Poster Spotlight 2: Molecular Evolution in Early and Late Stage Tumors

M. Antonini¹, A. Mattar², F. P. Cavalcante³, F. Zerwes⁴, E. C. Millen⁵, F. P. Brenelli⁶, L. A. Frasson⁷, L. X. Félix⁸, J. C. Vieira⁸, R. C. Campos⁸, M. Teixeira², M. Figueiredo², M. Madeira⁹, H. C. Lima¹⁰, G. Facina¹¹, R. Arakelian¹², A. D. Lima¹³, L. R. Soares¹⁴, R. J. Freitas¹⁵, G. N. Garcia¹⁶, F. Bagnoli¹⁷, G. T. Tosello¹⁸, L. H. Gebrim¹⁹; ¹Mastology, Hospital do Servidor Publico Estadual, Sao Paulo, SP, BRAZIL, ²Mastology, Hospital da Mulher – SP, Sao Paulo, BRAZIL, ³Mastology, Hospital Geral de Fortaleza, Fortaleza, BRAZIL, ⁴Mastology, Pontifícia Universidade Católica do Rio Grande do Sul, Porto Alegre, BRAZIL, ⁵Mastology, Americas Oncologia, Rio de Janeiro, BRAZIL, ⁶Mastology, Universidade Estadual de Campinas, Campinas, BRAZIL, ⁷Mastology, Hospital Israelita Albert Einstein, Sao Paulo, BRAZIL, ⁸Mastology, Hospital do Servidor Publico Estadual, Sao Paulo, BRAZIL, ⁹Mastology, Faculdade Israelita de Ciências da Saúde Albert Einstein, Sao Paulo, BRAZIL, ¹⁰Mastology, Redimama – Redimasto, Belo Horizonte, BRAZIL, ¹¹Mastology, Universidade Federal de São Paylo, Sao Paulo, BRAZIL, ¹²Oncology, Hospital da Mulher – SP, Sao Paulo, BRAZIL, ¹³Mastology, Universidade Federal do Rio de Janeiro, Rio de Janeiro, BRAZIL, ¹⁴Mastology, Universidade Federal de Goiás, Goiania, BRAZIL, ¹⁵Mastology, Universiade Federal de Goias, Goiania, BRAZIL, ¹⁶Mastology, Oncoclinicas, Sao Paulo, BRAZIL, ¹⁷Mastology, Santa Casa de Misericórdio de São Paulo, Sao Paulo, BRAZIL, ¹⁸Mastology, UNOESTE, Presidente Prudente, BRAZIL, ¹⁹Mastology, Hospital Beneficiência Portuguesa de São Paulo, Sao Paulo, BRAZIL.

Background: Neoadjuvant chemotherapy (NAC) may alter tumor biology, leading to immunohistochemical (IHC) subtype conversions between diagnosis and residual disease. The prognostic impact of such changes remains unclear. This systematic review and meta-analysis evaluated the effect of NAC-induced IHC subtype conversion on overall survival (OS) and disease-free survival (DFS) in patients with residual disease (non-pCR).Methods: This study was registered in PROSPERO (CRD42024558811). PubMed, EMBASE, and Cochrane Library were searched through 2025 for studies reporting IHC subtype conversions after NAC and associated survival outcomes in non-pCR patients. Eligible studies included randomized trials and observational cohorts of adult women with primary breast cancer. Two reviewers independently selected studies and extracted data. Risk of bias was assessed using the Newcastle-Ottawa Scale and Cochrane tool. Pooled risk ratios (RRs) were calculated using random-effects models. Heterogeneity was quantified with I², and publication bias was assessed by Egger’s test.Results: Twenty-five retrospective studies (n=5,263 non-pCR patients) were included. Mean patient age ranged from 46-54 years, and tumors were predominantly T2-T3. IHC subtype conversion occurred in 20% [95% CI: 0.17-0.24], with high heterogeneity (I²=89.8%). Common conversions included luminal-to-TNBC (35%), luminal-to-HER2+ (25%), HER2+-to-luminal (20%), and TNBC-to-luminal (20%).Patients without subtype conversion had significantly better OS (RR=1.13 [95% CI: 1.01-1.26], p=0.03; I²=62%). Luminal tumors showed the greatest OS benefit without conversion (RR=1.19 [1.01-1.41], p=0.04), while HER2+ (RR=1.11 [0.95-1.31], p=0.19) and TNBC (RR=0.98 [0.84-1.16], p=0.85) showed no significant OS impact. DFS was not significantly affected by conversion overall (RR=0.95 [0.80-1.14], p=0.58; I²=79%). Sensitivity analysis confirmed OS findings. Egger’s test revealed no publication bias for OS (p=0.33) but suggested bias for luminal DFS (p=0.044).Conclusion: IHC subtype conversion after NAC occurs in 1 in 5 patients with residual disease and is associated with worse OS—particularly in luminal breast cancer. Routine post-NAC IHC reassessment should be considered to improve prognostication and guide adjuvant treatment. Prospective studies with standardized reporting are needed to validate these findings and refine treatment algorithms.

