FFPE Samples Research Articles

5hmC-profiles in Puerto Rican Hispanic/Latino men with aggressive prostate cancer.

Puerto Rican (PR) Hispanic/Latino (H/L) men are an understudied population that has the highest prostate cancer (PCa) specific mortality among other Hispanic populations. Little information is known about the higher mortality in PR H/L men. It is thought that epigenetic changes in key genes may play a critical role in aggressive tumors. We aimed to identify key 5-hydroxymethylcytosine (5hmC) changes in PR H/L men with aggressive PCa. We performed sequencing analysis using the 5hmC-enriched DNA from 22 prostate tumors and 24 adjacent normal FFPE samples. We identified 808 differentially methylated genes (DMGs) in tumors compared to adjacent normal tissues (FDR<0.05, log2FC>|0.4|). Pathway analysis of DMGs demonstrated that DNA repair pathway was most upregulated in tumors. Since 5hmC abundance positively correlates with gene expression levels, we further investigated 808 DMGs in TCGA PCa gene expression data. Further, we identified 59 DMGs (80.1%, FDR<0.05, ΔGE (gene expression) >|1|) with significant gene expression changes in the same direction. Additionally, we identified 111 aggressiveness-related DMGs, of which, two hypomethylated genes ( CCDC122 , NUDT15 ) and four hypermethylated genes ( PVT1 , RPL30 , TRMT12 , UBR5 ) were found to be altered at transcriptomic level in a concordant manner in PR H/L PCa patients (N=86). The aberrant 5hmC (N=55) and GE (N=497) changes in these six genes were also associated with progression-free survival in the mixed PCa population. In conclusion, our study identified 59 DMGs showing concordant epigenetic and transcriptomic changes in tumor tissues and 111 DMGs showing association with aggressive PCa among PR H/L men.

Evaluation of blood MSI burden dynamics to trace immune checkpoint inhibitor therapy efficacy through the course of treatment.

Analysis of serial liquid biopsy (LB) samples has been found to be a promising approach for the monitoring of tumor dynamics in the course of therapy for patients with colorectal cancer (CRC). Currently, somatic mutations are used for tracing the dynamics of the tumor via LB. However, the analysis of the dynamic changes in the molecular signatures such as microsatellite instability (MSI) is not currently used. We hypothesized that changes in blood MSI burden (bMSI) could be registered using serial LB sampling in the course of immune checkpoint inhibitors (ICI), and that its changes could potentially correlate with treatment outcomes. We report the preliminary findings of the observational trial launched to study (NCT06414304) the dynamics of bMSI in 9 MSI-positive CRC patients receiving ICI. NGS-based MSI testing was performed on both pre-treatment FFPE and serial LB samples. For patients who had detectable bMSI burden in any of the LB samples (n = 8, 89%), median bMSI was 1.4% (range, 0.01-40%). Among patients with detectable MSI in available FFPE samples, median MSI burden was 29.3% (range, 10-40%). bMSI detected in baseline LB and FFPE samples were positively correlated (Pearson's R 0.47). Maximal variant allele frequencies of driver mutations observed in LB were also positively correlated with bMSI burden (Pearson's R 0.7). Patients who had clinical benefit had undetectable bMSI burden at follow-up. Our results provide the rationale for further validation of bMSI as a predictive biomarker of ICI in MSI-positive patients.

Medullary thyroid cancer: single-cell transcriptome and tumor evolution

BackgroundMedullary thyroid cancer (MTC) is a rare neuroendocrine tumor that originates from the parafollicular C cells of thyroid gland. Understanding the fundamental pathophysiology of MTC is essential for clinical management. Single-cell RNA sequencing (scRNA-seq) technology is a powerful tool for identifying distinct cell types, offering a new biological foundation for comprehending the MTC ecosystem and developing precise treatment.MethodsFormalin fixed and paraffin-embedded (FFPE) samples of primary and adjacent non-cancerous tissues of three MTC cases were collected, and single-cell transcriptome data of MTC were obtained by using scRNA-seq technology. Annotated cell subpopulations were categorized and functionally enriched by principal component analysis, differential gene expression, and cell clustering analysis, to explore the biological process of tumor evolution that may be involved in each cell subpopulation. The copy number variation (CNV) profile was used to distinguish the malignancy of parafollicular thyroid cells, and the evolutionary trajectories of normal cells and tumor cells were revealed by the proposed time series analysis. The highly expressed genes in each cell subpopulation were analyzed by the FindAllMarker function of Seurat software, and verified by immunohistochemistry and fluorescence in situ hybridization. The prognostic value of specific cell subtypes was validated using large-scale public datasets.ResultsA total of 32,544 cells were obtained from the MTC tissue samples and 11,751 cells from the adjacent non-cancerous samples, which were classified into 7 heterogenous subpopulations by using R package of Seurat module. Copy number variations (CNVs) were significantly higher in tumor tissues than in adjacent non-tumor samples, predominantly enriched in subtypes C2 and C4. In addition, the pseudo-time for trajectory analysis suggested that the evolution of MTC tumor cells might begin with the C2 subtype, then transition to the early cancer subgroup C3, and further differentiate into four major malignant cell subpopulations C0, C1, C5 and C6. Survival analysis of a thyroid cancer cohort using the TCGA dataset revealed that high expression of genes linked to the C0 subcluster was correlated with poorer overall survival compared to low expression. Immunohistochemical staining showed that MAP3K4 was highly expressed in MTC tissues compared to adjacent non-cancerous tissues. Fluorescence in situ hybridization also confirmed the amplification of these two genes in MTC samples.ConclusionsBy conducting scRNA-seq on FFPE samples, we mapped the single-cell transcriptome of MTC, uncovering the tumor heterogeneity and unique biological features of each cellular subpopulation. The biological roles of identified tumor cell subpopulations such as C0 and C3 subtypes of parafollicular cells suggested the potential to discover new therapeutic targets and biomarkers for MTC, providing valuable insights for future translational and clinical research.

