Oncogenic HrasG12V expression plus knockdown of Cdkn2a using ecotropic lentiviral vectors induces high-grade endometrial stromal sarcoma

The uterine corpus represents the most common site for tumour development in the female genital system. Uterine neoplasms are categorised as epithelial, mesenchymal, mixed epithelial-mesenchymal or trophoblastic tumours. In this study we employed a mouse genetic approach using the MuLE lentiviral gene regulatory system to functionally test the ability of ecotropic lentiviruses to model epithelial and mesenchymal uterine malignancies ex vivo and in vivo. We discovered that MuLE lentiviruses efficiently infect uterine stromal cells but not endometrial epithelial cells when injected into the uterus of cycling, pseudopregnant or ovarectomized mice. Consistent with this cellular infection spectrum, we show that intra-uterine injection of ecotropic MuLE viruses expressing oncogenic HrasG12V together with knockdown of Cdkn2a induce high-grade endometrial stromal sarcomas. These findings establish this approach as an efficient method of generating autochthonous mouse models of uterine sarcomas and in general for performing genetic manipulations of uterine stromal cells in vivo.


Introduction
Early-stage uterine tumours are associated with an excellent prognosis after surgical treatment. However, for patients with advanced, metastatic or recurrent uterine cancer, limited therapeutic options are available [1,2]. To improve understanding of these diverse diseases it would be desirable to have a series of representative mouse models that reflect not only the different histological types of uterine tumours but also the underlying molecular genetics of each of these different tumour types. In recent years there has been some progress towards this goal.
The most prevalent tumours of the female reproductive system are carcinomas of the endometrium, which are classified in a dualistic model based on histological appearance and a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 molecular differences [1]. Type I endometrioid tumours account for 80% of endometrial malignancies and are characterised by a relatively high survival rate, due partly to their frequent early detection following abnormal uterine bleeding. Several studies report that PTEN mutations occur early in the development of type I tumours [3][4][5]. In terms of generating rapid, flexible and accurate mouse models of this disease, it has been shown that the injection of adenoviruses expressing Cre into the uterine lumen of adult Pten fl/fl mice caused the development of uterine carcinomas [6]. Tumour onset and severity can be accelerated by combining Pten deletion with Kras activation [7]. This combination of germline-based generation of Cre-regulatable genetic alleles plus viral-based delivery of Cre to somatic endometrial cells represents one approach to speed up the generation of mouse models of different molecular forms of endometrial carcinoma. Type II endometrial tumours are more common in older, non-obese women and, despite their rarity, have a worse outcome due to their highly invasive and metastatic nature. They comprise three different subtypes: serous, clear cell and undifferentiated carcinomas, each representing approximately 5-10% of all endometrial carcinoma cases [1]. Type II tumours are associated with both overlapping and distinct sets of genetic alterations to those found in type I tumours. They exhibit extensive somatic copy number alterations. In contrast to type I tumours, type II tumours are frequently characterised by early mutations in TP53 [8][9][10]. The deletion of the Trp53 gene in the mouse endometrium results in the formation of all histological subtypes of type II disease after 14-16 months [11].
Uterine mesenchymal tumours account for 3% of uterine tumours and are derived from the soft tissue of the uterine corpus, which comprises endometrial stroma, smooth muscle and blood vessels [12]. Uterine sarcomas are broadly classified as leiomyosarcomas, accounting for 60% of cases, endometrial stromal sarcomas, and undifferentiated sarcomas. These categories are defined by the presence of specific molecular alterations as well as by tumour morphology and prognosis [13]. Leiomyosarcomas are composed of cells which demonstrate smooth muscle differentiation. The majority of uterine leiomyosarcomas occur sporadically. Commonly found alterations include MED12, HMGA2 and PTEN [14][15][16]. Patients with germline mutations in fumarate hydratase (FH) have an increased risk of developing uterine leiomyosarcomas as well as uterine leiomyomas [17].
Endometrial stromal sarcomas are composed of cells resembling the endometrial stroma and are far less frequent than smooth muscle tumours, such as leiomyosarcomas. They are characterized by chromosomal translocations, most commonly involving the zinc finger genes JAZF1 and SUZ1 [18,19]. Undifferentiated uterine sarcomas represent a very rare and aggressive category of uterine tumours. Considering their low incidence rate, there is little to no information about molecular alterations found in these tumours. They are diagnosed by exclusion after elimination of other high-grade uterine tumours with a sarcomatous component [20,21]. However, again using the approach of combined germline genetics plus viral delivery of Cre, injection of an adenovirus expressing Cre-recombinase into the uterus of mice harbouring conditional alleles of oncogenic Kras and Trp53 caused the development of undifferentiated pleomorphic sarcoma [22,23].
In order to accelerate the process of generating genetically-complex autochthonous mouse models of human tumours we recently developed the MuLE lentiviral gene regulatory system, that facilitates combinatorial somatic genetics in cultured cells and in vivo in mouse tissues. As a proof of principle, we showed that this system could be employed to generate new mouse models of muscle-derived undifferentiated sarcomas that harbour alterations in multiple signalling pathways [24]. In the current study we investigated whether the MuLE system could also be applied to the setting of modelling of different forms of uterine tumours. We show that the MuLE system can be used to model endometrial sarcomas but does not appear to be easily applicable to modelling of uterine carcinomas.

