npj Breast Cancer

ERK3/MAPK6 (MAPK6 gene) along with its paralog ERK4/MAPK4 (MAPK4 gene) define a distinct subfamily of atypical mitogen-activated protein kinases (MAPKs)1. Much remains to be learned about the substrates and biological functions of these signaling enzymes. Interestingly, recent work has suggested that ERK4 promotes prostate cancer progression via the non-canonical activation of AKT/mTOR signaling2,3. In their recent study, Wang et al.4 report that ERK4 is expressed in a subset of triple-negative breast cancer (TNBC) cell lines and that this expression is critical for AKT activation and for sustaining TNBC cell proliferation in vitro and tumor growth in mice. They also show that depletion of ERK4 sensitizes TNBC cells to phosphatidylinositol-3-kinase (PI3K) inhibitors. They conclude that ERK4 is a promising therapeutic target for TNBC and has potential for combination therapy with PI3K inhibitors. Here, we raise concerns about the cellular models and experimental approaches used in this study, which compromises the conclusions on the oncogenic role of ERK4 in TNBC.

Gene expression-based recurrence assays are strongly recommended to guide the use of chemotherapy in hormone receptor-positive, HER2-negative breast cancer, but such testing is expensive, can contribute to delays in care, and may not be available in low-resource settings. Here, we describe the training and independent validation of a deep learning model that predicts recurrence assay result and risk of recurrence using both digital histology and clinical risk factors. We demonstrate that this approach outperforms an established clinical nomogram (area under the receiver operating characteristic curve of 0.833 versus 0.765 in an external validation cohort, p = 0.003), and can identify a subset of patients with excellent prognoses who may not need further genomic testing.

1.Breast cancer is the most common cancer in women. A better understanding of risk factors plays a central role in disease prediction and prevention. We aimed to identify potential novel risk factors for breast cancer among post-menopausal women, with pre-specified interest in the role of polygenic risk scores (PRS) for risk prediction. We designed an analysis pipeline combining both machine learning (ML) and classical statistical models with emphasis on necessary statistical considerations (e.g. collinearity, missing data). Extreme gradient boosting (XGBoost) machine with Shapley (SHAP) feature importance measures were used for risk factor discovery among [~]1.7k features in 104,313 post-menopausal women from the UK Biobank cohort. Cox models were constructed subsequently for in-depth investigation. Both PRS were significant risk factors when fitted simultaneously in both ML and Cox models (p < 0.001). ML analyses identified 11 (excluding the two PRS) novel predictors, among which five were confirmed by the Cox models: plasma urea (HR=0.95, 95% CI 0.92-0.98, p < 0.001) and plasma phosphate (HR=0.67, 95% CI 0.52-0.88, p = 0.003) were inversely associated with risk of developing post-menopausal breast cancer, whereas basal metabolic rate (HR=1.15, 95% CI 1.08-1.22, p < 0.001), red blood cell count (HR=1.20, 95% CI 1.08-1.34, p = 0.001), and creatinine in urine (HR=1.05, 95% CI 1.01-1.09, p = 0.008) were positively associated. Our final Cox model demonstrated a slight improvement in risk discrimination when adding novel features to a simpler Cox model containing PRS and the established risk factors (Harrells C-index = 0.670 vs 0.665).

