MALAT1 Fusions and Basal Cells Contribute to Primary Resistance against Androgen Receptor Inhibition in TRAMP Mice

Simple Summary We deeply characterized a frequently used mouse model of prostate cancer and found cellular and molecular regulators of resistance against antihormonal treatment, such as basal cell function and MALAT1 gene fusions. As these mechanisms also occur in human disease, our findings highlight the importance of this model for human cancer and may be helpful for future research focusing on overcoming antihormonal treatment resistance. Abstract Targeting testosterone signaling through androgen deprivation therapy (ADT) or antiandrogen treatment is the standard of care for advanced prostate cancer (PCa). Although the large majority of patients initially respond to ADT and/or androgen receptor (AR) blockade, most patients suffering from advanced PCa will experience disease progression. We sought to investigate drivers of primary resistance against antiandrogen treatment in the TRAMP mouse model, an SV-40 t-antigen driven model exhibiting aggressive variants of prostate cancer, castration resistance, and neuroendocrine differentiation upon antihormonal treatment. We isolated primary tumor cell suspensions from adult male TRAMP mice and subjected them to organoid culture. Basal and non-basal cell populations were characterized by RNA sequencing, Western blotting, and quantitative real-time PCR. Furthermore, effects of androgen withdrawal and enzalutamide treatment were studied. Basal and luminal TRAMP cells exhibited distinct molecular signatures and gave rise to organoids with distinct phenotypes. TRAMP cells exhibited primary resistance against antiandrogen treatment. This was more pronounced in basal cell-derived TRAMP organoids when compared to luminal cell-derived organoids. Furthermore, we found MALAT1 gene fusions to be drivers of antiandrogen resistance in TRAMP mice through regulation of AR. Summarizing, TRAMP tumor cells exhibited primary resistance towards androgen inhibition enhanced through basal cell function and MALAT1 gene fusions.


Introduction
In developed countries, prostate cancer (PCa) is the most prevalent cancer and the second most common cause of cancer-related death in males [1]. Approximately one in seven men will be diagnosed with PCa throughout his lifetime and most patients exhibit good courses of disease with five-year survival rates of up to 99 percent [2]. Various factors contribute to whether prostate cancer patient prognosis is favorable or fatal, including comorbidities such as diabetes or hypertension [3][4][5]. However, today it is generally accepted that biology of the individual tumor and its unique molecular biological aspectssuch as genetical aberrations and potentially cellular origin and heterogeneity of the tumor-play a major role in influencing patient outcome [6].
Tumors from patients exhibiting disease progression after or upon systemic androgen deprivation treatment (ADT), known as castration resistant prostate cancer (CRPC), often show differentiation towards an aggressive phenotype of cancer, known as neuroendocrine prostate cancer or NE-PC. For this entity, reported to be primarily resistant towards further antihormonal treatment, novel preclinical models are dearly needed for the establishment of novel treatment options as well as diagnostic and prognostic biomarkers.
Recently, using a whole exome sequencing approach, Beltran and colleagues were able to show major genetic aberrations in NE-PC [7], highlighting its unique features distinct from adenocarcinoma and CRPC [8] on a molecular level. Further, according to findings published by Aparicio et al. [9], NE-PC is caused by dual loss of the tumor suppressor genes TP53 and RB1. It was shown to be able to clonally evolve from the basal stem cell compartment of the prostate epithelium and to display a small-cell phenotype also referred to as small cell prostate cancer (SCPC) [10].
In this study, we used the Simian-Virus 40 (SV-40) T-antigen driven model of the mouse prostate (TRAMP) as a model for studying primary resistance against second generation antiandrogen treatment [11,12]. As in human disease, TRAMP carcinogenesis is caused by the dual blockade of p53 and Rb1, specifically by SV-40 t-antigens under control of a rat probasin promoter exclusively expressed in the murine prostate [13]. In this model, SCPC, metastasis, castration resistance, and neuroendocrine differentiation have been observed primarily and upon both surgical and chemical castration [14][15][16], highlighting the advantage of intratumoral heterogeneity in TRAMP mice as seen in men [17,18]. While frequency of neuroendocrine differentiation was reported to be dependent on genetic background of the mice used [19,20], it is an event commonly observed in TRAMP mice. Using this model, we aimed to identify regulators of resistance towards androgen receptor inhibition.

