PAM50 Classifier Research Articles

Abstract PO-004: Intra-individual heterogeneity of prostate cancer gene expression signatures of luminal and basal subtypes: Implications for selection of biopsy cores for genomic testing

Abstract Background: Prostate cancer (PrCa) heterogeneity is a well-recognized challenge for the development of molecular biomarker tests for clinical applications. Prostate biopsy is a critical point for diagnosis and treatment planning; biopsy procedures typically produce 12-16 cores and multiple cancer-positive cores are common. For the commercial prognostic gene expression tests, current recommendations are to test only the highest grade core with the highest % tumor involvement. More recently, gene expression tests that classify PrCa according to luminal and basal subtypes have been proposed as predictors of response to specific therapeutic interventions. In PrCa, Luminal B (LumB) appears to be the most aggressive subtype and most sensitive to ADT while the Basal subtype includes a subset of PrCa’s that are relatively insensitive to ADT therapy (PMID: 28494073, 31515456). Clinical trials are underway to assess whether molecular stratification by luminal and basal signatures will identify subsets of PrCa patients who will derive the greatest benefit from ADT. In this pilot study, we compare the results of testing all biopsy cores with sufficient tumor vs testing only the highest grade/% core in the detection of PrCa luminal and basal subtypes. Study design: Gene expression profiles were obtained from 238 cancer-positive biopsy cores from 83 treatment-naïve PrCa patients. Areas of cancer delineated by the pathologist were macrodissected from formalin-fixed paraffin embedded sections and the RNA analyzed using an established clinical grade microarray platform (PMID: 23826159). Luminal A (LumA), LumB and Basal gene expression signatures were obtained using the PAM50 gene expression classifier (PMID: 19204204). Results: Of the cancers in the 238 biopsy cores, 106 were classified as LumA, 81 as LumB and 51 as Basal by the PAM50 classifier. As expected, the LumB cores had a higher proportion of higher grade cancer (grade group 2+) than the LumA and Basal cores (59% vs 27% and 24%, respectively, P < 0.01). Of the 55 patients with multiple cancer positive cores, 31 (56%) showed more than one luminal/basal subtype and 6 had at least one core of each of the three subtypes. Overall, at least one core tested positive for the LumA, LumB and Basal signatures in biopsies from 47, 37 and 34 patients, respectively. If only the core with the highest grade/% core had been tested, 12 patients would have had undetected LumA, 4 undetected LumB and 14 undetected Basal cancer foci, potentially confounding clinical trials seeking to develop biomarker guided interventions. Conclusion: PrCa heterogeneity may require that the current practice of testing only the biopsy core with the highest grade/% cancer be expanded for luminal/basal subtyping. Citation Format: Sandra M. Gaston, Sanoj Punnen, Elai Davicioni, Seagle Liu, Oleksandr N. Kryvenko, Benjamin O. Spieler, Alan Pollack, Radka Stoyanova. Intra-individual heterogeneity of prostate cancer gene expression signatures of luminal and basal subtypes: Implications for selection of biopsy cores for genomic testing [abstract]. In: Proceedings of the AACR Virtual Special Conference on Tumor Heterogeneity: From Single Cells to Clinical Impact; 2020 Sep 17-18. Philadelphia (PA): AACR; Cancer Res 2020;80(21 Suppl):Abstract nr PO-004.

Abstract PO-068: Intratumoral heterogeneity of prognostic multigene signatures for breast cancer

Abstract Background Multigene assays for breast cancer are becoming more commonly used for research, diagnostic, and treatment-selection purposes. These assays are commonly based on a single tumor sample, so this study assesses whether clinically-relevant multigene scores are subject to misclassification due to technical variation or intratumoral heterogeneity. Methods Using FFPE blocks, tumors from 37 patients were sampled at different pathologist-selected spatial locations, targeting regions with distinct histological appearance. A second pathologist assessed mitotic activity and presence of immune infiltration for each specimen. Gene expression was quantified using the PAM50 assay. Samples were classified with respect to intrinsic subtype (Luminal A, Luminal B, HER2-enriched, or Basal-like) and risk of recurrence with proliferation score (ROR-P, high vs. med/low). Euclidean distances between samples were calculated across the 50 genes, and the distribution of Euclidean distances were compared for samples that were concordant vs. discordant for each multigene classifier for PAM50 or ROR-P. Results Of the 37 tumors, 11 had samples with discordant PAM50 subtype and 7 had samples with discordant ROR-P group (75% and 83% agreement, respectively). The Euclidean distance between paired samples with discordant PAM50 intrinsic subtype or ROR-P score was significantly greater than that among concordant tumors. Samples discordant for ROR-P had more heterogeneity in histological characteristics, while discordance of PAM50 subtype call was generally not predicted by heterogeneity of these histological characteristics. Conclusion The concordance of PAM50 subtype calls and ROR-P, despite oversampling of heterogenous-appearing tumor regions, demonstrates reproducibility of the PAM50 classifier for breast cancer tumor intrinsic subtyping. However, assay sampling strategies for histologically heterogeneous tumors merit further consideration. Citation Format: Amber N. Hurson, Alina Hamilton, Linnea T. Olsson, Erin L. Kirk, Benjamin C. Calhoun, Joseph Geradts, Mark E. Sherman, Melissa A. Troester. Intratumoral heterogeneity of prognostic multigene signatures for breast cancer [abstract]. In: Proceedings of the AACR Virtual Special Conference on Tumor Heterogeneity: From Single Cells to Clinical Impact; 2020 Sep 17-18. Philadelphia (PA): AACR; Cancer Res 2020;80(21 Suppl):Abstract nr PO-068.

The New Paradigm of Network Medicine to Analyze Breast Cancer Phenotypes.

