Synthesis: Single-cell heterogeneity and clonal evolution in TNBC

Evidence grades (GRADE-adapted): A high — multiple well-conducted RCTs or systematic reviews converge. B moderate — single pivotal RCT or consistent observational evidence. C limited — single observational study, mechanistic, or expert consensus. D preclinical / hypothesis-generating.

Why single-cell methods matter for TNBC

Bulk-tumor transcriptomic profiling produces a single average expression profile per sample. This average can be misleading when a tumor contains multiple cell populations with distinct biology — a basal-like tumor by bulk PAM50 may contain luminal-expressing minority populations; an immune-hot tumor by bulk TIL scoring may have spatially organized immune-rich and immune-cold regions with different therapeutic implications.

Single-cell RNA sequencing (scRNA-seq) profiles individual cells (typically 1,000–50,000 cells per sample) and reveals the underlying heterogeneity that bulk averages over. In TNBC, scRNA-seq has documented:

• Intratumoral subclonal populations with distinct transcriptomic signatures
• Stem-like and proliferative subpopulations that may seed recurrence and metastasis
• Immune microenvironment composition at cellular resolution
• Treatment-induced shifts in cell-population balance under chemotherapy and IO
• Biology of paired primary and metastatic samples revealing clonal evolution

The Karaayvaz 2018 single-cell TNBC study

Karaayvaz and colleagues, in one of the earliest comprehensive scRNA-seq studies in TNBC, profiled 1,189 cells from 6 primary TNBC tumors^[1]B. Key findings:

• Substantial intratumoral heterogeneity in expression of major signaling pathways
• Distinct cellular subpopulations with different molecular subtype features within single tumors (e.g., basal-like and mesenchymal cells coexisting)
• Cancer stem cell-like populations defined by ALDH1A3, CD44, and stemness signatures
• Subclones with metastasis-associated transcriptomic features even in primary tumors not yet metastatic

This study established that the molecular subtype heterogeneity captured by bulk frameworks may underestimate the actual cellular diversity within individual tumors.

Treatment-induced clonal evolution

Multiple groups have profiled paired biopsies before and after chemotherapy (or chemo + IO) to characterize how cell populations evolve under treatment. Common findings:

• Selection of treatment-resistant subpopulations. Post-treatment tumors often show enrichment for stem-like, slowly proliferating, or DNA-damage-resistant cell populations that were minor components of the pre-treatment tumor.
• Shifts in immune-cell composition. Pembrolizumab + chemo treatment can reduce regulatory T cells, increase CD8 effector populations, and alter macrophage polarization — though responses vary substantially across patients.
• Increased EMT signatures in residual disease, with mesenchymal-like cell populations enriched.
• Stem-cell-like population enrichment, consistent with the "cancer stem cell" model where the small fraction of stem-like cells resistant to standard therapy seeds recurrence.

The clinical implication: monotherapy that effectively eliminates the proliferating cell population may not eliminate the slowly dividing stem-like population that drives long-term recurrence. Strategies targeting these resistant subpopulations (cancer stem cell-targeting agents, DNA-damage-response inhibitors that may be more active against quiescent cells) are under investigation.

Single-cell findings on the BRCA / HRD axis

Single-cell studies of BRCA1-mutated TNBC have characterized the spectrum of HR-deficient vs HR-competent subclones within tumors. Even tumors with biallelic BRCA1 loss can harbor minor subpopulations with partial HR competence that may explain initial chemo or PARP-inhibitor resistance. Reciprocally, tumors with intact BRCA1 may have minor HR-deficient subpopulations that produce transient PARP-inhibitor responses^[2]C.

This subclonal heterogeneity may explain why bulk BRCA mutation testing doesn't perfectly predict PARP inhibitor response — some BRCA-positive tumors respond minimally; some BRCA-negative tumors respond significantly. Functional HR assessment at single-cell or spatial-clonal resolution could refine prediction.

Spatial transcriptomics

Spatial transcriptomics methods (10x Genomics Visium, Nanostring GeoMx, GeoMx Digital Spatial Profiler, CODEX multiplexed imaging) preserve spatial information while measuring expression. In TNBC, spatial methods have documented:

• Tumor-immune microenvironment spatial organization. "Immune-inflamed" patterns (T cells contacting tumor cells throughout), "immune-excluded" patterns (T cells confined to tumor stroma, not within tumor nests), and "immune-cold" patterns (no T-cell infiltration). The patterns have different predictive associations with IO response.
• Tertiary lymphoid structures. Spatial methods identify and characterize organized immune-cell aggregates that resemble lymph node structure. TLS presence in TNBC predicts better IO response.
• Hypoxic vs normoxic regions. Tumor regions with hypoxia signatures have distinct biology, including increased treatment resistance and altered immune-microenvironment composition.
• Metabolic heterogeneity. Glycolytic vs oxidative metabolism gradients within tumors correlate with cellular phenotype and treatment response.