RNA Sequencing and Gene Ontology Enrichment Analysis
Total RNA was isolated using RNAzol RT (R4533, Sigma-Aldrich, St. Louis, MO, USA). Quality control and cDNA synthesis was performed by the Core Facility for Genomics at the Medical University of Vienna. Raw RNA-seq data were aligned using STAR algorithm to the mouse genome version mm10. Aligned .bam files were imported into SeqMonk software (Babraham Bioinformatics, Babraham, UK) for quality control followed by differential expression analysis between groups using DeSeq2An FDR < 0.01 and a logfold change < +/−3 were defined as cutoffs for differential gene expression. Three biological replicates per group were analyzed. Gene ontology enrichment analyses were performed using the shinyGO tool [21]. Gene fusion search was performed using [22]. Data visualization was done using the R package chimeraviz

TRAMPC1 Cell Culture, Lentiviral Knockdown, Treatment, and Viability Assays
Cytotoxicity of enzalutamide was measured using the CellTiter Blue assay (Promega Corporation, Fitchburg, WI, USA) by seeding 2500 cells into each well of a 96 well plate. Cells were treated at least in triplicates with 10 µM enzalutamide in 100 µL during the indicated time. After the addition of 10 µL CellTiter Blue reagent, fluorescence was measured on a Varioskan™ LUX multimode plate reader (Thermo Fisher Scientific Inc., Waltham, MA, USA).

Verification of MALAT1 Knockdown and Alterations in Target Gene Expression by RT-PCR
To confirm the knockdown of MALAT1 in TRAMPC1, RNA was isolated using RNAzol ® RT (R4533, Sigma-Aldrich) according to the manufacturer's protocol. RNA concentration was measured on a Nanodrop 8000 Spectrophotometer (Thermo Scientific) and reversely transcribed using Lunascript RT Supermix Kit (New England Biolabs, E3010L). RT-PCR was performed on a QuantStudio 7 Flex System (Applied Biosystems) using GoTaq 2× qPCR-Mix (Promega), 500 nM primers, and 10 ng RNA/reaction. Primer sequences used were in Table 2:

Validation of MALAT1-Fusions by PCR
To detect NCBP3-MALAT1 fusion in TRAMPC1 cells, total RNA was isolated with RNAzol RT (Molecular Research Center IncCincinnati, OH, USA) and reversely transcribed using LunaScript RT SuperMix Kit (NEB, Ipswich, MA, USA). Primers were designed applying Primer-Blast software with the forward primer targeting NCBP3 (ACAGCGTGGAAACAACCTCA) and the reverse primer annealing to MALAT1 sequence (ACTGCTCGCTCCATCAGAAA). PCR was performed with Hot Star Taq Polymerase Kit (Qiagen, Venlo, The Netherlands). PCR products were purified using Qiaquick Gel Extraction Kit (Qiagen, Venlo, The Netherlands). Lastly, sequencing was performed.

Statistical Analysis
Organoid cell culture experiments were performed as a pooled experiment of three different TRAMP tumors and repeated twice with 12 wells in each experiment, resulting in 24 replicates for both populations and treatments. Quantitative real-time PCR experiments were carried out in triplicates. Two outliers out of 72 measurements were identified and excluded from further analysis. Statistical calculations were performed using Graph-Pad PRISM 7 (GraphPad Software Inc, La Jolla, CA, USA). Welch's t-tests were used to assess statistical significance. A two-sided p-value lower than 0.05 was considered as statistically significant.

Basal and Luminal Cell-Derived Organoids from TRAMP Tumors Exhibit Heterogenous Cellular Architecture and Potency
To examine differences in organoid formation between basal and luminal cell populations within TRAMP mice, we isolated basal and luminal TRAMP tumor cells using flow cytometry as previously shown by our group [23] ( Figure 1A) and submitted them to organoid culture using a protocol adapted from the Shen laboratory [24] (Video S1). After sorting, we used RNA sequencing to show distinct expression patterns of cytokeratins (CK) 5 and 8 for luminal and basal cells ( Figure S3B).We ensured isolation of transduced cells by performing qRT-PCR for SV40 t-antigen ( Figure S3A) and injecting part of the cell suspensions obtained into NSG mice, where they grew out as allografts and formed pulmonary metastases, while retaining neuroendocrine marker expression ( Figure S1D). While we observed no differences in maximum organoid size between luminal and basal cell-derived organoids ( Figure 1B) we noted striking differences in cellular organoid architecture. Basal cell-derived organoids exhibited multiple layers of cells, while luminal cell-derived organoids had thinner walls that mainly consisted of one cell layer ( Figure 1C). Basal cell-derived organoids exhibited expression of CK 5 and 8, with CK5 being expressed in basal cells and CK8 in luminal cells, as expected. Luminal cell-derived organoids exhibited a lack of basal cells and CK5 expression on the protein level, as confirmed by immunoblotting. Both basal and luminal cell-derived organoid lines showed androgen receptor expression ( Figure 1D).