Dissecting AI-based mutation prediction in lung adenocarcinoma: A comprehensive real-world study

IntroductionMolecular profiling of lung cancer is essential to identify genetic alterations that predict response to targeted therapy. While deep learning shows promise for predicting oncogenic mutations from whole tissue images, existing studies often face challenges such as limited sample sizes, a focus on earlier stage patients, and insufficient analysis of robustness and generalizability. MethodsThis retrospective study evaluates factors influencing mutation prediction accuracy using the large Heidelberg Lung Adenocarcinoma Cohort (HLCC), a cohort of 2356 late-stage FFPE samples. Validation is performed in the publicly available TCGA-LUAD cohort. ResultsModels trained on the larger HLCC cohort generalized well to the TCGA dataset for mutations in EGFR (AUC 0.76), STK11 (AUC 0.71) and TP53 (AUC 0.75), in line with the hypothesis that larger cohort sizes improve model robustness. Variation in performance due to pre-processing and modeling choices, such as mutation variant calling, affected EGFR prediction accuracy by up to 7 %. DiscussionModel explanations suggest that acinar and papillary growth patterns are critical for the detection of EGFR mutations, whereas solid growth patterns and large nuclei are indicative of TP53 mutations. These findings highlight the importance of specific morphological features in mutation detection and the potential of deep learning models to improve mutation prediction accuracy. ConclusionAlthough deep learning models trained on larger cohorts show improved robustness and generalizability in predicting oncogenic mutations, they cannot replace comprehensive molecular profiling. However, they may support patient pre-selection for clinical trials and deepen the insight in genotype-phenotype relationships.

Novel liquid biopsy CNV biomarkers in malignant melanoma

Malignant melanoma (MM) is known for its abundance of genetic alterations and a tendency for rapid metastasizing. Identification of novel plasma biomarkers may enhance non-invasive diagnostics and disease monitoring. Initially, we examined copy number variations (CNV) in CDK genes (CDKN2A, CDKN2B, CDK4) using MLPA (gDNA) and ddPCR (ctDNA) analysis. Subsequently, low-coverage whole genome sequencing (lcWGS) was used to identify the most common CNV in plasma samples, followed by ddPCR verification of chosen biomarkers. CNV alterations in CDK genes were identified in 33.3% of FFPE samples (Clark IV, V only). Detection of the same genes in MM plasma showed no significance, neither compared to healthy plasmas nor between pre- versus post-surgery plasma. Sequencing data showed the most common CNV occurring in 6q27, 4p16.1, 10p15.3, 10q22.3, 13q34, 18q23, 20q11.21-q13.12 and 22q13.33. CNV in four chosen genes (KIF25, E2F1, DIP2C and TFG) were verified by ddPCR using 2 models of interpretation. Model 1 was concordant with lcWGS results in 54% of samples, for model 2 it was 46%. Although CDK genes have not been proven to be suitable CNV liquid biopsy biomarkers, lcWGS defined the most frequently affected chromosomal regions by CNV. Among chosen genes, DIP2C demonstrated a potential for further analysis.

Combination of MRI-based prediction and CRISPR/Cas12a-based detection for IDH genotyping in glioma

Early identification of IDH mutation status is of great significance in clinical therapeutic decision-making in the treatment of glioma. We demonstrate a technological solution to improve the accuracy and reliability of IDH mutation detection by combining MRI-based prediction and a CRISPR-based automatic integrated gene detection system (AIGS). A model was constructed to predict the IDH mutation status using whole slices in MRI scans with a Transformer neural network, and the predictive model achieved accuracies of 0.93, 0.87, and 0.84 using the internal and two external test sets, respectively. Additionally, CRISPR/Cas12a-based AIGS was constructed, and AIGS achieved 100% diagnostic accuracy in terms of IDH detection using both frozen tissue and FFPE samples in one hour. Moreover, the feature attribution of our predictive model was assessed using GradCAM, and the highest correlations with tumor cell percentages in enhancing and IDH-wildtype gliomas were found to have GradCAM importance (0.65 and 0.5, respectively). This MRI-based predictive model could, therefore, guide biopsy for tumor-enriched, which would ensure the veracity and stability of the rapid detection results. The combination of our predictive model and AIGS improved the early determination of IDH mutation status in glioma patients. This combined system of MRI-based prediction and CRISPR/Cas12a-based detection can be used to guide biopsy, resection, and radiation for glioma patients to improve patient outcomes.

A random forest model to predict the efficacy of anti-PD-1/PD-L1 monoclonal antibody in lung adenocarcinoma.

e20558 Background: Not all patients with lung adenocarcinoma (LUAD) could benefit from anti-PD-1/PD-L1 monoclonal antibody therapy, and existing biomarkers cannot explain all clinical benefits. Therefore, an effective model to predict the efficacy of anti-PD-1/PD-L1 monoclonal antibody is necessary. Methods: DNA methylation sequencing was used in LUAD patients’ FFPE samples, who received anti-PD-1/PD-L1 monoclonal antibody. The DNA methylation level of each gene was transformed into binary format according to its median value. The progression-free survival (PFS) was defined as the duration from the beginning of anti-PD-1/PD-L1 monoclonal antibody therapy to disease progression or death. Patients were randomly divided into training cohort (70%) and test cohort (30%). In the training cohort, uni-Cox regression was used to estimate the value in predicting PFS, and LASSO-Cox regression was performed to screened genes to build random forest model. The accuracy of this random forest model was validated in the test cohort. Finally, DNA methylation data were used to calculated the risk score of patients in the TCGA-LUAD cohort, and GSEA and GSVA were employed in RNA data to explore the mechanism and explain the efficacy of the prediction model. Results: In this study, 40 patients with LUAD were enrolled, and the median PFS of them was 7.57 months. A total of 29 genes with P &lt; 0.05 and C-index &gt; 0.7 were selected to perform LASSO-Cox regression, and 13 genes ( BTBD7, CAD, FBXL20, FKBP7, HBP1, IGFBP3, RAB5B, RASGRP1, RNF8, SOHLH1, TMEM223, UHRF1 and WWC2) were used to build prediction model. The C-index of risk score from this model was 0.96 and 0.93 in the training and test cohort, respectively. Patients were divided into the high and low risk groups according to the median risk score, and the median PFS of the high risk group was shorter than that of the low risk group (training cohort: 3.05 vs. 10.89 months, P &lt; 0.01, HR = 2.51, 95% CI: 1.50-4.22; test cohort: 1.64 vs. 11.31 months, P = 0.03, HR = 1.80, 95% CI: 1.04-3.14). In the low and high risk group of the TCGA-LUAD cohort, there were 92.2% (388/421) and 0.5% (2/421) significant enriched ( P &lt; 0.05 and FDR &lt; 0.25) immunity-related pathways from results of GSEA, respectively, meanwhile 94.3% (397/421) and 1.9% (8/421) significant enriched immunity-related pathways from results of GSVA, respectively. Therefore, the lower risk score of this random forest model accompanied with the higher activation of immunity-related pathways which might help anti-PD-1/PD-L1 monoclonal antibody to express better efficacy. Conclusions: A random forest model based on DNA methylation data was constructed and showed a potential value in predicting the PFS of anti-PD-1/PD-L1 monoclonal antibody therapy in LUADs, and the risk score of this model is negatively correlated to the activation of immunity-related pathways, which could help to identify LUAD patients that would benefit from immunotherapy.

