Additional associations between signatures. A. Distribution of CIN70 by HRD score and genomic scars (NtAI tertiles, AiCna tertiles and HLAMP). B Distribution of RPS by HRD score and genomic scars (NtAI tertiles, AiCna tertiles and HLAMP). C. Distribution of PARPi7 by HRD score and genomic scars (NtAI tertiles, AiCna tertiles and HLAMP). D. Distribution of TILs by HRD score, E. Distribution of ConcensusTME average score by HRD score F. ConsensusTME cell type estimates are highly correlated excluding fibroblasts.Correlation assessed using Spearman correlation. HRD<42 = HRD low; HRD≥42 = HRD high.

Supplementary table 1 - Biomarkers of interest

This is an example of DDR-deficient case with high TILs and low gene-expression measurements. These cases were confirmed to have high TIL content (black delineation) and are characterized by both high tumour area- stromal area ratio as well as a high tumour cell- stromal cell ratio. Moreover, all these cases were characterized by high grade features, such as necrosis (blue delineation), high mitotic activity (green arrow), and high levels of atypia (blue arrow), and all had a solid growth pattern, with no formation of glands.

Background: TNBC is molecularly heterogeneous, showing varied responses to therapeutic agents (Txt). We developed a 140-gene classifier (TNBC-ICR) based on random-forest (RF) to cateogorize TNBC into four biological subgroups. It was trained and tested on 551 TNBC (Zhu ESMO Congress 2023), and had better precision in predicting time to recurrence with anthracycline-based (A) txt compared to TNBC-Baylor subtypes (Burstein Clin Cancer Res, 2015) in TACT2 trial (NCT00301925). Here, we assessed the genomic characteristics and clinical value of TNBC-ICR classifier to predict response for standard of care and emerging Txt. Methods: TNBC-ICR, consisting of 1000 decision trees, assigned each case to immune-enriched (IM), luminal-AR (LAR), mesenchymal-like (MES) or highly proliferative basal-like (BL-Prolif) based on RF probabilities linked to these subgroup features. Association with genomic characteristics and prognosis were tested in The Cancer Genome Atlas (n=155, TCGA Nature, 2012), METABRIC (n=269, Curtis Nature, 2012), and SCAN-B (n=571, Saal Genome Med, 2015). Association with pathological complete response (pCR) was tested across 4 neoadjuvant (NAT) clinical studies: 1) a combined cohort of MDACC, ISPY-1, LBJ/IN/GEI and USO (MDACC, n=188, Hatzis JAMA 2011), 2) CALGB 40603 (n=389, NCT00861705), 3) BrighTNess (n=482, NCT02032277) and 4) ISPY-2 (n=364, NCT01042379). These studies, all including Paclitaxel (P), assessed the additional benefits of Bevacizumab (Bev), Carboplatin (Carbo), PARP inhibitor (V) or Pembrolizumab (Pem), followed by A and Cyclophosphamide (C). Gene expression data (GE) was available for all studies, with batch correction applied to reduce platform effects before using the TNBC-ICR classifier. Statisical analyses included multivariable Cox regression models to assess the hazard ratio (HR), chi-squared tests to compare genomic characteristics and pCR rates among the subgroups and logistic regression models to assess the significance and odds ratio (OR) of tumor probabilities to IM, LAR, MES and BL-Prolif related to pCR. Results: Using whole exome sequencing data from TCGA and METATRIC (n = 306), TP53 was the most frequently mutated gene, observed in IM (88%), MES (76%), BL-Prolif (85%) and LAR (68%, p = 0.008), with PIK3CA mutations in LAR (41%) and the others (7%, p < 0.001). The pCR rates for IM, LAR, MES and BL-Prolif were as follows: MDACC (P+AC), 33% (18/54), 10% (3/30), 25% (9/36) and 48% (29/60), p = 0.002; CALGB 40603 (P+Carbo+AC+Bev), 67% (76/103), 56% (30/54), 36% (31/86) and 54% (73/136), p < 0.001; BrighTNess P+Carbo+V arm: 69% (51/74), 29% (8/28), 62% (34/55) and 43% (34/80), p < 0.001; P+Carbo arm: 73% (29/40), 37% (7/19), 62% (13/21) and 48% (20/42), p = 0.03; in P arm: 43% (16/37), 17% (3/18), 27% (8/30) and 34% (13/38), p = 0.21; ISPY-2 P arm: 22% (6/27), 50% (5/10), 12% (3/26) and 14% (3/22), p = 0.06; P+Pem arm: 100% (8/8), 60% (3/5), 40% (2/5) and 55% (6/11), p = 0.098; In P arms, pCR rate was generally low but addition of either Carbo or V increased pCR rates in IM and MES subtypes and addition of Pem over P increased pCR rates for IM and BL. There was no consistent association of TNBC-Baylor subtypes to pCR in these studies. Statistical modelling of TNBC-ICR classifier on txt response showed that probability to IM-subgroup predicted pCR (OR = 3.52, p < 0.001) in NAT and better recurrence-free survival in adjuvant chemotherapy-treated subgroups of SCAN-B (HR = 0.38, p = 0.04) and METABRIC (HR = 0.3, p = 0.04). Conclusion: Tumors with higher IM-features had increased pCR rates after NAT. We validated the reproducibility and clinical validity of TNBC-ICR classifier in predicting differential response to therapies, including taxane and immunotherapy, demonstrating its potential as a practical clinically relevant integrative biology-driven machine learning algorithm. Citation Format: Xixuan Zhu, Orsolya Sipos, Katherine A. Hoadley, Jane Bayani, Lucy Kilburn, John M.S. Bartlett, Judith Bliss, David Cameron, Andrew Tutt, Maggie Chon U Cheang. A 140-gene Machine Learning Classifier Predicts Survival and Response to Chemotherapy and Immunotherapy in 2500 TNBC [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P2-01-23.

