Biomarker Quantification in Breast Cancer using Xenium In Situ
Advances in spatial transcriptomics enable high-throughput quantitation of both established and novel biomarkers at single cell resolution, offering the potential to transform diagnostics. Using Xenium in situ technology in FFPE human breast samples, we address two challenges in the cancer field: 1. achieving reliable normalization of gene expression across heterogeneous sample populations; 2. identifying biomarkers that predict invasion or metastasis. We describe a scalable approach to identify low-variation housekeeping (HK) genes within any given sample set, then use those HK genes for cross- and intra-sample normalization of biomarkers. Analyzing 12 FFPE human breast samples-primarily ductal carcinoma in situ (DCIS)-with a custom 280-gene panel, we identified four HK genes (EEF1G, EEF2, MALAT1, and RPLP0) that exhibited minimal variability in tumor cells, four tumor cell biomarkers (LDHA, SDC1, PIGR, SFRP1) that increased or decreased with tumor grade, and one tumor-associated myoepithelial biomarker (LAMC2). Normalizing biomarkers to the four HK genes preserved the dynamic range of expression necessary for distinguishing tumor grades, outperforming HKs from legacy RT-PCR diagnostic panels. Lastly, we employed a cell-agnostic approach in the tumor periphery to quantify MMP11, a biomarker correlated with proliferative and potentially pre-invasive ducts. Our results establish a single cell normalization method for spatial in situ transcriptomics and reveal and quantitate biomarkers relevant to DCIS risk and progression.