**Stroma Transcriptomic and Proteomic Profile of Prostate Cancer Metastasis Xenograft Models Reveals Prognostic Value of Stroma Signatures**

**Sofia Karkampouna 1, Maria R. De Filippo 1, Charlotte K. Y. Ng 2, Irena Klima 1, Eugenio Zoni 1, Martin Spahn 3, Frank Stein 4, Per Haberkant 4, George N. Thalmann 1,5,**† **and Marianna Kruithof-de Julio 1,5,\*,**†


Received: 1 December 2020; Accepted: 10 December 2020; Published: 15 December 2020

**Simple Summary:** Currently, there is a need for prognostic tools that can stratify patients, who present with primary disease, based on whether they are at low or high risk for drug resistant and hormone-independent lethal metastatic prostate cancer. The aim of our study was to assess the potentially added value of tumor microenvironment (stroma) components for the characterisation of prostate cancer. By utilising patient derived-xenograft models we show that the molecular properties of the stroma cells are highly responsive to androgen hormone levels, and considerable ECM remodelling processes take place not only in androgen-dependent but also in androgen-independent tumor models. Transcriptomic mechanisms linked to osteotropism are conserved in bone metastatic xenografts, even when implanted in a different microenvironment. A stroma-specific gene list signature was identified, which highly correlates with Gleason score, metastasis progression and progression-free survival, and thus could potentially complement current patient stratification methods.

**Abstract:** Resistance acquisition to androgen deprivation treatment and metastasis progression are a major clinical issue associated with prostate cancer (PCa). The role of stroma during disease progression is insufficiently defined. Using transcriptomic and proteomic analyses on differentially aggressive patient-derived xenografts (PDXs), we investigated whether PCa tumors predispose their microenvironment (stroma) to a metastatic gene expression pattern. RNA sequencing was performed on the PCa PDXs BM18 (castration-sensitive) and LAPC9 (castration-resistant), representing different disease stages. Using organism-specific reference databases, the human-specific transcriptome (tumor) was identified and separated from the mouse-specific transcriptome (stroma). To identify proteomic changes in the tumor (human) versus the stroma (mouse), we performed human/mouse cell separation and subjected protein lysates to quantitative Tandem Mass Tag labeling and mass spectrometry. Tenascin C (TNC) was among the most abundant stromal genes, modulated by androgen levels in vivo and highly expressed in castration-resistant LAPC9 PDX. The tissue microarray of primary PCa samples (*n* = 210) showed that TNC is a negative prognostic marker of the clinical progression to recurrence or metastasis. Stroma markers of osteoblastic PCa bone metastases seven-up signature were induced in the stroma by the host organism in metastatic xenografts, indicating conserved

mechanisms of tumor cells to induce a stromal premetastatic signature. A 50-gene list stroma signature was identified based on androgen-dependent responses, which shows a linear association with the Gleason score, metastasis progression and progression-free survival. Our data show that metastatic PCa PDXs, which differ in androgen sensitivity, trigger differential stroma responses, which show the metastasis risk stratification and prognostic biomarker potential.

**Keywords:** prostate cancer; stroma signature; patient-derived xenografts

#### **1. Introduction**

Bone metastases are detected in 10% of patients already at the initial diagnosis of prostate cancer (PCa) or will develop in 20–30% of the patients subjected to radical prostatectomy and androgen deprivation therapy and will progress to an advanced disease called castration-resistant prostate cancer [1]. Metastases are established when disseminated cancer cells colonize a secondary organ site. An important component of tumor growth is the supportive stroma: the extracellular matrix (ECM) and the nontumoral cells of the matrix microenvironment (e.g., endothelial cells, smooth muscle cells and cancer-associated fibroblasts). Upon interaction of the stroma compartment and tumor cells, the stroma responds by the secretion of growth factors, proteases and chemokines, thereby facilitating the remodeling of the ECM and, thus, tumor cell migration and invasion [2]. Therefore, tumor cell establishment requires an abnormal microenvironment. It is unclear whether the stroma is modulated by the tumor cells or by intrinsic gene expression alterations. Understanding the mechanisms of tumor progression to the metastatic stage is necessary for the design of therapeutic and prognostic schemes.

The bone microenvironment is favorable for the growth of PCa, as well as breast cancer, indicated by the high frequency of bone metastasis in these tumors. Studies have shown that cancer cell growth competes for the hematopoietic niche in the bone marrow with the normal residing stem cells [3], and depending on the cancer cell phenotype, this may lead to either osteoblastic or osteolytic lesions. The stroma signature of osteolytic PCa cells (PC-3) xenografted intraosseously in immunocompromised mice induce a vascular/axon guidance signature [4]. The stroma signature of osteoblastic lesions from human VCap and C4-2B PCa cell lines indicated an enrichment of the hematopoietic and prostate epithelial stem cell niche. A curated prostate-specific bone metastasis signature (Ob-BMST) implicated seven highly upregulated genes (*Aspn*, *Pdgrfb*, *Postn*, *Sparcl1*, *Mcam*, *Fscn1* and *Pmepa1)* [5], among which, *Postn* and *Fscn1* are bone-specific. Furthermore, *Aspn* and *Postn* expression is also increased in primary PCa cases [5], indicative of osteomimicry processes. The induction of osteoblastic genes in the stroma of primary tumors (PCa and breast), such as osteopontin and osteocalcin, has been suggested as a mechanism termed osteomimicry [6] to explain why the bone microenvironment is the preferential metastasis site. High stromal differences between benign, indolent and lethal PCa, combined with the enrichment of bone remodeling genes in high Gleason score cases [7], suggest that the stroma is an active player in PCa. During androgen deprivation, androgen-dependent epithelial cells will undergo apoptosis, while the supporting stroma is largely maintained or replaces the necrotic tissue areas [8]. Stromal cells do express androgen receptors (AR) and have active downstream signaling, while the absence of stromal AR expression is used as a prognostic factor of disease progression [9]. Furthermore, AR binds to different genomic sites in prostate fibroblasts compared to the epithelium [10] and to cancer-associated fibroblasts (CAFs) [11], indicating different roles of AR in epithelial or stroma cellular contexts. Prostate CAFs have tumor-promoting effects on marginally tumorigenic cells (LNCaP), irreversibly altering their phenotype and influencing their progression to androgen independence and metastasis [12,13].

In this study, we investigated whether metastatic PCa patient-derived xenograft models (PDXs) that differ in androgen sensitivity are triggering a differential stroma response. To elucidate the mechanisms of stroma contribution to tumor growth later on, we determined the unique gene

expression profile of the stroma compared to the tumor compartment, the proteome changes of the tumor versus stroma. We identified androgen-dependent stroma gene expression signatures with potential disease progression prognostic values for primary PCa.