Optimization of fusion detection in non-small cell lung cancer by amplicon-based RNA sequencing.

e20042 Background: In non-small cell lung cancer (NSCLC), more and more driver fusions are suitable for targeted therapy. Sequential or parallel DNA and RNA sequencing is an option, but it is costly and requires sufficiently high quality biopsy material. The purpose of this study was to confirm the consistency of fusion genes detected by DNA-based NGS and RNA-based NGS in the same sample, and to validate the performance of FFPE samples for RNA-based NGS. Methods: In this study, we evaluated the benefits of capture-based DNA sequencing and amplicon-based RNA sequencing in identifying gene fusion and exon-skipping events in 98 NSCLC patients. All cases were FFPE samples stored for 1-5 years. In addition, we investigated the limiting performance of RNA-based NGS for RNA initiation. Results: In the cases of positive DNA sequencing fusion samples, the fusion consistency detection rate of RNA sequencing was 95% (38/40). Two cases were not detected by RNA-based NGS, one was a five-year FFPE sample and the other was due to a probe not covered problem. In patients with negative DNA fusion genes, RNA sequencing could detect 12.07% more fusion-positive cases involving ALK (n = 1), ROS (n = 3), MET 14 exon skipping (n = 1), RET (n = 1), and BRAF (n = 1). Moreover, the FFPE samples stored for 3-5 years inevitably degrade the RNA, but could still be used for RNA-based sequencing. Among 66 FFPE samples stored for 3-5 years, the detection rate of fusion consistency between DNA sequencing fusion samples and RNA sequencing was 96.77%(30/31). In patients with negative DNA fusion genes, RNA sequencing was able to detect 14.29% (5/35) more fusion-positive cases. The performance verification of the other 8 samples with different RNA input amounts found that 87.5% (7/8) of the samples could meet the quality control of RNA sequencing and accurately detect fusion mutations as low as 1 ng. Conclusions: Molecular profiling based on mixed DNA and RNA sequencing improves the effectiveness of detection of fusion-driven NSCLC. The method was feasible on small tissue samples as well as FFPE samples stored for 3-5 years, and could greatly reduce the complexity and cost of molecular examination.

Expression of alternatively spliced BCMA RNA with skipping of transmembrane domain.

7514 Background: The TNFRSF17 gene, which encodes for the B cell maturation antigen (BCMA), is a small gene expressed mainly on the surface of plasma cells and some DLBCL cells. This gene is composed of three exons. Exon 1 is responsible for the extracellular domain, exon 2 for the transmembrane domain, and exon 3 for the TRAF binding domain. The BCMA is currently targeted by various types of immunotherapies as a major therapeutic approach for the treatment of multiple myeloma (MM). This includes CAR-T cells, bi-specific antibodies, and antibody-drug conjugates (ADC). However, emerging data indicates that alternative splicing in genes is an important mechanism for expression of different isoforms that may influence antibody-based therapies. We explored the potential of the presence of alternative splicing in BCMA transcripts via sequencing of BCMA RNA in patients with lymphoma or multiple myeloma. Methods: RNA was extracted from 587 fresh bone marrow samples with lymphoid/plasma cell neoplasms or from FFPE samples with lymphoma. In addition, cfRNA was extracted from 260 peripheral blood plasma samples from patients with lymphoma or multiple myeloma. RNA was sequenced using a hybrid capture-targeted RNA panel with analysis focused on TNFRSF17 (BCMA) gene transcript. Quantification of RNA transcript was done using Salmon algorithm. Results: Of the 587 lymphoma/plasma cell samples, 161 (27%) samples showed alternative splicing involving deletion of exon 2 (BCMA∆Ex2). Of the 260 cfRNA samples, 14 (6%) showed BCMA∆Ex2. The median percentage of BCMA∆Ex2 transcripts was 0.7% of total BCMA transcripts in cellular samples as compared with 9% in cfRNA samples. In cellular samples, there was a correlation between levels of BCMA and BCMA∆Ex2 (R=0.63). Cases with higher levels of BCMA had significantly higher levels of BCMA∆Ex2 (P&lt;0.0001, Kruskal-Wallis). In contrast, cfRNA showed no correlation between levels of BCMA and levels of BCMA∆Ex2 (R=021, Spearman) and mildly higher level of BCMA∆Ex2 in cases with higher levels of BCMA (P=0.002, Kruskal-Wallis). Conclusions: BCMA exon 2 skipping is detected in a significant number of patients with multiple myeloma and lymphoma. The percentage of skipping transcripts is low in cells and relatively higher in cfRNA. Since exon 2 skipping deletes the transmembrane domain, this BCMA protein may remain in the cytoplasm or perhaps is secreted as cell-free BCMA protein. The demonstration of relatively higher levels of BCMA∆Ex2 in cfRNA is likely reflecting higher turnover of cells and raises the possibility that cells with this isoform of BCMA are more aggressive than cells without the expression of this isoform. Further studies are needed to explore the clinical relevance of the expression of such abnormal BCMA protein on treatment with the various forms of anti-body-based therapy or CAR-T.