Breast cancer (BC) is a heterogeneous and complex disease as witnessed by the existence of different subtypes and clinical characteristics that poses significant challenges in disease management. The complexity of this tumor may rely on the highly interconnected nature of the various biological processes as stated by the new paradigm of Network Medicine. We explored The Cancer Genome Atlas (TCGA)-BRCA data set, by applying the network-based algorithm named SWItch Miner, and mapping the findings on the human interactome to capture the molecular interconnections associated with the disease modules. To characterize BC phenotypes, we constructed protein–protein interaction modules based on “hub genes”, called switch genes, both common and specific to the four tumor subtypes. Transcriptomic profiles of patients were stratified according to both clinical (immunohistochemistry) and genetic (PAM50) classifications. 266 and 372 switch genes were identified from immunohistochemistry and PAM50 classifications, respectively. Moreover, the identified switch genes were functionally characterized to select an interconnected pathway of disease genes. By intersecting the common switch genes of the two classifications, we selected a unique signature of 28 disease genes that were BC subtype-independent and classification subtype-independent. Data were validated both in vitro (10 BC cell lines) and ex vivo (66 BC tissues) experiments. Results showed that four of these hub proteins (AURKA, CDC45, ESPL1, and RAD54L) were over-expressed in all tumor subtypes. Moreover, the inhibition of one of the identified switch genes (AURKA) similarly affected all BC subtypes. In conclusion, using a network-based approach, we identified a common BC disease module which might reflect its pathological signature, suggesting a new vision to face with the disease heterogeneity.

Abstract 5468: Gene fusions in breast cancer in Nigerian women

Abstract Background: As a chromosomal rearrangement event, gene fusion plays a critical role in the pathogenesis of cancer by creating potentially oncogenic chimeric proteins. However, gene fusions have been understudied in breast cancer patients of African ancestry, who are often diagnosed at a younger age and with more aggressive molecular features of cancer than patients of other races/ethnicities. Methods: Ninety-six women diagnosed with invasive breast cancer were recruited from Ibadan, Nigeria with mean age at diagnosis of 51.6 ± 12.4 years. Primary tumors were collected, of which 62 (64.6%) were hormone receptor negative (HR-) by immunohistochemistry, and 31 (32.3%) were basal-like subtype by PAM50 classification. Paired-end reads from RNA-seq on these tumors were used for gene fusion detection by three programs, STAR-Fusion, STAR-SEQR, and Arriba. To increase specificity, we applied an ensembling method by selecting fusions identified by at least two of the three callers. Multiple filters were applied to fusion candidates to remove likely false positives, including fusions containing genes of mitochondrial origin, fusions consisting of pairs of paralogues or orthologs, fusions involving HLA genes, and fusions found in several databases of non-cancer tissues. To investigate potentially druggable fusion transcripts, we compared our call set with the OncoKB precision oncology knowledge base. Results: STAR-Fusion, STAR-SEQR, and Arriba identified 682, 4529, and 3056 unique fusion transcripts, respectively. Following application of the ensembling method and filtering to select final fusions, 709 unique fusions were identified. Comparison with 13 databases and published papers identified 62 of these fusions as previously reported. The mean fusion burden per sample was 7.7 ± 7.4. The fusion burden per sample was significantly smaller for tumors classified as Luminal A subtypes by PAM50 classification than Luminal B, Basal, and HER2 subtypes (P < 0.03) as well as normal like (P < 0.01). Fusion burden was highest in HER2 tumors (NS), consistent with previous reports. The number of fusion transcripts per sample did not differ significantly according to the patient's age. Ninety-six percent of fusion transcripts were only identified in a single sample, including the ETV6-NTRK3 fusion, previously reported in secretory breast carcinoma, and BCR-ABL1, both of which are targeted by a drug identified in OncoKB. The most commonly observed fusion, EDDM13–ZNF71, appeared in 4.2% (4/96) of samples. Conclusion: The vast majority of breast cancer samples from Nigerian women demonstrate unique fusion transcripts in expressed RNA. Fusion burden per sample was related to PAM50 classification. Future work will incorporate functional studies of the recurrent gene fusion events identified in our cohort. We will additionally focus on validation and refinement of our approach to detecting gene fusions in this understudied population, in order to better identify patients who can benefit from new therapies. Citation Format: Anna Elizabeth Woodard, Toshio F. Yoshimatsu, WABCS Working Group, Jason J. Pitt, Yonglan Zheng, Olufunmilayo I. Olopade. Gene fusions in breast cancer in Nigerian women [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5468.

Abstract 4428: Characterizing the tumor-infiltrating T-cell repertoire of breast cancer in indigenous African patients from Southwest Nigeria

Abstract Background: Women of African ancestry have a higher risk of young onset breast cancer than their European ancestry counterparts, with often more aggressive molecular tumor subtypes: these patients have fewer treatment options available to them, leading to higher mortality rates. Immunotherapy has recently garnered significant interest as a possible treatment for aggressive tumor subtypes (such as triple-negative breast cancer), but our lack of understanding of the tumor immune micro-environment has made it difficult to find more adaptive treatments for such cancers. Methods: We studied 96 invasive breast cancers collected from Nigerian women (mean age at diagnosis, 51.6±12.4 years), of which 62 (64.6%) were hormone receptor negative (HR-) by IHC, and 31 (32.3%) were basal-like subtype by PAM50 classification. Paired-end reads from bulk transcriptome sequencing were aligned to the human reference genome hg19 using STAR, and two algorithms (TRUST and MiXCR) were used to characterize the T-cell diversity in the tumors. To overcome the difficulty of extracting T-cell receptor (TCR) sequence information due to V(D)J recombination during T-cell maturation, we built a pipeline that incorporated data from public TCR databases (VDJdb and McPAS-TCR) to ensure a more accurate characterization of TCR diversity. Each potential TCR was first truncated to the appropriate start and end amino acids. Fragments of TCRs were discarded, and pairwise alignment of each output sequence was used to remove any duplicates. Results: Despite the many differences between TRUST and MiXCR, there was a 48.5% overlap in TCR sequences between the two methods, providing confirmation of the validity of these two algorithms. The intersection of the results from the two algorithms was then used for downstream analysis, which allowed for an expansive characterization of the tumor immune micro-environment, with 3,127 TCR sequences extracted over all 96 samples. We observed higher TCR diversity in basal-like tumors as compared to luminal-A/B tumors (p=0.011), whereas we saw no significance between HER2-enriched vs. luminal-A/B tumors (p=0.23). This is in accordance with other studies, which have shown that the most inflamed tumors are triple-negative/basal-like. Conclusion: Overall, this novel pipeline harnesses the power of two algorithms to effectively study the diversity of T-cells in the tumor, which, combined with further analyses of the tumor micro-environment, would allow for a more detailed understanding of the tumor immune infiltrates as potential therapeutic targets. Citation Format: Jean-Baptiste Reynier, Toshio F. Yoshimatsu, WABCS Working Group, Yonglan Zheng, Olufunmilayo I. Olopade. Characterizing the tumor-infiltrating T-cell repertoire of breast cancer in indigenous African patients from Southwest Nigeria [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4428.

