Tumor ecology and the paradox of clean margins in multicentric breast cancer: an exploratory organoid-based functional study

Abstract

The gene-centric view of cancer has provided important mechanistic insights, but does not fully account for the clinical heterogeneity observed in complex cases. An eco-evolutionary framework instead conceptualizes tumors as dynamic systems shaped by interactions among malignant clones, stromal components, immune populations and the surrounding microenvironment. Multicentric/multifocal breast cancer (MMBC) provides a relevant context in which to explore this biological complexity. Using patient-derived organoids, we observed that organoids derived from histologically tumor-free margins can display stem-like and basal-like features under defined culture conditions, whereas those from primary tumor regions may exhibit more differentiated phenotypes. These observations suggest that histological margin status alone may not fully capture the functional heterogeneity of peritumoral tissue. Our findings support the hypothesis that biological influences within the margin region may contribute to shaping cellular phenotypes beyond the presence of overt malignant cells. Overall, this study highlights the potential value of integrating organoid-based functional approaches with eco-evolutionary concepts to further investigate the biological landscape of the peritumoral field in MMBC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-026-02301-z.

Introduction

Over the last several decades, the study of cancer has been dominated by a gene-centric paradigm. Catalogs of driver mutations, oncogenic signaling pathways, and molecular subtypes have advanced our understanding of tumor biology and inspired the development of targeted therapies and precision medicine approaches [1]. These molecularly guided strategies have extended survival in many cancer types, including breast cancer [2].

Despite these achievements, major clinical challenges remain unresolved. Local recurrence, therapeutic resistance and heterogeneous treatment responses still undermine durable cures, and such phenomena are not fully explained by genetic alterations alone. Increasingly, tumors are being recognized not merely as clonal proliferations of mutated cells [3] but as evolving ecosystems composed of malignant cells, stromal fibroblasts, immune infiltrates, endothelial cells, nerve fibers and extracellular matrix. Within this eco-evolutionary framework, cancer progression reflects not only the intrinsic properties of malignant clones but also the dynamic interactions between cellular populations and their microenvironment [4, 5]. It further highlights that spatially separated tissue compartments may possess distinct ecological dynamics, contributing to unpredictable treatment outcomes.

This perspective aligns tumor behavior with principles observed in natural ecosystems. Competition, cooperation, niche construction and predator–prey dynamics govern cellular interactions, while therapies act as perturbations that reshape selective pressures [6]. This lens helps explain why initially effective treatments often fail: the tumor ecosystem reorganizes, favoring resistant phenotypes or supportive niches [7]. Importantly, ecological and evolutionary dynamics unfold on clinically relevant timescales, meaning that therapeutic perturbations can rapidly alter tumor composition [8]. Understanding these dynamics may uncover functional vulnerabilities not apparent from static genetic or histological analysis.

Multicentric/multifocal breast cancer (MMBC), which accounts for 10–25% of breast cancer presentations [9], provides a clinically relevant setting in which an eco-evolutionary view may be particularly informative. Unlike unifocal breast cancer (UFBC), which is limited to a single tumor mass, MMBC involves the presence of two or more distinct foci within the same breast. Multicentric disease refers to spatially separated foci in different quadrants, whereas multifocal disease refers to multiple foci within the same quadrant. However, since these definitions do not follow any internal anatomical boundary, the term MMBC usually denote any ipsilateral, synchronous tumors presenting with separate lesions. Historically, MMBC was considered a contraindication to breast-conserving therapy because of concerns about local recurrence [10]. Although subsequent studies suggest that breast conservation with radiation can be feasible in selected patients, outcomes remain inconsistent [11, 12]. Some series report worse local control and survival compared to UFBC, while others find no significant differences when adjuvant therapy is optimized [13]. These discrepancies may reflect limitations of both current staging and margin assessment in capturing MMBC complexity. To note, TNM classification typically records only the largest lesion, ignoring heterogeneity among co-existing foci [14, 15]. Regarding margin assessment, surgical success is defined by “negative margins,” conventionally considered as the absence of malignant cells at the inked resection edge. This metric implicitly assumes that tissue beyond the tumor border is biologically inert. However, the biological behavior of histologically normal tissue adjacent to these tumors remains poorly understood. The concept of “field cancerization” describes the presence of molecular or functional alterations in histologically normal tissue surrounding tumors, which may predispose to tumor initiation or progression [16, 17]. Functional heterogeneity within the peritumoral field may therefore represent an overlooked source of recurrence risk, especially in MMBC.

To address these limitations, we used patient-derived organoids as a functional platform to interrogate margin biology [18]. While organoid models do not recapitulate the full complexity of the tumor microenvironment, they provide a tractable system to explore self-renewal, differentiation bias and collective behavior ex vivo. In doing so, we observed distinct functional patterns between disease subtypes. In UFBC, tumor-derived organoids consistently exhibited basal-like and sustained growth characteristics, whereas non-tumor-derived organoids were largely more differentiated and showed limited long-term proliferative capacity. In MMBC, however, organoids derived from histologically tumor-free margins displayed stem-like and basal-type features in vitro, while tumor-derived organoids more frequently exhibited differentiated phenotypes and reduced long-term growth capacity. These observations raise the possibility that functional properties of peritumoral tissue may differ between UFBC and MMBC. While this study is exploratory and hypothesis-generating in nature, it provides a basis for further investigation into the biological heterogeneity of tumor-adjacent tissues in MMBC.