Expression of alternatively spliced PD-L1 RNA with skipping of transmembrane domain.

e14596 Background: Programmed death-ligand 1 (PD-L1) is a transmembrane ligand for the programmed cell death protein 1 (PD-1), a receptor that inhibits T-cell activity. The PD-L1/PD-1 immune checkpoint axis has been successfully targeted to enhance antitumor immune responses. This gene is composed of seven exons. Exons 1-4 are responsible for the extracellular domain, exon 5 for the transmembrane domain, and exons 6-7 for the intracellular domain. Emerging data indicates that alternative splicing in genes is an important mechanism for expression of different isoforms that may influence antibody-based therapies. We explored the potential of the presence of alternative splicing in PD-L1 transcripts via sequencing of PD-L1 RNA in patients with various type of solid tumor cancers. Methods: RNA was extracted from 3443 FFPE solid tumor samples including lung, colon, pancreas, breast, and other cancer types. In addition, cfRNA was extracted from 532 peripheral blood plasma samples from patients with solid tumor cancers. RNA was sequenced using a hybrid capture targeted RNA panel on FFPE samples and peripheral blood with analysis focused on PD-L1 gene transcript. Quantification of RNA transcript was done using Salmon algorithm. Results: Of the 3443 FFPE samples, 38 (1.2%) samples showed alternative splicing involving deletion of exon 5 (PD-L1∆Ex5). The alternative splicing was in 2.10% of lung cancer, 2% of colon cancer, 0.70% of pancreatic cancer, 1.40% of gastric cancer, 0.38% of breast cancer, 0.30% of ovarian cancer, 1.10% of sarcoma cases, 1.50% of glioblastomas, 3.30% of thyroid cancer and 2% of urothelial cancer. Of the 532 cfRNA samples, 20 (3.7%) showed PD-L1∆Ex5. The median percentage of PD-L1∆Ex5 transcripts was 2.2% of total PD-L1 transcripts in tissue samples as compared with 6.6% in cfDNA samples. In tissue samples, there was a correlation between levels of PD-L1 and PD-L1∆Ex5 (r = .567, p &lt; .001). In contrast, cfRNA showed no significant correlation between levels of PD-L1 and levels of PD-L1∆Ex5 (r = .268). Conclusions: PD-L1 exon 5 skipping was detected in a number of patients with different types of solid tumor cancers. The percentage of skipping transcripts is low in cells and relatively higher in cfRNA. Since exon 5 skipping deletes the transmembrane domain, the PD-L1 protein may remain in the cytoplasm or perhaps is secreted as cell-free PD-L1 protein. The demonstration of relatively higher levels of PD-L1∆Ex5 in cfRNA is likely reflecting higher turnover of cells and raises the possibility that cells with this isoform of PD-L1 are more aggressive than cells without the expression of this isoform. Further studies are needed to explore the clinical relevance of the expression of such abnormal PD-L1 protein on treatment with immune checkpoint inhibitors.

Genomic instability in advanced non-small cell lung cancer (NSCLC) treated with maintenance durvalumab in the UNICANCER SAFIR02-Lung/ IFCT1301 trial.

2632 Background: Genomic instability is a hallmark of cancer that has been associated to immune evasion, cancer progression and multidrug resistance. In SAFIR02-Lung, advanced NSCLC pts were enrolled to receive a platinum-based chemotherapy. After 4 cycles, pts with no targetable alterations were oriented to the immunotherapy sub-study, and if stable or in response, were randomized between maintenance durvalumab or Standard of Care (SoC) (PMID 35802649). Durvalumab prolonged survival only in the PD-L1 ≥ 1% subgroup. In this study we investigated a genomic instability score based on somatic copy number alterations as determined by Comparative Genomic Hybridization (CGH) array as a biomarker for maintenance immunotherapy. Methods: SAFIR02-Lung (NCT02117167) is a multicentric, randomized phase 2 clinical trial. Tissue samples were obtained before induction chemotherapy. Extracted DNA from fresh frozen tumor was analyzed using high-resolution Affymetrix Cytoscan arrays, DNA from FFPE samples was analyzed using the Affymetrix Oncoscan CNV platform. The absolute copy-number (ACN) profile was generated by ASCAT v3.1 through EaCoN v0.3.6 ( https://github.com/gustaveroussy/EaCoN , PMID 36669143 ) on R v4.1.1 and then the genomic instability score was estimated as the sum of the segment length multiplied by the absolute difference between segment copy number and the ASCAT estimated sample ploidy. This sum is reported to the reference genome length. Progression Free Survival (PFS) was the primary endpoint, defined as the time from randomization until the date of objective radiological disease progression, clinical progression or death. The secondary endpoint was Overall Survival (OS). Results: CGH array data were available for 79 patients out of 183, 56 randomized to durvalumab and 23 to SoC. 47 were men, 72 had non-squamous histology and 12 had brain and 10 had liver metastasis. We divided patients in tertiles according to instability (INS) score. No correlation was seen between INS score and PD-L1 positivity (p = 0.5), or liver and brain metastasis. Pts with high INS had lower median PFS of 2.5 months (m) (95% CI 1.4 – 5.2) vs 4.3m (95% CI 2.6 – 13.7) for those with intermediate INS and 6.1m for those with low INS (95% CI 1.9 – 21.2), p 0.012. Similarly, low INS patients had longer OS of 29.6 m (95% CI 15.4 – NR) vs 17.9m for intermediate INS (95% CI 10 – NA) and 12.1m for high INS (95% CI 10.2 – 19.5), p = 0.039. No difference was seen in the control group (p=0.25 for PFS and p=0.4 for OS). A multivariable model showed that high INS was independently associated worse PFS (HR 2.60, 95% CI 1.36 - 4.98, p 0.004) and OS (HR 2.90, 95% CI 1.33 - 6.35, p 0.008). Conclusions: Genomic instability score based on CGH array-based copy number alterations profile is associated with worse PFS and OS in patients under response to chemotherapy who received durvalumab maintenance. Clinical trial information: NCT02117167 .

Defining low-HER2 cancers using RNA quantification by next generation sequencing.

