Breast cancer is the second most prevalent cancer globally. In 2022, approximately 2.3 million newly diagnosed cases of female breast cancer occurred worldwide, and more than 665 thousand people lost their lives to this disease1.
Understanding of breast cancer began with clinical observations and pathological assessments. Various histologic types, sizes, nodal status, metastasis, grades, and stages predict different clinical outcomes and influence the selection of treatment options. In addition, molecular profiling offers another approach for classifying breast malignancies, thus enabling exploration of the biological factors underlying cancer development. Because each subtype is characterized by distinct oncogenic features that promote breast cancer onset and progression, targeting these biological processes has provided new opportunities for precision treatment. Consequently, molecular subtyping has become an integral and indispensable aspect of breast cancer taxonomy, and is critical in diagnosing breast cancer and determining subsequent precise systemic treatment.
This editorial provides an overview of advancements in molecular subtyping for breast cancer with evolving molecular profiling methods, and ongoing efforts to integrate leading-edge subtyping concepts into precise clinical practice. We hope that this editorial will inspire researchers and medical professionals to recognize the importance of molecular classification and effectively apply these insights to clinical work.
From receptor-based subtyping to comprehensive molecular subtyping of breast cancer
On the basis of the status of the estrogen receptor (ER), progesterone receptor (PgR/PR), and human epithelial growth factor receptor 2 (HER2, also known as Erb-b2 receptor tyrosine kinase 2, ERBB2), breast tumors are generally classified into 3 receptor-based subtypes: hormone receptor (HR)-positive breast cancer, HER2-positive breast cancer, and triple-negative breast cancer (TNBC). Patients with HR+ tumors should be treated with endocrine therapy, whereas patients with HER2+ tumors should receive targeted therapy with anti-HER2 agents2. However, the lack of feasible molecular targets has resulted in a preference for less specific chemotherapy as a systemic treatment option for TNBC2. Although the use of the traditional 3-class classification according to receptor status has increased patient survival over the past several decades, several challenges persist in enhancing patient prognosis3.
To better determine the molecular landscape of breast cancer and identify potentially actionable targets, transcriptomic profiling was first used to systemically reveal the complexity of breast cancer. This technique provides a new perspective for stratifying patients beyond clinicopathological assessments, and enables the administration of treatment strategies with enhanced precision.
The first systemic analysis of transcriptomic molecular subtypes was performed by Perou and colleagues4,5, who robustly constructed a clinically relevant classification system of intrinsic subtypes. After a decade, the intrinsic subtypes—widely known as luminal A, luminal B, Erb-B2 overexpression, and basal-like—were officially used by the St. Gallen International Consensus Guidelines for guiding systemic treatment2,6. Although molecular profiling to determine the intrinsic subtype can delineate the disease with enhanced accuracy, more affordable surrogate molecular subtyping has been widely used in clinical practice. This surrogate approach is based on immunohistochemical measurements of ER, PgR, Ki-67, and HER2, together with evaluation of recurrence risk through multigene expression assays (if available)6, and involves the following subtypes: (1) luminal A-like, (2) luminal B-like (HER2 negative), (3) luminal B-like (HER2 positive), (4) HER2 positive (non-luminal), and (5) triple negative (ductal).
The rapid advancement of high-throughput technologies at lower cost has enabled large-scale, comprehensive, and multi-layer molecular profiling of breast cancer extending beyond gene expression. The Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) conducted an integrated analysis of approximately 2,000 breast cancer cases, including both genomic and transcriptomic data for novel biological subgroups7. Breast cancer was classified into 10 integrative clusters, each characterized by unique patterns of copy number alterations. A study based on The Cancer Genome Atlas (TCGA) invasive breast cancer dataset has further characterized the molecular landscape by using data generated from 6 platforms including DNA copy number arrays, DNA methylation, exome sequencing, messenger RNA arrays, microRNA sequencing and reverse-phase protein arrays8. That study has provided new insights into the multi-level characterization of breast cancer, thereby comprehensively revealing the heterogeneity of breast cancer.