Supplemetary table 3 - Change in signature scores from primary to recurrence

Changes in A. CIN70, B. RPS and C. PARPi7 from treatment naïve primary tumours to post-treatment metastatic samples in an independent dataset. Multiple metastatic samples are included for some patients. 𝛃 coefficients and p-values presented for timepoint term from linear regression models with a random effect for patient to account for multiple samples from patients.

Heatmap showing clustering of all module scores filtered for a significant interaction with treatment. Our original clusters are shown against the new clusters at the top of the heatmap.

Supplementary table 2 - Model estimates for logistic regression models of objective response rate and linear regression models of restricted PFS by treatment group

Heatmap of signatures used for clustering by the resulting novel clusters.

Supplementary Table 4 - Representativeness of Study Participants

Abstract Background: Mechanisms of resistance to endocrine therapy are not well understood within ER+HER2+ breast cancer (BC). Our prior work suggested that intrinsic HER2-Enriched (HER2E) molecular subtype predicts early resistance to aromatase inhibitors (AI) (Bergamino eBioMedicine 2022) and high on-treatment (on-Txt) Ki67 levels predict poor survival (Smith Lancet Oncol 2020). Improved early detection of persistent proliferating tumor cells with endocrine resistance pathways could be targeted by pre-emptive personalized therapy and reduction in recurrence. In this study, we proposed to further identify additional alterations/features from genomic and spatial data to provide unprecedented new insight into intrinsic and adaptive resistant pathways in tumor cells that may assist to identify molecular targets for treatment. Materials: POETIC was a phase III trial of post-menopausal patients with ER/PR+ invasive BC (n = 4480) randomized 2:1 to 2 weeks of peri-operative AI (POAI) vs control, followed by standard-of-care treatment. Ki67 was assessed by IHC and intra-tumor heterogeneity was evaluated (5-15 regions) for all the POETIC POAI samples (N = 2487). ER+HER2+ samples were classified as good responders (GR) or poor responders (PR) based on a reduction in Ki67 between pre-treatment (pre-Txt) and 2-week on-Txt samples. Tumor-infiltrating lymphocytes were assessed; multiplex Immunofluorescence (mIF) was performed to measure immune cell densities in tumor and stroma compartments (CD3, CD20, CD68, FOXP3, and CD3 FOXP3 co-expression). Gene expression profiles by BC360™ (Nanostring) on all 210 pairs of POAI treated ER+/HER2+; whole exome sequencing (WES, 100X) were performed on pre-Txt tumor and blood samples from 13 GR, 17 PR, and 9 HER2E GR. We performed GeoMx Whole Transcriptome on 4 pairs (pre-Txt and on-Txt) of GR and GeoMx Proteins (77 including IO proteins) on 6 pairs of GRs and 6 pairs of PRs. Results: The most frequently mutated genes were TP53, PIK3CA, GATA3, and CHD4. Only TP53 was associated with PR (Fisher’s exact p=0.01). TP53 mutated cases had higher expression of TP53 mutant-like gene expression signature compared to wild-type cases (Wilcoxon test p=0.001), mIF FOXP3 (Wilcoxon test p = 0.0005), and