Since 2019, breast cancer has surpassed lung cancer as the malignant tumor with the highest global incidence (1), and its incidence is closely related to socioeconomic development, with the highest risk of disease in economically transformed regions and the lowest survival rates in economically underdeveloped regions (2). Underdiagnosis, misdiagnosis and lack of effective treatments are the most important reasons for the huge differences in prevalence and survival rates worldwide (3). The current treatment options for breast cancer include various modalities such as surgery, radiation, immunization, endocrine, targeted,chemotherapy and Chinese medicine. Immunotherapy has made significant breakthroughs in recent years, with immune checkpoint inhibitors in particular boosting survival in many patients. However, this approach has shown significant results in only a subset of patients, with relatively low success rates. Objective remission rates in the vast majority of clinical studies were less than 20%, while median progression-free survival and overall survival were less than 3 and 24 months (4). Therefore, there is a need for in-depth research on new methods and ideas for the treatment of breast cancer. TLSs are a class of ectopic lymphoid organs that form in non-lymphoid tissues and are common in conditions such as tumors, chronic inflammation, and autoimmune diseases. Because TLSs are formed in non-lymphoid tissues, they have a more flexible ability to respond to local abnormal lesions, thus playing an important immunomodulatory role in chronic inflammation and tumor development and contributing to the efficiency of immune response and local therapeutic efficacy. The presence of TLSs has been observed in a variety of malignant tumors such as non-small cell lung cancer, colorectal cancer, breast cancer, melanoma, sarcoma, and renal cell carcinoma (5). Recent findings indicate that researchers are gradually recognizing the positive role of TLSs in treating breast cancer patients. These insights have provided clinicians with new perspectives to reformulate breast cancer treatment protocols from an immune perspective (6). Machine learning belongs to a kind of artificial intelligence, with the arrival of the era of big data, the ability of computers to quickly process complex and huge data is increasingly dominant, and in recent years the increasing popularity of high-throughput histological data and the success of artificial intelligence technology has made machine learning in a variety of fields widely used (7–9). Machine learning is more efficient than humans in predicting and prognosticating treatment for some cancers. This review summarizes the current understanding of TLSs in breast cancer and the potential application of machine learning in the study of TLSs and breast cancer.

The most common techniques for detecting TLSs are multiple immunofluorescence, hematoxylin and eosin (HE) staining, of which multiple immunofluorescence is difficult to generalize in research due to its high cost, small field of view, and high complexity. In contrast, HE staining is easier and remains the clinical standard in histopathology. TLSs have been detected by pathologists on HE slides, but manual detection is time-consuming and laborious, and results vary according to the level of expertise.Li, Z et al. (10) developed a machine-learning-based computational tool for the automatic detection and quantitative assessment of TLSs on routine HE slides. They confirmed its independent prognostic value in an international multicenter cohort of 1924 patients with six common gastrointestinal cancers. An important advantage of this method is the automated enumeration and quantitative characterization of TLSs. This study, the largest to date, confirms the association of TLSs with the survival of patients with gastrointestinal cancers.

TLSs contain within them T cells, B cells, dendritic cells, and high endothelial venules (HEV) (11). Similar to SLOs, both share the components of B cells, T cells, and dendritic cells. However, compared to SLOs, TLSs have a simpler organization with scattered lymphocyte aggregates and lack the connective tissue membrane of SLOs (12). The non-enveloped structure of TLSs allows the immune cells to fully interact with the surrounding microenvironment, and, unlike SLOs, TLSs do not persist in a specific location, but rather are formed under specific pathologic conditions, independent of secondary lymphoid organ regions. TLSs are formed under specific pathological conditions, independent of secondary lymphoid organ regions, and trigger immune responses under the regulation of clear inflammatory signals. However, the mechanism that triggers the formation of TLSs remains unclear (13).