S. Krishnamurthy, R. S. Tidwell, M. Kai, L. Villareal, H. Lopez, B. Lim, A. Nazrazadani, R. Layman, S. Saleem, S. X. Sun, A. Lucci, V. Valero, W. Woodward, MDACC Inflammatory Breast Cancer team; Pathology, The University of Texas MD Anderson Cancer Center, Houston, TX.

Stability of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) in residual invasive tumor after neoadjuvant therapy in inflammatory breast cancer Background: Current guidelines do not mandate routine evaluation of ER, PR, and HER2 in residual tumor after neoadjuvant therapy (NAT), including chemotherapy, targeted therapy, and immune checkpoint inhibitors, in breast cancer. There are no studies that evaluated the stability of these biomarkers in inflammatory breast cancer (IBC) after NAT. We examined expression of ER, PR, and HER2 by immunohistochemistry (IHC) in residual tumor after NAT in IBC and compared the values with baseline values to determine changes in biomarker status. Methods: We included 81 IBC patients with residual tumor in mastectomy specimens after NAT. IHC analysis was performed on routinely processed tissue sections of the residual invasive breast tumor using antibodies against ER (6F11, Leica), PR (16, Leica), and HER2 (4B5, Roche/Ventana). The American Society of Clinical Oncology/College of American Pathologists criteria for scoring ER/PR/HER2 were followed. The associations between expression of ER, PR, and HER2 (positive vs. not) at baseline in the core biopsy specimen obtained as standard of care and in residual tumor in the mastectomy specimen was measured with the kappa statistic and 95% confidence intervals (CIs), using the weighted kappa for the combination of ER/PR and HER2.Results: Of the 81 patients in the study, 63 (78%) had invasive ductal tumors, and 18 (22%) had other tumor types (mixed ductal and lobular, lobular, micropapillary, mucinous, and metaplastic). The IHC findings at baseline and after NAT and alterations in invasive tumor subtypes are summarized in Table 1. Alterations in ER, PR, ER/PR, and HER2 status occurred in 10%, 22%, 12%, and 12%, respectively, of IBC patients treated with NAT. After NAT, 15 patients had low and 14 had ultralow HER2 expression (29/81, 36%). Conclusions:1.Alterations in ER/PR/HER2 status occurred in 22% that resulted in a change in invasive tumor subtype in 22% of IBC patients.2. Low levels of HER2 expression (HER2 low and ultralow) in 36% of IBC patients may provide options for targeted therapy.3.Evaluation of ER/PR/HER2 status in residual tumor after NAT in IBC may be considered to determine the altered expression of biomarkers, which has implications for prognosis and response to adjuvant therapy.

P. D. Rädler¹, B. M. Felsheim¹, A. Fernandez-Martinez², A. D. Pfefferle¹, M. C. Hayward¹, W. Sikov³, L. A. Carey¹, C. M. Perou¹; ¹Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, ²Drug Development Department (DITEP), Inserm UMR981, IHU-National PRecISion Medicine Center in Oncolo, Gustave Roussy Cancer Campus, Villejuif, FRANCE, ³Warren Alpert Medical School, Brown University, Providence, RI.

