor cell function, this process bypasses direct cell-to-cell contact and signaling interactions. Moreover, the current co-culture system oversimplifies the high complexity of the TME, which includes diverse cell types, spatial architecture, and dynamic extracellular matrix – all factors that which may significantly influence the efficiency and regulatory mechanisms of mitochondrial transfer. To address these limitations, future studies will employ advanced techniques such as real-time live-cell imaging with mitochondrial-specific fluorescent markers to directly visualize and quantify intercellular mitochondrial transfer. Additionally, we plan to develop more physiologically relevant 3D co-culture models incorporating stromal components and immune cells, which will better mimic the spatial constraints and paracrine signaling of the in vivo microenvironment. These approaches will help elucidate the precise molecular mechanisms governing mitochondrial transfer while maintaining physiological relevance. Experimental studies employing targeted genetic or pharmacological manipulation of mitochondrial dynamics could yield critical insights into the causal relationships that underpin our observations. Moreover, exploring the synergistic effects of mitochondrial modulation in combination with existing therapeutic modalities represents a promising direction for innovative treatment strategies.

Conclusion

The TMT score provides a novel perspective on the interplay between mitochondrial function and tumor dynamics in breast cancer. Our study not only underscores the critical role of mitochondria in shaping the immune and metabolic landscape of tumors but also opens new pathways for therapeutic intervention. Continued exploration of mitochondrial dynamics in cancer will undoubtedly enhance our understanding and treatment of this complex disease, paving the way for more effective and targeted therapeutic strategies.

Methods

Data sources and study design

The open breast cancer gene expression dataset with complete prognostic and clinicopathological information annotation was downloaded from The Cancer Genome Atlas (TCGA; n = 1118) databases (https://www.cancer.gov/ccg/research/genome-sequencing/tcga). Single-cell RNA-sequencing data (GEO: GSE176078) of breast cancer samples were used to detect immune and metabolic differences in TMT subgroups. The GSE173991 was used to analyze the metabolic reprogramming of the TMT subgroups. Gene mutation information was obtained from the cBioPortal database (https://www.cbioportal.org/).

TMT scores

A 17-gene set, containing ABI1, ARF6, F11R, BAIAP2L1, MYADM, ACTR3, ARHGEF5, SRC, CAPZA2, GOLPH3, RAB10, ARF4, CAP1, PSEN1, YWHAG, YWHAZ, and HSPA4, was reported by Zhang et al^[20]. The TMT scores were calculated using the GSVA algorithm with method = “gsva” and kcdf = “Gaussian” parameter setting and default values for all other parameters^[61].

Functional and pathway enrichment analysis

DEGs between the high- and low-TMT score groups were identified using the limma R package^[62]. To further explore the potential functions of DEGs, gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed using the clusterProfiler R package (version 4.0.5)^[63]. Moreover, to determine the different pathways for analyzing the differences in biological function between high- and low-score groups, the gene set variation analysis (GSVA) package (version 1.40.1) was used to calculate GSVA scores of hallmark gene sets from the Molecular Signatures Database (MSigDB)^[64]. P values less than 0.05 were considered to indicate significant differences.

Comprehensive analysis of immune characteristics in different TMT subgroups

To identify the immune characteristics of breast cancer patients in the TCGA cohort, their expression data were imported into the xCell function, and analysis with 1000 resamples was performed to estimate the relative proportion of 22 types of immune cells^[33]. Then, we calculated the Pearson’s correlation coefficient of 22 types of immune cells between TMT score, and the results are presented in a heatmap. Tumor Immune Dysfunction and Exclusion (TIDE) analysis (http://tide.dfci.harvard.edu/) was performed to predict immunotherapeutic response^[34].

Metabolic analysis

To determine the metabolic profile of breast cancer patients in the TCGA cohort, we downloaded 86 metabolically relevant pathways from the KEGG database and evaluated them using the GSVA R package to determine their metabolic changes. We also calculated their Pearson correlation coefficients between the TMT score and the metabolic pathways. Moreover, metabolomics data was included in this study, and analyzed using MetaboAnalyst 6.0 (https://www.metaboanalyst.ca/)^[65].

Quality control and cell type recognition

An analysis of scRNA-seq data from breast cancer samples (GSE176078) was performed using the Seurat R package (version 5.1.0)^[66]. According to previously reported quality control criteria, single cells with <200 genes or UMI count <1000 or the percent of mitochondrial genes over 20% of total expressed genes were screened as low-quality cells and eliminated. Thereafter, the major cell types were recognized based on previously reported markers. After preprocessing, normalization, and batch correction, T cells were reclustered.