e13044 Background: The recent success of treatment with HER2-directed antibody-drug conjugate T-DXd (trastuzumab-deruxtecan) in treating low-HER2 expression breast cancer raises the possibility of similar success in other tumors that express similarly low levels of HER2. HER2 amplification has been reported in various types of tumors at relatively low frequency. However, “low-HER2” expression in various types of tumors remains undefined. In this study, we utilized next generation sequencing to explore the levels of ERBB2 (HER2), ERBB3, ERBB4, PIK3CA, and ERS1 mRNA in various types of cancers and attempted to set limits reflecting distinct biology of tumors based of level of ERBB2 expression. Methods: RNA was extracted from FFPE samples from patients with cancers of breast (N=203), ovary (N=95), endometrium (N=75), and esophagus/stomach (N=46). RNA was sequenced via a 1408 gene panel using a standard hybrid capture approach. Results: Based on previous data correlating gene amplification with gene expression (DOI: 10.1002/imed.1051), we defined ERBB2 expression as high, low, and very-low when ERBB2 expression measured &gt;1500 TPM (transcripts per million), 800-1500 TPM, and &lt;800 TPM, respectively. ERBB2 high, low, and very-low were detected in 20%, 32%, and 48% of breast cancer; 5%, 29%, and 65% of endometrial cancer; 9%, 26%, and 65% of ovarian cancer; 12%, 16%, and 72% of esophageal/stomach cancer; and 8%, 19%, and 73% of lung cancer. With minor exceptions, cases classified as ERBB2 high and low from the five tumor types showed no correlation with levels of PIK3CA, ESR1, or ERBB4 mRNA. In contrast, cases with relatively very-low ERBB2 showed significant correlation with PIK3CA, ESR1, and ERBB4 mRNA. Endometrial cancer was an exception and showed no correlation with PIK3CA mRNA in very-low ERBB2 cases. Activation of PIK3CA has been reported to be associated with resistance to anti-HER2 therapy. These findings suggest that cases showing high or low ERBB2 are independent of PIK3CA and likely less influenced by the negative effects of PIK3CA on response to HER2 inhibitors. Cases classified as ERBB2 high or low showed significantly higher (P&lt;0.0001) PIK3CA mutations as compared with ERBB2 very-low cases (27% vs 14%). ERBB2 mutations were relatively rare, detected in only 7%, 8%, 14%, 1%, and 4% of breast, endometrial, esophageal/stomach, ovarian, and lung cancers, respectively. However, significantly (P=0.001) higher mutation rate is detected in ERBB2-high and -low as compared with very-low ERBB2 cases (8% vs 6%). Conclusions: ERBB2/HER2 RNA quantification in cancers can define a biologically distinct group of patients. Patients defined as very-low ERBB2 as classified in this approach appear to be biologically distinct and more likely to be resistant to anti-HER2 therapy. Clinical trials using this approach in selecting patients for treatment with antibody-drug conjugates are justified and needed.

CLDN18/CLDN18.2 IHC assay comparison (SP455, 43-14A, EPR19202) and co-prevalence expression with other biomarkers in gastric carcinoma.

4089 Background: Despite recent advances in treatment, gastric cancer (GC) remains the fourth leading cause of cancer-related mortality worldwide. Claudin 18.2 (CLDN18.2) has emerged as a promising therapeutic target as its expression is mostly restricted to the gastric epithelium and persists in a significant fraction of GC and other upper gastrointestinal tumors (e.g. pancreatic ductal adenocarcinoma). The goal of this study was to compare the analytical performance of multiple CLDN18/CLDN18.2 IHC assays and understand the prevalence of CLDN18.2 with other GC-relevant biomarkers (PD-L1, HER2) and the tumor immune microenvironment. Methods: Cell lines (n=11) and primary resection specimens from patients with GC (n=91) were studied. To compare the analytical performance of three CLDN18/CLDN18.2 IHC assays, FFPE samples were stained with pan-CLDN18 (43-14A) and CLDN18.2-specific (SP455 and EPR19202) clones. Staining results were validated using orthogonal methods i.e., flow cytometry and mass spectrometry. The relationship of CLDN18.2 expression with PD-L1 (SP263) and HER2 (HercepTest) as well as T cell contexture (CD3/CD8/PanCK; CD3/PD-L1/Ki67) were explored by IHC and chromogenic multiplex IHC. Association of these markers with clinicopathological features was also investigated. Results: Comparison of the three CLDN18/CLDN18.2 IHC assays showed similar analytical performance across cell lines and GC samples. Using orthogonal methods, we determined that the limit of detection and linear range was comparable amongst assays. In GC, samples classified as CLDN18/CLDN18.2 positive at a cutoff of 75% with 2-3+ intensity included 22% (SP455), 22% (43-14A), and 15% (EPR19202) of samples. Classification at other CLDN18/CLDN18.2 cutoffs were also explored and will be presented. Only one case (1%) showed significantly discrepant staining patterns between pan-CLDN18 and CLDN18.2-specific assays (H score 235 vs. 105, respectively). CLDN18.2 positivity did not correlate with HER2 expression, PD-L1 expression, T cell densities (cells/mm2), or any of the tested clinicopathological characteristics i.e., tumor grade, histologic subtype, or TNM stage. Conclusions: We demonstrated a concordant analytical performance between the tested CLDN18/CLDN18.2 IHC assays in GC. Their implementation in the clinical setting might help identify patient candidates that could benefit from treatment with CLDN18.2 targeted therapies. Additional research is needed to confirm similar analytical performance between the three IHC assays in other indications.

Implementation of a comprehensive genomic profiling (CGP) panel for precision oncology at a cancer center in Brazil.

e23520 Background: Assessing molecular alterations in solid tumors is essential for providing the best care in precision oncology. CGP tests utilize next-generation sequencing (NGS) to detect clinically relevant alterations and molecular signatures that are increasingly utilized in cancer careand can be performed in-house, which decreases test turnaround time (TAT) and costs. A paucity of studies have documented the experience of in-house testing at academic centers in Brazil. Methods: The TruSight Oncology (TSO) 500 assay/TSO 500 HRD are research use only (RUO) CGP assays that were validated for the detection of small variants (SVs), copy number variants (CNVs), microsatellite instability (MSI), tumor mutation burden (TMB) and genomic instability score (GIS) in DNA and gene fusions/splice variants in RNA in the NextSeq 500 system. Secondary analysis was performed with Illumina Connected Analytics. Tertiary analysis was performed with Illumina Connected Insights (ICI) and Sophia DDM. TMB ≥ 10mut/Mb was defined as TMB-High and GIS ≥ 42 was homologous recombination deficient (HRD). Results: We performed CGP on 40 FFPE samples (16 with HRD) with variable quality and block age most of which had known alterations detected by other genomics/molecular tests: 17 TNBC (16 paired DNA and RNA, 1 DNA only), 18 lung cancer (13 paired DNA and RNA, 5 DNA only), 1 endometrium cancer (DNA only, POLE mutated, 4 replicates), 1 colorectal cancer (DNA only, MSI unstable) and 11 sarcomas (RNA only). 84%, 84%, 90% and 89% of samples with DNA integrity number (DIN) &gt; 2.5 were qualified for analysis of SVs/TMB, MSI, CNVs, and GIS, respectively, while for samples with DIN &lt; 2.5 these values felt to 47%, 74%, 68% and 43%, respectively. RNA isolated from blocks dated up to 5 y.o. were qualified in 90% of cases while from older blocks only 6%. The 4 replicate (single sample) analysis showed 82% and 83% overall variant concordance in the ICI and Sophia DDM software, respectively. All previously known oncogenic alterations (SVs ≥5% VAF, fusions/splice, MSI, TMB) were detected (qualified samples) and with similar performance in ICI and Sophia, respectively: concordance 0.91/p &lt; 0.0001 vs 0.94/p &lt; 0.0001; sensitivity 100% vs 100%; specificity 98% vs 100%. TMB had 100% concordance with previous TMB results. Of note, 73% (8/11) TNBC samples analyzed for GIS were HRD and 75% (6/8) explained by BRCA1 germline mutation (25%) and BRCA1or RAD51C somatic promoter methylation (75%). Conclusions: This study demonstrates the utility of CGP and HRD testing in identifying multiple molecular alterations in FFPE tissue samples with high sensitivity and specificity. Pre-analytical variables directly impact data generation. BRCA1/RAD51C promoter methylation showed to be a potential biomarker for selecting metastatic TNBC patients for PARP inhibitor therapies. Offering CGP can potentially bring precision oncology to more patients with cancer in Brazil.