These integrative analyses have incorporated genomic information and phenotypic variations into the intrinsic subtypes, thus providing an effective method for molecular typing of breast cancer and paving the way to further in-depth studies.
Subclassification of receptor-based subtypes
Clinical diagnosis and treatment based on receptor-based subtypes and intrinsic molecular subtypes have achieved substantial success. The current focus of molecular profiling of breast cancer is on deciphering the heterogeneity within each receptor-based subtype. Thus, while retaining previously effective treatments, new drugs targeting newly revealed therapeutic targets can further enhance treatment effectiveness and provide clinical translation value.
TNBC subclassification
TNBCs make up approximately 15% of all breast cancers and are usually associated with younger age; these cancers are prone to early recurrence and metastasis9. The existing treatment options for unselected patients with TNBC have limited efficacy: 70% of patients with TNBC do not survive 5 years after diagnosis10. Additionally, TNBC is not a single disease but a heterogeneous entity with different subtypes11. Consequently, substantial efforts have focused on molecular subtyping of TNBC to identify subtype-specific treatment approaches (Table 1).
Comprehensive molecular subtyping of TNBCs suggests subtype-specific treatment strategies
Lehmann subtyping
In 2011, Lehmann et al., in pioneering research, analyzed the gene expression profiles of 587 TNBC cases and identified 6 subtypes with individual biological characteristics, comprising, 2 basal-like subtypes (BL1 and BL2), an immunomodulatory subtype (IM), a mesenchymal subtype (M), a mesenchymal stem–like subtype (MSL), and a luminal androgen receptor (LAR) subtype12. The BL1 and BL2 subtypes are characterized by elevated expression of cell cycle and DNA damage response genes, which are potentially responsive to cisplatin-related agents. The M and MSL subtypes are distinguished by upregulation of epithelial–mesenchymal transition and growth factor pathways, and should be treated with Src/Abl inhibitors in combination with PI3K inhibitors. The LAR subtype is enriched in genes associated with androgen receptor signaling, and is uniquely sensitive to AR antagonists. Genes associated with immune processes are upregulated in the IM subtype.
Five years later, Lehmann refined the TNBC molecular subtypes from 6 to 4 (i.e., TNBCtype-4), comprising BL1, BL2, M, and LAR13. Each TNBCtype-4 subtype exhibits large differences in terms of age at diagnosis, histopathology, tumor grade, disease progression course, and response to neoadjuvant chemotherapy.
In another study14, Bareche et al. conducted a global replication of Lehmann’s TNBC classification in a cohort of 550 TNBCs from TCGA and METABRIC, thus demonstrating the ability to segment TNBC into 5 distinct and stable transcriptional subtypes: BL1, IM, LAR, M, and MSL. Furthermore, TNBC subtypes have been shown to exhibit unique somatic mutation profiles and copy number alteration profiles, particularly regarding specific driver cancer genes such as TP53 and PIK3CA. These findings have been instrumental in identifying potential targets for genomic-driven therapies for TNBC, thus offering hope for more personalized and effective treatment strategies in the future.
Burstein subtyping
In 2015, Burstein and colleagues15 conducted molecular profiling of 198 TNBC tumors, identified 4 TNBC subtypes, and confirmed their results by using external datasets. They clustered TNBC into 4 subtypes: LAR, mesenchymal (MES), basal-like immunosuppressed (BLIS), and basal-like immune-activated (BLIA). Each subtype has a distinct prognosis and specific targetable molecules, including AR and MUC1 for LAR; growth factor receptors for MES; immunosuppressive molecules for BLIS; and Stat signaling-associated molecules and cytokines for BLIA.
FUSCC subtyping
In 2019, Jiang et al. performed transcriptome-based clustering in the Fudan University Shanghai Cancer Center (FUSCC) TNBC multi-omics cohort16, and classified TNBC into 4 subtypes: IM, LAR, MES, and BLIS. This cohort, consisting of 465 patients, is the largest single-center multi-omics cohort of TNBC in an Asian population to date.