BackgroundThe HOXB13/IL17BR gene expression biomarker has been shown to predict response to adjuvant and extended endocrine therapy in patients with early-stage ER+ HER2- breast tumors. HOXB13 gene expression is the primary determinant driving the prognostic and endocrine treatment-predictive performance of the biomarker. Currently, there is limited data on HOXB13 expression in HER2+ and ER- breast cancers. Herein, we studied the expression of HOXB13 in large cohorts of HER2+ and ER- breast cancers. MethodsWe investigated gene expression, genomic copy number, mutational signatures, and clinical outcome data in the TGGA and METABRIC breast cancer cohorts. Genomic-based gene amplification data was validated with tri-colored fluorescence in situ hybridization. ResultsIn the TCGA breast cancer cohort, HOXB13 gene expression was significantly higher in HER2+ versus HER2- breast cancers, and its expression was also significantly higher in the ER- versus ER+ breast cancers. HOXB13 is frequently co-gained or co-amplified with ERBB2. Joint copy gains of HOXB13 and ERBB2 occurred with low-level co-gains or high-level co-amplifications (co-amp), the latter of which is associated with an interstitial deletion that includes the tumor suppressor BRCA1. ERBB2/HOXB13 co-amp tumors with interstitial BRCA1 loss exhibit a mutational signature associated with APOBEC deaminase activity, and copy number signatures associated with chromothripsis and genomic instability. Among ERBB2-amplified tumors of different tissue origins, ERBB2/HOXB13 co-amp with a BRCA1 loss appeared to be unique to breast cancer. Lastly, patients with ERBB2/HOXB13 co-amplified and BRCA1 lost tumors displayed a significantly shorter progression-free survival (PFS) than those with ERBB2-only amplifications. The difference in PFS was restricted to the ER- subset patients and this difference in PFS was not solely driven by HOXB13 gene expression. ConclusionsHOXB13 is frequently co-gained with ERBB2 at both low-copy number level or as complex high-level amplification with relative BRCA1 loss. ERBB2/HOXB13 amplified, BRCA1-lost tumors are strongly enriched in breast cancer, and patients with such breast tumors experience a shortened PFS.

Recurrence scores based on a 21-gene assay are clinically useful for predicting prognosis and chemotherapy benefit in postmenopausal node-positive breast cancer patients, but its performance in premenopausal patients is inconsistent. Here, we evaluated Ataraxis Breast RISK (ATX), an AI test that predicts recurrence risk, and compared it with the genomic assay. ATX identified high risk patients misclassified as low risk by the genomic assay and therefore may refine selection of patients for adjuvant chemotherapy.

BackgroundHER2 expression level is a key factor in determining the optimal treatment course for breast cancer patients. Roughly 15% of breast cancers are HER2(+), and determination of HER2 status is routinely assessed by immunohistochemistry (IHC). Accurate assessment of the HER2 IHC score by pathologists is therefore critical, especially in light of novel therapeutic approaches demonstrating efficacy in the HER2-low setting. However, there is an opportunity to improve inter-pathologist agreement at the lower levels of HER2 scoring (0, 1+, and 2+). MethodsA machine learning model (AIM-HER2) was developed to generate accurate, slide-level HER2 scores aligned with ASCO-CAP guidelines in clinical breast cancer HER2 IHC specimens. AIM-HER2 was assessed as an AI-assist tool in a retrospective reader study, where 20 HER2-trained pathologists scored breast cancer cases (N=200) with and without AIM-HER2 assistance using a 2-cohort crossover design with a 3-week washout. A separate panel of 5 expert HER2 pathologists read all 200 cases manually to establish reference scores. ResultsIn a significant fraction of cases examined, less than 70% inter-pathologist agreement was observed. When used as an AI assist tool, AIM-HER2 improved inter-rater agreement overall and specifically at the 0/1+ and 1+/2+ cutoffs. Similarly, AIM-HER2 AI-assist significantly increased PPA at the 0/1+ and 1+/2+ cutoffs. When interacting with the AI-assist tool, pathologists displayed a wide range of override rates, and the quality of a pathologists overrides was correlated with their manual accuracy. Lastly, the impact of the reference panel on AIM-HER2 accuracy metrics was assessed, revealing that measurements of model accuracy are highly dependent on reference panel composition. ConclusionsThe use of AIM-HER2 as an AI-assist tool for scoring HER2 IHC in breast cancer may improve pathologist reproducibility and accuracy, particularly at the 0/1+ and 1+/2+ cutoffs.