Results

Organoid biobank development and patient cohort

We established a biobank of patient-derived breast cancer organoids in collaboration with Hospital La Paz (Madrid). In this exploratory study, 12 samples were collected from both tumor regions and adjacent non-tumor tissue beyond resection margins, in patients who underwent either upfront surgery or surgery after neoadjuvant therapy, for UFBC or MMBC (Fig. 1a–c). The cohort was heterogeneous, including hormone receptor–positive, HER2-positive, and triple-negative tumors across a range of histologic grades (Fig. 1c). This diversity allowed us to examine functional patterns across different molecular subtypes and histological contexts.

Fig. 1.

Long-term lifespan of patient-derived breast tumor organoids is not associated with tumor subtype or histological classification. a Schematic overview of the workflow used for organoid-based functional analyses. Following surgical resection, small fragments of tumor and adjacent non-tumor tissue were immediately processed to generate organoids. Distinct organoid morphology was observed as early as 2–5 days after seeding. b Representative bright-field images of cystic and basal organoids, along with the most commonly observed migratory behaviors of patient-derived organoids in culture (illustrative examples from independent biological replicates, expanded in Supplementary Fig. 3; n = 12). c Clinicopathological characteristics of 12 breast tumor samples used for organoid generation, including molecular subtype (luminal A, luminal B, HER2-positive, or triple-negative breast cancer [TNBC]), histological type (invasive ductal carcinoma [IDC], invasive lobular carcinoma [ILC], ductal carcinoma in situ [DCIS], or adenoid cystic carcinoma [ACC]), tumor grade, receptor expression, proliferation rate (KI67 staining), axillary node status, age at diagnosis, receipt of neoadjuvant treatment (if any), type of surgery, tumor focality, and cell culture lifespan of tumor-derived organoids (days in culture) [1, 2]. For two cases, independent sets of samples were obtained from the same patient presenting a unifocal tumor in one breast and a multicentric tumor in the contralateral breast. d–g Duration of organoid lifespan for each patient-derived line shown in (c), illustrating no correlation with molecular subtype (d), histological type (e), proliferation rate (f), or tumor grade (g); n = 12 biological replicates

Organoid cultures were successfully established from surgical samples, both tumor and non-tumor compartments, in > 95% of cases, reflecting robust tissue viability and reproducibility of the method. These organoids recapitulated key features of epithelial organization and lineage bias observed in the original tissues (Supplementary Fig. 1 and 2a), and selectively enriched for tumor cells, excluding stromal and immune components (Supplementary Fig. 2b), therefore enabling precise interrogation of tumor-intrinsic capacities for self-organization, self-renewal, migration and adaptability under controlled ex vivo conditions (Supplementary Fig. 2c). Importantly, these stem-like and EMT signatures, while prominent in the organoid transcriptomic data, were not detectable in the corresponding primary tissue sections (Supplementary Fig. 2c). This discrepancy underscores a significant capability of functional organoid models: they may serve to unmask or amplify latent molecular traits that remain below the detection threshold of standard bulk tissue analysis. It should be noted that these phenotypes may be at least partially driven or selectively enriched by the stem cell-supportive factors present in the organoid culture media, which provide an environment conducive to the expansion of these specific cellular subpopulations.

Of interest, the long-term functional lifespan of organoid cultures varied considerably among patient samples, independently of canonical tumor subtype, histological type, proliferation rate or tumor grade (Fig. 1d–g). This observation supports the view that plastic potential is not strictly tied to conventional classifications, but rather reflects an intrinsic adaptability trait of tumor cells. Although the limited cohort size precludes definitive conclusions regarding clinical predictors, these findings align with previous observations that breast cancer cells can undergo flexible subtype transitions in response to environmental pressures [19].

Differential organoid phenotypes in unifocal breast cancer (UFBC)

In UFBC cases, tumor-derived organoids consistently displayed basal-like phenotypes (Fig. 2 and Supplementary Fig. 3a). Collective migration was commonly observed, with cohesive clusters extending across culture surfaces (Fig. 2a, b, d and Supplementary Fig. 3a). Morphologically, they formed large grape-shaped structures and, at the molecular level, showed high expression of CK14, CD44 and SOX10, and low expression of luminal markers such as CK8 and CD24 (Fig. 2b, d). These features are consistent with prior reports linking basal-like signatures to enhanced plasticity and invasive potential in vitro [20, 21]. Together, these observations confirm that UFBC tumor-derived organoids retain expected functional characteristics of aggressive basal-like epithelial cells.

Fig. 2.