Basal Cell-Derived Organoids Display Resistance towards Androgen Receptor Inhibition
As a next step, we aimed to identify differences in responses to antiandrogen treatment of basal and luminal-cell derived organoids using the androgen receptor inhibitor enzalutamide. Interestingly, both organoid lines were able to grow under enzalutamide treatment ( Figure 2). Luminal cell-derived organoids, however, exhibited significantly smaller organoid size and viability when treated with enzalutamide ( Figure 2B,D). Basal cell-derived organoids on the other hand showed a small, but statistically significant decrease in size, but no decrease in viability as determined by an ATP-dependent assay (Figure 2A,C). Interestingly, no difference between luminal and basal-cell derived organoids was observed upon treatment with abiraterone ( Figure S3C).

Gene Expression Profiling Reveals Distinct Epithelial and Neuroendocrine Signatures for Basal and Luminal TRAMP Tumor Cells
Intrigued by our finding of primary resistance of basal cell-derived organoids against enzalutamide treatment, we set out to investigate gene expression signatures of basal and luminal cell populations determined by RNA sequencing of three individual TRAMP tumors. Interestingly, we found that basal and luminal cell populations exhibited distinct molecular gene expression signatures ( Figure 3A). Overall, 575 and 714 basal and luminal-specific genes were identified, respectively. Gene ontology enrichment analyses of these genes showed enrichment of genes involved in response to wounding, cell migration and proliferation, epithelial differentiation, and keratinocyte differentiation in basal cells ( Figure 3B), while luminal cells exhibited enrichment of genes involved mainly in developmental processes including neuronal development ( Figure 3B), such as NTRK1. We confirmed the relative upregulation of NTRK1 in luminal cells compared to basal cells using qRT-PCR and the existence of NTRK1-positive cells within two individual TRAMP tumors using flow cytometry ( Figure S2A).

MALAT1-Fusions Are Abundant and Regulate Resistance towards Androgen Receptor Inhibition in TRAMP Tumor Cells
Since gene fusions-including ETS transcription factors and NTRK1-are common in human prostate cancer, we investigated whether TRAMP tumors would harbor fusion genes, again using RNA sequencing. Interestingly, we found that TRAMP tumors cells including the commercially available TRAMPC1 cell line harbored various fusion genes of MALAT1. These fusions were detectable in both luminal and basal cell-derived organoids and independent from treatment with enzalutamide ( Figure 4A, Supplementary Table S1), having said that fusions were not detectable in all luminal organoid specimen upon enzalutamide treatment, partially due to low RNA yield. We confirmed the most common of these fusions, NCBP3-MALAT1, in the TRAMPC1 cell line using RT-PCR ( Figure S1A). A breakpoint graph depicting this gene fusion is provided in Figure S2B. To evaluate the effects of such MALAT1 fusions on cancer cell biology in the TRAMP mouse model, we knocked down MALAT1 expression in TRAMPC1 cells using lentiviral shRNA delivery ( Figure 4B). Interestingly, downregulation of MALAT1 led to downregulation of the androgen receptor on the RNA ( Figure 4B, lower panel) and protein ( Figure 4C) level, which correlated with higher resistance towards enzalutamide treatment ( Figure 4D). Original Western blot can be found at File S1. (D) MALAT1 knockdown sensitizes TRAMPC1 cells to enzalutamide when compared to scrambled control, as seen through lowered relative absorbance in viability assays. Representative experiment with three technical replicates shown.

