SOX9 is a driver of aggressive prostate cancer by promoting invasion, cell fate and cytoskeleton alterations and epithelial to mesenchymal transition

Aggressive lethal prostate cancer is characterised by tumour invasion, metastasis and androgen resistance. Understanding the mechanisms by which localised disease progresses to advanced lethal stages is key to the development of effective therapies. Here we have identified a novel role for the transcription factor, SOX9, as a driver of aggressive invasive prostate cancer. Using genetically modified mouse models, we show that increased Sox9 expression in the prostate epithelia of animals with Pten loss leads to a highly invasive phenotype and metastasis. In depth analysis of these mice and related in vitro models reveals that SOX9 acts a key regulator of various processes that together promote tumour progression. We show that this factor promotes cell lineage plasticity with cells acquiring properties of basal stem cells and an increase in proliferation. In addition, increased SOX9 leads to changes in cytoskeleton and adhesion, deposition of extracellular matrix and epithelia to mesenchyme transition, properties of highly invasive cells. Analysis of castrated mice showed that the invasive phenotype driven by SOX9 is independent of androgen levels. Our study has identified a novel driver of prostate cancer progression and highlighted the cellular and molecular processes that are regulated by Sox9 to achieve invasive disease.


INTRODUCTION
Prostate cancer is a leading cause of cancer mortality in men in the western world. Lethal aggressive disease is characterised by tumour invasion, metastasis and castration resistance [1]. Identifying and understanding the processes that drive prostate cancer progression is an important challenge. The PTEN (phosphatase and tensin homolog deleted on chromosome 10) tumour suppressor gene has been implicated in prostate cancer with latestage disease showing loss of function in over 60% of samples [2,3]. Mice lacking Pten in the prostate show prostate intraepithelial neoplasia (PIN) that progresses to adenocarcinoma in older animals with minimal invasive properties [4]. These animals have been used to identify factors that cooperate with Pten loss to drive prostate cancer progression to more aggressive stages [5,6].
The transcription factor SOX9 has been shown to be a key regulator in various processes during embryogenesis, stem cell commitment and cancer. In the prostate, we and others have shown that it is expressed at early stages of organogenesis and required for prostate development [7, www.impactjournals.com/oncotarget/ Oncotarget, 2018, Vol. 9, (No. 7), pp: 7604-7615 Research Paper www.impactjournals.com/oncotarget 8]. SOX9 has been implicated in prostate cancer, with high levels of SOX9 found in early stages of prostate neoplasia and high grade PIN in mice and associated with increasing Gleason grade in humans [9]. Proposed oncogenic mechanisms to achieve increased SOX9 include transcriptional activation by ERG [10], which is highly expressed in tumours with TMPRSS2:ERG fusions, and loss of Zbtb7a, encoding a protein that antagonizes SOX9 activity [11]. in vivo functional studies have shown that overexpression of Sox9 in prostate epithelia in transgenic mice induced increased proliferation and high grade PIN in mice that also had a heterozygous Pten deletion [9]. Moreover, Sox9 loss inhibited tumour formation in two prostate cancer mouse models (TRAMP and Myc overexpression) [7].
Here we identify a novel role for SOX9 at later stages of prostate cancer progression. Analyses of genetically modified mice show that overexpression of Sox9 in the prostate of mice with a homozygous Pten deletion leads to highly invasive prostate cancer and metastasis. Our studies reveal that Sox9 regulates various processes that together drive increased proliferation and concerted stromal cell invasion and metastasis, hallmarks of aggressive disease.