Clinicopathological data and organoid-based functional assessment of breast cancer and adjacent non-tumor tissues. a Summary of clinicopathological characteristics (molecular subtype, histologic type, and tumor grade), surgery type, tumor focality, histopathological diagnosis of every tissue sample analyzed, and their corresponding organoid culture outcomes derived from breast tumor and matched adjacent non-tumor tissues. The table includes organoid lifespan (days in culture), morphology, and invasiveness, as well as the functional interpretation of organoid behavior relative to histologically clean surgical margins. A healthy tissue control, derived from a prophylactic mastectomy, is included for comparison. Functional classifications are based on qualitative and semi-quantitative phenotypic assessments. b Representative confocal immunofluorescence images of tumor UFBC-derived organoids. UFBC non-tumor derived organoids were exhausted and therefore not included in further analyses. CK14 (green) and CK8 (red); CD44 (green) and CD24 (red); and SOX10 (white) are shown. Nuclei are counterstained with DAPI. Scale bar, 100 μm (n = 6 independent biological replicates). c Representative confocal immunofluorescence images of tumor and non-tumor MMBC-derived organoids. CK14 (green) and CK8 (red); CD44 (green) and CD24 (red); and SOX10 (white) are shown. Nuclei are counterstained with DAPI. Scale bar, 100 μm (n = 6 independent biological replicates). All images were captured with identical acquisition settings to allow direct comparison between samples. d Violin plots showing SOX10 nuclear positivity (proportion of DAPI-positive nuclei) and CK8/CK14, CD24/CD44 relative fluorescence intensity per area, quantified from immunofluorescence images. **p < 0.01; ****p < 0.0001; ns, not significant (one-way ANOVA). The number of imaging planes evaluated per condition (derived from the analyses of 6 independent biological replicates) is indicated in the figure panel. Effect size (Cohen’s d) was also calculated to further assess biological relevance (d < 0.2 = small/not relevant; 0.2–0.5 = moderate; >0.5 = large)

By contrast, organoids derived from non-tumor tissue in UFBC patients exhibited luminal phenotypes (Fig. 2 and Supplementary Fig. 3b). These cultures frequently exhausted after few passages, reflecting reduced stemness and self-renewal capacity. They typically formed cystic structures with very limited proliferative potential (Fig. 2b, d and Supplementary Fig. 3b). This dichotomy—basal-like functional traits confined to tumor-derived organoids and more differentiated phenotypes in margin-derived organoids—was expected and aligned with conventional oncologic assumptions. It also underscores that, in UFBC, functional assays largely corroborate histological observations.

Inverted organoid phenotypes in multicentric/multifocal breast cancer (MMBC)

In MMBC cases, this pattern was unexpectedly inverted. Tumor-derived organoid cultures often showed early exhaustion, suggesting impaired stem-like capacity (Fig. 2 and Supplementary Fig. 3c). These organoids displayed luminal-like characteristics, including cystic morphology (Fig. 2c, d and Supplementary Fig. 3c), low CK14 versus high CK8 expression, low CD44 versus high CD24 expression and cytoplasmic (rather than nuclear) SOX10 localization (Fig. 2c, d). This functional exhaustion occurred despite the tumor tissue appearing histologically active, indicating a disconnect between histology and functional potential.

In contrast, organoids established from histologically normal, margin-derived tissue exhibited basal-like and stem-associated functional features (Fig. 2 and Supplementary Fig. 3d). These margin-derived organoids demonstrated robust self-renewal over multiple passages. They were morphologically grape-shaped (Fig. 2c, d), expressed high CK14 and low CK8, showed nuclear SOX10, and harbored a CD44+/CD24– stem-like profile (Fig. 2c, d). Collective migration, rarely observed in luminal organoids or healthy tissue-derived cultures (Supplementary Fig. 3), was a dominant feature in MMBC margin-derived organoids (Fig. 2 and Supplementary Fig. 3d), which generated migratory structures with lamellipodia-rich leading edges. Although these features do not directly establish clinical aggressiveness, they suggest that functional potential in MMBC margins may be independent of tumor mass and influenced by local ecological conditions.

Traditional views may attribute this behavior to fixed, patient-specific characteristics. However, our data suggest that such plasticity may emerge from local tissue context and prior ecological perturbations. In a representative patient presenting a unifocal tumor in one breast and multifocal tumors in the contralateral breast, organoid phenotypes mirrored the patterns observed in tumor- and adjacent non-tumor-derived samples when analyzed by functional organoid assays (Fig. 3). These findings suggest that the tumor and its surrounding tissue may jointly shape the capacity of the system to adapt and reorganize under selective pressures, consistent with a field effect rather than fixed clonal properties alone. Overall, these observations support the concept of the peritumoral ecosystem as a functional entity, with potential implications for recurrence and therapy response.

Fig. 3.