Multi-omics profiling the genomic origin and adeno-squamous transdifferentiation mechanism in adenosquamous carcinoma.

e20032 Background: Adenosquamous carcinoma (ASC) represents a rare subtype (0.4-4%) of non-small-cell lung cancer (NSCLC), with its multi-omic landscape yet to be fully elucidated. This study aims to explore the genomic origins and molecular mechanisms of adeno-squamous transdifferentiation in lung ASC through comprehensive multi-omics. Methods: We analyzed 69 samples from 31 lung ASC patients, including 16 FFPE samples with microdissected paired adenocarcinoma (AC) and squamous cell carcinoma (SCC) components, and additionally, incorporated pure Lung Adenocarcinoma (LUAD, n = 51) and pure Squamous Cell Carcinoma (LUSC, n = 28) for comparative analysis. Techniques included whole-exome sequencing, whole-genome methylation profiling, transcriptomics, single-cell RNA sequencing, and spatial transcriptomics. Results: AC and SCC exhibited similar genomic mutation landscapes, indicating a common genomic origin for AD and SCC. The SCC component in ASC displayed higher homologous recombination repair deficiency (HRD) and a notable 16p13.3 deletion in 31% of ASC transformations. Differential methylation analysis highlighted TP63's lower methylation in SCC, correlating negatively with RNA expression (P &lt; 0.001), indicative of methylation-driven regulation. DMR binding motif analysis revealed hypo-methylation in p63, p53, and AP-1 motifs, and hyper-methylation in MKX2-1, FOXA1/2 motifs (P &lt; 0.05). The transition from AC to SCC in ASC was marked by upregulated MYC signaling, proliferation, and DNA repair, with EMT changes not methylation-driven. Metabolic pathways were upregulated, and EMT/KRAS pathways downregulated in SCC compared to LUSC. Immune infiltration in LUAD-AC-ASC transformation showed an initial increase and subsequent decrease, transitioning from immune-infiltrated and immune-excluded to immune-desert phenotypes. IL36G of the IL1 family was highly expressed in the SCC component and strongly correlated with the squamous marker P63 (Pearson r = 0.72, P &lt; 0.001). Single-cell RNA-seq data demonstrated a significant increase in IL36G in squamous cells of lung ASC. In vivo models showed adeno-squamous differentiation in adenocarcinoma cells expressing IL36G. Conclusions: This study provides a comprehensive multi-omic insight into the complex genomic and epigenomic landscape of ASC, highlighting a shared genomic origin and distinct molecular mechanisms in the adeno-squamous transdifferentiation process. The findings, particularly the role of TP63 methylation and IL36G expression in the SCC component, offer novel insights into the molecular heterogeneity of ASC. These results not only deepen our understanding of ASC but also pave the way for future investigations into targeted therapies that could potentially exploit these molecular distinctions for more effective treatment strategies in ASC.

First-line sintilimab combined with platinum-based chemotherapy in extensive stage small cell lung cancer: A phase II study and post-hoc biomarker analysis.

e20127 Background: Immune checkpoint inhibitors (ICIs) combined with etoposide and platinum-based chemotherapy improves prognosis of patients with extensive-stage small-cell lung cancer (ES-SCLC). However, nearly 40% of patients do not benefit from Immunochemotherapy. Herein, this phase II study of first-line immunochemotherapy was conducted aiming to investigate potential biomarkers associated with therapeutic efficacy. Methods: A multi-center, single-arm, prospective phase II trial (ChiCTR2000038354) was designed and conducted to evaluate the efficacy and safety of sintilimab combined with platinum-based chemotherapy in patients with ES-SCLC who had not previously received treatment. The primary endpoint of this study is progression-free survival (PFS), with patients being defined as disease progression according to RECIST 1.1. Pre-treatment FFPE samples were analyzed using whole exome sequencing (WES) and whole transcriptome sequencing (WTS) to investigate biomarkers. Results: A total of 44 patients were included in the study, with a median follow-up of 10 months and a median PFS of 7.57 months. 8 patients with a PFS greater than 6 months were defined as the responder group (R group) and 5 patients with a PFS less than 6 months belonged to the non-responder group (NR group). WES and WTS were performed on the pre-treatment samples of these 13 patients. WES results revealed mutation frequencies of TP53 (69%), RB1 (31%) and FAT1 (31%) aligning with prior research. Significantly, ZFHX4 mutations correlated with R group (p = 0.05). Genomic instability, as defined by Homologous Recombination Deficiency (HRD) scores, displayed higher scores in NR group than in R group (p = 0.023), which suggests potential benefits of PARP inhibition in the NR group. WTS data revealed that higher M2/M1 macrophage ratio (p = 0.028) in NR group. Additionally, the NR group demonstrated significant VEGFA overexpression. Besides, NOTCH-signaling (p = 0.034), KRAS-signaling-DN (p&lt;0.001), Angiogenesis (p&lt;0.001) and EMT (p&lt;0.001) pathway were also enriched in NR group. Conversely, M1 macrophages (p = 0.079), CD4+T cells (p = 0.093), Mast cells (p = 0.045), NK cells (p = 0.045) and neutrophils (p = 0.06) was higher in R group. Conclusions: This study highlights genomic instability as well as activation of cancer hallmarks such as angiogenesis, EMT may result the resistance to immunotherapy in ES-SCLC patients. These findings may shed light on combinational strategy overcoming therapeutic resistance. Clinical trial information: ChiCTR2000038354.