Discussion
To our knowledge, our study is the first to provide deep understanding of primary androgen receptor inhibition resistance in TRAMP mice. Primary resistance towards androgen receptor inhibition is a major problem in PCa treatment. In our study, we used the transgenic adenocarcinoma of the mouse prostate model (TRAMP mice), to identify regulators of primary resistance against androgen receptor inhibition. Advantages of this model lie in its ability to exhibit both castration resistance and primary neuroendocrine differentiation. Other murine models of PCa exhibiting neuroendocrine phenotypes and/or resistance to antiandrogen treatment [25][26][27], partially through transdifferentiation [28], exist. While these models often restrict carcinogenesis to either luminal or basal cell lineages, we found that TRAMP tumor cells of both basal and luminal phenotypes gave rise to cancer organoids in vitro (Figure 1, Figure S1D). This finding is in line with publications of both the Shen and Witte laboratories, who previously described luminal and basal cells to be potential cells of origin for prostate cancer [24,29,30]. Stoyanova et al. described that in the case of basal cells of origin, luminal cells were responsible for further tumor progression [31], which is reflected in our results, as basal cell populations vanished upon passaging in NSG mice ( Figure S1B). Accordingly, we found no basal cell marker expression in subcutaneous allografts and lung metastases from TRAMP tumor cells ( Figure S1D).
Interestingly, both luminal and basal organoids exhibited primary resistance towards androgen receptor inhibition. While this effect was more pronounced in basal cell-derived cancer organoids, both organoid lines were able to form and grow out under the presence of the androgen receptor inhibitor enzalutamide (Figure 2). The same was true for treatment with abiraterone, having said that, for these treatments, no differences in viability between basal cell-and luminal cell-derived organoids were observed ( Figure S3C). It is known that both basal and luminal cells respond to androgen withdrawal and stimulation and express the androgen receptor [32]. To find factors that contributed to this resistance, we performed RNA sequencing (RNAseq) experiments and provide gene expression data for both luminal and basal TRAMP tumor cells ( Figure 3A). Interestingly, ETS family members such as ERG, ETV1, or ETV4 were upregulated in basal cells as compared to luminal cells ( Figure S1C).
A potential weakness of our study is the cell isolation protocol used, which did not allow sorting of different luminal cell populations. This is of some concern, as recently, Karthaus et al. [33] as well as Crowley et al. [34] reported heterogeneity within luminal, but not basal cells of the murine prostate using single-cell RNA sequencing approaches. More research is needed to understand the function of these subsets within the luminal cell compartment of murine prostate cancer models such as TRAMP mice.
Another potential weakness of our methodology may be caused by selection of epithelial (wild-type) cells through organoid culture per se. We tried to minimize this error by ensuring that cell solutions used for organoid cultures contained tumor-initiating cells ( Figure S1) and that organoids expressed the SV40 t-antigen ( Figure S3A). Still, we cannot completely rule out contamination of our TRAMP tumor organoid lines with benign epithelial cells.
Furthermore, organoid experiments in our laboratory were conducted with EGFcontaining cell culture medium and TRAMP cells known to express wild-type PTEN only. As a previously described crosstalk between AR and EGF [35] will potentially influence the response to AR inhibition, we argue that future experiments with EGF-deprived cell culture medium and/or organoids gained from mouse models exhibiting loss of PTEN are needed.
As fusion genes were shown to be of paramount importance in human prostate cancer and androgen signaling [36], we searched for fusion genes using RNAseq. We found fusions of MALAT1 with various genes in organoid lines of both luminal and basal origin as well as in the TRAMPC1 cell line and confirmed their existence using RT-PCR ( Figure 4). MALAT1, a long non-coding RNA often expressed and thought to deregulate RNA splicing in CRPC patients [37], was shown to be a regulator of androgen receptor expression, to mediate cancer cell growth, invasion and migration and to correlate with PSA values and Gleason grading [38][39][40] in prostate cancer. Of note, a similar downregulation of the AR upon knockdown of MALAT1 was shown by Dai and colleagues in LnCAP cells [40]. Furthermore, studies found MALAT1 to be a potential diagnostic or prognostic biomarker in PCa [41][42][43][44] and a potential therapeutic target in various cancer entities [45,46]. Recently, in prostate cancer, MALAT1 was proposed to be a mediator of enzalutamide resistance through its indispensable role in AR-splice variant 7 (AR-V7) formation [47], and hence a potential target for pharmacological intervention in prostate cancer [48]. According to preclinical research by Chou et al., Cis-and Carboplatin-mediated suppression of the MALAT1/SF2 RNA splicing complex may lead to degradation of AR-V7, potentially resensitizing AR-V7 expressing PCa cells to enzalutamide [49]. Docetaxel treatment, on the other hand, may increase the generation of AR-V7 via altering the MALAT1/SF2 complex, potentially causing enzalutamide resistance [50]. Further, targeting MALAT1 was shown to change PCa cell metabolism towards a more glycolytic phenotype and to decrease the expression of oxidative phosphorylation enzymes causing cell arrest and death by Nanni et al. [51]. Among others, underlying mechanisms for these effects of MALAT1 in PCa were shown to be upregulation of miR-140 [52], deregulation of miR-1 and KRAS [53], enhancing function of EZH2 [54], as well as association with estrogen receptor subunits on the chromatin level [55]. The aforementioned research efforts dealing with MALAT1 and its role in regulating AR-V7, all mainly carried out in vitro, shows a need for PCa animal models expressing MALAT1.
Of note, two of the genes we found to be forming fusion genes with MALAT1, namely MVP and NCBP3, were shown to play a role in nucleocytoplasmic transport [56]. MVP, the major vault protein, was shown to be linked to multidrug resistance in a series of cancers [57] including PCa [58], and was recently proposed as a biomarker for lethal outcomes in PCa by Ramberg and colleagues [59]. To our current knowledge, it is unknown whether MALAT1 fusions occur in men, and whether they would play a role in resistance to enzalutamide or development of neuroendocrine phenotypes such as NEPC in PCa. We therefore highlight the importance of future studies evaluating the clinical impact of MALAT1 overexpression and/or gene fusions in human PCa.

Conclusions
We highlight the role of TRAMP mice as a model of studying MALAT1-driven prostate cancer and primary second-generation antiandrogen resistance.  Table S1: List of gene fusions found, Video S1: Video showing growth of TRAMP organoids, File S1: Original western blots.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy issues.