Tumor cell plasticity and local dissemination are consistent with intrinsic properties rather than patient-specific effects. a Schematic of mammary gland tissue samples from a representative patient, with a unifocal tumor in the left breast and multiple tumor foci in the contralateral breast. b Representative bright-field images of organoids from tumor (T1, T2) and adjacent non-tumor regions (D1–D3) of the right (multifocal) breast after 5 days in culture. Scale bar: 100 μm. c Representative bright-field images of organoids from tumor (T1, T2) and adjacent non-tumor regions (D1, D2) of the left (unifocal) breast after 5 days in culture. Scale bar: 100 μm. d, e Representative confocal immunofluorescence images of non-tumor organoids (multifocal, right breast, d) and tumor organoids (unifocal, left breast, e). CD44 (green), CD24 (red); CK14 (green), CK8 (red) are shown. Nuclei are stained with DAPI. Scale bar: 100 μm. f Violin plots showing immunofluorescence quantifications from paired non-tumor (multifocal, right breast) and tumor (unifocal, left breast) organoids from the same patient. SOX10 nuclear positivity was measured as proportion of DAPI-positive nuclei, while CD44, CD24, CK14 and CK8 were measured as relative fluorescence intensity per area. *p < 0.05; ****p < 0.0001; ns, not significant (Student’s t-test). The number of imaging planes analyzed per condition is indicated in the figure panel. Effect size (Cohen’s d) was also calculated to further assess biological relevance (d < 0.2 = small/not relevant; 0.2–0.5 = moderate; >0.5 = large). g–j Violin plots showing organoid lifespan (g), organoid size (h), number of organoids (i), and percentage of organoids exhibiting collective migration (j) at day 20, for the four conditions tested. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant (two-way ANOVA)

Discussion

Our findings suggest distinct functional differences in organoid phenotypes between multicentric/multifocal breast cancer (MMBC) and unifocal breast cancer (UFBC). In UFBC-derived organoid cultures, tumor-derived organoids more frequently exhibited basal-like features, whereas margin-derived organoids tended to show reduced proliferative capacity in vitro. In contrast, MMBC-derived organoids displayed a different distribution of functional states, with tumor-derived organoids more often showing differentiated phenotypes and limited long-term growth, while margin-derived organoids more frequently exhibited stem-like and basal-like characteristics. Organoids were derived from surgical specimens obtained either after neoadjuvant therapy or from upfront resections; therefore, these observations may reflect a combination of intrinsic tumor properties and treatment-associated biological adaptation, although this cannot be formally dissected within the current study.

Overall, these results highlight potential differences in functional phenotypes between tumor- and margin-derived tissues in MMBC compared with UFBC. While these observations remain exploratory, they provide a basis for further investigation into their generalizability and underlying mechanisms and are intended to stimulate sustained discussion in the field.

The paradox of clean margins

Margin assessment is an established component of surgical pathology in breast cancer, typically based on histological evaluation of tumor involvement at resection borders. This approach assumes that histologically tumor-free tissue is functionally homogeneous; however, emerging molecular and experimental studies have suggested that adjacent peritumoral tissue may harbor significant biological heterogeneity [22, 23].

In this study, functional assays using patient-derived organoids demonstrate that margin-derived cells from MMBC can exhibit stem-like and migratory features in vitro, even when appearing histologically normal in the original surgical specimens. These observations align with the concept of a “field effect”, where non-malignant-appearing tissue displays molecular or functional alterations compared with distant normal tissue [22].

Our findings suggest that peritumoral tissue in MMBC may harbor distinct functional states that are not captured by standard histopathological evaluation. In this context, we propose functional organoid-based assays as a potential complementary research tool to uncover these ‘invisible’ biological features, rather than a replacement for current gold-standard diagnostic methods.

However, it is critical to emphasize that our study was not designed to assess clinical outcomes or recurrence; therefore, no causal relationships can be inferred. The functional behaviors observed should be interpreted as ex vivo properties of isolated cell populations under defined conditions, rather than direct predictors of in vivo tumor behavior. Factors such as patient heterogeneity, prior treatments and the complex microenvironmental context must be integrated into future large-scale analyses, to fully elucidate the clinical relevance of these findings.

Therapy as perturbation with eco-evolutionary consequences

An eco-evolutionary framework may provide a conceptual context to interpret how tumor and margin-associated tissues respond to environmental and therapeutic perturbations. Systemic therapies, including chemotherapy and radiation, affect not only malignant cell populations but also surrounding non-malignant compartments and the tumor microenvironment.

In this context, repeated perturbations such as hypoxia, extracellular matrix remodeling and immune modulation may influence the functional states of cells within the peritumoral field. Rather than being static, these cellular populations may exhibit phenotypic plasticity in response to changing microenvironmental conditions.

This interpretation is consistent with evolutionary models in which selection pressures can shape the distribution of cellular phenotypes within heterogeneous tissues [24]. Overall, the peritumoral ecosystem can be considered dynamically responsive to both intrinsic and extrinsic influences, including therapy-associated changes, but this remains a conceptual framework requiring experimental validation.

Future studies using approaches such as single-cell lineage tracing or multi-omics profiling may help determine whether stem-like traits in margin-associated cells are pre-existing, selected or induced under specific conditions.