On the basis of genomic and transcriptomic analysis, potential therapeutic subtype-specific targets have been identified. Compared with other TNBC subtypes, the LAR subtype shows enrichment in ERBB2-activating mutations, activating cell-cycle signaling, and androgen receptor overexpression, thus indicating potential sensitivity to HER2-targeting therapy, CDK4/6 inhibitors, and AR inhibitors, respectively. Additionally, LAR is frequently observed in Asian patients, thus underscoring its importance. For the IM subtype, expression profiles have indicated enrichment in both immune-activated cells and immune-stimulating factors, thereby supporting the rationale for using immune checkpoint blockade to treat this subtype. The BLIS subtype, which has relatively poor prognosis and minimal immune activation, is further divided into high-homologous recombination deficiency (HRD) and low-HRD subgroups according to HRD scores. The BLIS subgroup with high HRD scores might substantially benefit from treatment with DNA-damaging agents. However, patients in the low-HRD subgroup have poor prognosis, with a 5-year recurrence-free survival (RFS) rate of 73%. In the MES subtype, gene expression analysis has highlighted features associated with cancer stem cells, including JAK/STAT3 pathway activation for cancer stem cell maintenance. These findings suggest that STAT3 inhibitors might provide a promising therapeutic option for the MES subtype.
HR+/HER2− subclassification
The subclassification of HR+/HER2− breast cancer, in contrast to TNBC, remains an understudied area because of this cancer’s favorable prognosis with endocrine therapy17. Nevertheless, crucial problems remain, such as the persistent risk of long-term recurrence, resistance to endocrine therapy, and the precise administration of novel medications such as CDK4/6 inhibitors.
Given these challenges and HR+/HER2− breast cancer heterogeneity7,8, molecular profiling of this disease is imperative to develop novel subtype-specific precision treatment strategies. Recently, Jin et al., using multi-omics data for 579 patients, have classified HR+/HER2− breast cancer into 4 subtypes—cranial luminal (SNF1), immunogenic (SNF2), proliferative (SNF3), and receptor tyrosine kinase (RTK)-driven (SNF4)—each with distinct biological and clinical characteristics18. The SNF subtypes have provided ideas for new therapeutic regimens for HR+/HER2− breast cancers. Endocrine therapy is likely to be most suitable for the SNF1 subtype, whereas SNF2 tumors are immune activated and consequently may be targeted by immune checkpoint blockade. The SNF3 subtype is potentially sensitive to CDK4/6 inhibitors and PARP inhibitors, because of its proliferative features and high homologous recombination deficiency scores, respectively. SNF4 tumors have the poorest outcomes among the 4 subtypes; these tumor cells are characterized by RTK pathway upregulation, thereby indicating sensitivity to RTK inhibitors.
HER2+ subclassification
Anti-HER2 targeted therapy has achieved remarkable success in treating HER2+ breast cancer. However, resistance to therapy is nonetheless observed, and revealing the heterogeneity within HER2+ breast cancer is essential. In 2016, Ferrari et al.19 conducted genome-wide RNA sequencing of 99 selected HER2+ breast tumors and clustered them into 4 subgroups: A, B, C, and D. In addition to the different degree of HER2 amplification among the four subtypes, each exhibited distinctive genomic features. Furthermore, Li et al.20, on the basis of a multi-omics cohort of 180 patients with HER2+ breast cancer, have proposed a four-subtype strategy and further suggested possible subtype-dependent precision treatment regimens with patient derived organoids: anti-HER2 targeted therapy for the classical HER2-enriched (HER2-CLA) subtype and the immunomodulatory (HER2-IM) subtype; standard targeted treatment with endocrine therapy and CDK4/6 inhibitors for the luminal-like (HER2-LUM) subtype; and adjunct treatment with EGFR, PDGFR, or VEGFR inhibitors with anti-HER2 targeted therapy for the basal/mesenchymal-like subtype (HER2-BM). Molecular subtyping of HER2+ breast cancer has received relatively little attention, but accumulating evidence supports its benefits for patients with this type of breast cancer.