SOX9 promotes highly invasive tumour formation in Pten deficient prostates
In an effort to determine the role of SOX9 in aggressive prostate cancer we generated genetically modified mice that express extra levels of Sox9 in prostate epithelia in addition to a homozygous Pten deletion. For this, mice containing a construct where Sox9 is conditionally expressed under the control of a ubiquitous promoter in tissues where Cre is expressed (Z/Sox9, [9]) were bred with mice with a homozygous Pten loxP containing allele (Pten fl/fl , [12]) and a transgene where Cre is driven by a Probasin derived promoter (PbCre4 [13]) (the Z/Sox9 tg/+ ; Pten fl/fl ; PbCre tg/+ mice are referred to as Sox9;Pten;PBCre mutants throughout the manuscript). This Sox9 containing transgene also drives expression of GFP, allowing us to specifically follow prostate epithelial cells that overexpress Sox9. As expected, Sox9;Pten;PBCre mutant prostates were positive for GFP staining ( Figure 1A). Initial analysis of prostates from 3 month old mice showed that Sox9;Pten;PBCre mutants had a larger size compared to Pten;PBCre mutants ( Figure  1A and 1B). Sections of these prostates revealed prostatic ducts filled with cells and high grade PIN, as seen in Pten;PBCre mutant animals. In contrast to Pten;PBCre mutant animals, Sox9;Pten;PBCre prostates had abundant evidence of a highly invasive phenotype that was focal in nature and associated with an increase in stromal tissue ( Figure 1C). The epithelial cells invading into the surrounding stroma expressed high levels of SOX9 ( Figure  1D). As Sox9 is expressed in normal prostate, we used GFP staining to reveal transgenic Sox9 expression. This analysis confirmed that GFP positive cells show elevated levels of SOX9 and that the invading epithelial cells express transgenic Sox9 ( Figure 1E and Supplementary Figure 1 [9]). Consistent with the invasive phenotype we observe metastatic GFP positive cells in lymph nodes and lungs of 6 month old Sox9;Pten;PBCre mutants, a process that we do not observe in Pten;PBCre mutant animals at this stage (Supplementary Figure 2).

Increased proliferation in Sox9;Pten;PBCre mutant prostates
The larger size of Sox9;Pten;PBCre mutant prostates prompted us to look at proliferation in this tissue. To determine the specific effect of transgenic Sox9 we stained prostate sections with Ki67, a marker of proliferation, and GFP antibodies. These studies showed that Sox9;Pten;PBCre mutant prostates have a significant increase in Ki67 positive epithelial cells that were also positive for GFP, compared to GFP negative cells or to Pten;PBCre mutant animals (Figure 2A and 2B). To understand the mechanisms relating to proliferation regulated by Sox9, we stained Sox9;Pten;PBCre mutant prostates with several markers and observed an increase in the expression of Cyclin D1, particularly in invasive regions ( Figure 2C). Pten;PBCre mutant prostates have been reported to show senescence that is bypassed when other oncogenic insults such as Tp53 or Zbtb7a loss are present [11,14]. Due to the focal nature of the phenotype, we used antibody staining to investigate senescence associated genes and found no major difference in expression of p53, p27, p16 and p19 between Sox9;Pten;PBCre and Pten;PBCre mutant prostates ( Figure 2C). Based on the study of mice with Zbtb7a and Pten loss, which show a decrease in PTEN induced senescence (PIS) and an increase in SOX9 transcriptional activity [11], we investigated the expression of the tumours suppressor retinoblastoma (RB). This analysis showed decreased RB expression in Sox9;Pten;PBCre mutant prostates compared to Pten;PBCre mutants ( Figure  2C). In support of this, direct evidence that RB loss can overcome PIS has been shown in prostates that have both Rb and Pten deleted [15].

Adhesion phenotype in Sox9;Pten;PBCre mutant prostates
Further analysis of the invasive phenotype in Sox9;Pten;PBCre mutant prostates showed that it is focal and found in all prostatic lobes although a lower incidence is seen in the ventral prostate, possibly due to variation in the expression of Cre in different lobes ( Figure 3A and 3B, [13]). It is characterised by the disruption of the stromal barrier, as shown by the interrupted smooth-muscle (SMA) actin staining pattern surrounding the duct ( Figure 3C). An increase in stromal tissue is observed in the invasive regions as well as high levels of extra cellular matrix markers such as fibronectin and collagen when compared to Pten;PBCre mutant animals ( Figure 2C). These studies indicated that high levels of Sox9 lead to a change in adhesive properties. To investigate this hypothesis we overexpressed SOX9 in LNCaP cells, which do not express PTEN. LNCaP cells with high levels of SOX9 showed a difference in cellular morphology with a flatter appearance than control cells ( Figure 3D). This difference was highlighted with Phalloidin staining, which showed an increase in F-actin in SOX9 expressing cells (Figure   3D). Consistent with an effect of Sox9 on adhesion, genes associated with adhesomes, GSN, MACF1 and PALLD, were found to be increased in cells expressing high levels of SOX9 ( Figure 3E). Moreover, these cells showed increase adhesion to fibronectin but not to collagen ( Figure  3F).