CD68 (Wilcoxon test p = 0.019) density score. However, within the HER2-E subset, we found that TP53 mutations were associated with GR (Fisher’s exact p=0.02). We found spatial heterogeneity of Ki67 IHC levels across POAI samples. Examining IHC, while there was higher heterogeneity of Ki67 in the ER+HER2- samples (n = 2264) with 3% of pre-Txt and 9% on-Txt, 6% of ER+HER2+ samples (13/223, 6 LumA, 5 LumB, and 2 HER2E) showed heterogeneity of Ki67 exclusively on-Txt. Even in GR tumors with Ki67 < 10% on-Txt, we identified hotspots with retained proliferating Ki67+ cells after 2 weeks of POAI. The lobular tumors were GR and had characteristic CDH1 mutations. Importantly, cases with persistent areas of Ki67+ cells, regardless of Her2 status, were associated with late relapse. To further explore intratumoral heterogeneity, we performed spatial whole transcriptomics profiling on 95 regions from 4 pairs of GR samples (Ki67 > 10% at baseline and Ki67 < 10% on-Txt) and found low intratumoral heterogeneity in the pre-Txt samples that increased at 2 weeks on-Txt. In a larger set of samples including both GR and PR with the GeoMx protein method, we found increased intratumoral heterogeneity in the PR vs GR. Conclusion: While TP53 mutation was generally a predictor of poor response; in HER2-E it paradoxically was associated with a good early response to aromatase inhibitor which warrants further investigation. Ki67 levels in ER+HER2+ showed higher intratumoral heterogeneity in a subset of patients on treatment suggesting the potential of persistent, proliferating cells leading to later recurrence. Our spatial RNA and protein data further observe the intratumoral heterogeneity that identifies pathways for use as potential spatial biomarkers. Citation Format: Maggie Chon U Cheang, Xixuan Zhu, Orsolya Sipos, Anastasia Alataki, Mikayla Feldbauer, Elena López-Knowles, Holly Tovey, Lucy Kilburn, Milana Bergamino Sirvén, Dhrusti Patel, Hui Xiao, Perry Maxwell, Anthony Skene, Chris Holcombe, Manuel Salto-Tellez, Nicholas Turner, Andrew Dodson, Ian Smith, John Robertson, Judith Bliss, Gene Schuster, Roberto Salgado, Mitch Dowsett, Katherine A Hoadley. Genomic characterization of endocrine resistance in ER+HER2+ breast cancers in the POETIC Trial [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PS09-08.

Association of biomarkers of interest with clinical outcomes by treatment group. A, Association of biomarkers with objective response. Odds ratios for each biomarker are presented from univariable logistic regression models. B, Association of biomarkers with PFS. Model coefficients of each biomarker are presented from linear regression of restricted mean PFS. Ninety-five percent confidence intervals are shown. Response rates are also presented by Baylor subtypes (C) and TNBC subtypes (D).

Overview of the characterization of DNA damage/repair and immune features in TNT samples and their association with each other. A, Distribution of DNA damage repair features by BRCA1/2 status and Baylor subtype. B, Distribution of average immune infiltration features by BRCA1/2 status and Baylor subtype. C, Correlation matrix of all signatures of interest.