Hsa-miR-155-5p Up-Regulation in Breast Cancer and Its Relevance for Treatment With Poly[ADP-Ribose] Polymerase 1 (PARP-1) Inhibitors.

miR-155-5p is a well-known oncogenic microRNA, showing frequent overexpression in human malignancies, including breast cancer. Here, we show that high miR-155-5p levels are associated with unfavorable prognostic factors in two independent breast cancer cohorts (CSS cohort, n = 283; and TCGA-BRCA dataset, n = 1,095). Consistently, miR-155-5p results as differentially expressed in the breast cancer subgroups identified by the surrogate molecular classification in the CSS cohort and the PAM50 classifier in TCGA-BRCA dataset, with the TNBC and HER2-amplified tumors carrying the highest levels. Since the analysis of TCGA-BC dataset also demonstrated a significant association between miR-155-5p levels and the presence of mutations in homologous recombination (HR) genes, we hypothesized that miR-155-5p might affect cell response to the PARP-1 inhibitor Olaparib. As expected, miR-155-5p ectopic overexpression followed by Olaparib administration resulted in a greater reduction of cell viability as compared to Olaparib administration alone, suggesting that miR-155-5p might induce a synthetic lethal effect in cancer cells when coupled with PARP-1-inhibition. Overall, our data point to a role of miR-155-5p in homologous recombination deficiency and suggest miR-155-5p might be useful in predicting response to PARP1 inhibitors in the clinical setting.

Frequent Molecular Subtype Switching and Gene Expression Alterations in Lung and Pleural Metastasis From Luminal A-Type Breast Cancer.

PURPOSEConversion of tumor subtype frequently occurs in the course of metastatic breast cancer but is a poorly understood phenomenon. This study aims to compare molecular subtypes with subsequent lung or pleural metastasis.PATIENTS AND METHODSIn a cohort of 57 patients with breast cancer and lung or pleural metastasis (BCLPM), we investigated paired primary and metastatic tissues for differential gene expression of 269 breast cancer genes. The PAM50 classifier was applied to identify intrinsic subtypes, and differential gene expression and cluster analysis were used to further characterize subtypes and tumors with subtype conversion.RESULTSIn primary breast cancer, the most frequent molecular subtype was luminal A (lumA; 49.1%); it was luminal B (lumB) in BCLPM (38.6%). Subtype conversion occurred predominantly in lumA breast cancers compared with other molecular subtypes (57.1% v 27.6%). In lumA cancers, 62 genes were identified with differential expression in metastatic versus primary disease, compared with only 10 differentially expressed genes in lumB, human epidermal growth factor receptor 2 (HER2)–enriched, and basal subtypes combined. Gene expression changes in lumA cancers affected not only the repression of the estrogen receptor pathway and cell cycle–related genes but also the WNT pathway, proteinases (MME, MMP11), and motility-associated cytoskeletal proteins (CK5, CK14, CK17). Subtype-switched lumA cancers were further characterized by cell proliferation and cell cycle checkpoint gene upregulation and dysregulation of the p53 pathway. This involved 83 notable gene expression changes.CONCLUSIONOur results indicate that gene expression changes and subsequent subtype conversion occur on a large scale in metastatic luminal A–type breast cancer compared with other molecular subtypes. This underlines the significance of molecular changes in metastatic disease, especially in tumors of initially low aggressive potential.

Abstract B37: Understanding DCIS initiation and progression using PDX and genetically engineered mouse and rat models

Abstract Ductal carcinoma in situ (DCIS) is a starting lesion in the milk duct of the breast, which accounts for 25% of all “breast cancers” detected. DCIS is usually treated by surgery combined with radiotherapy, which can have a large impact on the life of patients. However, there is little to no evidence that treatment of low- and intermediate-grade DCIS reduces mortality, while women diagnosed with DCIS do perceive their risk of dying to be the same as patients with invasive disease. To reduce the negative perception and the overtreatment of DCIS, but assure proper treatment for high-risk DCIS, it is crucial to understand the biology underlying DCIS. To model DCIS and the progression to invasive ductal carcinoma (IDC), we took two approaches. 1) We established patient-derived xenograft (PDX) models by intraductally injecting fresh patient DCIS material into the mammary ducts of female immunocompromised mice. To date, 100 human patient DCIS samples were engrafted and incubated in vivo for a period of 12 months. The first 25 of these PDX models had a 72% take rate and show both indolent and invasive tumor growth. Histology and molecular subtyping by PAM50 classification are well preserved in the outgrown PDX lesions compared to the original human counterpart, ensuring that our models are a good representation of the patient disease. On top of this, we profiled copy numbers for both PDX and original patient lesions. Finally, using whole mount imaging, we were able to create 3D sections of the injected mammary glands extracted from our PDX models, showing two distinct growth patterns correlated with invasion. 2) We generated genetically engineered mouse and rat models by intraductal injection of lentiviral vectors (over)expressing known breast cancer driver genes and mutations, or combinations thereof, which are suspected to play a role in DCIS initiation and/or its progression. We were able to successfully and selectively transduce the mammary epithelial cell layer in both mice and rats and are therefore able to somatically engineer the mammary gland in vivo. DCIS formation presented in mice with overexpression of Ccnd1 and Myc only. However, the in situ stage could be observed in rat models with overexpression of the latter genes, as well as oncogenic point mutations in Pik3ca and p53 and all combinations of the tested genetic perturbations. These models will eventually allow us to study the biologic processes underlying DCIS progression. This knowledge could then be used to predict DCIS evolution and distinguish patients with high-risk DCIS from those with low-risk DCIS, which will aid in better care. Citation Format: Stefan Hutten, Catrin Lutz, Colinda Scheele, Madelon Badoux, Timo Eijkman, Stefano Annunziatio, Jelle Wesseling, Jos Jonkers. Understanding DCIS initiation and progression using PDX and genetically engineered mouse and rat models [abstract]. In: Proceedings of the AACR Special Conference on the Evolving Landscape of Cancer Modeling; 2020 Mar 2-5; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2020;80(11 Suppl):Abstract nr B37.