Epithelia to mesenchyme transition in Sox9;Pten;PBCre mutant prostates
The highly invasive phenotype observed in Sox9;Pten;PBCre mutant prostates suggested an increase in epithelia to mesenchyme transition (EMT), which has been associated with SOX9 activity in other tissues. To address this possibility sections from Sox9;Pten;PBCre mutant prostates were stained with antibodies to E-cadherin or Vimentin together with GFP, to mark the cells expressing transgenic Sox9. Cells undergoing EMT have been shown to downregulate the expression of the epithelial marker E-cadherin and upregulate the mesenchymal marker vimentin. Our studies show that some GFP positive cells do show this phenotype and are particularly abundant in regions of high invasion ( Figure  4A and 4B). Several key regulators of EMT have been shown to promote the loss of epithelial characteristics and the gain of mesenchymal traits [16]. Analysis of several of these regulators showed that there is an increase in SLUG (encoded by Snai2) and ZEB1 positive cells in Sox9;Pten;PBCre mutant prostates ( Figure 4C). To investigate whether EMT is a direct effect of increased Sox9 in prostate cells, we analysed this process in LNCaP cells that had high levels of SOX9. These cells had an increase in Vimentin positive cells when compared to control cells, however, E-cadherin was not found to be reduced ( Figure 4D, 4E and 4F). Analysis of the expression of EMT regulators revealed an increase in ZEB1 but not SNAIL, SLUG or TWIST in LNCaP cells with high SOX9 ( Figure 4G) suggesting SOX9 can promote mesenchymal features in prostate cells, possibly through ZEB1.

Cell fate identity of invasive cells in Sox9;Pten;PBCre mutant prostates
Prostate epithelial ducts are formed of basal and luminal cells. We therefore used antibody staining to investigate the cell fate identity of the invasive cells in Sox9;Pten;PBCre mutant prostates. Our studies revealed that invasive cells within the stroma express both luminal (CK8) and basal (CK5 and p63) markers ( Figure 5A). Co-staining of Sox9;Pten;PBCre mutant prostates with GFP showed that some invasive regions had a high number of basal cells that coexpressed transgenic Sox9 while others regions lacked basal cells ( Figure 5B). The prostate basal cell population has been proposed to contain a high number of progenitor/stem cells [17,18]. The relative increase in basal markers in Sox9;Pten;PBCre mutant prostates and the association of SOX9 with stem cell activity in other tissues, prompted us to look at stem cell markers. Antibody staining revealed an increase in BMI1, SOX2 and TCF4 in neoplastic regions of Sox9;Pten;PBCre prostates, including the invasive regions, relative to Pten;PBCre mutants ( Figure 5C). This result was confirmed in LNCaP cells expressing high levels of Sox9 that showed an increase in BMI1, SOX2 and TCF4 proteins (Supplementary Figure 3), although we could not detect p63 in these cells.

Effect of castration on Sox9;Pten;PBCre mutant prostates
Castration is the first line of therapy for prostate cancer patients in the clinic and SOX9 has been proposed to interact with AR [19]. Therefore we wanted to investigate whether the phenotype observed in Sox9;Pten;PBCre mutants was dependent on AR levels. Analysis of castrated mice showed that Sox9;Pten;PBCre mutant prostates were similar in size to Pten;PBCre mutant prostates, which were both smaller than their intact counterparts. However, the invasive phenotype was still observed in the Sox9;Pten;PBCre mutant prostate highlighted by the lack of stromal SMA staining surrounding the ducts (Figure 6A and 6B). Consistent with the lack of size difference, the levels of castration resistant proliferation, as measured by Ki67 staining, and apoptosis, as measured by Caspase 3 staining, were found to be similar between Sox9;Pten;PBCre and Pten;PBCre mutants ( Figure 6B and 6C). Levels of nuclear AR staining were low in both mutants, a phenotype that was found in Sox9;Pten;PBCre mutant prostates that had not been castrated ( Figure 6B and Supplementary Figure 4).