Background: Despite therapeutic advances in the early-stage triple-negative breast cancer (TNBC) setting, residual disease (RD) following neoadjuvant therapy remains a key predictor of poor prognosis. In this study, we performed transcriptomic profiling on 340 pre-treatment and 70 matched post-treatment RD samples from the CALGB 40603 Phase II trial, which evaluated the effects of adding carboplatin or bevacizumab to a neoadjuvant chemotherapy (NACT) regimen of weekly paclitaxel, followed by dose-dense doxorubicin and cyclophosphamide, in patients with stage II/III TNBC. RNAseq data from these samples were used to assess 959 gene expression signatures and their association with outcomes. Results:The prognostic gene expression signatures identified in RD specimens differed substantially from those in pre-treatment samples for both event-free survival (EFS) and overall survival (OS), including features that are predictors of outcome in the pre-treatment setting. These include immune expression signatures that are prognostic in the pre-treatment setting but lack prognostic value when measured in RD specimens. Conversely, PAM50 molecular subtype and proliferation signatures, which are not prognostic in the pre-treatment setting, were prognostic when measured in RD. The RD specimens were further classified based on their residual cancer burden (RCB), which we observed to be a strong prognostic feature of OS. Supervised analysis comparing paired pre- and post-treatment samples show that RCB-I RD specimens had high fibroblast signatures and low tumor/proliferation signatures. We therefore focused on comparing pre-treatment vs. RCB II+III RD, where tumor cell content remains high. TNBC patients whose tumors were basal-like both pre- and post-treatment by PAM50 subtyping, showed the worst OS compared to other RD subtypes (Univariate Cox; HR = 3.51; 95% CI: 1.72 to 7.15, p < 0.001). Comparing the pre-treatment samples of these “basal-basal” tumors (n=21) to the other pre-treatment tumors (n = 102), we identified elevated chromosomal amplifications of oncogenic drivers (i.e., MYC, CDK6, and CCND1) and lower B-cell and T-cell gene expression. Moreover, we show that adding RD basal-like status to a Cox model of RCB significantly improves prognostic value for OS (HR = 3.46; 95% CI: 1.50 to 7.97; p = 0.004). Paired analyses of basal-like RD and matched pre-treatment specimens revealed even lower lymphocyte levels following NACT, along with downregulation of MHC Class I and interferon signaling gene signatures, indicating an immune-cold environment in basal-like RD. We next explored potential drug targets in these immune-cold, basal-like RD tumors and found moderate to high mRNA expression of TROP2 and/or HER2 (i.e., HER2-low). We therefore tested the antibody-drug conjugates (ADCs) sacituzumab govitecan and trastuzumab deruxtecan on a basal-like patient-derived xenograft model established from a pre-treatment biopsy of a “basal-basal” TNBC patient who ultimately had RD following conventional NACT. Both ADCs demonstrated activity in this chemo refractory model, suggesting we may see a benefit of ADCs in ongoing studies of TNBC patients with RD following standard NACT. Conclusion:We show that in early-stage TNBC, prognostic features measured in RD specimens following NACT may be different from those in the pre-treatment setting, providing a rationale for prognostic biomarker development in the RD setting. Substantial immune depletion in basal-like RD may have implications for treatments with immune checkpoint inhibition. However, our preclinical data suggests that this high-risk group of “basal-basal” patients may benefit from use of ADCs. Support: U10CA180821, U24CA196171; https://acknowledgments.alliancefound.org.

E. F. Schuster¹, I. T. Kleijn¹, K. Dunne¹, R. Sardinha¹, E. O’Neill¹, G. Latifi², C. Kennedy-Dietrich², N. B. Jamieson², J. Cuzick³, A. R. Brentnall³, M. Dowsett¹, S. R. Johnston⁴, N. C. Turner⁴; ¹Division of Breast Cancer Research, The Institute of Cancer Research, London, UNITED KINGDOM, ²School of Cancer Sciences, University of Glasgow, Glasgow, UNITED KINGDOM, ³Wolfson Institute of Population Health, Queen Mary University of London, London, UNITED KINGDOM, ⁴Breast Unit, The Royal Marsden NHS Foundation Trust, London, UNITED KINGDOM.

Aims: Primary breast cancer (BC) is highly heterogeneous and this drives divergent clinical outcomes. Many molecular features have significant prognostic value in hormone receptor positive (HR+) BC, and several commercial multigene prognostic assays currently guide chemotherapy decision making. The translational study from the monotherapy arms of the Arimidex, Tamoxifen, Alone or in Combination trial (transATAC), has been pivotal for the validation and regulatory approval of multiple prognostic assays. Spatial single cell analysis may lead to new insights into molecular features and tumor microenvironment (TME) that associate with prognosis and allow the future development of the next generation of prognostic assays. We present preliminary analysis of the transATAC spatial single-cell atlas of BC heterogeneity. We assess whether single cell analysis of a tissue micro-assay (TMA) core, sampling of ~1000 cells per tumor, is sufficient to recapitulate the major features of the tumor. Methods: TMA blocks of 0.6mm diameter cores were generated from tumors of 1710 patients. Cores were taken from two or more distinct tumor regions in 27 patients. Every region/tumor had 3 spatially distant cores taken and spread across separate TMA blocks. We present initial analysis of single cell spatial transcriptomic analysis of 6108 genes using a CosMx spatial molecular imager from two TMAs. The preliminary analysis was generated data for 160 cores from 155 transATAC patients with data combined per patient to generate 155 tumor samples that would better represent molecular data used in prognostic assays. OncotypeDX recurrence scores on a surgery excision were available for 133 HR+ tumors. Cell type was determined by semi-supervised clustering with InSituType using BC specific reference profiles. Pseudo-bulk gene expression profiles were created by summing the total gene expression from all cells and separately all tumor cells. Large-scale chromosomal copy number alterations (CNA) in individual tumor cells were detected by InferCNV and used to understand tumor clonality. Results: We analysed 155 transATAC HR+ and HR- tumours. A mean of 952 cells were analysed, with 967 transcripts from 603 genes. Across the full dataset, cell typing assigned 59% of cells to tumor cells, 21% to immune cells (T-cells, B-cells and Myeloid cells) and 20% other cells (Endothelial, Fibroblast and Pericytes). Single cell CNA analysis of individual tumors cells suggested only one patient had multiple clones (with >50 tumors cells per clone) and this patient had cores taken from 2 regions of the tumor. Recurrence scores generated from single cell pseudo-bulk data were significantly correlated with Oncotype Dx assay recurrence scores generated from surgery excisions (Pearson r=0.76; p<2.2e-16). Across all 155 tumors, 97% of ESR1 transcripts were expressed in tumor cells. Pseudo-bulk ESR1 expression from tumors cells was significantly correlated with ESR1 expression in the Oncotype assay (Spearman ρ=0.66, p<2.2e-16) and with the H score of ER protein expression (ρ=0.5, p=1.3e-9) taken from excision specimens. There was a significant inverse correlation between ESR1 expression in tumor cells and the percentage of immune cells in a tumor across all 155 samples (ρ =-0.27, p=0.0006) and for the HR+ subset with OncotypeDX scores (ρ = -0.28, p=0.001). Conclusions: Single cell spatial analysis of transATAC TMA cores reproduced key features of tumor biology assessed in resection specimens. Single cell analysis revealed an inverse association between ESR1 expression in tumor cell cells and immune cell infiltration in the tumour. Analysis of transATAC TMAs is ongoing, and updated analysis will be presented. Single cell spatial analysis is a powerful and robust technique to analyse single cell gene expression and may build a better understanding of BC biology and prognosis.