Western blot

Total protein was extracted in RIPA buffer with 1% PMSF (Beyotime, China). The samples were separated using 4–12% SDS PAGE gels and subsequently transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA). After blocking, the membranes were incubated with primary and secondary antibodies, and protein expression was detected using an enhanced chemiluminescence (ECL) kit (Tanon, China). The antibodies used for Western blot were as follows: anti-GAPDH antibody (1:20 000, Proteintech, China), anti-DAP3 antibody (1:200, Santa Cruz, China) and HRP-conjugated goat anti-mouse IgG antibody (1:1000, Beyotime, China).

RNA preparation and RT-qPCR

Total RNA was isolated with TRIzol reagent (Invitrogen, USA), reverse transcribed into cDNA with a Revert Aid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, USA), and finally, SYBR reagent (Bio-Rad, China) and specific primers (Generay, China) were used for quantification. After normalization to GAPDH, the relative expression was calculated.

Mitochondrial isolation

Mitochondria were isolated from cultured cells using a Mitochondria Isolation Kit (Beyotime Biotechnology, China; Cat# C3601) following the manufacturer’s protocol with minor modifications. Briefly, cells were washed twice with ice-cold phosphate-buffered saline (PBS) and harvested by gentle scraping. The cell pellet was resuspended in 1x mitochondrial isolation buffer. After incubation on ice for 20 minutes, cells were homogenized using a pre-chilled Dounce homogenizer (30 strokes). The homogenate was centrifuged at 600g for 10 minutes at 4 °C to remove nuclei and unbroken cells. The supernatant was further centrifuged at 11 000g for 10 minutes at 4 °C to pellet the mitochondrial fraction. The mitochondrial pellet was washed twice with isolation buffer and resuspended in base medium.

CCK-8 assay

A total of 1500 transfected cells were seeded in 96-well plates and incubated overnight. The cells were subsequently incubated with CCK-8 solution (MCE, USA) for 2 hours, and the optical density (OD) value was subsequently measured at 450 nm using a plate reader (Thermo, USA) every 24 hours.

Transwell assay

Transwell assays were conducted using the concentration difference of serum. The chamber (Costar, USA) and Matrigel (BD, USA) were used for invasion assays, whereas migration assays were performed with no gel. After 48 hours, the cells were fixed with 4% paraformaldehyde (Beyotime, China), stained with 0.1% crystal violet (Sigma, USA), and photographed under a microscope (Olympus, Japan). The cells in random fields were counted and analyzed using GraphPad Prism 8.0.

Statistical analysis

Statistical analyses were performed using R software (version 4.2.1). All P-values less than 0.05 (P < 0.05) were considered statistically significant.

Ethical approval

This study adhered to the International Ethical Guidelines for Health-related Research Involving Humans (https://cioms.ch/publications/product/international-ethical-guidelines-for-healthrelated-research-involving-humans/) and was approved by the Ethics Committee of Fudan University Shanghai Cancer Center (FUSCC; Shanghai, China; approval No. 050432-4-2307E).

Consent

Not applicable.

Sources of funding

This work was funded by the Foundation of Shanghai Anticancer Association EYAS PROJECT (SACACY22B08), Beijing Xisike Clinical Oncology Research Foundation (Y-Young2024-0374) and National Natural Science Foundation of China (No. 82173177 and No. 82472955).

Author contributions

Z.Q.: conceptualization, data curation, methodology, resources, software, writing – original draft; Y.Y.: conceptualization, data curation, resources, software; Z.W.: resources: S.L.: resources; D.L.: formal analysis, project administration; Y.G.: supervision, resources; L.C.: formal analysis; M.F.: data curation, project administration; C.Z.: data curation; X.C.: supervision; B.H.: validation; X.W.: conceptualization, investigation, writing – review & editing; L.L.: conceptualization, data curation, funding acquisition, investigation, methodology, validation, writing – review & editing; Y.C.: conceptualization, funding acquisition, validation, writing – review & editing.

Conflicts of interest disclosure

The authors declare that they have no competing interests.

Research registration unique identifying number (UIN)

Not applicable.

Guarantor

Yiqun Cao.

Provenance and peer review

Not invited.

Data availability statement

The data sets used in this investigation are available through public repositories.