Clinical heterogeneity in MMBC

The clinical literature on MMBC reports heterogeneous findings regarding outcomes compared with unifocal disease, with studies describing variable differences in recurrence and prognosis across cohorts [25, 26]. It is important to note that in the modern era, clinical management—supported by prospective data such as the ACOSOG Z0011 trial [27] and advanced MRI-based mapping [28]—has enabled breast-conserving approaches in appropriately selected patients, and improved the delineation of disease extent and surgical planning. Within this clinical landscape, our results provide a potential biological framework that may contribute to understanding the remaining heterogeneity, by highlighting functional differences between tumor-derived and margin-derived organoids in MMBC. In particular, peritumoral tissue may represent a biologically heterogeneous compartment that is not fully represented by analyses focused exclusively on the dominant tumor lesion.

Within this context, functional characteristics of margin-associated cells may represent one of several biological variables contributing to tumor heterogeneity in MMBC. While our study does not establish a new clinical need for margin assessment or allow inference regarding prognostic value yet, future studies may resolve whether functional phenotypes observed in organoid models are associated with clinical outcomes such as disease recurrence or therapeutic response, in appropriately designed prospective cohorts.

Organoids as functional research tools

Histopathology remains a fundamental component of breast cancer diagnosis, providing essential morphological and architectural information. However, it is inherently static and does not capture functional properties of living cells. Although organoid models do not recapitulate the full tumor microenvironment, including immune and systemic influences, they provide a dynamic experimental system to assess epithelial cell behaviors such as self-renewal, differentiation bias and migratory capacity.

In our study, margin-derived organoids exhibited distinct functional characteristics within a short time frame after culture establishment, which were not evident from histological evaluation of the corresponding tissue. These observations suggest that organoid-based assays may provide complementary functional information to conventional histopathology in experimental research settings.

Future work will be required to determine the extent to which organoid-derived readouts reflect clinically relevant biological variation and to explore their utility as research tools for hypothesis generation in breast cancer biology.

Future perspectives

Collectively, these insights suggest several avenues for future biological investigation and hypothesis generation (overview in Fig. 4):

Fig. 4.

Conceptual framework derived from this study. Schematic representation summarizing the conceptual framework emerging from our observations. Key elements include: (i) consideration of intra-tumor and peritumoral heterogeneity beyond the dominant lesion; (ii) integration of molecular and functional (organoid-based) approaches to complement histopathological assessment; (iii) application of eco-evolutionary perspectives to explore tumor–microenvironment interactions; (iv) conceptual exploration of cellular plasticity and state modulation as areas for future research; and (v) the importance of interdisciplinary dialogue across experimental and clinical domains. This figure is intended as a conceptual framework to support future hypothesis-driven research, rather than to imply direct clinical application

Comprehensive lesion characterization:

Pathological assessment in MMBC typically focuses on the dominant lesion; however, this approach may not fully capture the heterogeneity of the disease. A more systematic evaluation of multiple lesions, together with adjacent peritumoral tissue, may provide a more comprehensive biological description of disease complexity and support future research into intra-patient heterogeneity.

Margin biology as a research question:

Histological criteria alone define margin status in current clinical practice, but this definition does not account for potential functional heterogeneity. Future studies could investigate whether molecular and functional characteristics of margin-associated cells, including stemness- or differentiation-related markers, vary across MMBC cases and how these relate to underlying biological diversity. Additional studies may include multi-omics approaches, such as single-cell and spatial transcriptomic analyses, to further resolve the cellular heterogeneity within margin-derived cultures and help assess their relationship to the corresponding tissues of origin.

Organoids as functional complements to histopathology:

Histopathology remains the cornerstone of breast cancer diagnosis and staging, providing essential morphological information. As a complement to also capture dynamic cellular behaviors, here we propose that patient-derived organoids may offer an experimental platform for studying epithelial cell plasticity in controlled ex vivo conditions. Within this framework, organoid models may be useful for generating hypotheses about functional heterogeneity in tumor and margin-derived tissues. Integration with molecular profiling could help characterize biological variability across samples, although any relationship with clinical outcomes remains to be established.

Eco-evolutionary modeling:

Applying mathematical and evolutionary models to tumor–stroma dynamics could provide a theoretical framework to explore how peritumoral tissue acquires stem-like traits. Such models may identify eco-evolutionary “tipping points” in silico, generating testable hypotheses. Integration of functional organoid data with these models could enable patient-specific simulations of tumor–margin interactions and predict potential recurrence hotspots.

Exploration of cellular plasticity:

Beyond cytotoxic approaches, the modulation of cellular plasticity and phenotypic states represents an area of growing conceptual interest in cancer biology [21, 29]. In this setting, the ability of tumor- and margin-derived cells to adopt either stem-like or more differentiated states may reflect underlying biological plasticity within the peritumoral ecosystem. Accordingly, approaches aimed at modulating differentiation or cellular identity may provide a useful framework for future research. Organoid-based platforms offer an opportunity to explore how different perturbations influence cellular phenotypes and state transitions, thereby supporting hypothesis generation regarding the role of plasticity in tumor and margin biology.

Prospective clinical trials:

Future prospective studies will be required to determine whether molecular and functional features of margin-associated tissues are associated with clinical outcomes, including recurrence or treatment response in MMBC. Such studies should also assess the reproducibility, robustness, and biological relevance of organoid-based functional assays across larger and independent patient cohorts, encompassing diverse molecular subtypes and clinical settings.