BackgroundPrimary luminal breast cancer cells lose their identity rapidly in standard tissue culture, which is problematic for testing hormone interventions and molecular pathways specific to the luminal subtype. Breast cancer organoids are thought to retain tumor characteristics better, but long-term viability of luminal-subtype cases is a persistent challenge. Our goal was to adapt short-term organoids of luminal breast cancer for parallel testing of genetic and pharmacologic perturbations. MethodsWe freshly isolated patient-derived cells from luminal tumor scrapes, miniaturized the organoid format into 5 {micro}l replicates for increased throughput, and set an endpoint of 14 days to minimize drift. Therapeutic hormone targeting was mimicked in these "zero-passage" organoids by withdrawing {beta}-estradiol and adding 4-hydroxytamoxifen. We also examined sulforaphane as an electrophilic stress and commercial neutraceutical with reported anti-cancer properties. Downstream mechanisms were tested genetically by lentiviral transduction of two complementary sgRNAs and Cas9 stabilization for the first week of organoid culture. Transcriptional changes were measured by RT-qPCR or RNA sequencing, and organoid phenotypes were quantified by serial brightfield imaging, digital image segmentation, and regression modeling of cellular doubling times. ResultsWe achieved >50% success in initiating luminal breast cancer organoids from tumor scrapes and maintaining them to the 14-day zero-passage endpoint. Success was mostly independent of clinical parameters, supporting general applicability of the approach. Abundance of ESR1 and PGR in zero-passage organoids consistently remained within the range of patient variability at the endpoint. However, responsiveness to hormone withdrawal and blockade was highly variable among luminal breast cancer cases tested. Combining sulforaphane with knockout of NQO1 (a phase II antioxidant response gene and downstream effector of sulforaphane) also yielded a breadth of organoid growth phenotypes, including growth inhibition with sulforaphane, growth promotion with NQO1 knockout, and growth antagonism when combined. ConclusionsZero-passage organoids are a rapid and scalable way to interrogate properties of luminal breast cancer cells from patient-derived material. This includes testing drug mechanisms of action in different clinical cohorts. A future goal is to relate inter-patient variability of zero-passage organoids to long-term outcomes.

For older patients with competing comorbidities, optimizing oncologic therapies is of paramount importance. Circulating tumor DNA (ctDNA) is a validated prognostic factor across solid tumors and may provide a strategy to identify patients for whom safe de-escalation of certain therapies is possible. In this prospective, hybrid-decentralized trial (n = 43 patients; NCT05914792) that integrated clinical outcomes, patient- and caregiver-reported outcomes, and correlative tissue analysis, the primary objective was to determine if ctDNA levels were associated with tumor progression in older patients who opted to forgo breast cancer surgery in favor of primary endocrine therapy (pET). ctDNA levels were highly concordant with imaging findings, and a lack of ctDNA clearance at 6 months was associated with tumor progression. In a competing risk regression adjusted for patient age, tumor stage, tumor grade, and tumor Ki-67, pre-treatment ctDNA positivity was associated with a significant risk of tumor progression (HR 30, 95% CI 4.4-209; p = 0.0005). No patients with pre-treatment ctDNA negativity experienced tumor progression. In correlative analyses examining ctDNA-positive tumors progressing on pET, we identified populations of CD11+ T cell-interacting macrophages that upregulate CD109 and CD89 and secrete immunosuppressive chemokines to create a favorable environment for cancer epithelial cell proliferation. These findings suggest that ctDNA may be a surveillance modality for older patients who receive pET, warranting future evaluation in a randomized setting. STATEMENT OF SIGNIFICANCEClinical management of older women with breast cancer can be challenging, and some women may opt to forego surgery through shared decision making with their physicians. Use of ctDNA for these patients may identify those at an increased risk for progression as well as those with endocrine-sensitive tumors who are good candidates for surgical de-escalation.