Molecular subtyping of breast cancer from bench to bedside
Clinical trials
Molecular profiling of breast cancer to discover new taxonomies is ultimately aimed at improving treatment efficacy and patient prognosis through precision-targeted therapy. Therefore, clinical trials are crucial in promoting the application of molecular subtyping.
Clinical trials have been designed to validate the effectiveness of precision treatment strategies, represented by a series of trials based on TNBC subclassification. Researchers from Spain21 have discovered that, according to Lehmann’s refined classification of TNBC, significant differences exist in the pathological complete response (pCR) rates of patients with TNBC treated with neoadjuvant carboplatin and docetaxel. In comparison to an overall pCR rate of 44.7%, the BL1 and BL2 subtypes had higher pCR rates of 65.6% and 47.5%, respectively, thus indicating that patients with BL subtypes might benefit from neoadjuvant chemotherapy treatment. In addition, The TBCRC 023 study by Lehmann et al. has revealed enhanced clinical benefit with combined treatment with an AR antagonist and a PI3K inhibitor for AR+ TNBCs, particularly those classified as LAR subtype22.
The FUTURE studies performed by FUSCC are among the best-known series trials evaluating the efficacy of subtype-specific treatment strategies for TNBC. On the basis of FUSCC-TNBC subtypes, clinicians initially conducted the FUTURE umbrella clinical trial23. The final results indicated a 29.8% objective response rate for patients with advanced TNBC who had failed multiple lines of treatment. Among the 7 arms, the immune therapy arm (arm C) targeting the IM subtype achieved an objective response rate of 43.5%, surpassing the efficacy of conventional chemotherapy in such patients (10%–15%)23. By building on the promising results of the immunotherapy arm (arm C) in the FUTURE trial, researchers designed the FUTURE-C-Plus trial24. In that study, immunotherapy was prioritized as the first-line intervention, and the objective response rate notably increased, to 81.3%. In the subsequent FUTURE-SUPER trial25, Patients with TNBC received first-line therapies according to their molecular subtypes and genomic markers. The median progression-free survival was significantly longer in the pooled subtyping-based group than the control group, thus demonstrating possible therapeutic benefits of optimizing treatment according to comprehensive molecular subtypes for patients with TNBC.
The efficacy of using the SNF subtypes of HR+/HER2− breast tumors to guide precise treatment is also being evaluated in BCTOP-L-M05 (NCT05594095), in which patients with HR+/HER2− advanced breast cancer who had previously received CDK4/6 inhibitors were recruited. After implementation of second-line and first-line clinical trials, the team is currently conducting a neoadjuvant platform trial based on molecular subtyping of breast cancers, including the FUSCC-TNBC subtypes and HR+/HER2− SNF subtypes (FASCINATE-N)26.
Feasible tools
Current molecular subtyping methods for breast cancer heavily rely on high-throughput molecular biological techniques. Cost-effective and convenient approaches have been developed to widely implement these advancements in breast cancer treatment. Consequently, researchers have begun to seek alternative methods for molecular subtyping through pathology and medical imaging.
For instance, a practical clinical IHC-based subtyping strategy based on 4 markers (AR, CD8, FOXC1, and DCLK1) was designed to identify FUSCC-TNBC subtypes27. This IHC-based classification approach for TNBC has demonstrated a significant correlation with transcriptome-based classifications, and has facilitated broader application of molecular subtyping in clinical practice.
Digital pathology involves the conversion of histopathological slides into digital formats through whole-slide scanners, for enhanced objective, quantitative analysis of the digitized images28. Artificial intelligence techniques have been applied to analyze digital images, thus making digital pathology an efficient tool for the molecular classification of breast cancer. Couture et al.29 have analyzed digital slides with deep learning to predict ER status, intrinsic subtypes (basal vs. nonbasal), and risk of recurrence score (ROR-PT), with a high overall accuracy of approximately 75%–80%. Jin et al.18 have developed convolutional neural network models based on digital pathology to infer HR+/HER2− breast cancer subtypes (SNF1, SNF2, SNF3, and SNF4), which have been used in the FASCINATE-N trial and BCTOP-L-M05 trial.