EGFR-mutant NSCLC may remodel TME from non-inflamed to inflamed through acquiring resistance to EGFR-TKI treatment

BackgroundEGFR-TKI represent the standard first-line therapy for advanced NSCLC harboring EGFR mutations. However, resistance to EGFR-TKI inevitably develops in nearly all patients. Previous clinical study have demonstrated that, some patients that failed EGFR-TKI therapy show a benefit outcome from immunotherapy. Our objective is to explore the immune microenviroment remodeling induced by EGFR-TKI treatment in EGFR mutant lung cancer patients and to investigate the immune cell types and potential molecular signatures involved. MethodsA cohort of 37 EGFR mutant advanced-stage NSCLC patients, who are resistant to at least one type of TKI treatment, was retrospectively established. Both pre-treatment and TKI resistance tumor FFPE samples of each pairs were collected. Transcriptional profiling and bioinformatics analysis were employed to evaluate the change of immune associated hallmarks before and after EGFR-TKI therapy. ResultsTumor samples after EGFR-TKI treatment displayed enrichment of proinflammatory signaling like interferon-γ, allograft rejection and inflammatory response. Of note, cytotoxic factor granzyme A as well as PD-L1 were found to be more expressed in EGFR-TKI resistance samples. Approximately 33.3 % (11/33) of EGFR-TKI treated samples were classified as “hot” tumor, especially for EGFR L858R mutated NSCLC patients (46.7 %,7/15). Effector cells were significantly overexpressed in ‘hot’ tumors feature following TKI resistance. In addition, we found that four effector genes (CD8A, CDB8, GZMB, GZMK) showed higher expression in ‘hot’ tumors post-TKI resistance, and its 4-gene effector cell signature was found to have a good correlation with survival benefit in external immunotherapy database. ConclusionsTKI treatment may initiate immune activation in EGFR mutant NSCLC, leading to changes in immune cell infiltration following TKI resistance. We mechanistically explored that this might be due to an increased immune response caused by the rise in effector cells post-TKI resistance.

Abstract RF01-05: Functional assessment of RAD51 foci and replication fork dynamics in PARPi resistant BRCA1/2 mutated breast cancer

Abstract Background: Although platinum salts (Pt) or Poly (ADP-Ribose) Polymerase inhibitors (PARPi) are effective in treating homologous recombination defective (HRD) breast cancer, resistance often emerges, especially in advanced disease. Predicting response and relapse is complex, even in patients with germline BRCA1/2 mutations (gBRCA1/2m). Clinically approved HRD detection methods are limited to identification of pathogenic mutations in HR genes or mutational signatures in the genome of tumors caused by HRD. In PDX models derived from HRD breast cancer, the restoration of nuclear RAD51 foci formation, a key feature of functional HR, can predict resistance to HRD-targeted treatment.1 In addition, the restoration of replication fork stability, despite PARPi or induction of Pt adducts that induce replication fork arrest and collapse, confers resistance in pre-clinical models2; to date this resistance mechanism has not been clinically validated. Here we analyse RAD51 foci in FFPE samples and DNA replication fork dynamics, using DNA fibre combing assays, in patient derived organoids (PDO) relating these to clinical response to illustrate how such assays might predict clinical HRD-targeted therapy resistance in metastatic breast cancer (MBC). Patients and Methods: We used immunofluorescent detection of RAD51 foci as a marker for HR proficiency (HRP) in tumors from 29 patients with gBRCA1/2m with MBC treated with HRD-targeted treatment (n=6 PARPi, n=2 Pt and n=21 both agents in sequence). All patients developed resistance that was either de novo or acquired. RAD51 and BRCA1 foci were scored in a minimum of 50 geminin (a marker of S/G2 phase) positive tumor cells; cells with ≥5 foci were classified as positive, and tumours where ≤10% or &amp;gt;10% of cells were positive were considered HRD or HRP, respectively. DNA replication fork dynamics were assessed in with or without PARPi using thymidine analogue labelling. DNA fibre analysis was performed in PDOs developed from fresh tumor sampling of HRD-targeted treatment sensitive or resistant tumors, to determine the relationship between replication fork dynamics, stability and PARPi sensitivity. Results: RAD51 analysis was performed on 9 treatment naïve samples (n=9 patients), 8 samples obtained after HRD-targeted treatment resistance (n=8 patients) and 27 samples obtained pre and post HRD-targeted treatment resistance (n=12 patients). Functional HRD by RAD51 in treatment naive samples was seen in 100% (n=17) of patients with acquired resistance and 66% of patients with de novo resistance. All patients, whether with de novo or acquired resistance, exhibited high RAD51 scores in post-resistance tumour samples, suggesting restoration of HR function is the dominant mechanism of PARPi resistance. As such, RAD51 analysis shows potential as a biomarker of clinical PARPi resistance. Replication fork stability fibre was analysed after exposure to potent PARP1 trapping inhibitors or an ATRi control in 3 PDOs (2 gBRCA1m, 1 BRCAwt). As controls we used an isogenic 2D cell line with and without a CRISPR engineered BRCA1 reversion mutation (SUM149 parental BRCA1m and SUM149BS*1 revertant) or with a PARP1 mutation (SUM149 TR2 clone) that prevents PARP1 being trapped on DNA. These experiments indicate that replication fork dynamics can be assessed in “patient derived” models of breast cancer and that PARPi sensitivity was associated with PARPi induced replication fork instability (median ratio IdU:CldU in resistant lines: 0.98, sensitive lines: 0.60). Conclusion: We show that HRP restoration and RAD51 foci in advanced BRCA1/2m breast cancers is the dominant form resistance to HRD-targeted treatment. We also demonstrate for the first time that analysis of DNA replication fork dynamics can be carried out in breast cancer PDOs and could be further explored as a functional predictive biomarker of PARPi resistance. 1.Pellegrino et al, 2022 (PMID: 35425960) 2.Chaudhuri et al, 2016 (PMID: 27443740) Citation Format: Elizabeth Harvey-Jones, Maya Raghunandan, Luisa Robbez-Masson, Thanussuyah Alaguthurai, Alba Llop-Guevara, Jennifer Trendell, Olga Rodriguez, Marta Guzman, Daniel Weekes, Christelle deRenty, Eleanor Knight, Rebecca Marlow, Christopher Starling, Ruifang Liu, Nivedita Ravindran, Ioannis Roxanis, Natalia Lukashchuk, Violeta Serra, Stephen Pettitt, Christopher Lord, Andrew NJ Tutt. Functional assessment of RAD51 foci and replication fork dynamics in PARPi resistant BRCA1/2 mutated breast cancer [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 RF01-05.