Primary-metastatic clonal evolution

Paired primary and metastatic samples profiled by single-cell or single-cell-derived methods reveal:

• Metastatic samples are usually clonally narrower than primary tumors. Metastasis selects for specific subclones that successfully completed the metastatic cascade; metastatic samples often contain a single or a few dominant clones rather than the broad clonal diversity of the primary.
• Metastatic clones often pre-existed in the primary as minor populations. The metastasis-seeding clone is sometimes detectable as a small subpopulation in the primary tumor; this raises the possibility of identifying high-metastatic-risk primary tumors by detecting metastasis-seeding signatures.
• Treatment between primary and metastasis further shapes the clonal landscape. Chemotherapy, IO, and targeted therapy each select for resistant subclones, producing metastatic samples with treatment-history-shaped genetic and transcriptomic features.

Implications for biomarker development

Single-cell findings have several implications for how clinical biomarkers should be developed:

• Bulk-based biomarkers may underestimate response heterogeneity. If a tumor contains a small treatment-resistant subpopulation, bulk biomarker assays may classify the tumor as treatment-sensitive but the patient will progress because the resistant clone expands.
• Functional readouts may complement molecular markers. Drug sensitivity testing on tumor-derived organoids or patient-derived xenografts integrates the actual functional response of the cell populations present, potentially capturing what static molecular markers miss.
• ctDNA tracking with serial sampling. Liquid biopsy can detect emerging resistant clones earlier than tissue biopsy. ctDNA-based detection of BRCA reversion mutations, acquired resistance mutations, and emerging subclones is being deployed clinically.
• Spatial-aware biomarkers. TILs Working Group scoring captures spatial information (stromal vs intratumoral); future biomarkers may incorporate more sophisticated spatial features (tertiary lymphoid structures, immune-inflamed-vs-excluded patterns) as routinely measurable.

Evidence table

Open questions and active investigation

• Clinical translation of single-cell findings. Single-cell methods are research-grade; routine clinical deployment requires lower cost, faster turnaround, and standardized analysis pipelines. Whether single-cell biomarkers will enter routine clinical use over the next 5 years depends on these operational considerations.
• Cancer stem cell targeting. If stem-like subpopulations drive recurrence, drugs targeting stem-cell biology (notch pathway inhibitors, hedgehog inhibitors, BMI1 inhibitors) might augment standard therapy. Multiple trials testing in TNBC.
• Functional vs molecular biomarkers. Tumor-derived organoid drug sensitivity testing and patient-derived xenograft assays provide functional readouts. Whether functional biomarkers complement or replace molecular markers for treatment selection is being investigated.
• Single-cell ctDNA / serial liquid-biopsy monitoring. Could resistant subclones be detected earlier via single-cell or high-sensitivity ctDNA monitoring? Whether early detection enables proactive treatment switching that improves outcomes is being tested.
• Spatial heterogeneity-aware treatment selection. If a tumor has immune-rich and immune-cold regions, should the treatment be selected for one or the other? Could intratumoral heterogeneity inform sequential or combination therapy choices?
• Integration with machine learning. Single-cell datasets are large and complex; machine learning approaches that integrate scRNA-seq, spatial, and clinical data to predict treatment response are being developed (see ML multimodal fusion synthesis when available).

For the molecular subtyping frameworks that single-cell methods complicate, see the intrinsic subtypes synthesis, the Lehmann/Pietenpol synthesis, and the Burstein synthesis. For TIL biology relevant to immune microenvironment heterogeneity, see the TME and TILs synthesis.

References

Each citation links to the original publication via DOI. The same records are searchable in the evidence library by title or DOI.

1. Karaayvaz M, Cristea S, Gillespie SM, et al. Unravelling subclonal heterogeneity and aggressive disease states in TNBC through single-cell RNA-seq. Nat Commun. 2018;9(1):3588. doi:10.1038/s41467-018-06052-0. ↩
2. Bareche Y, Venet D, Ignatiadis M, et al. Unravelling triple-negative breast cancer molecular heterogeneity using an integrative multi-omic analysis. Ann Oncol. 2018;29(4):895–902. doi:10.1093/annonc/mdy024. ↩
3. Wu SZ, Al-Eryani G, Roden DL, et al. A single-cell and spatially resolved atlas of human breast cancers. Nat Genet. 2021;53(9):1334–1347. doi:10.1038/s41588-021-00911-1. ↩
4. Bassez A, Vos H, Van Dyck L, et al. A single-cell map of intratumoral changes during anti-PD1 treatment of patients with breast cancer. Nat Med. 2021;27(5):820–832. doi:10.1038/s41591-021-01323-8. ↩

Last reviewed: 2026-06-04. Researcher-layer synthesis page. Evidence grades follow the GRADE-adapted rubric defined at the top of this page.