Abstract B080: Assessment of the molecular mechanisms of a protective variant for breast cancer in Latinas

Abstract The incidence of breast cancer is lower in women of Latin American origin in the U.S. compared to European American and African American women. Among Latinas, the rs140068132A>G variant, which is common in women with IAA, has been associated to a lower risk of breast cancer. The frequency of the G allele is 0% for non-Latinos while 12% in Latinos, being highest in the Peruvian population (23%). This variant is located on chromosome 6 near the Estrogen Receptor 1 gene (ESR1) and even though experimental evidence suggests that this variant might be functional, the molecular mechanisms that explain its protective effect are unknown. We hypothesize that the rs140068132-G variant decreases ESR1 expression, which affects the expression or function of genes involved in associated pathways. We aim to test the association of the rs140068132 variant and gene expression in breast cancer tumors from patients with high IAA. We collected 47 breast tumors and blood samples from the Instituto Nacional de Enfermedades Neoplásicas in Lima, Peru. These patients were genotyped for IAA estimation and determination of the rs140068132 genotype. Total RNA was extracted from tumor samples and used for a paired-end sequencing (2 × 75bp paired-end,100 million reads per sample) in the Illumina NextSeq500. Differential gene expression between genotypes was performed by DEseq2 R package and statistical significance was determined using FDR<0.05 for samples with at least log2 1.5-fold change. Differentially expressed isoforms were detected by EBseq R package using FDR<0.05. Tumor intrinsic subtypes were obtained using PAM50 as implemented in the genefu R package. The average IAA for the 47 Peruvian patients was 77% (SD=0.17). Twenty-seven patients were homozygous AA, 19 heterozygous AG and 1 GG for the rs140068132 variant. According to PAM50 classification, 10 tumors were Luminal A, 12 Luminal B, 15 HER2+ and 10 Basal. Among luminal tumors there was a suggestive trend towards lower expression of the ESR1 gene in patients carrying the protective allele (p=0.16). Including all subtypes, 27 genes were differentially expressed according to the rs140068132 genotype. Four of these genes are ER dependent or associated with ER status. The expression of the top gene, which is a transcriptional target of ER, is lost in patients with the protective allele. This association is mainly driven by its expression in luminal tumors and remained significant after adjusting for IAA. ESR1 isoforms were not differentially expressed by genotype, however significant differences were detected in the expression of isoforms for 115 genes, of which 18% have been reported to be transcriptional targets of ER or functionally related. Our preliminary results suggest that the rs140068132 variant decreases ER expression and affects the expression of functionally associated genes in luminal tumors. Further allele-specific expression analysis will elucidate if this variant is part of a cis-regulatory module. Note: This abstract was not presented at the conference. Citation Format: Valentina A. Zavala, Tatiana Vidaurre, Katie Marker, Sandro Casavilca, Lizeth Tamayo, Carlos Castañeda, Jeannie Vásquez, Fernando Valencia, Zaida Morante, M. Calderon, J. Abugattas, H. Gomez, H. Fuentes, C. Monge-Pimentel, Silvia Neciosup, Jovanny Zabaleta, Laura Fejerman. Assessment of the molecular mechanisms of a protective variant for breast cancer in Latinas [abstract]. In: Proceedings of the Twelfth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2019 Sep 20-23; San Francisco, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(6 Suppl_2):Abstract nr B080.

Ribonucleotide reductase small subunit M2 is a master driver of aggressive prostate cancer.

Although there are molecularly distinct subtypes of prostate cancer, no molecular classification system is used clinically. The ribonucleotide reductase small subunit M2 (RRM2) gene plays an oncogenic role in many cancers. Our previous study elucidated comprehensive molecular mechanisms of RRM2 in prostate cancer (PC). Given the potent functions of RRM2, we set out to determine whether the RRM2 signature can be used to identify aggressive subtypes of PC. We applied gene ontology and pathway analysis in RNA‐seq datasets from PC cells overexpressing RRM2. We refined the RRM2 signature by integrating it with two molecular classification systems (PCS and PAM50 subtypes) that define aggressive PC subtypes (PCS1 and luminal B) and correlated signatures with clinical outcomes in six published cohorts comprising 4000 cases of PC. Increased expression of genes in the RRM2 signature was significantly correlated with recurrence, high Gleason score, and lethality of PC. Patients with high RRM2 levels showed higher PCS1 score, suggesting the aggressive PC feature. Consistently, RRM2‐regulated genes were highly enriched in the PCS1 signature from multiple PC cohorts. A simplified RRM2 signature (12 genes) was identified by intersecting the RRM2 signature, PCS1 signature, and the PAM50 classifier. Intriguingly, inhibition of RRM2 specifically targets PCS1 and luminal B genes. Furthermore, 11 genes in the RRM2 signature were correlated with enzalutamide resistance by using a single‐cell RNA‐seq dataset from PC circulating tumor cells. Finally, high expression of RRM2 was associated with an immunosuppressive tumor‐immune microenvironment in both primary prostate cancer and metastatic prostate cancer using CIBERSORT analysis and LM22, a validated leukocyte gene signature matrix. These data demonstrate that RRM2 is a driver of aggressive prostate cancer subtypes and contributes to immune escape, suggesting that RRM2 inhibition may be of clinical benefit for patients with PC.

