Abstract Background Trastuzumab is the cornerstone of adjuvant therapy for HER2-positive breast cancer (BC), but due to de novo or primary resistance, >20% of early-stage patients become resistant. Thus, predicting the mechanisms of trastuzumab resistance would facilitate the planning of specific therapeutic strategies. We performed a case-control study of 26 HER2-positive BC patients who relapsed within 5 years of starting adjuvant treatment and 26 controls, no relapsed, analyzing an extensive gene expression profile. Methods Total RNA was isolated from formalin-fixed paraffin-embedded primary tumors with AllPrep DNA/RNA FFPE Kit (Qiagen) and quantified by Nanodrop, before performing Nanostring gene expression profiling. nCounter Breast Cancer 360 Panel (Nanostring Technologies) was used according to the manufacturer’s instructions to analyze the expression of 770 genes and important signatures for BC (e.g. PAM50). For the statistical analyses, genes were normalized using a ratio of the expression value to the geometric mean of all housekeeping genes on the panel. Housekeeper-normalized data were then log(2) transformed. These data were evaluated by unsupervised selection of genes using consensus clustering with the Partitioning Around Medoids algorithm followed by the association of cluster “representative” genes (i.e. the medoids) with respect to the presence of relapse. The optimal number of clusters was determined by inspecting the consensus matrix cumulative distribution. Given the small sample size, the association between the "representative" genes, demographic and clinical covariates and the risk of relapse was evaluated using a logistic regression model with ELASTIC-NET regularization with α and λ tuned by 5-fold cross-validation. Penalized regression coefficients β are reported. The analyses were performed using R version 4.0.4 and package “ConsensusClusterPlus” and “glmnet”. Results Patient clinical characteristics are reported in Table 1. Relapsed patients showed a higher tumor stage, larger tumor size and greater lymph node involvement than controls (Table 1). Six samples failed the quality control check and so the analyses on gene expression data were performed on 46 subjects who had a distribution of clinical covariates similar to that reported in Table 1. The PAM50 intrinsic subtype, Risk Of Recurrence (ROR) score and ROR category were not associated with relapse status (P = 0.5, 0.5 and 0.1, respectively). The consensus clustering analysis revealed that the optimal number of clusters was 5, and the corresponding medoids were: HDAC6 for cluster 1 (n1=192 genes), WNT4 for cluster 2 (n2=102 genes), BMPR2 for cluster 3 (n3=162 genes), PALB2 for cluster 4 (n4=191 genes), and PARP1 for cluster 5 (n5=84 genes). HDAC6 and BMPR2 showed the highest correlation (r = 0.53, P<0.001) among the 5 genes. HDAC6 showed prognostic potential in the penalized logistic regression analysis, independently of tumor stage and Body Mass Index (β: 0.57, 1.49, and 0.11, respectively). Higher HDAC6 expression was associated with a higher risk of relapse, all other conditions being equal. Conclusions We are now planning to validate these results in a larger case series and in in vitro models to verify whether HDAC6, in addition to being an unfavorable prognostic factor, could also be a new therapeutic target for trastuzumab resistance. Table 1.Patient characteristics in relation to Trastuzumab outcome.VariableAllCasesControlsP(n=52)(n=26)(n=26)n(%)n(%)n(%)Age at start adiuvant therapyMean ± sd53.4 ± 11.953.1 ± 12.653.7 ± 11.40.854Menopausal StatusPre-menopause21(40.4)10(38.5)11(42.3)0.777Post-menopause31(59.6)16(61.5)15(57.7)Stage of DiseaseI9(18.8)2(8.0)7(30.4)0.001II20(41.7)7(28.0)13(56.5)III19(39.6)16(64.0)3(13.0)Unknown413Lymph Node StatusNegative18(36.0)4(16.7)14(53.9)0.008Positive32(64.0)20(83.3)12(46.2)Unknown220Tumor SizeT128(53.9)7(26.9)21(80.8)<0.001T213(25.0)8(30.8)5(19.2)T36(11.5)6(23.1)0(0.0)T45(9.6)5(19.2)0(0.0)Histologic Grade12(4.1)1(4.2)1(4.0)0.889218(36.7)8(33.3)10(40.0)329(59.2)15(62.5)14(56.0)Unknown321Ki-67<20%17(32.7)8(30.8)9(34.6)0.768≥20%35(67.3)18(69.2)17(65.4)ERNegative14(26.9)7(26.9)7(26.9)1.000Positive38(73.1)19(73.1)19(73.1)PRNegative22(43.1)11(44.0)11(42.3)0.903Positive29(56.9)14(56.0)15(57.7)Unknown110Vascular invasionNo13(56.5)3(37.5)10(66.7)0.221Yes10(43.5)5(62.5)5(33.3)Unknown291811Adjuvant ChemotherapyAnthracyclines and taxanes24(46.2)7(26.9)17(65.4)<0.001Taxanes only6(11.5)4(15.4)2(7.7)Anthracyclines only10(19.2)3(11.5)7(26.9)Other12(23.1)12(46.2)0(0.0)BMI at start of Adjuvant ChemotherapyMean ± sd24.8 ± 4.025.9 ± 3.223.6 ± 4.40.039sd: standard deviation BMI: Body Mass Index Citation Format: Sara Ravaioli, Roberta Maltoni, Elisabetta Petracci, Michela Palleschi, William Balzi, Francesca Pirini, Maria Maddalena Tumedei, Michele Zanoni, Michela Cortesi, Giovanni Martinelli, Andrea Rocca, Sara Bravaccini. HDAC6 is an unfavorable prognostic factor in HER2-positive breast cancer patients treated with adjuvant trastuzumab [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-13-44.