HEV is a unique vascular structure in the lymphatic system, which helps the formation of TLSs through its unique morphology and function, and contributes to lymphocyte colonization and immunoregulation in the lymphatic system. Most of the current opinions believe that the formation of TLSs is highly dependent on SLOs and HEV (25–28). Some of the lymphocytes in SLOs spread to tumor tissues through lymphatic vessels or HEV when stimulated by inflammatory factors for a long period of time, thus initiating the formation of TLSs (29). In a retrospective cohort study, Martinet analyzed data containing 146 patients with invasive breast cancer and showed that the density of HEV was positively correlated with patients’ disease-free survival, metastasis-free survival, and overall survival, demonstrating their important role in the formation of TLSs (30). Therefore, tumor HEV may be potential therapeutic targets in cancer diagnosis and treatment. For example, Colbeck and his team effectively killed tumor cells in the Foxp3DTR mouse model by depleting Treg cells to promote the self-expanding circuit of T-cell activation as well as the formation of HEV in the tumor, which in turn induced the formation and maturation of TLSs in the tumor (31).

Existing research suggests that machine learning is increasingly used for clinical cancer diagnosis, grading, genetic alteration prediction, and disease prognosis, and that it can identify molecular markers for cancer treatment (62–67), predict surgical outcomes (68), and interpret electrocardiograms. Liu, X et al. developed a novel breast cancer recurrence and metastasis risk assessment framework from histopathology images using image features and machine learning techniques (69), The study is expected to reduce the workload of pathologists and improve the chances of survival for breast cancer patients. Sammut, S.-J. analyzed breast tumors from patients with primary invasive cancers who participated in the TransNEO study at Cambridge University Hospitals NHS Foundation Trust between 2013 and 2017, and found that machine-learning models combining clinical, molecular, and numerical pathology data for predicting response to treatment significantly outperformed models based on clinical variables, emphasizing the The importance of data integration for response prediction and can be used to generate similar predictors for other cancers (70). Machine learning can not only diagnose and assess the prognosis of breast cancer from a pathologic perspective, but also help clinicians choose breast cancer treatments from an imaging perspective. Yu, Y. et al. (71) then developed an efficient preoperative magnetic resonance imaging imaging histology for assessing axillary lymph node status in breast cancer using machine learning techniques, which can identify patients with axillary lymph node metastasis in early-stage invasive breast cancer preoperatively. While machine learning has made significant progress in the diagnosis, treatment, and prognosis assessment of breast cancer, it also has limitations. The performance of machine learning algorithms highly depends on the quality and quantity of input data. Therefore, incomplete or biased datasets may affect the accuracy of the model.

The human microbiota refers to the collection of microorganisms that inhabit the human body. While most microorganisms are beneficial to the human body, an imbalance in the microbial community, where harmful bacteria outnumber beneficial ones, can lead to various diseases, including cancer. In recent years, researchers have uncovered previously unrecognized connections between immune microenvironment dysregulation and breast cancer (72–76). There is limited knowledge about the microbial composition associated with normal breast tissue and breast-related diseases. Human breasts are not sterile; they harbor diverse bacterial communities. Studies have confirmed that, besides the skin, some microbial communities in breast tissue can also translocate from the gastrointestinal tract through the nipple, possibly facilitated by breastfeeding and/or oral contact through sexual activity (77). This breast microbiota stimulates resident immune cells to maintain healthy breast tissue. Additionally, the types of bacteria present and their metabolic activities, such as their ability to degrade carcinogenic substances, may also contribute. Xuan, C. observed that the baseline expression of antimicrobial response genes in tumors is lower than in healthy breast tissue. Microbial DNA is present in the breast, suggesting that bacteria or their components may influence the local immune microenvironment (78). Banerjee et al. examined the specific and common viral, bacterial, fungal, and parasitic features of each breast cancer subtype. They identified distinct patterns in triple-negative and triple-positive breast cancer samples, while ER-positive and HER2-positive samples exhibited similar microbial characteristics. These features, unique or shared among different breast cancer types, offer a new research avenue for gaining further insights into the treatment and prognosis of breast cancer. This provides a novel understanding of the role of the microbiota in breast cancer (79). Research has found that postmenopausal women newly diagnosed with breast cancer exhibit less diversity and compositional differences in their fecal microbiota compared to women without breast cancer. This discovery suggests that the gut microbiota may influence the risk of breast cancer occurrence and could potentially do so through estrogen-independent pathways (80). The stability of the microbiota in the breast is closely intertwined with the immune environment. Clinical practitioners can develop tailored treatment strategies based on the mechanisms of action of the microbiota in the breast in improving the prognosis of breast cancer.