A. R. Michmerhuizen¹, B. M. Felsheim¹, G. L. Wheeler², B. J. Kelly², M. Singha², T. Hinoue³, W. C. Nenad⁴, A. C. Garrido-Castro⁵, J. M. Balko⁶, K. Giridhar⁷, U. Chandran⁸, A. V. Lee⁹, L. A. Carey¹, T. A. King¹⁰, K. A. Hoadley¹¹, P. W. Laird³, E. R. Mardis², C. M. Perou¹¹, AURORA US Metastasis Network; ¹Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, ²The Steve and Cindy Rasmussen Institute for Genomic Medicine, Nationwide Children’s Hospital, Columbus, OH, ³Department of Epigenetics, Van Andel Institute, Grand Rapids, MI, ⁴Bioinformatics and Computational Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, ⁵Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA, ⁶Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, ⁷Mayo Clinic Comprehensive Cancer Center, Mayo Clinic, Rochester, MN, ⁸Department of Biomedical Informatics, University of Pittsburgh, Pittsburgh, PA, ⁹UPMC Hillman Cancer Center, University of Pittsburgh, Pittsburgh, PA, ¹⁰Department of Surgical Oncology, Dana-Farber Cancer Institute, Boston, MA, ¹¹Department of Genetics, Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC.

Background: Patients with metastatic breast cancer (MBC) often have short survival, and clinical disease management is challenging due to tumor heterogeneity, treatment resistance, and unique contributions of organ-specific metastatic microenvironments. Methods: The retrospective and prospective phases of the AURORA US Metastasis Project have profiled samples from 220 MBC patients using RNA-seq, DNA-seq, and DNA methylation arrays. RNA-seq has been performed on 392 specimens from 143 primary tumors and 249 metastases. In total, there are 162 primary-metastasis pairs from 134 patients (13 patients have 2+ metastases profiled). The 249 metastatic samples include 116 liver, 29 lymph node, 17 lung, 14 bone, 10 brain, and 8 skin metastases. Each specimen was assayed to determine its clinical and molecular subtypes. Here we investigate the molecular causes of clinical and PAM50 subtype switching in primary tumor-metastasis pairs using multi-platform data. Results: Of the patients with paired data in the AURORA dataset, 38/134 (28%) had clinical subtype (HR+/HER2+, HR-/HER2+, HR+/HER2-, TNBC) switches between the primary tumor and paired metastasis. The most common clinical subtype switch was loss of hormone receptor expression (16/38 patients) in the metastatic tumor. Many of these observations have been confirmed by concomitant RNA and/or DNA changes. In addition, 59/134 (44%) patients had PAM50 molecular subtype switches. 26/59 (44%) patients had a Luminal A primary tumor which switched either into Luminal B (16/26) or HER2-Enriched (7/26) subtypes, suggesting recurrent progression mechanisms during metastasis. For example, patient ARRP had a Luminal A primary tumor and a HER2-Enriched bone metastasis. Using DNA clonality data, we identified a clonal mutation in the DNA binding domain of FOXA1 (H247L) as well as mutations in CDH1 (frameshift) and NLRC3 (Q650H) that originated in the primary tumor. A distinct, dominant clone arose in the metastasis, harboring oncogenic activating mutations in PIK3CA (H1047R) and ERBB2 (L869R). There was also an increase in RNA expression of ERBB2 and FGFR4 but no change in ERBB2 copy number. This is in alignment with previous publications demonstrating a shift to the HER2-Enriched subtype, often without gain of HER2 amplification. Interestingly, there were 21 patients that had both clinical and PAM50 subtype switches. Patient ANFC had a HR+/HER2- primary tumor (Luminal A) and a TNBC liver metastasis (HER2-Enriched). Notably, the metastasis lost clinical expression of ER and PR (by IHC), and this was also observed through a decrease in ESR1 and PGR mRNA expression. Using DNA clonality data, we identified 5 new clones in the metastasis that were not present in the primary tumor. Finally, we observed that basal-like primary tumors with a PAM50 subtype switch in the paired metastasis had a significantly lower correlation to the basal-like centroid compared to primary tumors with a basal-like paired metastasis (p=0.0053). This suggests that either these tumors had significant intra-tumoral heterogeneity in the primary tumor and the metastasis originated from a non-basal-like portion of the tumor, or the switches from a basal-like primary tumor are a byproduct of the PAM50 bioinformatics algorithm (i.e. the tumor is very close to 2 subtype centroids). Ongoing analyses will precisely determine which subtype switches can be attributed to genetic or epigenetic causes, organ site biases, and/or bioinformatic methodologies. Conclusions: Using multi-platform molecular analyses, we identified cases of clinical and molecular subtype switching in primary-metastasis pairs. These findings begin to elucidate the underlying molecular drivers of tumor progression and subtype switching in patients with MBC, which may have implications for clinical treatment.