Ovarectomy
Adult, cycling female B6.Cg-Gt(ROSA)Sor tm14(CAG-tdTomato)Hze /J mice were anaesthetized with isoflurane and the ovaries were exposed. Both uterine horns were ligated and the ovaries were removed. 2 weeks following surgery lentiviruses were injected into the uterus.

Intrauterine viral injections
Approximately 4 week-old female mice were anaesthetized using isoflurane and the uteri were exposed by making a small incision. 10 μL of concentrated lenti-and adeno-viruses expressing different genetic manipulations were injected into one uterine horn with a 30G insulin syringe. The animals were euthanised 1-6 months after injection depending on the experiment. The uteri were harvested, fixed in 10% formalin, paraffin-embedded and cut in 5 μm thick sections. For the identification of the Cre-infected cells, B6.Cg-Gt(ROSA)Sor tm14(CAG-tdTomato)Hze /J uteri were fixed in 4% paraformaldehyde for 4-6 h, transferred to 30% sucrose overnight, embedded in OCT and cut in 5 μm thick sections.

Allograft studies
For allograft experiments, 8 week-old female SCID/beige mice were anaesthetized by the inhalation of 2.5% isoflurane and 1x10 6 cells resuspended in 50% Matrigel (BD, no. 354230, Eysins, Switzerland) were injected subcutaneously in a total volume of 150 μL. In vivo imaging as well as a caliper were used to follow tumour development over time.

In vivo imaging studies
Noninvasive in vivo fluorescence and bioluminescence imaging was performed using the IVIS Spectrum (Perkin Elmer, Waltham, MA) together with the Living Image software (version 4.4). Mice were anaesthetized by using 2.5% isoflurane. During imaging, the isoflurane levels were reduced to 1.5%. All fluorescence measurements were performed in epi-fluorescence mode. For bioluminescence imaging, mice were injected subcutaneously with 150 mg/kg Dluciferin (Caliper, no. 122796, Waltham, MA) and imaged 15 min after injection. For the quantification of the total radiant efficiency, a region of interest was drawn around the mouse and the total radiant efficiency was automatically detected.

Isolation and maintenance of pEEC
We used previously published protocols [26,27] with minor modifications for the isolation of primary endometrial epithelial cells from mice. Five to ten 3-week-old C57BL/6 mice were sacrificed by cervical dislocation and uterine horns were taken out, dissected in Hanks Balanced Salt Solution (HBSS, ThermoScientific, Waltham, MA) and cut open lengthwise. The fragments were then pooled and incubated with 0.5% Trypsin (Sigma Aldrich, St. Louis, MO)/ 2.5% pancreatin (Thermo Fisher Scientific, Waltham, MA) in HBSS for 60 min at 4˚C and 60 min at 22˚C. Digested uteri were transferred to ice-cold HBSS and vortexed for 30 sec to release epithelial cells. Uterine tissues were filtered through a 40 μm nylon mesh and epithelial cells were collected and centrifuged at 1.000 g for 5 min. Cells were then resuspended in basal medium containing DMEM/F12 (Sigma Aldrich, St Louis, MO) supplemented with 1 mmol/l HEPES (Thermo Fisher Scientific, Waltham, MA) and 1% penicillin/streptomycin (Sigma Aldrich, St Louis, MO). Cells were mechanically disrupted by pipetting until clumps of cells are observed under the microscope. Cells were then seeded in basal medium supplemented with 5 ng/ml EGF (PeproTech, Hamburg, Germany), 1 mg/ml insulin (Sigma Aldrich, St Louis, MO), 0.55 mg/ml transferrin (Sigma Aldrich, St Louis, MO), 0.5 μg/ml sodium selenite (Sigma Aldrich, St Louis, MO) (BIE-basal medium, ITS supplement and EGF) and 0.5% dextran-coated charcoal-stripped FCS (Biochrom AG, Berlin, Germany) and plated into cultures dishes. Cells were cultured in BIE plus 0.5% stripped FCS in an incubator at 37˚C with saturating humidity and 5% O 2 .