PurposeThis study evaluated the prognostic performance of RlapsRisk BC, a multimodal deep learning tool designed to predict distant recurrence-free interval (dRFI) in early-stage, ER-positive, HER2-negative breast cancer using routine H&E-stained whole-slide images (WSIs) and standard clinicopathologic features. MethodsRlapsRisk BC was developed and internally validated on seven retrospective cohorts totaling 6,039 patients. Its ability to stratify patients into high- and low-risk groups for dRFI was then assessed in a blinded, retrospective validation across three independent international cohorts (UK, France, USA), including 591 patients with non-metastatic ER+/HER2- breast cancer treated with adjuvant endocrine therapy alone. ResultsAcross all validation cohorts, RlapsRisk BC showed strong prognostic performance, stratifying patients into distinct low- and high-risk groups with hazard ratios from 3.93 to 9.05. At 5 years, distant recurrence ranged from 0.85%-4.7% in low-risk vs. 6.39%-34.8% in high-risk groups. This separation remained robust across subgroups, including grade 2 tumors, menopausal status, and nodal involvement. RlapsRisk BC was also an independent prognostic factor and improved performance when combined with clinical variables (age, tumor size, nodal status). It increased the c-index by 0.08, 0.19, and 0.20 across the three cohorts, with significant improvement in two. Compared to genomic assays, RlapsRisk BC showed complementary--and sometimes superior--performance, particularly for identifying low-risk patients. At matched specificity, it achieved higher sensitivity: 0.85 vs. 0.33 (Oncotype DX) and 0.74 vs. 0.49 (EndoPredict). ConclusionRlapsRisk BC demonstrates strong, independent prognostic value and may offer a scalable, accessible alternative to genomic assays. Further studies are needed to confirm clinical utility and support integration into treatment decisions. ContextO_ST_ABSKey ObjectiveC_ST_ABSThis study aimed to assess whether RlapsRisk BC, a digital pathology-based AI model, can provide clinically meaningful prognostic stratification in ER-positive, HER2-negative early breast cancer. The AI model combines standard histology exhibited in whole slide images of hematoxylin and eosin stained tissue sections and clinical data. The prognostic performance of RlapsRisk BC was evaluated across multiple independent patient cohorts. Knowledge GeneratedThe study demonstrates that RlapsRisk BC offers independent prognostic value beyond established clinical variables and genomic assays. It consistently stratifies patients into high- and low-risk groups for distant recurrence, with reproducible performance across diverse and independent cohorts, supporting its potential integration into routine clinical decision-making. RelevanceRlapsRisk BC may serve as a scalable alternative or adjunct to molecular assays, supporting more personalized and accessible treatment decisions in breast cancer, particularly in settings where genomic testing is unavailable, limited, or yields intermediate-risk results.

BackgroundGenomic assays such as Oncotype DX, MammaPrint, and Prosigna have transformed risk stratification and treatment selection in early-stage, estrogen receptor-positive (ER+), HER2-negative breast cancers by enabling more precise prognostication and chemotherapy de-escalation in selected patients. However, their clinical utility is limited in lymph nodes positive disease. A major unmet need is the development of compact, mechanistically grounded biomarkers that extend risk and treatment-response prediction to clinically challenging ER+/HER2- subgroups, including lymph node-positive patients. MethodsBuilding on a cell-of-origin framework, we previously established a 70-gene triple hormone receptor (THR; ER, AR, VDR) signature (THR-70) that reflects luminal epithelial differentiation programs and is prognostic across breast cancer subtypes. Here, we refined this framework using interactome-guided clustering to derive a six-gene cell-of-origin signature (THR-6E: KIF4A, KIF2C, CDC20, FAM64A, TPX2, and LMNB2). We evaluated the prognostic and predictive performance of THR-6E across >7,000 breast cancer cases from multiple independent cohorts, assessed treatment-response prediction using endocrine- and chemotherapy-annotated datasets, and performed independent validation in the I-SPY2 adaptive clinical trial. FindingsTHR-6E robustly stratifies relapse-free survival (RFS) in ER+/HER2- breast cancer independent of tumor grade, proliferation status, and subtype. Hazard ratios for RFS were 2.41 (p<1x10-{superscript 1}), 1.61 (p=4.9x10-), and 1.50 (p=6.2x10-3) for grades 1, 2, and 3, respectively, and 2.16 and 1.33 for Luminal A and Luminal B subtypes. THR-6E maintained predictive value across endocrine- and chemotherapy-treated, untreated, lymph node-positive, and lymph node-negative subgroups. Beyond prognosis, THR-6E predicted endocrine therapy response in ER+/HER2-, node-negative disease and chemotherapy response in ER+/HER2-, node-positive disease, achieving approximately 70% sensitivity and specificity (AUC=0.714, p=3.6x10-), with consistent performance across taxane-, anthracycline-, and FEC-based regimens (AUCs 0.71-0.72). Single-cell transcriptomic and proteomic analyses demonstrated that THR-6E expression is specific to normal and malignant breast glandular epithelium, preserved during transformation, and further enriched in metastatic disease. Consistent with a cell-of-origin program, THR-6E genes were rarely mutated in breast cancer and retained normal tissue-like co-expression patterns. In the I-SPY2 trial, THR-6E achieved robust prediction of pathologic complete response across multiple treatment arms (AUCs 0.72-0.94), with an overall AUC of 0.741. InterpretationThese results support a cell-of-origin-anchored approach to biomarker development and challenge purely tissue-agnostic models of therapeutic response. THR-6E represents a compact, biologically interpretable signature that extends prognostic and predictive assessment to clinically relevant ER+/HER2- subgroups, including lymph node-positive disease. Its mechanistic grounding and stable performance across cohorts position THR-6E as a complementary tool to existing multigene assays, warranting prospective diagnostic accuracy studies to define its clinical utility and impact on treatment decision-making.