DISCUSSION
SOX9 has been implicated in cancer in many tissues and its role has been shown to be varied and context dependent. We have been investigating the function of this factor in the normal and neoplastic prostate. Our previous work had shown that Sox9 can promote tumour initiation in mice with a heterozygous Pten mutation. This study identifies a novel role for SOX9 as a driver of aggressive late stage invasive prostate cancer in homozygous Pten mutant animals. In addition, we show that its mechanism of action is to regulate several key processes that together promote highly invasive disease. Our study reveals that high Sox9 levels lead to Pten mutant prostate epithelial cells increasing proliferation and acquiring properties of basal stem cells; to the induction of EMT, the deposition of extracellular matrix and changes in cytoskeleton and adhesion, properties that promote a highly invasive phenotype. Our analysis is consistent with SOX9 being a key regulator of these processes rather than regulating one event that brings about these phenotypes. A similar www.impactjournals.com/oncotarget role has recently been proposed for this factor in basal cell carcinoma (BCC) [20].
Lineage plasticity has been proposed to be a mechanism not only of tumour progression to aggressive stages but also resistance to therapy [15,21]. This change in cell fate is thought to occur through the increase in cells with stem like properties [22]. Recent transcriptome analysis has identified an association between human basal cells and aggressive prostate disease [18]. Our studies reveal that Sox9 can induce a more basal-like phenotype in Pten mutant prostates. In addition, we identified an increase in the stem cell markers Sox2, Tcf4 and Bmi1 in both human and mouse prostate cells overexpressing SOX9. Direct regulation of Bmi1 expression by SOX9 has been shown in mouse embryonic fibroblasts and proposed to occur in colorectal cancer [23]. A more indirect mechanism could give rise to the high levels of SOX2 in Sox9;Pten;PBCre mutants as loss of RB has been proposed to lead to an increase in SOX2 expression, which was greatly augmented with concomitant TP53 loss [15].
A striking effect of high levels of SOX9 on LNCaP cells is their change in morphology and adhesion properties. In a study on BCC, SOX9 has been shown to directly regulate genes involved in adhesion, cytoskeleton remodelling and extracellular matrix [20]. Consistent with this, we observe an increase in genes involved in cell shape and adhesion. Our work indicates that through the regulation of cytoskeleton dynamics and adhesion, SOX9 contributes to Pten mutant cells altering their epithelial properties and becoming more motile. The actin remodelling genes MACF1 and PALLD that we show are increased in SOX9 overexpressing cells have been shown to promote cell migration [24,25]. Antibody staining on Sox9;Pten;PBCre mutant prostates revealed groups of invasive cells, suggesting a mechanism of collective cell migration. Presumably this process was further influenced by the ECM changes in the stroma.
SOX9 has been associated with the process of EMT in many tissues including embryonic neural crest [26]. Evidence of this process was found in Sox9;Pten;PBCre mutant prostates with a subset of GFP positive cells showing Vimentin staining. Increased staining was also observed in LNCaP cells expressing increased SOX9, although not in all cells. Moreover, an increase in ZEB1, an EMT regulator, was observed in both models. EMT is thought to be a transient process, therefore the sporadic staining we observed could reflect cells at different stages of the pathway.
Consistent with the highly invasive phenotype observed in Sox9;Pten;PBCre mutant prostates, we did find evidence of metastasis in tissues such as lymph nodes and lungs. However, in most cases these are single or small cell clusters suggesting that high levels of SOX9 promote cell migration to secondary sites but other factors are required to establish colonisation and growth at these locations. A similar effect was observed in a breast cancer study where overexpression of Sox9 led to micrometastasis in the lung but macrometastases were only observed when cells had increased levels of Slug [27].
Our results show directly that increased levels of SOX9 are required to drive both tumour initiation and invasive advanced disease and that it does this in association with other oncogenic pathways such as loss of Pten. SOX9 is expressed in the normal prostate and mechanisms to achieve increased levels in tumours have been proposed. These include regulation of SOX9 expression by ERG [10], which is found in tumours with TMPRSS2:ERG fusions, and loss of Zbtb7a, which encodes a factor proposed to antagonize SOX9 function rather than expression [11]. This dependence on levels might explain the focal nature of the phenotype in Sox9;Pten;PBCre mutant animals as only some cells achieve the required amount of protein to drive aggressive disease.
In conclusion, we identified a novel role of SOX9 as a driver of invasive prostate cancer, a property of aggressive disease. Our study shows that SOX9 is a key regulator of various cellular and molecular processes that together act to drive tumour progression. The genetically modified mice in this study are an ideal model to study the process of cell invasion and disease progression in prostate cancer and identify biomarkers of this process.

Mouse strains
The PBCre transgenic mice (ARR2PBCre) have been described previously [13]. Mice with the conditional allele of Pten were obtained from The Jackson Laboratory [28]. The Z/Sox9 transgenic mice were kindly provided by Kathryn Cheah [9]. These male animals were bred on a mixed genetic background. All mouse work was carried out in accordance with the Institute of Cancer Research guidelines and with the UK Animals (Scientific Procedures) Act 1986.