Molecular Biological Features of Nottingham Histological Grade 3 Breast Cancers.

Cancer biology dominates the behavior and prognosis of a tumor. Although Nottingham histological grade is a subjective pathological determination, it has been accepted as a surrogate model for cancer biology. As such, histologic grade was incorporated into the latest 8th edition of the American Joint Committee on Cancer breast cancer staging system. In this study, we hypothesized that grade 3 breast cancers demonstrate aggressive molecular biological profiles, reflecting worse biology and possible underlying immunogenicity. Transcriptomic and clinical data were obtained from the Molecular Taxonomy of Breast Cancer International Consortium, and the findings were validated by The Cancer Genome Atlas breast cancer cohort and GSE25066. Overall, 2876 patients were analyzed in this study. Grade 3 tumors were more common in estrogen receptor (ER)-negative, advanced-stage patients, and were associated with human epidermal growth factor receptor 2 and basal subtypes by the PAM50 classifier, as well as with increased MKI67 expression (all p <0.001). Disease-free survival was significantly worse in grade 3 tumors (all cohorts). Gene set enrichment analysis demonstrated that grade 3 tumors were significantly enriched with not only cell proliferation and cell cycle-related gene sets but also immune activity-related gene sets. CIBERSORT confirmed that grade 3 tumors were infiltrated with macrophage M1, follicular helper T cells, and activated natural killer cells (all p <0.001). Furthermore, grade 3 tumors were associated with more diverse T cell receptors (p =0.001) and increased cytolytic activity (p <0.001). Lastly, major T-cell exhaustion markers were significantly elevated in grade 3 breast cancers (p <0.001). Grade 3 breast cancers demonstrated aggressive transcriptomic features with enhanced immunogenicity and elevated T-cell exhaustion markers.

Molecular determinants of prostate specific antigen (PSA) kinetics and clinical response to apalutamide (APA) in patients (pts) with nonmetastatic castration-resistant prostate cancer (nmCRPC) in SPARTAN.

5521 Background: In SPARTAN, APA + androgen deprivation therapy (ADT) prolonged metastasis-free survival (MFS) and improved PSA kinetics over placebo (PBO) + ADT in high-risk nmCRPC. All molecular subtypes derived benefit in MFS from APA (Feng FY, et al. ASCO GU 2019; abstract 42). We evaluated the association of PSA decline and efficacy outcomes in SPARTAN pts with different molecular subtypes. Methods: Gene expression from archival primary tumors (biomarker population) was assessed with the DECIPHER platform (Decipher Biosciences, Inc.) and stratified into genomic classifier (GC) high- and low-to-average risk using GC score &gt; 0.6 and ≤ 0.6, respectively, and ADT-resistant or -sensitive basal or luminal A/B (PAM50 classifier) subtypes. PSA nadir and confirmed PSA decline (Table) were assessed in APA pts overall and at 3, 6, and 12 mo. Associations between molecular subtypes and outcomes were assessed. Results: Of 233 available samples, 154 were from APA pts; 49% of APA pts had high GC score and 66% had basal subtype. PSA levels at baseline were similar across all subtypes. Regardless of GC score or basal/luminal subtype, &gt; 50% of patients achieved ≥ 90% reduction in PSA with APA. PSA declined faster and PSA reduction was deeper at 6 mo (Table) in GC low to average vs GC high risk and luminal vs basal subtypes. Overall, only luminal vs basal subtypes had a significantly higher % of pts with ≥ 90% PSA decline (Chi square p = 0.037). In luminal pts, deeper PSA decline with APA was consistent with improved MFS vs basal pts. In GC high pts, MFS benefit with APA was similar to that in GC low to average pts despite lower PSA decline. Although GC low to average and luminal pts had more rapid and deeper PSA responses than GC high or basal pts, respectively, all pts derived MFS benefit. Association of long-term outcomes with PSA decline in these molecular subtypes will be presented. Conclusions: In SPARTAN, all molecular subtypes of pts with nmCRPC treated with APA + ADT had MFS benefit and rapid and sustained PSA decline. PSA responses were deepest and most rapid in GC low to average and luminal subtypes. Clinical trial information: NCT01946204 . [Table: see text]

Molecular determinants of outcome for metastatic castration-sensitive prostate cancer (mCSPC) with addition of apalutamide (APA) or placebo (PBO) to androgen deprivation therapy (ADT) in TITAN.

5535 Background: In TITAN, addition of APA to ADT improved radiographic progression-free survival (rPFS) and overall survival (OS) versus PBO plus ADT in patients (pts) with mCSPC. In this post hoc analysis, we performed transcriptome-wide profiling of tumor samples and assessed association of molecular subtypes with rPFS. Methods: The DECIPHER platform (Decipher Biosciences, Inc.) was used to assess gene expression in archival primary prostate tumors from TITAN. Samples were classified into high versus low to average risk of metastases (DECIPHER genomic classifier [GC] &gt; 0.6 and ≤ 0.6, respectively), basal and luminal A/B (PAM50 classifier), and androgen receptor activity (AR-A) signature high and low. Associations between subtypes with rPFS were assessed with Cox proportional hazards model. Results: The biomarker population included 222 pts (APA, 110; PBO, 112). Benefit in rPFS from APA in the biomarker population (HR [95% CI]; p value; 0.49 [0.31-0.78]; 0.002) resembled that in the overall study population (0.49 [0.40-0.61]; &lt; 0.0001). The majority of TITAN pts had GC high scores (n = 166, 75%). GC high risk subtype in the PBO group had poorer prognosis for rPFS than GC low to average risk subtype (median rPFS 18.2 mos for GC high vs not reached [NR] for GC low to average, 0.28 [0.11-0.69]; 0.006), but there was no difference in prognosis between high and low to average GC risk subtypes in the APA group (GC high NR vs GC low to average NR; 0.81 [0.35-1.89]; 0.625). Pts were further stratified based on basal/luminal and AR-A signatures. Basal (n = 112, 50%) and AR-A low (n = 96, 43%) subtypes, known to be nonresponsive to ADT, both showed significant benefit from APA vs PBO (0.30 [0.16-0.57]; &lt; 0.001 and 0.25 [0.12-0.52]; &lt; 0.001, respectively). The majority of AR-A low subtype (74%, 71/96) overlapped with basal subtype. Further conclusions for risk of rPFS in GC low, luminal, and AR-A high subtypes and OS across all subtypes will be assessed as more events occur. Conclusions: In TITAN, addition of APA to ADT improved rPFS for all subtypes of pts with mCSPC. APA overcame the poor prognosis of GC high risk subtype and prolonged rPFS in ADT-resistant AR-A low and basal molecular subtypes, suggesting APA is beneficial especially for the highest risk molecular subtypes. Clinical trial information: NCT02489318 .