There has been a long-standing interest in targeting the insulin-like growth factor-1 receptor (IGF-1R) signaling system in breast cancer due to its key role in neoplastic proliferation and survival. However, no IGF-1R targeting agent has shown substantial clinical benefit in controlled trials, and no treatment predictive biomarkers for IGF-1R targeting agents exist. IGFBP7 is an atypical insulin-like growth factor binding protein as it has a higher affinity for the IGF-1R than IGF ligands. We report that low IGFBP7 gene expression identifies a subset of breast cancers for which the addition of ganitumab (an anti-IGF-1R monoclonal antibody) to chemotherapy substantially improved the pathological complete response rate compared to neoadjuvant chemotherapy alone. Furthermore, high IGFBP7 expression predicted increased distant metastasis risk. If our findings are confirmed, decisions to halt the development of IGF-1 targeting drugs, which were based on disappointing results of prior trials that did not use predictive biomarkers, should be reviewed.

IntroductionMetastatic breast cancer (MBC) remains an incurable disease with a 5-year overall survival rate below 25%. Metastases often emerge from subclinical, disseminated tumor cells that persist despite systemic therapy of primary disease - referred to as minimal residual disease (MRD). Detecting MRD is critical for identifying patients at high risk of recurrence and enabling timely intervention. MethodsIn this study, we developed MammaTrace, a plasma-only cell-free DNA (cfDNA) methylation-based MRD assay, informed by differentially methylated regions (DMRs) identified in MBC using whole genome bisulfite sequencing. MammaTrace was evaluated in an independent longitudinal cohort of early-stage breast cancer patients treated with curative intent. ResultsMammaTrace achieved a sensitivity of 91% and specificity of 83%, with a median follow-up of 12.4 months. A positive MammaTrace score, indicative of MRD, preceded clinical or radiologic recurrence by a median of 457 days, providing a substantial lead time for therapeutic intervention to prevent progression to metastatic disease. ConclusionsMammaTrace enables detection of minimal residual disease in breast cancer patients, offering a substantial lead time before clinical recurrence. This approach may improve risk stratification and guide early therapeutic strategies to delay or prevent metastatic progression.