M. Benelli¹, A. Guerrero-Zotano², F. Santaniello³, D. Cameron⁴, A. Llinas-Bertran⁵, M. Dadiani⁶, M. Paoli⁷, L. Ferrando⁸, D. Romagnoli¹, M. Oliveira⁹, C. Caballero³, T. Crestani³, E. Agostinetto¹⁰, D. Martins-Branco¹⁰, E. Nili Gal-Yam¹¹, F. Hilbers¹², M. Balic¹³, F. Cardoso¹⁴, C. Sotiriou¹⁵, G. Curigliano¹⁶, B. Linderholm¹⁷, E. de Azambuja¹⁰, S. Knox¹⁸, V. Adam³, E. Ciruelos¹⁹, N. Davidson²⁰, G. Viale²¹, F. Duhoux²², B. Rojas²³, S. Servitja²⁴, J. Albanell²⁵, M. Colleoni²⁶, C. Duhem²⁷, N. Harbeck²⁸, S. Loibl²⁹, S. Marreaud³⁰, S. Fielding³¹, P. Bedard³², O. Johannsson³³, J. Bliss³⁴, S. Loi³⁵, A. Raimbault³⁶, T. Goulioti³⁷, C. Dauccia³⁸, A. Vingiani³⁹, D. Venet¹⁵, G. Zoppoli⁴⁰, J. Seoane⁴¹, P. Aftimos¹⁵, M. Piccart¹⁵; ¹Department of Experimental and Clinical Biomedical Sciences, University of Florence, Florence, ITALY, ²n/a, Instituto Valenciano de Oncología, Valencia, SPAIN, ³Scientific team, Breast International Group, Brussels, BELGIUM, ⁴CRUK Scotland Centre, Institute of Genetics & Cancer, University of Edinburgh, Edinburgh, UNITED KINGDOM, ⁵Cancer Computational Biology Group, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, SPAIN, ⁶Cancer Research Center, Sheba Medical Center, Ramat Gan, ISRAEL, ⁷CIBIO – Department of Cellular, Computational and Integrative Biology, University of Trento, Trento, ITALY, ⁸Department of Internal Medicine, University of Genoa, Genoa, ITALY, ⁹Medical Oncology Department, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, SPAIN, ¹⁰Academic Trials Promoting Team, Université libre de Bruxelles (ULB), Hôpital Universitaire de Bruxelles (H.U.B), Institut Jules Bordet, Anderlecht, BELGIUM, ¹¹Institute of Breast Oncology, Sheba Medical Center, Ramat Gan, ISRAEL, ¹²Department of Molecular Pathology, Netherlands Cancer Institute (NKI), Amsterdam, NETHERLANDS, ¹³n/a, Medical University of Graz, University of Pittsburgh, Graz, AUSTRIA, ¹⁴President, Advanced Breast Cancer (ABC) Global Alliance, Lisbon, PORTUGAL, ¹⁵n/a, Institut Jules Bordet, Hôpital Universitaire de Bruxelles (H.U.B), Université Libre de Bruxelles (ULB), Anderlecht, BELGIUM, ¹⁶Department of Oncology and Hemato-Oncology, European Institute of Oncology, IRCCS and University of Milano, Milano, ITALY, ¹⁷Department of Oncology, Sahlgrenska Academy at Gothenburg University and Sahlgrenska University Hospital, Gothenburg, SWEDEN, ¹⁸Department of Experimental and Clinical Biomedical Sciences, Europa Donna- The European Breast Cancer Coalition, Milano, ITALY, ¹⁹n/a, University Hospital 12 de Octubre, Madrid, SPAIN, ²⁰n/a, Fred Hutchinson Cancer Research Center and University of Washington, Seattle, WA, ²¹n/a, European Institute of Oncology IRCCS, Milano, ITALY, ²²Department of Medical Oncology, Cliniques universitaires Saint-Luc, Brussels, BELGIUM, ²³Medical Oncology Department, Centro Integral Oncológico Clara Campal, Madrid, SPAIN, ²⁴Medical Oncology Department, Hospital del Mar, Barcelona, SPAIN, ²⁵Medical Oncology Department, Hospital del Mar, Barcelona; GEICAM; CIBERONC, Barcelona, SPAIN, ²⁶Division of Medical Senology, European Institute of Oncology, Milano, ITALY, ²⁷n/a, Centre Hospitalier Luxembourg, Luxembourg, BELGIUM, ²⁸Breast Center, Dept. OB & GYN, LMU Hospital and Breast Center, LMU University Hospital and West German Study Group, Munich, GERMANY, ²⁹n/a, GBG Forschungs GmbH, Neu-Isenburg, GERMANY, ³⁰Medical Department, EORTC, Brussels, BELGIUM, ³¹n/a, Frontier Science Scotland, Kincraig, UNITED KINGDOM, ³²n/a, Princess Margaret Cancer Centre, Toronto, ON, CANADA, ³³n/a, The National University Hospital of Iceland, Reykjavík, ICELAND, ³⁴n/a, The Institute of Cancer Research, London, UNITED KINGDOM, ³⁵n/a, Peter MacCallum Cancer Centre, Melbourne, AUSTRALIA, ³⁶Operational, Breast International Group, Brussels, BELGIUM, ³⁷n/a, Breast International Group, Brussels, BELGIUM, ³⁸Academic Trials Promoting Team, Institut Jules Bordet, Hôpital Universitaire de Bruxelles (H.U.B), Université Libre de Bruxelles (ULB),, Anderlecht, BELGIUM, ³⁹n/a, Fondazione IRCCS Istituto Nazionale dei Tumori, Milano, ITALY, ⁴⁰n/a, Università degli Studi di Genova and IRCCS Policlinico San Martino and Gruppo Oncologico Italiano di Ricerca Clinica (Parma, IT), Genoa, ITALY, ⁴¹Cancer Computational Biology Group, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, SPAIN.