Background The TNT trial (NCT00532727) showed no evidence of carboplatin (C) superiority over docetaxel (D) overall in metastatic triple negative breast cancers (TNBC), but a C benefit was observed in the pre-specified sub-group analysis in patients with a gBRCA1/2 mutation (Tutt et al, Nat Med 2018). Given only ~30% of patients have a gBRCA1/2 mutation, broader predictive biomarkers of response are needed. In this cohort we previously found that DNA Damage Response (DDR) signatures were associated with improved C response in chemotherapy (CT) naïve patients only (Tovey et al, ASCO 2020). Since DDR activities influence tumour immune-microenvironment, we explored the predictive ability of immune cell markers and performed integrative analyses on multi-omics features to identify novel TNBC subgroups. Patients and Methods Tumour infiltrating lymphocytes (TILs) were evaluated on haematoxylin and eosin stained primary tumour (PT) slides for 222/376 TNT patients. Formalin-fixed paraffin-embedded PT tissues from 186/376 TNT patients were successfully profiled using total RNA-sequencing. Matched recurrence (REC) was also sequenced for 13 patients. Twenty-five immune signatures were assessed. Logistic regression and restricted mean progression free survival (PFS) were applied to delineate the relationship of these features with treatment outcomes. Random forest clustering of multi-omics DDR and immune biology markers, including gene expression signatures and mutation/methylation status, was applied to identify subgroups. We further molecularly characterised these clusters through supervised clustering of 693 gene expression “modules” (sets of co-expressed genes), immune cell deconvolution and genomic scars. Results Immune gene expression signatures and TILs were highly correlated. Average immune infiltration based on ConsensusTME was lower in mutated/methylated tumours compared with BRCA1 wildtype tumours (p=0.04). Immune signature score markers decreased from PT to REC, demonstrating a dynamic immune microenvironment. In the overall population and when restricting to prior CT treated patients, high immune infiltration (gene expression based & TILs) was associated with response to D while C response rates were not associated with immune scores (interaction p-values< 0.05). This did not translate to a PFS benefit. Multi-omics clustering identified 6 biological subgroups including immune enriched, immune depleted, DDR deficient and proficient clusters as well as 2 small clusters with no obvious distinguishing features. Immune enriched TNBC were predominantly basal-like immune activated with high B-cell/T-cell diversity. Immune depleted TNBC showed higher activity of proliferation and DDR pathway modules. DDR proficient tumours had low expression of immune markers and enrichment for ESR1/PGR expression, markers of extra cellular formation, cell structure, lipid metabolism and proliferation. The DDR deficient cluster was enriched for proliferation and demonstrated high number of TILs despite no apparent enrichment for gene expression-based immune modules. In the prior CT treated cohort, the immune enriched cluster had preferential response to D (62.5% (D) vs. 29.4% (C); p=0.02). The immune depleted cluster had preferential response to C (8.0% (D) vs. 40.0% (C); p=0.01). Numbers were too small to assess differential response within the other clusters or in the CT naïve cohort. Conclusions Tumours with high immune features have high response to D while those with low immune features have preferential response to C in advanced TNBC. Combining multi-omics markers of DDR deficiency and immune biology can identify clusters of patients with distinct biological profiles and differential treatment specific response rates. Citation Format: Holly Tovey, Orsolya Sipos, Katherine A Hoadley, Joel S Parker, Jelmar Quist, Sarah Kernaghan, Lucy Kilburn, Roberto Salgado, Sherene Loi, Richard D Kennedy, Ioannis Roxanis, Patrycja Gazinska, Sarah E. Pinder, Judith Bliss, Charles M. Perou, Syed Haider, Andrew Tutt, Anita Grigoriadis, Maggie Chon U Cheang. Histopathological and molecular immune landscape and DNA damage response signatures to predict response to carboplatin and docetaxel in TNT trial TNBC cohort [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr PD9-06.

AbstractPurpose: The TNT trial (NCT00532727) showed no evidence of carboplatin superiority over docetaxel in metastatic triple-negative breast cancer (mTNBC), but carboplatin benefit was observed in the germline BRCA1/2 mutation subgroup. Broader response-predictive biomarkers are needed. We explored the predictive ability of DNA damage response (DDR) and immune markers. Experimental Design: Tumor-infiltrating lymphocytes were evaluated for 222 of 376 patients. Primary tumors (PT) from 186 TNT participants (13 matched recurrences) were profiled using total RNA sequencing. Four transcriptional DDR-related and 25 immune-related signatures were evaluated. We assessed their association with objective response rate (ORR) and progression-free survival (PFS). Conditional inference forest clustering was applied to integrate multimodal data. The biology of subgroups was characterized by 693 gene expression modules and other markers. Results: Transcriptional DDR-related biomarkers were not predictive of ORR to either treatment overall. Changes from PT to recurrence were demonstrated; in chemotherapy-naïve patients, transcriptional DDR markers separated carboplatin responders from nonresponders (P values = 0.017; 0.046). High immune infiltration was associated with docetaxel ORR (interaction P values < 0.05). Six subgroups were identified; the immune-enriched cluster had preferential docetaxel response [62.5% (D) vs. 29.4% (C); P = 0.016]. The immune-depleted cluster had preferential carboplatin response [8.0% (D) vs. 40.0% (C); P = 0.011]. DDR-related subgroups were too small to assess ORR. Conclusions: High immune features predict docetaxel response, and high DDR signature scores predict carboplatin response in treatment-naïve mTNBC. Integrating multimodal DDR and immune-related markers identifies subgroups with differential treatment sensitivity. Treatment options for patients with immune-low and DDR-proficient tumors remains an outstanding need. Caution is needed using PT-derived transcriptional signatures to direct treatment in mTNBC, particularly DDR-related markers following prior chemotherapy.