Immunohistochemistry
Immunohistochemical analysis was performed on formalin-fixed paraffin sections after antigen retrieval (

Establishment of a primary endometrial epithelial cell culturing system
To study the ability of the MuLE system to model uterine malignancies, we first focused on the most common uterine tumour type, endometrial carcinoma. We established a murine endometrial epithelial cell culture model to allow us to study endometrial tumour development ex vivo. Freshly isolated cultures of primary endometrial epithelial cells (pEECs) showed almost uniform postive staining for E-CADHERIN, ß-CATENIN and CYTOKERATIN-7 and, with the exception of a small number of cells, negative staining for VIMENTIN, establishing the relative purity of the epithelial isolation ( Fig 1A). Next, we optimised cell culture conditions to allow the proliferation of these cells. The addition of 0.5% charcoal-stripped FCS to the medium allowed expansion of the cells into larger clusters in which cells maintained their epithelial cobblestone morphology. If FCS was not added the cells maintained their epithelial morphology but did not expand and passaging them was not possible (Fig 1B). Based on these results, we decided to culture the cells with 0.5% FCS.
The extended replicative life span which results from the inactivation of the p53 pathway is thought to represent an important event in the multistep process of endometrial carcinoma development. Given this genetic link to human endometrial carcinogenesis and additionally since pEECs fall into senescence with ongoing cultivation on plastic dishes [28], we tried to immortalise them by using primary epithelial cultures prepared from Trp53 fl/fl uteri. Trp53 was deleted ex vivo by treating the cells with a Cre-expressing adenovirus. Additionally, in a parallel approach, we infected pEECs from wildtype mice with a MuLE vector expressing a miR-30-based shRNA against Trp53 together with drug resistance. Genetic deletion or knockdown of Trp53 allowed the cells to be expanded and passaged, but passaging still caused a change in cellular morphology to a more fibroblast-like appearance (Fig 1C), meaning that these cultures do not represent a system that can be used for long-term studies of endometrial epithelial cells.
We next sought to determine whether primary endometrial epithelial cells could be readily infected by MuLE viruses. For these experiments we isolated endometrial epithelial cells from uteri of Pax8-rtTA; LC1; ROSA26-lox-stop-lox tdTomato mice [25,29]. PAX8 is a transcription factor which is essential for the development of the female genital tract, including luminal and glandular endometrial epithelial cells [30]. In this mouse model, the reverse tetracyclinecontrolled transactivator (rtTA) is expressed under the control of the Pax8 promotor (Pax8-rtTA). In the presence of doxycycline, rtTA can bind to a tetracycline response element (TRE) which leads to the expression of Luciferase and Cre, specifically in endometrial epithelial cells, which induces the expression of tdTomato (Fig 2A). Following the administration of female mice with doxycycline to activate Cre-mediated recombination, uteri were harvested and Pax8-driven tdTomato expression in glandular and luminar epithelial cells was confirmed by in vivo imaging (Fig 2B) and microscopy. We isolated pEECs from the uteri of these mice and  infected them with ecotropic MuLE lentiviruses expressing GFP to label infected cells, After 48 hours only a few Pax8-tdTomato-expressing cells were GFP-positive but the majority of GFPexpressing cells were Pax8-tdTomato-negative (Fig 2C, top half), demonstrating that infection of epithelial cells by MuLE viruses is rare.
We nonetheless reasoned that it might be possible to transform epithelial cells and potentially create an ex vivo engineered endometrial carcinoma model by infecting pEECs with an ecotropic MuLE virus expressing shCdkn2a plus Hras G12V , taking advantage of the fact that largely mutually exclusive genetic alterations in KRAS, HRAS, NRAS, BRAF and NF1 occur in about 32% of human endometrial carcinomas (www.cbioportal.org) and loss of function of Cdkn2a in mouse cells allows escape from RAS-induced senescence and transformation [31,32]. Infection of the above-described genetically labelled pEEC cultures with the shCdkn2a and Hras G12V -expressing lentivirus led to the emergence of a population of apparently transformed cells that proliferated rapidly, could be sequentially passaged and formed stream-like structures in culture, however these cells were negative for tdTomato expression (Fig 2C,  lower half), indicating that they are not epithelial in origin. It is known that endometrial epithelial cell cultures can be contaminated by small numbers of endometrial stromal cells and indeed, when these transformed cells were injected subcutaneously into SCID/beige mice they formed tumours with a sarcoma-like morphology. Histologically, the tumours stained negatively for CD31, DESMIN, MYOD1, MYOGENIN, alpha-SMA and positively for VIMENTIN ( Fig 2D). This molecular phenotype is consistent with a diagnosis of undifferentiated sarcoma of likely mesenchymal cell origin. We conclude that while endometrial epithelial cells can be infected by MuLE viruses, albeit at low efficiency, transformation of other stromal cell types present in these cultures occurs efficiently by Hras G12V overexpression and Cdkn2a knockdown.