Background ER+/HER2− breast cancer with luminal primary tumors (P) can undergo intrinsic subtype switching, with metastases (M) displaying non-luminal intrinsic subtypes associated with a more aggressive behavior and poorer therapeutic responses. We investigated the molecular characteristics of intrinsic subtype switching from luminal P to non-luminal M among patients (pts) with ER+/HER2- breast cancer. Methods The AURORA study (NCT02102165) collected multi-omics data from paired P and M and plasma samples from 1148 pts. We focused on paired RNA-seq data from pts with ER+/HER2- LumA or LumB (L) P. PAM50 intrinsic subtype assignments were verified for reliability. Overall survival (OS) was calculated from the date of diagnosis of metastatic disease. Fisher’s exact, comet, Wilcoxon, and log-rank tests were used to assess statistical associations where appropriate. Results Of 131 pts, 16% (n=21) had a non-luminal M subtype, mainly HER2-enriched (LH, 14%, n=18), while 84% (n=110) remained luminal (LL). IHC subtype switching occurred in 55% of LH, with 33% losing ER and 22% acquiring HER2. Pts with LH tumors had a higher frequency of relapse while on adjuvant endocrine therapy (ET) or within 1 year of ET completion (p<0.001) compared to pts with LL tumors. Pts with LH tumors had worse OS than pts with LL tumors, even after excluding pts with de novo disease (p<0.001, HR=2.7, 95% CI: 1.5-4.9). LH tumors showed a higher frequency of TP53 mutations in both P and M compared to LL (p<0.001), with no significant difference observed for PIK3CA mutations. No ESR1 mutations (ctDNA or M) were detected in LH, while 18% of LL were ESR1 mutated (p=0.02). MAPK pathway mutations were more frequent in LH compared to LL (p=0.07 in M), particularly when compared to LL with detected ESR1 mutation (p=0.05). Gene expression analysis showed less similarity between P and M pairs in LH tumors (median Pearson’s correlation R=0.70) compared to LL (R=0.80) (p=0.002). LH showed decreased ESR1 (paired p<1e-5) and increased ERBB2 (p<1e-4) expression in M compared to P counterparts. P from LH had higher levels of epithelial-mesenchymal transition and immune (q<1e-15) and lower estrogen receptor response (q<1e-10) signatures compared to LL. Discussion ER+/HER2- tumors that switched from luminal P to HER2-enriched subtype at metastasis had worse clinical outcomes and distinct molecular profiles, including TP53/MAPK mutations and immune related signatures already present in primary tumors, and reduced ER signaling and higher ERBB2 expression in metastases. ESR1 mutations were found only in tumors that remained luminal, suggesting that the luminal to HER2-enriched subtype switch may represent a distinct mechanism of ET resistance. Our findings highlight key features of primary tumors that may predict this subtype switching, guiding future research toward tailored (neo)adjuvant strategies.