BackgroundThe management of residual axillary disease after neoadjuvant therapy (NAT) remains controversial, as current recommendations often treat ypN1 breast cancer as a homogeneous entity despite potential prognostic heterogeneity. Evidence supporting uniform axillary surgical strategies across different levels of residual nodal burden is limited. We investigated whether survival associations related to axillary surgical evaluation differ according to residual nodal burden in ypN1 disease, using an adjuvant cohort to validate a SEER-based proxy for surgical extent. MethodsPatients with 1-3 positive lymph nodes were identified in the SEER database (2000-2022) and stratified into neoadjuvant (NAT; n=30,560) and adjuvant (AT; n=197,586) cohorts. Axillary surgical evaluation was categorized as limited (2-3 examined nodes) or extensive ([&ge;]10 examined nodes). Survival was analyzed using Kaplan-Meier methods and log-logistic accelerated failure-time models, adjusted with inverse probability of treatment weighting. ResultsIn the ypN1 cohort, limited axillary evaluation was not associated with inferior overall survival among patients with a single residual positive node (IPTW-adjusted HR: 1.15, p=0.134; time ratio [TR]: 0.86, p=0.184). In contrast, limited evaluation was associated with worse survival in patients with two positive nodes (HR: 1.70, 95%CI 1.54-1.87; TR: 0.58, 95%CI 0.53-0.64). The findings were similar when using breast cancer-specific survival as the endpoint. ConclusionsSurvival associations related to axillary surgical evaluation after NAT vary according to residual nodal burden. Axillary de-escalation appears feasible in patients with a single residual positive node but cannot be extrapolated to those with multiple residual nodes, underscoring heterogeneity within ypN1 disease.

Circulating tumour DNA (ctDNA) detection in liquid biopsy is an emerging alternative to tissue biopsy, but its utility in treatment response monitoring and prognosis in triple negative breast cancer (TNBC) is not yet well understood. In this study, we determined the presence of ctDNA detectable actionable mutations with a clinically validated hotspot treatment indication panel in early stage TNBC patients, without local recurrence or metastatic disease at the time of evaluation. Sequencing of plasma DNA and validation of variants from 130 TNBC patients collected within 7 months of primary treatment completion revealed that 7.7% had detectable residual disease with a hotspot panel. Among neoadjuvant treated patients, we observed a trend where patients with incomplete pathologic response and positive ctDNA within 7 months of treatment completion were at much higher risk of reduced progression free survival. We propose that a high risk subset of early TNBC patients treated in NAT protocols may be identifiable by combining tissue response and sensitive ctDNA detection.

BackgroundOncogene-induced replication stress characterizes many aggressive cancers, including triple-negative breast cancer (TNBC). Several drugs are being developed that target replication stress, although it is unclear how tumors with high levels of replication stress can be identified. We aimed to develop a gene expression signature of oncogene-induced replication stress. MethodsTNBC and non-transformed RPE1-TP53wt and RPE1-TP53mut cell lines were engineered to overexpress the oncogenes CDC25A, CCNE1 or MYC. DNA fiber analysis was used to measure replication kinetics. Analysis of RNA sequencing data of cell lines and patient-derived tumor samples (TCGA n=10,592) was used to identify differential gene expression. Immunohistochemical validation was conducted on breast cancer samples (n=330). ResultsRNA sequencing revealed 52 commonly upregulated genes after induction of CDC25A, CCNE1 or MYC in our cell line panel. Integration with gene expression data of TGCA samples with amplification of replication stress-inducing oncogenes (CDC25A, CCNE1, MYC, CCND1, MYB, MOS, KRAS, ERBB2, and E2F1), yielded a six-gene signature of oncogene-induced replication stress (NAT10, DDX27, ZNF48, C8ORF33, MOCS3, and MPP6). Expression of NAT10 in breast cancer samples was correlated with phospho-RPA (R=0.451, p=1.82x10-20) and {gamma}H2AX (R=0.304, p=2.95x10-9). Applying the oncogene-induced replication stress signature to patient samples (TCGA n=8,862 and GEO n=13,912) defined the replication stress landscape across 27 tumor subtypes, and identified diffuse large B cell lymphoma, ovarian cancer, TNBC and colorectal carcinoma as cancer subtypes with high levels of oncogene-induced replication stress. ConclusionWe developed a gene expression signature of oncogene-induced replication stress, which may facilitate patient selection for agents that target replication stress.