Results of a novel clustering. A, Response rates by novel clusters and treatment groups. P values are presented from Fisher exact tests. Clusters 4 and 6 are not shown due to small numbers. B, Heatmap showing biological features of each novel cluster. PGA, percentage of genome altered; SI, Shannon diversity index; NtAI, number of telomeric allelic imbalances; AiCna, allelic imbalanced CNA; AbCna, allelic balanced CNA; CnLOH, copy number neutral loss of heterozygosity.

The ability to identify regulatory interactions that mediate gene expression changes through distal elements, such as risk loci, is transforming our understanding of how genomes are spatially organized and regulated. Capture Hi-C (CHi-C) is a powerful tool to delineate such regulatory interactions. However, primary analysis and downstream interpretation of CHi-C profiles remains challenging and relies on disparate tools with ad-hoc input/output formats and specific assumptions for statistical modeling. Here we present a data processing and interaction calling toolkit (CHiCANE), specialized for the analysis and meaningful interpretation of CHi-C assays. In this protocol, we demonstrate applications of CHiCANE to region capture Hi-C (rCHi-C) and promoter capture Hi-C (pCHi-C) libraries, followed by quality assessment of interaction peaks, as well as downstream analysis specific to rCHi-C and pCHi-C to aid functional interpretation. For a typical rCHi-C/pCHi-C dataset this protocol takes up to 3 d for users with a moderate understanding of R programming and statistical concepts, although this is dependent on dataset size and compute power available. CHiCANE is freely available at https://cran.r-project.org/web/packages/chicane .

1074 Background: In the Triple Negative Trial we observed no improved response rate (RR) to C over D in aTNBC [Tutt et al, Nat Med 2018], but we did in BRCA1/2 mutated (mut) patients (pts). We hypothesise tumors with other aberrant DNA damage response (DDR) characteristics having higher RR to DNA damage inducing C than D. Methods: We tested the predictive value of DDR process related gene expression signatures (PARPi7, chromosomal instability CIN70, TP53 & DDR Deficiency (DDRD)) on 192 treatment naïve primary tumours (PT) by total RNA-sequencing. Odds ratio (OR) for RR are reported. Paired PT & recurrent (REC) signature scores were compared. Results: Unexpectedly, high DDRD and PARPi7 were associated with higher RR to D than C ( p =0.01 & 0.06). No effect was observed for CIN70 or TP53 signature. To assess whether the unexpected results were due to biological changes 12 PT-REC pairs were available from pts who received chemotherapy (CT) between PT & REC. CIN70 increased from PT to REC, DDRD (non-significantly) & PARPi7 decreased. 4/5 TP53 wildtype classified PT samples classified as mut in REC. The BRCA1/2 & DDRD-treatment interactions only held in pts who received CT before trial entry (table). The PARPi7-treatment interaction only held in CT naïve pts. In CT naïve pts, high CIN70 tumors suggested higher C RR as hypothesized. Restricted to the 149 PAM50 basal-like pts, results were non-significant but similar trends seen. Conclusions: In this trial of aTNBC, DDRD high pts with prior CT had better RR to D than C. A possible explanation for this unexpected result is selective pressure of adjuvant DNA damaging CT and selection for relative taxane sensitivity in those who recur despite a high DDRD score. The hypothesised CIN70 treatment interaction was observed in CT naïve pts. Our results suggest care is required in application of signatures to initial diagnostic material when predicting response to DNA damaging agents at REC particularly in pts with prior CT. [Table: see text]