Immunogenicity and transcriptomic biology in Nottingham histologic grade 3 breast cancer.

e12563 Background: Cancer biology dictates disease progression and prognosis of a tumor. Histological grade has been used as a proxy of tumor biology in various cancer types, although it depends on morphological analysis of the tissue. As such, Nottingham histological grade is recently incorporated into the 8th edition of the AJCC Breast Cancer Staging. Herein, we hypothesized that Nottingham histological Grade3 breast cancer (BC) demonstrate strong immunogenicity in response to aggressive phenotypes, reflecting worse clinical outcomes. Methods: Clinicopathological data and transcriptomes were obtained from Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), The Cancer Genome Atlas (TCGA) Invasive breast carcinoma (BRCA) cohort, and GSE25066. While METABRIC was used as a training cohort, the other two cohorts served as validation cohorts. As TCGA did not contain pathological information, Nottingham histological grade was manually collected through the Text Information Extraction Systems (TIES). Results: 2876 patients were analyzed in this study. Grade1 tumors were 274 patients (9.5%), Grade2 tumors 1185 patients (41.2%), and Grade3 tumors 1417 patients (49.3%). Grade3 tumors were common in patients with Estrogen receptor (ER) negative (p &lt; 0.001), with Her2 and Basal molecular subtypes by the PAM50 classifier (p &lt; 0.001), and associated with increased MKI-67 expression (p &lt; 0.001). Disease-free survival (DFS) was significantly worse in Grade3 tumors in all cohorts. Gene set enrichment analysis (GSEA) demonstrated that Grade3 tumors significantly enriched immune activity related gene sets as well as cell proliferation and cell cycle related gene sets. CIBERSORT also demonstrated that Grade3 BCs were infiltrated with anti-cancer macrophage M1, follicular helper T cells, and activated NK cells (all p &lt; 0.001). Furthermore, Grade3 tumors were associated with more diverse T cell receptor (p = 0.001) and increased cytolytic activity (p &lt; 0.001). Lastly, T cell exhaustion markers, including PD-L1, PD-1, and CTLA4, were significantly elevated in Grade3 BCs (all p &lt; 0.001) as a negative feedback loop. Conclusions: Grade3 BCs demonstrated enhanced immunogenicity as well as aggressive transcriptomic features.

A comprehensive analysis of IDO1 expression with tumour-infiltrating immune cells and mutation burden in gynaecologic and breast cancers.

Gynaecologic and breast cancers share some similarities at the molecular level. The aims of our study are to highlight the similarities and differences about IDO1, an important immune‐related gene in female cancers. The NGS data from TCGA of cervical squamous cell carcinoma (CESC), ovarian serous cystadenocarcinoma (OV), uterine corpus endometrial carcinoma (UCEC), uterine carcinosarcoma (UCS) and breast invasive carcinoma (BRCA) were analysed to identify molecular features, and clinically significant and potential therapeutic targets of IDO1. We found IDO1 was significantly up‐regulated in four gynaecologic cancers and breast cancer. According to breast cancer PAM50 classification scheme, IDO1 expression was higher in tumours of basal than other subtypes and showed better survival prognosis in BRCA and OV. Through immune infiltration analysis, we found a strong correlation between IDO1 and immune cell populations especially for dendritic cells and T cells. In addition, we investigated the association between IDO1 and tumour mutation burden (TMB) and found that IDO1 was significantly correlated with TMB in BRCA and CESC. GSVA revealed that hallmarks significantly correlated with IDO1 were involved in interferon gamma response, allograft rejection and inflammatory response. We also found PD‐L1 and LAG3 were highly positive related to IDO1 in gynaecologic cancers when comparing with their corresponding normal tissues. Our results indicated that IDO1 participated in anti‐tumour immune process and is correlated with mutation burden. These findings may expand our outlook of potential anti‐IDO1 treatments.

The LSD1 inhibitor iadademstat (ORY-1001) targets SOX2-driven breast cancer stem cells: a potential epigenetic therapy in luminal-B and HER2-positive breast cancer subtypes.

SOX2 is a core pluripotency-associated transcription factor causally related to cancer initiation, aggressiveness, and drug resistance by driving the self-renewal and seeding capacity of cancer stem cells (CSC). Here, we tested the ability of the clinically proven inhibitor of the lysine-specific demethylase 1 (LSD1/KDM1A) iadademstat (ORY-100) to target SOX2-driven CSC in breast cancer. Iadademstat blocked CSC-driven mammosphere formation in breast cancer cell lines that are dependent on SOX2 expression to maintain their CSC phenotype. Iadademstat prevented the activation of an LSD1-targeted stemness-specific SOX2 enhancer in CSC-enriched 3-dimensional spheroids. Using high-throughput transcriptional data available from the METABRIC dataset, high expression of SOX2 was significantly more common in luminal-B and HER2-enriched subtypes according to PAM50 classifier and in IntClust1 (high proliferating luminal-B) and IntClust 5 (luminal-B and HER2-amplified) according to integrative clustering. Iadademstat significantly reduced mammospheres formation by CSC-like cells from a multidrug-resistant luminal-B breast cancer patient-derived xenograft but not of those from a treatment-naïve luminal-A patient. Iadademstat reduced the expression of SOX2 in luminal-B but not in luminal-A mammospheres, likely indicating a selective targeting of SOX2-driven CSC. The therapeutic relevance of targeting SOX2-driven breast CSC suggests the potential clinical use of iadademstat as an epigenetic therapy in luminal-B and HER2-positive subtypes.