IntroductionMore than 150,000 women die worldwide every year of Triple-Negative Breast Cancer (TNBC). There are a range of treatment options, but no good way to match patients to their optimal treatment. For most newly diagnosed patients with early TNBC, the current standard of care is neoadjuvant chemo/immunotherapy before surgery, with patients who achieve a pathological complete response (pCR) having a better prognosis. We have developed a Functional Precision Medicine (FPM) test that uses a fresh biopsy, dissociates the cells, embeds them in a 3D hydrogel matrix, cultures them in a microfluidics device and tests them against a range of systemic therapies, while using a computer vision pipeline to measure responses to therapies ex vivo. MethodsWe designed and conducted an observational multi-centre clinical trial to assess the feasibility of using our FPM assay in patients with newly diagnosed TNBC undergoing neoadjuvant therapy. Patients underwent an additional core needle biopsy followed by systemic therapy as part of routine care. We assessed the response in our assay against whether patients achieved pCR or not at the time of definitive surgery, and calculated Receiver-Operating Characteristic curves (ROC) to optimize cut-offs. In patients who did not achieve pCR, we explored whether there were other regimens that had a better in-assay performance. ResultsIn cohort A, we recruited 34 patients, of whom 12 are evaluable as of 31st July 2024 All were female. Nine patients achieved a pCR. Specificity was 100%, sensitivity 78%, p = 0.0455 and the AUC for the ROC for predicting pCR vs. non-pCR was 0.78. In the 3 patients who did not achieve a pCR, one patient had a regimen that performed better in assay than the treatment they received, and where the response was greater than the cut-off that predicted pCR in other patients. ConclusionWe have presented interim results from a novel FPM assay in patients with early stage TNBC. Our test demonstrates good performance in predicting pCR. The trial continues to accrue data, and Cohort B continues to recruit (PEAR-TNBC; NCT05435352). CoI statementWilliams, Peerani, Tham, Iori, de Fraine, Loughrey, Kaffa, Richardson, Liberal, Velentza-Almpani, Wiskerke, Sangkolah, Crawley, Kearney, Bah, Ranatunga are employees of Pear Bio, with salary, stock options and IP. Hall has received honoraria from Pfizer, Eisai, MSD, Seagen, Exact Sciences, Gilead, AstraZeneca and conference expenses from Lilly and Novartis. Williams has research funding or agreements from Cancer Research UK, Breast Cancer Now, The Brain Tumour Charity, Brain Tumour Research and Novocure. Tasoulis has received honoraria from the BMJ and Company: BMJ and IntegraConnect Kirwan reports no Conflict of Interest

Breast cancer (BC) is the most common cancer affecting women in the United States. Ductal carcinoma in situ (DCIS) is the earliest identifiable pre-invasive BC lesion. Estimates show that 14 to 50% of DCIS cases progress to invasive BC. Our objective was to identify nuclear matrix proteins (NMP) with specifically altered expression in DCIS and later stages of BC compared to non-diseased breast reduction mammoplasty and a contralateral breast explant using mass spectrometry and RNA sequencing to accurately identify aggressive DCIS. Sixty NMPs were significantly differentially expressed between the DCIS and non-diseased breast epithelium in an isogenic contralateral pair of patient-derived extended explants. Ten of the sixty showed significant mRNA expression level differences that matched the protein expression. These 10 proteins were similarly expressed in non-diseased breast reduction cells. Three NMPs (RPL7A, RPL11, RPL31) were significantly upregulated in DCIS and all other BC stages compared to the matching contralateral breast culture and an unrelated non-diseased breast reduction culture. RNA sequencing analyses showed that these three genes were upregulated increasingly with BC progression. Finally, we identified three NMPs (AHNAK, CDC37 and DNAJB1) that were significantly downregulated in DCIS and all other BC stages compared to the isogenically matched contralateral culture and the non-diseased breast reduction culture using both proteomics and RNA sequencing techniques.