$$MISSING OR BAD TABLE SPECIFICATION {BA1E0DE4-757E-4CE1-B482-09CB905C1B6E}$$

N. Pasha¹, R. Cutts¹, A. Amenssag¹, S. Hrebien¹, N. Cunningham¹, C. Swift², P. Proszek³, A. Gulati⁴, G. Qiong⁴, K. Dunne¹, R. Roylance⁵, A. tutt⁶, R. Baird⁷, I. MacPherson⁸, A. Ring⁹, S. Johnston⁹, A. Okines⁹, M. hubank³, S. Haider⁴, I. Garcia-Murillas¹, S. Filippi¹⁰, N. Turner¹; ¹Molecular oncology, Institute of Cancer Reasearch, London, UNITED KINGDOM, ²Ralph Lauren Centre for Breast Cancer Research, The Royal Marsden Hospital, London, UNITED KINGDOM, ³Clinical Genomics, The Royal Marsden Hospital, London, UNITED KINGDOM, ⁴Bioinformatic core, Institute of Cancer Reasearch, London, UNITED KINGDOM, ⁵Breast unit, University College London Hospitals NHS Foundation Trust, London, UNITED KINGDOM, ⁶BCN, Institute of Cancer Reasearch, London, UNITED KINGDOM, ⁷Experimental Cancer Therapeutics, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UNITED KINGDOM, ⁸Wolfson Wohl Cancer Research Centre, The Beatson Cancer Centre, UK, UNITED KINGDOM, ⁹Breast Unit, The Royal Marsden Hospital, London, UNITED KINGDOM, ¹⁰Department of Mathematics, Imperial College London, London, UNITED KINGDOM.

Background: Subtype switching in metastatic breast cancer (BC), for example switching between ER positive and triple negative subtypes, is a well-recognized change in tumor phenotype which can be a major source of treatment failure if not identified. Guidelines^1,2 recommend repeat tissue biopsy of metastasis to retest ER+/PR+/HER2+ status at recurrence. Yet, such biopsies may not always be clinically feasible, single-site tumour biopsies sample only a single metastasis, and subtype may change later during metastatic therapy. We developed a non-invasive blood test to monitor the evolution of metastatic BC subtype based on circulating tumor DNA (ctDNA) methylation profiling.Methods: A capture based enzymatic ctDNA based methylation profile previously described³ was used to develop three independent binary Machine learning classifiers to discriminate between ER positive v ER negative, and HER-2 positive v HER-2 negative and TNBC vs non-TNBC from ctDNA methylation profiles. A 10-fold stratified nested cross-validation was used to assess classifier characteristics. Liquid biopsy subtype was compared with metastatic tissue biopsy subtype, in independent test folds, split by whether the subtype was the same or different in the original primary cancer. Subtype was defined by immunohistochemistry (IHC).Results: There was a total of 191 metastatic liquid biopsy samples taken from 86 patients where multiple biopsies had been taken, with a median 63 months between first and repeat tissue biopsy (range 20-203 months). Of these 37/191 (19%) samples were from 17 patients that had switched subtype based on tissue. The liquid biopsy correctly identified the subtype switch in 89.2% (33/37) of samples, with plasma samples taken a median of 7 months from the tissue biopsy.Focussing on the samples that were initially ER+HER2-. Of the cancers that were ER+HER2- on both tissue biopsies, 96% (129/134) were liquid ER+HER2- and 0% (0/134) were liquid TNBC. Conversely of the cancers that were ER+HER2- on initial biopsy and TNBC on the later biopsy, only 9% (1/11) were liquid ER+HER2-, and 91% (10/11) were liquid TNBC including one sample that was mixed ER+/TNBC liquid subtype (p<0.0001).Focussing on the samples that were initially TNBC. Of the cancers that were TNBC on both tissue biopsies, 88% (8/9) were liquid TNBC and 11% (1/9) were liquid ER+HER2-. Conversely of the cancers that were TNBC on initial biopsy and ER+HER2- on the later biopsy, 0% were liquid TNBC and 100% (10/10) were liquid ER+HER2- (p=0.0002).Similar results were obtained for other subtype switching, including robust identification of cancers that acquired HER2 positivity. Furthermore in 8% (3/37) of samples both the original and the new subtype were co-detected in liquid biopsy, suggesting both subtypes may co-exist.Conclusion: We demonstrate identification of tumors that switched subtype using a liquid biopsy and reveal that multiple subtypes may co-exist in the same tumor/patient. Subject to further validation, clinical trials are warranted to assess whether liquid biopsy subtyping has the potential to direct metastatic BC treatment.