Luminal B subtype as a predictive biomarker of docetaxel benefit for newly diagnosed metastatic hormone sensitive prostate cancer (mHSPC): A correlative study of E3805 CHAARTED.

162 Background: Through gene expression profiling (GEP), the PAM50 classifier demonstrates prognostic value in localized prostate cancer (PCa). Pre-clinical drug response models predict increased taxane sensitivity in luminal subtypes compared to basal subtype. Men with mHSPC and high-risk features have greatest benefit from androgen deprivation therapy (ADT) plus docetaxel (D) vs ADT alone. We therefore sought to test the prognostic and predictive value of PAM50 in pre-ADT specimens from E3805 CHAARTED. Methods: Whole transcriptomic profiling of formalin-fixed, paraffin-embedded primary PCa biopsies from pts enrolled in the E3805 CHAARTED trial of ADT vs ADT+D was performed using the Human Exon 1.0 ST microarray platform (Decipher Biosciences). Normalized gene expression was used to classify subjects as luminal A, luminal B or basal subtype. Multivariable analyses (MVA) adjusted for ECOG status, de novo metastasis vs prior local therapy and volume of disease. The primary endpoint was overall survival (OS). Secondary endpoint was time to castration resistant PCa (TTCRPC). Results: Successful GEP was completed in 160 of 198 pts with available specimens. Eighty (50%), 77 (48%) and 3 (2%) pts were classified as luminal B, basal and luminal A, respectively. High volume disease was similarly present in luminal B (79%) and basal (78%) subtypes. In the ADT arm, luminal B subtype was associated with shorter OS vs basal (HR 1.75, p=0.05); consistent in MVA. Pts with luminal B subtype treated with ADT+D showed significant improvement in TTCRPC and OS (Table). By contrast, basal subtype showed no OS benefit from ADT+D even in pts with high volume disease. Conclusions: We demonstrate that GEP identifies tumor subtypes associated with differential benefit from chemohormonal therapy for mHSPC. Luminal B subtype is associated with poorer OS with ADT alone and benefits from addition of D. Basal subtype shows a lack of OS benefit from upfront ADT+D. We plan to validate these findings in independent trial cohorts.[Table: see text]

Multiregion expression profiling of prostate cancer from men randomized in the STAMPEDE trial: Stage I results of a multistage biomarker analysis.

153 Background: The PAM50 gene expression classifier stratifies localized prostate cancer into luminal A/B and basal subtypes: luminal A have superior prognosis compared to luminal B and basal. Drug response scores built using the NCI-60 cell lines predict luminal subtypes are more taxane sensitive compared to basal; luminal B is hypothesized to be the most proliferative and hormone responsive compared to the basal subtype. The STAMPEDE trial provides a framework to validate prognostic and predictive associations of the PAM50 test in tumors from men randomized to androgen deprivation therapy (ADT) alone or with docetaxel or abiraterone. We here present multi-region whole-transcriptome expression array data at completion of Stage I designed to confirm the feasibility of molecular subtyping STAMPEDE tumor blocks. Methods: In collaboration with Decipher Biosciences we generated clinical-grade whole-transcriptome expression array data (Human Exon 1.0 ST GeneChip) performed to CLIA standards on mRNA (from three 10 micron slides) from cancer-enriched areas and applied the PAM50 classifier. Results: As of January 2019, we retrieved blocks from 2012 men from an ITT population of 3879. For Stage I feasibility assessment, we sectioned diagnostic trans-rectal biopsies of the prostate from 50 randomly selected men treated with ADT. We extracted mRNA from 109 cores; &gt; 92% cores (101) from &gt; 94% patients (47/50) passed quality control. The prevalence of PAM50 subtypes was: 31 basal cases (62%), 18 luminal B (36%) and 1 luminal A (2%). 26 cases (52%) were identified as ETS-related gene (ERG) positive. More than one core was interrogated from 35/50 cases (range: 1-6 cores). ERG positive cores were homogeneous across cores from the same patient but PAM50 subtyping identified intra-patient variability across cores from the same case. Conclusions: We confirm the feasibility of expression profiling STAMPEDE tumor blocks and identify a low prevalence of luminal A subtype. Whereas ERG status is consistent across multiple cores, basal and luminal subtypes co-exist in the same prostate. Multi-region analysis will enable further refinement for individual patient classification.

Abstract P4-06-14: Heterogeneity of multigene signatures in breast cancer among multiple specimens taken by core needle biopsy