CDK4/6 inhibitors (CDK4/6i) have transformed the treatment of hormone receptor-positive (HR+), HER2-negative (HR+) breast cancers as they are effective across all clinicopathological, age, and ethnicity subgroups for metastatic HR+ breast cancer. In metastatic ER+ breast cancer, CDK4/6i lead to strong and consistent improvement in survival across different lines of therapy. To improve understanding of how metastatic HR+ breast cancers become refractory to CDK4/6i, we have created a multimodal and longitudinal tumor atlas to investigate therapeutic adaptations in malignant cells and in the tumor immune microenvironment. This atlas is part of the NCI Cancer Moonshot Human Tumor Atlas Network and includes seven pairs of pre- and on-progression biopsies from five metastatic HR+ breast cancer patients treated with CDK4/6i. Biopsies were profiled with bulk genomics, transcriptomics, and proteomics as well as single-cell ATAC-seq and multiplex tissue imaging for spatial, single-cell resolution. These molecular datasets were then linked with detailed clinical metadata to create an atlas for understanding tumor adaptations during therapy. Analysis of our atlas datasets suggests a diverse set of tumor adaptations to CDK4/6i therapy. Malignant cells may adapt to therapy via mTORC1 activation, cell cycle bypass, and increased replication stress. The tumor immune microenvironment displayed evidence of both immune activation and immune suppression during therapy. Together, our metastatic ER+ breast cancer atlas represents a rich multimodal resource to better understand HR+ breast cancer tumor therapeutic adaptations to CDK4/6i therapy.

Cancer stem cells lie at the heart of progression and relapse for many solid tumours including breast cancers. The Breast Cancer Stem Cell (BCSC) population is typically isolated via a combination of markers utilizing various staining techniques which prevents the ability to track dynamic changes in expression and to dissect the role in pathogenesis overtime. Here we report the development of a reporter for the expression of Aldehyde Dehydrogenase 1A3 (ALDH1A3), a marker of high clinical importance in many breast cancers, and other solid tumours. BCSCs displaying increased transcriptional activation of ALDH1A3 demonstrate an increase in self-renewal capabilities. This tool improves the ability to reliably follow select cancer stem cell populations over time.

BackgroundThe ability to predict future risk of cancer development in non-malignant biopsies is poor. Cellular senescence has been associated with cancer as either a barrier mechanism restricting autonomous cell proliferation or a tumor-promoting microenvironmental mechanism that secretes pro-inflammatory paracrine factors. With most work done in non-human models and the heterogenous nature of senescence the precise role of senescent cells in the development of cancer in humans is not well understood. Further, more than one million non-malignant breast biopsies are taken every year that could be a major source of risk-stratification for women. MethodsWe applied single cell deep learning senescence predictors based on nuclear morphology to histological images of 4,411 H&E-stained breast biopsies from healthy female donors. Senescence was predicted in the epithelial, stromal, and adipocyte compartments using predictor models trained on cells induced to senescence by ionizing radiation (IR), replicative exhaustion (RS), or antimycin A, Atv/R and doxorubicin (AAD) exposures. To benchmark our senescence-based prediction results we generated 5-year Gail scores, the current clinical gold standard for breast cancer risk prediction. FindingsWe found significant differences in adipocyte-specific IR and AAD senescence prediction for the 86 out of 4,411 healthy women who developed breast cancer an average 4.8 years after study entry. Risk models demonstrated that individuals in the upper median of scores for the adipocyte IR model had a higher risk (OR=1.71 [1.10-2.68], p=0.019), while the adipocyte AAD model revealed a reduced risk (OR=0.57 [0.36-0.88], p=0.013). Individuals with both adipocyte risk factors had an OR of 3.32 ([1.68-7.03], p<0.001). Alone, 5-year Gail scores yielded an OR of 2.70 ([1.22-6.54], p=0.019). When combining Gail scores with our adipocyte AAD risk model, we found that individuals with both of these risk predictors had an OR of 4.70 ([2.29-10.90], p<0.001). InterpretationAssessment of senescence with deep learning allows considerable prediction of future cancer risk from non-malignant breast biopsies, something that was previously impossible to do. Furthermore, our results suggest an important role for microscope image-based deep learning models in predicting future cancer development. Such models could be incorporated into current breast cancer risk assessment and screening protocols. FundingThis study was funded by the Novo Nordisk Foundation (#NNF17OC0027812), and by the National Institutes of Health (NIH) Common Fund SenNet program (U54AG075932).