*2.3. Di*ff*erential Expression Analysis Reveals Androgen-Dependent Stromal Gene Modulation in Androgen-Independent PDX Model*

The relative ratio of human and mouse transcript reads reflected a higher stroma content in the BM18 compared to LAPC9 and significantly reduced human tumor content with enriched stroma content in the BM18 castrated group (Figure S2B). No major differences were observed in the LAPC9 castrated group (Figure S2C). We demonstrated that the human (tumor), as well as the mouse (stroma), transcriptomes follow androgen-dependent transcriptomic changes in the BM18 groups (intact versus castrated versus replaced) (Figure 2A,E). Venn Euler diagrams illustrate androgen level-dependent stromal gene expression modulation not only in the BM18 (Figure S2D and Table S1) but, also, in the androgen-independent (in terms of tumor growth) LAPC9 model (Figure S2E,F and Table S1). To identify the top-most significant AR-regulated stromal genes, we performed a differential expression analysis of BM18 tumors (Figure 4A) from castrated hosts and compared it to BM18 intact (the replaced tumors were not included here due to higher variability). Of the top-most variable genes, 50 were highly upregulated in BM18 tumors (z-score >1) and downregulated upon castration (Figure 4A). A differential expression analysis of LAPC9 tumors from castrated/replaced tumors versus intact tumors revealed the top-most differentially regulated genes: the 27 most upregulated genes in intact, which were downregulated in the castrated groups (Figure 4B). Among the 50 mouse genes that were highly upregulated in the intact BM18, and significantly modulated by castration, were 23 genes implicated in cell cycle/mitosis, 10 implicated in ECM and 3 related to spermatogenesis/hormone regulation, according to the Gene Ontology terms (Figure 4C). Two of these genes, *Tnc* and *Crabp1*, were also detected in the proteomic data (Figure 4C, highlighted in bold) and in both PDXs (Figure 4C,D, highlighted in red). Among the 27 mouse genes that were highly upregulated in the intact LAPC9, and significantly modulated by castration, seven genes were implicated in ECM/cell adhesion/smooth muscle function, and 14 were implicated in non-smooth muscle function and metabolism based on the Gene Ontology terms (Figure 4D). In the LAPC9 proteomic data, we detected 14 genes out of the 27 to be expressed in the mouse fractions (Figure 4D, bold), indicative of potential functional values. Of interest in potentially mediating tumor stroma extracellular interactions are a neural

adhesion protein (*CD56*), implicated in cell–cell adhesion and migration by homotypic signaling, as well as Tenascin C (*Tnc*), an extracellular protein that is found abundantly in the reactive stroma of various cancer types, yet not expressed in normal stroma. Both genes were expressed at the protein level, exclusively in the mouse compartment of the BM18 and LAPC9, at all states (intact, castrated and replaced). Furthermore, Tnc was detected in both BM18 and LAPC9 at the transcriptional and proteomic levels and was reactivated after 24 h of androgen replacement (Figure 4B), indicative of AR-direct target gene modulation.