N. Priedigkeit¹, M. E. Hughes², C. Weipert³, A. Lebrón-Torres⁴, K. Santos¹, S. Morganti¹, K. Smith⁵, C. E. Stever⁵, G. L. Suggs⁵, A. Patel⁵, G. Kirkner⁵, M. E. Skeffington⁵, H. A. Parsons⁵, D. L. Abravanel⁵, C. Snow⁵, E. P. Winer², S. L. Sammons⁵, S. M. Tolaney⁵, R. M. Jeselsohn⁵, N. U. Lin⁵; ¹Medical Oncology, Dana-Farber Cancer Institute / Broad Institute, Boston, MA, ²Yale Cancer Center, Yale, New Haven, CT, ³Research, Guardant Health, Palo Alto, CA, ⁴Cancer Program, Broad Institute of MIT & Harvard, Cambridge, MA, ⁵Medical Oncology, Dana-Farber Cancer Institute, Boston, MA.

Background: Despite therapeutic advances, hormone receptor-positive/HER2-negative (HR+/HER2-) metastatic breast cancer (MBC) remains largely incurable. Genomic characterization of MBC has historically relied on samples collected early in the disease course; however, a lethal phenotype—characterized by accelerated disease progression and profound therapeutic resistance—is clinically well-recognized. The biological mechanisms driving this terminal transition remain largely unknown. By performing a longitudinal ctDNA analysis at the start and end of this disease continuum, we aimed to characterize genomic alterations driving lethal HR+/HER2- MBC. Methods: Serial plasma samples, linked to detailed clinical annotation, were collected from patients with MBC enrolled in the EMBRACE cohort study. We included patients with HR+/HER2- subtype who had paired plasma collection within 180 days of MBC diagnosis (early MBC) and within 180 days of death (lethal MBC). Plasma-derived ctDNA was analyzed using a >700-gene liquid biopsy assay (Guardant360) that assesses single nucleotide variants (SNVs), indels, fusions, copy number alterations (CNAs), and a methylation-derived tumor fraction (TF) score. To limit variation due to tumor shed, comparisons of somatic alterations were restricted to patients with TF > 1% at both time points. Significance of enrichment was determined using Fisher’s exact tests with FDR corrections. Results: 56 patients with paired specimens from early and lethal MBC timepoints were included with a median interval of 819 days (range 98-2464) between samplings. Median TF was 0.7% (range 0% to 76%) in early MBC and 29.7% (range 0% to > 90%) in lethal MBC draws. In 25 patients with sufficient TF at both collections, lethal MBC exhibited a substantial enrichment of alterations in genes mediating therapy resistance. Patients received a median of 6 lines of therapy (range 3-16), including chemotherapy (100%), CDK4/6i (92%), and antibody-drug conjugates (16%) prior to lethal MBC sampling. Alterations in cell cycle genes were strikingly enriched with a greater than three-fold increase in RB1 alterations (from 12% to 46%), doubling of PTEN alterations (12% to 27%) and significant gains in CDKN2A (4% to 31%), CCND2 (12% to 35%), and CCNE2 (15% to 35%) alterations. Additionally, we found a marked increase in mutations mediating endocrine resistance and growth factor bypass mechanisms, with the prevalence of ESR1 alterations rising from 27% to 65%, ERBB2 from 15% to 42%, NF1 from 12% to 50%, and KRAS from 15% to 35%. Enrichments for ESR1, ERBB2, NF1, RB1, and CDKN2A alterations were statistically significant (FDR q < 0.25). Critically, CNAs seemed to be a dominant driver of this evolution, accounting for 64.5% of acquired alterations. Specifically, amplifications drove the rise in CCND2, and CCNE2 alterations, while deletions were the primary mechanism for the increased frequency of NF1, RB1, CDKN2A, and PTEN alterations. Conclusions: We identify dramatic shifts in the prevalence of resistance-driving alterations from early to lethal HR+/HER2- MBC. The striking enrichment of RB1, CDKN2A, PTEN, and cyclin alterations underscore a near-universal development of resistance to CDK4/6 inhibitors in the terminal phase of disease. Furthermore, the prevalence of structural variation as the primary driver suggests a significant degree of genomic instability that precedes death. Collectively, these findings suggest that: (1) routine ctDNA re-evaluation in advancing disease is critical to identify actionable resistance mechanisms, particularly those not present at metastatic diagnosis; and (2) despite inter-patient heterogeneity, lethal HR+/HER2- MBC exhibits convergent evolution towards common resistance pathways, presenting a formidable clinical challenge that demands more innovative therapeutic strategies.