Conceptual framing:

This work is intended as a conceptual framework, aimed at stimulating interdisciplinary dialogue. Understanding the complexity of multicentric breast cancer will likely require closer integration of experimental biology, pathology and clinical practice, as each perspective captures different dimensions of disease behavior. Rather than providing definitive conclusions, our findings are intended to highlight a potentially underexplored aspect of tumor biology and to encourage further investigation by the broader research community. In this context, we view progress in the field as an iterative process in which preliminary observations, when rigorously sustained and expanded across independent studies, may contribute to refining current models and uncovering new biological insights.

Methods

Patient cohort and sample collection

Inclusion criteria required participants to be over 18 years of age and to have tumor samples obtained from one of the following: (i) surgeries performed after neoadjuvant therapy showing a partial clinical response, (ii) surgeries or biopsies conducted following early local recurrence, or (iii) primary surgical excisions. Patients with prior contralateral breast cancer were excluded to avoid confounding effects on margin biology.

Because therapeutic interventions can modify tumor and peritumoral biology, the inclusion of neoadjuvant-treated cases could introduce biological heterogeneity and was acknowledged as a potential confounding factor in the interpretation of functional assays. Nonetheless, comparison of organoids derived from upfront surgeries versus post-neoadjuvant surgeries revealed no observable differences in morphology, cell culture lifespan, or basal/luminal phenotypes, suggesting that in this study, such therapeutic perturbations do not substantially impact the reliability of patient-derived organoids as a functional assay platform. This observation supports the robustness of organoid assays in capturing intrinsic functional traits independent of prior therapy.

The use of samples in this study did not interfere with routine pathology assessments. Sex was recorded from clinical records at the time of sample collection. Gender identity, race, ethnicity, and other socially defined categories were not collected, as they were not relevant to the biological objectives of the study. All participants provided written informed consent prior to sample collection. Tumor specimens were collected, processed, and stored in accordance with the study protocol approved by the Clinical Research Ethics Committee of Hospital La Paz, following ICH-GCP guidelines and the principles of the Declaration of Helsinki. All patient identifiers were removed prior to analysis to ensure confidentiality.

Subjects were allocated to the unifocal or multicentric groups based on their pathological diagnosis. No random assignment was performed, as group classification reflected clinical and pathological characteristics rather than experimental allocation. All samples used in this study were histologically verified, and correct sample labelling was confirmed for every case. All tissue samples were reviewed and histologically confirmed by board-certified pathologists prior to organoid derivation. Tumor samples were defined based on histopathological evaluation and contained ≥ 80% tumor cells, while adjacent normal tissues were confirmed to be free of malignant cells by standard H&E examination. In addition, adjacent samples were collected at a sufficient distance from the tumor mass according to routine pathological assessment to ensure they represented non-malignant tissue.

Generation and culture of breast cancer patient-derived organoids

Tumor samples from the previously described patient cohort were processed to establish organoid cultures, following the protocol previously described [21]. No attrition or dropout of participants occurred during the study, as all collected tumor samples were successfully processed according to the study protocol. While organoid cultures represent an epithelial-enriched system and do not retain the full stromal, immune, or vascular components of the tumor microenvironment, they provide a robust functional platform to assess properties such as self-organization and self-renewal that are not readily observable in the tissue of origin using standard histology. Organoids were also monitored for viability and morphological consistency across passages to ensure reproducibility. Briefly, samples were mechanically and enzymatically dissociated at 37 °C under constant agitation, followed by sequential centrifugation to enrich viable cells. The resulting cell suspensions were resuspended in basement membrane extract (BME) and plated as 20 µL domes in pre-warmed 24-well plates. After polymerization, domes were overlaid with breast cancer organoid expansion medium consisting of basal medium supplemented with 20% homemade R-spondin-1 conditioned medium, 1× B27 supplement, 1× N2 supplement, 10 mM nicotinamide, 1.25 mM N-acetylcysteine, 10 nM gastrin I, 100 ng/mL FGF10, 100 ng/mL Noggin, 50 ng/mL EGF, 5 nM neuregulin 1, 5 µM Y-27,632, 500 nM A83-01, 500 nM SB202190, and Primocin. All reagents were sterile-filtered and freshly prepared before use. The medium was refreshed every two days, and organoids were passaged every 7–10 days. Cultures were monitored daily, and bright-field images were captured every two days. In general, basal and luminal organoid phenotypes could be undoubtedly distinguished as early as five days post-seeding. Organoid culture exhaustion (defined as either: (i) culture ceased proliferation and could no longer be passaged, or (ii) persisted morphologically but exhibited exhausted growth with minimal expansion) was monitored and evaluated. Organoid health, size, number and morphology were documented for each passage to allow longitudinal analysis of functional changes. Although intrinsic differences in the starting cellular composition cannot be fully excluded, all samples were cultured under identical experimental conditions, allowing for comparison within a controlled assay framework.

