Multi-protein spatial signatures in ductal carcinoma in situ (DCIS) of breast

BACKGROUND: There is limited knowledge about DCIS cellular composition and relationship with breast cancer events (BCE). METHODS: Immunofluorescence multiplexing (MxIF) was used to image and quantify 32 cellular biomarkers in FFPE DCIS tissue microarrays. Over 75,000 DCIS cells from 51 patients (median 9 years follow-up for non-BCE cases) were analysed for profiles predictive of BCE. K-means clustering was used to evaluate cellular co-expression of epithelial markers with ER and HER2. RESULTS: Only ER, PR and HER2 significantly correlated with BCE. Cluster analysis identified 6 distinct cell groups with different levels of ER, Her2, cMET and SLC7A5. Clusters 1 and 3 were not significant. Clusters 2 and 4 (high ER/low HER2 and SLC7A5/mixed cMET) significantly correlated with low BCE risk (P = 0.001 and P = 0.034), while cluster 6 (high HER2/low ER, cMET and SLC7A5) correlated with increased risk (P = 0.018). Cluster 5 (similar to cluster 6, except high SLC7A5) trended towards significance (P = 0.072). A continuous expression score (Escore) based on these 4 clusters predicted likelihood of BCE (AUC = 0.79, log-rank test P = 5E-05; LOOCV AUC = 0.74, log-rank test P = 0.006). CONCLUSION: Multiplexed spatial analysis of limited tissue is a novel method for biomarker analysis and predicting BCEs. Further validation of Escore is needed in a larger cohort.

Automated summary

Limited knowledge about DCIS cellular composition and its relationship with breast cancer events (BCE) was investigated using immunofluorescence multiplexing (MxIF). A total of 32 cellular biomarkers in FFPE DCIS tissue microarrays from 51 patients were analyzed, revealing significant correlations between ER, PR, HER2, and BCE. Clustering analysis identified 6 distinct cell groups, with clusters 2, 4 showing low BCE risk associations and cluster 6 showing increased risk. The continuous expression score (Escore) based on these clusters predicted likelihood of BCE, indicating the potential for novel biomarker analysis and prediction of BCE using this method. Further validation of Escore in a larger cohort is necessary.