Abstract Use of multigene predictive signatures pre-chemotherapy in core needle biopsy specimens in low-resource systems in not applied due to multiple reasons. We are seeking a method of cost-effective integration of gene expression analysis into initial breast cancer subtype assessment pre-chemotherapy.The aim of the analysis was to assess the heterogeneity of independent cores taken from primary tumor in the context of prediction of its clinical parameters by gene expression profiling.Material and methods. The study group consisted of 65 women with high-risk breast cancer, before preoperative chemotherapy, median age at the start of therapy 41,2 years (range 29-62 yr). Among the diagnosed breast cancers, 71% were grade 3 (G3), while G1 only in 3% of individuals. The most common breast cancer biological subtypes were triple negative breast cancer (38%), luminal B HER2-negative (defined as Ki67&amp;gt;20% or PR-negative, 24%) and HER2-positive tumors (23%) of tumors were of lobular histology. In core needle biopsies taken before preoperative breast cancer chemotherapy gene expression profiling was carried out by oligonucleotide microarrays (U133A/U133 Plus 2.0), three independent cores were analyzed (192 microarrays). Analysis was carried out by JMP Genomics (SAS Institute, Cary, USA) and R/Bioconductor genefu package.At first step, the overall tumor profile was analysed by Principal Component Analysis. Based on first two principal components, 44 patients (68%) showed highly concordant profile in all three cores, whereas in 21 tumors (32%) larger scatter was observed. We analysed research versions of commercially available multigene predictors, PAM50 are presented in abstract. Prediction was done independently for each core. In 40% of patients all three cores provided identical predictions for all analysed signatures (mainly triple-negative tumors). In 17% of patients we observed minor discordances (in one core 1 or 2 signatures discordant), in 10% discordances in 2 cores. Profound differences were found in 33% of patients (similar to overall profile scatter seen by PCA), mainly in luminal and HER2-positive tumors. Subtypes predicted by genomic testing were - as expected - slightly different from immunohistochemistry prediction. The most often occurring was Basal subtype (87% of TNBC patients, but occurring also in HER2-positive tumors, 20-25% and luminal B tumors, 26.7%). The second molecular subtype was LumB. These tumors constituted 60% of clinically diagnosed luminal B patients, molecular LumB subtype was also frequent among clinically luminal B HER2-positive individuals (30%). Genomic LumA subtype was found in all clinical subgroups, in 10-25% of these patients. Her2 subtype by PAM50 classification was occurring less frequently, in 50% of HER2-nonluminal and 40% luminal B HER2-positive tumors. The concordance of results improved, when classification included both immunohistochemistry results as well as the results of genomic measurements. ER and HER2 showed excellent concordance between three tumor cores and showed bimodal expression, while MKI67 and PGR although concordant showed continuous pattern of expression rather than any visible subgroups. Conclusions. Significant heterogeneity is seen in genomic profile of at least 1/3 of tested breast cancers and shall be taken into account. The error of sampling was lowest in triple negative subtype Citation Format: Michal Jarzab. Heterogeneity of multigene signatures in breast cancer among multiple specimens taken by core needle biopsy [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P4-06-14.

Abstract P4-07-06: Breast cancer subtype classification using NanoString and RNAseq technologies

Abstract Background: Immunohistochemistry (IHC) is conventionally used in clinical practice to define Breast Cancer (BC) subtypes, while PAM50 is useful in specific cases. The former is partly subjective and semi-quantitative, the latter is based on gene expression profile. We compared gene expression data obtained by NanoString BC 360™ panel with those obtained by RNAseq on 12 Formalin-Fixed Paraffin-Embedded (FFPE) BC samples. Methods: RNA was isolated from FFPE tumors and the quality was checked before performing NanoString BC 360™ assay and preparing RNAseq library (NEBNext Ultra II RNA Library Prep Kit, Illumina) following the manufacturers’ instructions. Libraries were sequenced on NextSeq500 (Illumina). Reads were aligned using Kallisto and raw read counts were normalized as Transcripts Per Million (TPM). PAM50 and Tumor Inflammation Signature (TIS) were determined on NanoString data. Two-sided t-test on TPM and on normalized NanoString counts was used to compare luminal and Triple Negative (TN) BCs (p &amp;lt;0.05 statistically significant), considering the 734 genes shared by the 2 assays. Deseq2 (1.22.1) was used to perform a differential gene expression analysis on raw RNAseq counts. Enrichment tests were performed on KEGG pathways (FDR &amp;lt; 0.05). Python and R softwares were used for statistical analyses. Results: Of the 12 BC samples, 3 were luminal A-like, 3 luminal B (HER2+)-like and 6 TN by IHC (Table 1). Pearson’s correlation coefficient between NanoString counts and RNAseq TPM was 0.72. Out of the 734 genes shared by both methods, those differentially expressed between luminal and TN BCs, defined based on IHC, were 135 according to RNAseq TPM, 155 according to NanoString counts and 88 according to both methods. These 88 genes belonged to “cell cycle”, “pathways in cancer”, “HIF-1 signaling”, “progesterone-mediated oocyte maturation”, “prostate cancer” and “basal cell carcinoma” KEGG pathways. When considering the whole transcriptome analyzed by RNAseq, the differential gene expression analysis between luminal and TN/basal BCs was performed according to both IHC and PAM50 classifications. Considering the former, there were 1143 differentially expressed genes out of a total of 20936 (5.5%) and the pathways involved were “fatty acid elongation”, “cell cycle”, “oocyte meiosis”, “p53 signaling” and “alcoholism”. The analysis on the latter classification showed 1608 differentially expressed genes out of 21048 (7.6%), and the enriched pathways were “cell cycle”, “p53 signaling”, “DNA replication”, “alcoholism” and “systemic lupus erythematosus”. Regarding the TIS, no significant differences were found between TN/basal and luminal BCs, comparing both IHC and the PAM50 subtypes. An expected heterogeneity was found among TNBCs, that showed 3 distinct signatures: basal-like immune activated, basal-like immune-suppressed and luminal androgen receptor. Conclusions: This preliminary analysis shows a good concordance between NanoString and RNAseq data. The choice of the method must take into account important factors such as the costs, the number of genes analyzed, the requirement of bioinformatic support, the working time. Further analyses on pathway enrichment will be presented. Table 1. Breast cancer samples characteristics.Sample ID Age Grade T N IHC subtype PAM50 Subtype TIS scoreTNBC signature 0172222aLumALumB7,0NA026031c2aTNBasal7,2BLIA037221cxTNLumA7,4LAR045231c0LumALumA7,5NA053531c1aTNBasal5,5BLIS0675320TNBasal6,2BLIS077931cxLumB HER2+LumB6,7NA0878220LumALumB7,7NA097431a2aTNBasal8,5BLIA106031c1LumB HER2+LumA8,2NA116621b0LumB HER2+LumA6,1NA1241321aTNBasal7,5BLIATIS: Tumor Inflammation Signature; BLIA: Basal-like Immune Activated; BLIS: Basal-like Immune-Suppressed; LAR: Luminal Androgen Receptor. Citation Format: Andrea Rocca, Sara Ravaioli, Eugenio Fonzi, Iros Barozzi, Ylenia Perone, Luca Magnani, Francesca Pirini, Giovanni Martinelli, Sara Bravaccini. Breast cancer subtype classification using NanoString and RNAseq technologies [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P4-07-06.