Immunofluorescence staining

Organoids were seeded onto chamber slides, fixed in 10% formalin, permeabilized, and blocked in BSA-containing buffer. Primary antibodies against lineage (KRT14, KRT8), stemness (SOX10, CD24/CD44), proliferation (KI67) or membrane receptors (PR/HER2) markers were applied overnight at 4 °C. Antibody dilutions and incubation times were optimized for 3D organoid structures to ensure penetration and reproducibility. The specifications for primary antibodies are as follows: KRT14 (RRID: AB_2811031): rabbit monoclonal EPR17350 (Abcam, ab181595), dilution 1:100; KRT8 (RRID: AB_531826): mouse monoclonal TROMA-I (DSHB, ab531826), dilution 1:100; SOX10 (RRID: AB_2650603): rabbit monoclonal EPR4007 (Abcam, ab155279), dilution 1:400; CD24-Alexa FluorTM 647 (RRID: AB_10894010): mouse monoclonal ML5 (BD Pharmigen, 561644), dilution 1:20; CD44 (RRID: AB_493686): rat monoclonal IM7 (Biolegend, 103023), dilution 1:20. KI67 (RRID: AB_393778): mouse monoclonal B56 (BD Biosciences, 550609), dilution 1:400; PR (RRID: AB_2797144): rabbit monoclonal D8Q2J (Cell Signalling, 8757T), dilution 1:1000; HER2 (RRID: AB_627013): mouse monoclonal 3B5 (Santacruz, sc-33684), dilution 1:100. After washing, fluorophore-conjugated secondary antibodies were added, followed by DAPI nuclear counterstaining. Confocal imaging was performed on a Leica TCS SP8 microscope with z-stack acquisition (2.5 μm steps). Imaging parameters were kept constant across samples to allow quantitative comparison.

Immunofluorescence analysis and quantification

Immunofluorescence images were analyzed using ImageJ to quantify expression levels of multiple markers in organoids. Progesterone receptor (PR), HER2, and Ki67 were quantified as the proportion of marker-positive cells relative to the total number of cells, with total cell counts determined via nuclear detection using DAPI staining. For each organoid, the entire structure was analyzed, and nuclei were identified based on the DAPI signal. SOX10 expression was quantified similarly, as the proportion of nuclei positive for SOX10 relative to total DAPI-positive nuclei. Nuclear detection was performed for each imaging plane, and analysis was conducted across the full organoid structure. CK14, CK8, CD44 and CD24 expression levels were quantified based on fluorescence intensity per stained area. For each imaging plane, regions positive for each marker were identified, and the mean fluorescence intensity was measured within these regions. Intensity values were normalized to the area of staining to provide relative marker expression. All measurements were performed across entire organoid structures, and individual imaging planes were analyzed to capture spatial heterogeneity in marker expression. Analyses were performed in a blinded manner, and technical duplicates were averaged for each biological replicate.

RNA sequencing and transcriptomic analysis

Total RNA was extracted from tumor tissue and matched tumor-derived organoids from a representative Luminal A patient using Nucleozol™, following the manufacturer’s instructions. RNA extraction was performed under RNase-free conditions, and RNA integrity number (RIN) was confirmed > 9 for all samples. RNA quality was assessed with the Agilent Bioanalyzer 2100, and only high-quality samples (triplicates per condition) were used for library preparation and sequencing, which were performed by Novogene. Messenger RNA was isolated using poly-T oligo-attached magnetic beads, fragmented, and reverse-transcribed into cDNA. Both non-stranded and strand-specific libraries were prepared according to standard Illumina protocols, and library quality was evaluated using Qubit, qPCR, and Bioanalyzer. Sequencing was performed on an Illumina platform using sequencing-by-synthesis technology to generate high-throughput reads. Raw reads underwent quality control with FastQC, and clean reads were aligned to the reference genome using HISAT2. Gene counts were obtained and imported into R for analysis. Low-expression genes (total counts < 10 across all samples) were removed prior to differential expression analysis with DESeq2 [30]. Differentially expressed genes were identified using thresholds of adjusted p ≤ 0.05 and |log₂ fold change| ≥ 1. For gene set enrichment analysis (GSEA), genes were ranked according to the Wald test statistic from DESeq2. All analyses were performed in R v4.3.1 with Bioconductor packages to ensure reproducibility. Custom gene signatures for “Stemness” were obtained from the ARCHS4 database, and the “Epithelial-to-Mesenchymal Transition (EMT)” hallmark gene set was retrieved from MSigDB Hallmark Gene Sets (H, Homo sapiens). GSEA plots were generated using the Broad Institute GSEA software. Enrichment plots display the enrichment score, normalized enrichment score, nominal p-value, and FDR q-value for each gene set. Because each gene set was tested individually rather than as part of a full collection, FDR values are not considered significant. To visualize expression patterns, the most strongly differentially expressed genes were selected based on the largest absolute log₂ fold change and plotted as heatmaps using the pheatmap R package (https://CRAN.R-project.org/package=pheatmap). Expression values correspond to variance-stabilized counts from DESeq2 and were scaled across genes using a z-score transformation (mean-centered and divided by the standard deviation per gene) to highlight relative expression differences between tumor and organoid samples. Heatmaps were generated for the gene sets to show relative expression differences between tumor tissue and tumor-derived organoids. The RNA-seq data in this studio represent an exploratory analysis of a single matched pair, providing supportive transcriptomic evidence for the observed phenotypes rather than definitive population-level conclusions.