Abstract PO1-02-01: T Cell Receptor Sequencing to Monitor Pelareorep-Induced Expansion of Tumor Infiltrating Lymphocytes

Abstract Background: Tumor infiltrating lymphocytes (TILs) represent a major immunological tumor control mechanism that is associated with better prognosis in breast cancer. We have previously reported the effect of pelareorep (pela), an intravenously delivered, non-engineered oncolytic reovirus, on a composite measurement of TILs and tumor cellularity (CelTIL) from the AWARE-1 window-of-opportunity study in patients with early breast cancer (eBC). These results showed treatment increased in CelTIL scores especially in the atezolizumab group. To confirm and extend these findings we applied T cell receptor sequencing of matched tumor tissue and whole blood pre and post-treatment from the AWARE-1 study to further explore the effects of pela therapy on TILs. Methods: Newly diagnosed HR+/HER2- eBC patients were enrolled into two cohorts: Cohort 1: pela + letrozole (n=10); and Cohort 2: pela + letrozole + atezolizumab (n=10). Pela was administered on days 1, 2 and 8, 9, and atezolizumab was given on day 3. Tumor biopsies (FFPE samples) collected pre-treatment (~D-23) and on day ~21 when tumors were surgically removed. T cell fraction analysis and T cell receptor sequencing were performed by Adaptive Biotechnology (Seattle, Washington) Immunoseq protocol. Results: As was shown for the CelTIL scores, an increase in the tumor T cell fraction, a direct measurement of TILs, was observed at 1-month post-treatment in both cohorts with a mean percent increase of 21% for Cohort 1 and 67% for Cohort 2. Through TCR-sequencing of tumor tissue, we identified TIL clones and tracked their differential abundance in tumor tissue and blood post-treatment. The results of this analysis at Day 21 showed a mean post-treatment expansion of 10 tumor specific clones in the blood for Cohort 1, and a mean post-treatment increase of 40 tumor specific clones for Cohort 2. Conclusions: These results confirm the previously reported CelTIL results from AWARE-1 on pela-induced increases in TILs and demonstrate the expansion of these clones in both tumor and blood as a consequence of pela therapy for HR+/HER2- breast cancer. Citation Format: Richard Trauger, Houra Loghmani, Matt Coffey, Fernando Salvador, Joaquín Gavilá, Luis Manso, Tomás Pascual, Aleix Prat, Thomas Heineman. T Cell Receptor Sequencing to Monitor Pelareorep-Induced Expansion of Tumor Infiltrating Lymphocytes [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 PO1-02-01.

Abstract PO3-16-03: Investigating HER2DX genomic assay concordance within HER2-positive breast cancer

Abstract Background: The HER2DX genomic assay, encompassing a 27-gene panel, provides prognostic (risk-score) and predictive (pathological complete response [pCR]-score) information while quantifying ERBB2 mRNA expression (ERBB2-score). This study aims to explore the variability of HER2DX scores within tumor samples and between paired diagnostic biopsies and surgical specimens. Methods: The standardized HER2DX assay was conducted on pretreatment FFPE biopsies from two distinct geographical areas within tumors (i.e., two different core biopsies) of patients enrolled in the 14-409 phase II study. This study assessed the impact of HER2 heterogeneity on treatment response in stage II-III HER2+ breast cancer patients treated with neoadjuvant T-DM1 and pertuzumab. Additionally, HER2DX was evaluated on paired biopsy-surgery FFPE samples from 6 patients who underwent primary surgery without pre-operative treatment. Variability and concordance of the three HER2DX groups scores were calculated, and Pearson's correlation coefficient determined the correlation between scores at different sites/time points. Paired t-tests were performed between tumor samples for the HER2DX scores. Results: The HER2DX assay was evaluated in two different diagnostic pre-treatment core biopsies from 106 patients (n=212 HER2DX assays). Intra-tumor concordance rates of the three HER2DX groups (risk-score, pCR-score, and ERBB2-score) were 94.3%, 87.7%, and 91.5%, respectively. No significant differential expression of HER2DX scores was observed across paired core biopsies (p-values=0.327 [risk-score], 0.415 [pCR-score], and 0.451 [ERBB2-score]). The average score differences (core 1 - core 2 [scale from 0 to 100]) were 0.57 (risk-score), 0.63 (pCR score), and -0.63 (ERBB2-score). Correlation coefficients across paired core biopsies were 0.96 (risk-score), 0.91 (pCR-score), and 0.90 (ERBB2-score). Additionally, intra-patient variability of the HER2DX scores was evaluated. Average variability in risk-score was 3.38% (95% CI: 2.76-4.01), in pCR-score was 3.53% (95% CI: 2.84-4.23) and in ERBB2 was 3.87% (95% CI: 2.97-4.77). In 6 patients with cT1-2 cN0 disease, HER2DX was evaluated in a diagnostic core biopsy and a surgical specimen, with an average time interval of 8.5 weeks (range 4 to 12 weeks) between time points. The intra-patient concordance rates of the three HER2DX groups were 83.3%, 83.3%, and 100% for risk-score, pCR-score, and ERBB2-score, respectively. No significant differential expression of HER2DX scores was found across paired biopsy-surgery samples (p-values=0.814 [risk-score], 0.128 [pCR-score], and 0.607 [ERBB2-score]). The average score differences (core biopsy - surgical specimen [scale from 0 to 100]) were -0.67 (risk-score), -11.5 (pCR score), and 2 (ERBB2-score). Correlation coefficients across paired HER2DX assays were 0.97 (risk-score), 0.87 (pCR-score), and 0.94 (ERBB2-score). Conclusion: The HER2DX genomic assay demonstrates low variability within HER2+ breast cancer, supporting its reliability as a consistent tool for assessing prognosis, the probability of pCR and ERBB2 quantitative expression. Citation Format: Fara Brasó-Maristany, Laia Paré, Guillermo Villacampa, Mercedes Marín-Aguilera, Paul Spellman, Adrienne Waks, Ian Krop, Esther Sanfeliu, Zhequi Li, Patricia Galván, Eva Hernández-Illán, Pedro Jares, Joan Antón Puig-Butillé, Ana Vivancos, Patricia Villagrasa, Joel S. Parker, Charles M. Perou, Sara Tolaney, Otto Metzger, Kornelia Polyak, Aleix Prat. Investigating HER2DX genomic assay concordance within HER2-positive breast cancer [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 PO3-16-03.