Mouse prostate histology
Histological phenotype of samples was assessed on haematoxylin and eosin stained sections. Serial sections were then stained for immunohistochemical analysis. Histological assessment was based on published guidelines and assisted by a pathologist [29]. An unpaired t-test was performed to test if there is a significant difference between the wet weight of prostate tumours.

Quantification of cell proliferation
Proliferation was quantified by immunofluorescent stain of sections with an antibody against Ki67. The number of proliferating cells was calculated by counting the number of nuclear Ki67 stained cells with and shown as a percentage of the total number of cells stained with nuclear DAPI. Sox9;Pten mutant tumours were also stained with an antibody against GFP, and the number of Ki67 single positive cells and Ki67/GFP double positive cells were counted. Cells from at least 4 high power fields were counted per animal, which totalled more than 900 cells per animal. Three animals of each genotype were analysed. Randomly selected fields were counted for control analysis. An ANOVA or t-test was used to test if there was a significance difference in the number of proliferating cells between each group.

Immunohistochemistry, immunofluorescence and western blotting
Mouse tissues were fixed overnight in 4% paraformaldehyde (PFA), dehydrated in an ethanol gradient series, washed in Histoclear and embedded in wax. Antigen retrieval was obtained by boiling sections in citrate buffer (0.1 M sodium citrate pH6 and 0.05% Tween) and sections were treated with 3% H 2 O 2 to block endogenous peroxidase activity. Sections were blocked in 10% sheep serum and then incubated with primary antibodies in 1% sheep serum overnight. For DAB chromogen (Agilent) staining the ABC vector kit (Vector Laboratories) was used with biotinlyated secondary antibodies according to manufacturer's instructions and were counterstained with haematoxylin. Masson's stain was performed using the Trichrome light green stain from TCS Biosciences (HS773-LG). For all mouse tumour stains, sections were processed from at least three animals of each genotype. For immunofluorescence cell staining, LNCaP cells were plated on coverslips, fixed for 20 minutes in 4% PFA and washed in PBS containing 0.01% Triton. Cells were then blocked in 10% sheep serum and incubated with primary antibodies in 1% sheep serum overnight. DAPI was used as a nuclear counterstain. TRITC labelled phalloidin (Thermo Fisher Scientific) was used to stain F-actin. Secondary fluorescent antibodies were obtained from Molecular Probes and were used at a 1:1000 dilution. Fluorescent images were visualized and collected on a Leica TCS-SP2 confocal microscope. Western blotting was performed using standard protocols, with cells lysed in RIPA buffer. The following antibodies were used; SOX9 (Millipore AB5535), GFP (Abcam

Cell line work and lentivirus production and infection
LNCaP and 293T cells were STR profiled to confirm their identity. Both cell lines were maintained in RPMI1640 with 10% FCS and used at low passage number (less than 20). www.impactjournals.com/oncotarget Sox9 overexpression was achieved by infecting cells with Lentiviral particles generated from the construct pLenti-GIII-CMV-SOX9-GFP-2A-Puro (LV318630, Applied Biological Materials Inc) and control cells infected with Lentivirus particles from the plasmid pLenti-GIII-CMV-RFP-2A-Puro (Applied Biological Materials Inc) or GIPZ-GFP (GE Dharmacon). Lentivirus particles were made in 293T cells by transfecting the Lentiviral plasmid with Lipofectamine 3000 (Thermo Fisher Scientific), together with the packaging plasmids psPAX2 and pMD2.G. Viral supernatants were collected, filtered, supplemented with 8 μg/ml polybrene and used to infect LNCaP cells.

Cell adhesion assay
96-well plates were coated with 10 μg/ml Fibronectin (Sigma, F1141), 10 μg/ml Collagen I (BD Biosciences, 354236) or 10 mg/ml heat denatured BSA for 2 hours at room temperature and then washed. LNCaP cells were dissociating in TrypLE (Thermo Fisher Scientific) and seeded at 10,000 cells per well in 100 μl of media. Cells were incubated at 37°C for 1 hour, washed three times with PBS and fixed with 4% PFA for 20 minutes at room temperature. To quantify adhered cells, cells were stained with DAPI and the total number of cells per well were counted using the Celigo Imaging Cytometer (Nexcelom) with the Direct Cell Counting analysis.

Author contributions
AS, JF and JB conceived and designed the experiments; JF, AC, JN and EK performed the experiments; AS, JF and AC analysed the data; AS and JF wrote the manuscript.