PAM50 subtype analysis

PAM50 subtype similarity [31] was assessed using RNA-seq data from the patient tumor (histopathologically defined as Luminal A) and matched tumor-derived organoids. Normalized DESeq2 counts were used to extract PAM50 gene expression values. For each sample, a centroid profile was generated with the PCAPAM50 R package (https://cran.rstudio.com/package=PCAPAM50), and Pearson correlation coefficients were calculated against published PAM50 subtype centroids to quantify subtype association. This allowed us to confirm that organoids retained transcriptional features consistent with the original tumor subtype.

Tumor purity and microenvironment assessment

Tumor purity and stromal/immune content were assessed using RNA-seq data from both tumor and organoid samples with the ESTIMATE algorithm [32], (https://r-forge.r-project.org/projects/estimate/). DESeq2-normalized counts were used as input. ESTIMATE computes three scores: Stromal Score (stromal cell content), Immune Score (immune cell infiltration), and ESTIMATE Score (combined contribution to infer tumor purity). Tumor purity was derived from the ESTIMATE Score as described in the original publication. Scores were calculated for both tumor and organoid samples, and biological triplicates were used to compute mean ± standard deviation values. These analyses confirmed that organoids were highly enriched for epithelial tumor cells and largely depleted of stromal or immune components, supporting their use as a tumor-intrinsic functional platform.

Statistical analysis

All experiments were performed with at least three independent biological replicates. Samples from unifocal and multicentric cases were processed and analyzed in a fully blinded manner: investigators performing organoid experiments and subsequent quantifications were unaware of the patients’ diagnoses and clinical information until all analyses were completed. Group sizes were determined by the number of eligible tumor samples available during the study period; therefore, no formal power calculation was performed. Data are presented as mean ± SD. Statistical comparisons between two groups were performed using two-tailed Student’s t-tests, whereas comparisons involving more than two groups were analyzed using one-way ANOVA. Experiments including multiple variables or conditions were analyzed using two-way ANOVA. Pearson’s correlation coefficient was used to assess associations between continuous variables. Because a large number of organoids were quantified in immunofluorescence analyses, p-values may become statistically significant even for small differences. To better assess the magnitude and biological relevance of these differences, effect sizes were also calculated using Cohen’s d for pairwise comparisons (R package https://cran.rstudio.com/web/packages/effectsize/index.html). Given the limited number of patient samples and the exploratory nature of several comparisons, statistical analyses should be interpreted as hypothesis-generating rather than confirmatory. The study was not powered to establish direct clinical associations. All analyses were conducted using GraphPad Prism (RRID: SCR_002798). Graphs include individual data points to visualize variability, and outliers were assessed but not removed.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

NGM-I, EM-Z and MS-R conducted most of the experiments, with assistance from CQ-G. AP-P collaborated on cancer ecology conceptualization. LF-A assisted in preparing the necessary documents and protocols for patient enrollment at the hospital, recruited patients, and offered clinical advice. LY assisted in preparing samples from Pathology Unit. MS-R and NGM-I jointly contributed to experimental design, data evaluation, formal analysis and investigation. As project leader, MS-R was responsible for the conceptualization, administration of the study, resources and funding acquisition, as well as for supervision and manuscript preparation. The original draft was written by MS-R, with the contribution of the main authors. All authors reviewed and approved the final manuscript.

Funding

The work has been funded by the Spanish Ministry of Science, Innovation and Universities, supported with European Regional Development funds: CNS2022-135364 and PID2022-136508OA-I00 to MS-R. Additional institutional support for infrastructure and consumables was provided by UCM. MS-R was also supported by AECC (INVES18005SALA) and a Ramón y Cajal contract from the Ministry of Science, Innovation and Universities (RYC2020-028929-I). NGM-I and CQ-G were supported by AECC (PRDMA19003GARC and PRDMA258136QUEZ, respectively) and NGM-I also received a Fulbright fellowship (PS00380044).

Data availability

All data generated or analyzed in this study are included in the article. RNAseq data has been deposited in the SRA repository under BioProject accession number PRJNA1442725. Raw data are available from the corresponding author upon reasonable request, except for confidential patient information, which cannot be shared to protect participant privacy.

Declarations

Ethics approval and consent to participate

All procedures were conducted in accordance with the Declaration of Helsinki, national regulations, and ICH-GCP guidelines. This compliance ensures that the study adhered to internationally recognized standards for ethical research involving human subjects. The study protocol, informed consent forms, and patient information materials were approved by the Clinical Research Ethics Committee of Hospital La Paz. All participants provided written informed consent prior to sample collection. Sex was recorded from clinical records; gender identity, race, ethnicity, and other socially defined categories were not collected, as they were not relevant to the biological focus of the study.

Consent for publication

Written informed consent for publication was obtained from all patients. Patients were specifically informed that de-identified data and images derived from their samples could be included in scientific publications and presentations.

Competing interests

The authors declare no competing interests.