Radiomics transforms radiographic images into multiple quantitative image features, such as shape, size, and textural patterns30. Recent progress in AI technology has significantly improved the ability to automatically measure and analyze patterns found in radiomics. Li et al.31 have characterized 91 breast tumors through quantitative radiomics. The image-based phenotypes were able to discriminate prognostic biomarkers, such as receptor status, and were able to predict breast cancer molecular subtypes. Su et al.32, on the basis of a breast cancer radiomic cohort of 860 patients, have successfully built a radiomic model for discriminating TNBC from other subtypes and further inferring molecular subtypes within TNBC. Additionally, radiomic features reflecting TNBC peritumoral heterogeneity among patients have prognostic value, and their associations with metabolic reprogramming and immune infiltration have been elucidated.
In summary, these results support a promising future of clinically applicable methods for the molecular classification of breast cancer, thus enhancing precision treatment.
Discussion and perspectives
The landscape of breast cancer treatment has been revolutionized by the advent of molecular profiling, which has shifted the therapeutic paradigm from a one-size-fits-all approach to subtype-specific precision therapy. In general, to implement molecular subtyping in clinical applications and improve treatment efficacy for patients with breast cancer, further efforts are necessary in the following areas.
Insights into the molecular nature of subtypes Further investigation of molecular subtypes may reveal novel actionable targets. For instance, given that metabolic reprogramming and immune escape are emerging hallmarks of malignancy33, analyzing the dysregulation of metabolism and the tumor microenvironment of breast cancer has substantial potential for identifying such targets. Gong et al.34 have elucidated the metabolic heterogeneity among TNBCs by classifying them into 3 metabolic pathway-based subtypes. Yang et al.35 have discovered that the combination of ferroptosis inducers with immune checkpoint inhibitors may serve as an innovative treatment strategy for LAR-like TNBCs, by reprogramming the tumor microenvironment.
Digital pathology, radiomics, and AI Both digital pathology and radiomics can acquire high-dimensional information. After modeling, in an ideal scenario, rapid molecular subtyping and further diagnosis may be achieved at lower cost. In developing models that link image features and molecular subtyping, AI is playing an increasingly integral role. Thus, development of AI models suitable in the medical field is urgently needed. Additionally, leveraging clinical information and molecular multi-omics data to enhance model interpretability would increase the credibility of the results and prevent erroneous medical decisions.
Noninvasive molecular subtyping Currently, the precise classification of breast cancer presents challenges because of the methods’ invasiveness, particularly for patients requiring frequent monitoring, such as those with advanced breast cancer receiving neoadjuvant therapy. Noninvasive molecular subtyping is an emerging method that can enhance patient compliance with treatment and decrease patient discomfort. In addition to the use of radiomics as described above, the successful application of liquid biopsies in predicting molecular subtyping in non-Hodgkin’s lymphoma36 suggests this method’s application potential in breast cancer.
Prospective clinical trials and translational studies Molecular subtyping of breast cancer can reveal potential biological events for clinical targeting, and may support the development of novel medications for currently undruggable targets. The value of the targets must be validated in prospective clinical trials. Translational studies based on clinical trials can further aid in the development of precise treatments and suggest new therapeutic methods to address drug resistance in patients receiving traditional one-size-fits-all therapies. In conclusion, a continuing need exists for a cycle of molecular subtyping in the laboratory—clinical trials or translational research—further refinement of subtyping a cycle, from molecular subtyping in the laboratory, followed by clinical trails or translational research, and then back to the laboratory for refinement of subtyping. In this way, the continually evolving molecular subtyping of breast cancer could better guide precision treatment for patients.
Conflict of interest statement
No potential conflicts of interest are disclosed.
Author contributions
Conceived and designed the analysis: Yi-Zhou Jiang and Zhi-Ming Shao.
Wrote the paper: Rui Shan and Lei-Jie Dai.
Footnotes
↵*These authors contributed equally to this work.
- Received June 17, 2024.
- Accepted August 9, 2024.
- Copyright: © 2024 The Authors
This work is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.