Inflammation-related necrosis and macrophage infiltration may be associated with the formation of TLSs. Tumor necrosis is often accompanied by the onset of an inflammatory response in which damaged tissue releases inflammatory factors and cytokines in response to external aggression. In some cases, cell necrosis within the tumor results in the formation of foci of inflammatory necrosis. These necrotic foci become one of the underlying conditions for the formation of TLSs, which attract immune cells (especially macrophages) by releasing a large number of signaling molecules that accumulate in the damaged area (81). The high degree of plasticity of macrophages can adapt to various microenvironmental changes in tissues, and M1-type macrophages, which are usually activated by toll-like receptors, mostly exhibit anti-pro-inflammatory effects in the immune response (82–85). Olson demonstrated that modulation of phagocytosis in macrophages effectively inhibits tumor progression and improves the prognosis of cancer patients (86). Under inflammatory conditions, macrophages activate and release more inflammatory mediators and chemokines, which further direct immune cells to migrate to the region of TLSs, contributing to the structural formation and maintenance of TLSs (87).

TLSs have been shown to have a favorable prognostic function in a variety of malignancies, but have been less well studied in breast cancer. In the available studies, significant differences were found between TLSs-positive and TLSs-negative subgroups within the same molecular subtype of breast cancer, with significant effects on breast cancer recurrence, lymphovascular infiltration and perineural infiltration (88). TLSs in the breast cancer stroma have been associated with activation of tumor angiogenesis, suggesting that this may be a factor favoring breast cancer metastasis, but most of these studies have been performed in animal models and there are no data on human tissues (89).TLSs were detected in 60% of breast cancer tumors and correlated with higher infiltration of TILs by Buisseret. PD-1 and PD-L1 expression were also associated with higher density of TILs and TLSs, in addition TILs density, TLSs and PD-L1 expression were associated with more aggressive tumor characteristics (90), on the basis of this study, researchers found a positive correlation between the expression of immune checkpoint molecules and baseline TILs and TLSs suggesting that assessing these parameters in breast cancer patients may identify immunomodulatory therapies in tumors that are responsive to them (91).

In summary, the in-depth investigation of the relationship between breast cancer and TLSs has brought new ideas and prospects for the clinical diagnosis and treatment of this type of malignant tumor. Machine learning demonstrates high efficiency in the detection of TLSs and the diagnosis and prognosis of breast cancer.Immune checkpoint inhibitors based on TLSs have become an important breakthrough in breast cancer treatment, but not all patients can benefit from them. In the future, we should focus on researching the distribution and activity of TLSs, making effective use of machine learning to identify beneficiary populations, and ensuring more precise selection before treatment. Furthermore, by exploring the molecular signaling pathways influencing the formation and function of TLSs, as well as the microbial characteristics of each subtype of breast cancer, identifying intervention targets, and developing novel treatment strategies, it is possible to formulate personalized treatment plans based on individual patients. This approach has the potential to significantly enhance treatment efficacy and reduce unnecessary medical interventions.

XinL: Writing – review & editing, Writing – original draft, Validation, Supervision, Methodology, Investigation, Formal analysis, Conceptualization. HX: Writing – review & editing, Visualization, Validation, Supervision, Conceptualization. ZD: Writing – review & editing, Validation, Supervision, Investigation. QC: Writing – review & editing, Supervision, Resources, Investigation, Conceptualization. XiaL: Writing – review & editing, Writing – original draft, Supervision, Resources, Investigation, Funding acquisition.