Stromal cells are preferentially transduced by lentiviruses following in vivo injection
To study the efficacy of ecotropic lentiviral infection in vivo, we first injected mice with GFP-, tdTomato-or Luciferase-expressing ecotropic lentiviruses, but we were unable to detect the expression of any of these markers by immunofluorescence staining. However, using R26-lox-STOP-lox-tdTomato mice [29] as a reporter system (Fig 3A) to test cellular infection, we discovered that the injection of ecotropic lentiviruses expressing Cre into the uterus of cycling (n = 3), pseudopregnant (n = 1) and ovarectomized (n = 1) mice induced tdTomato expression in cells in the uterine stroma but did not lead to infection of endometrial epithelial cells ( Fig  3B, first column). Since it has been shown that adenoviruses can infect stromal and epithelial cells in vivo [6,22,33], we injected adenovirus expressing Cre into the uterus of cycling (n = 2), pseudopregnant (n = 1) and ovarectomized (n = 1) mice as a positive control (Fig 3B, second  column). In ovarectomized mice we were indeed able to detect tdTomato-positive epithelial cells but again there were many more positive stromal cells. We conclude that lentiviral and adenoviral intra-uterine injections strongly favour infection of stromal cells rather than epithelial cells, even in mice in which endometrial cycling was hormonally blocked (ovarectomy, pseudopregnancy) to prevent the loss of potentially infected epithelial cells. Intrauterine injection of MuLE viruses expressing shCdkn2a plus Hras G12V causes sarcoma development Given the data obtained above, we reasoned that it might be possible to generate a mouse model of uterine sarcomas. Since MuLE lentiviruses can be used to model undifferentiated pleomorphic sarcomas in vivo [24], a tumour of mesenchymal origin [22], concentrated ecotropic MuLE lentiviruses which cause either Cdkn2a silencing together with activation of Hras G12V or Pten or Trp53 silencing alone, were injected into the uterus of 4-week-old SCID/ beige mice (Fig 4A). Additionally, the MuLE vectors carried a coding sequence for firefly luciferase in order to label infected cells and to trace tumour development over time. The injection of the shCdkn2a plus Hras G12V expressing ecotropic MuLE lentivirus into the uterus of five SCID/beige mice caused the development of high-grade endometrial stromal sarcomas with 80% (n = 4 tumours) penetrance after less than two months (median overall survival 49 days) (Fig 4B, 4D and 4E). This process was monitored by an increase in luciferase expression over time (Fig 4C). Tumours had maximum diameters between 3-6 mm at the time of sacrifice of the mice. Tumours expressed high levels of H-RAS (S1 Fig) compared to normal uterine tissue,   (Fig  4G), diffusely positively for CD10 ( Fig 4H) and VIMENTIN (Fig 4I) and negatively for CD31 (Fig 4J), DESMIN (Fig 4K), MYOGENIN (Fig 4L) and alpha-SMA (Fig 4M). These cellular and molecular features are consistent with a diagnosis of high grade endometrial stromal sarcoma. Lesions showed an absence of mucin inclusions (S2C Fig), which ruled out an adenocarcinoma phenotype. Tumours did not show staining for smooth muscle cells or collagen by van Gieson's staining (S2D Fig) further indicating that these tumours are not leimyosarcomas. None of the other injected vectors expressing knockdown constructs for two of the most commonly altered genes in endometrial carcinoma patients, namely Pten or Trp53, were sufficient to cause any large increases in luciferase signal over time and to induce tumour formation within 6 months after injection (Fig 4B). The development of endometrial stromal sarcoma by the injection of shCdkn2a plus Hras G12V expressing ecotropic lentiviruses serves as a technical proof-of-principle that this approach can be used to model uterine sarcomas.
To further investigate the potential utility of this approach for modelling other types of uterine malignancies we generated a series of five different MuLE lentiviruses designed to reproduce the genetics of uterine leiomyomas and uterine leiomyosarcomas, namely i) overexpression of HMGA2, ii) expression of MED12 G44S , iii) expression of MED12 G44S plus shRNA against Trp53, iv) expression of shRNA against Fh or v) expression of shRNA against Fh plus shRNA against Cdkn2a. Each vector also expressed luciferase. Each of these vectors were injected into uteri of ten or eleven mice for the HMGA2 and MED12 experiments and three mice for the Fh experiments and the mice were monitored by luciferase imaging for 9-11 months. None of these animals showed any increase in luciferase signal nor any histological abnormalities, arguing that this approach is not likely to be useful for modelling uterine leiomyomas.

Discussion
In this study we employed a mouse genetic approach using the MuLE lentiviral gene regulatory system [24] to functionally test the ability of ecotropic lentiviruses to model epithelial and mesenchymal uterine malignancies ex vivo and in vivo.
A major problem in ex vivo endometrial cell culture-based studies are contaminating stromal cells, which are a byproduct of the isolation. While epithelial cells lose their proliferative capacity during ongoing cultivation on plastic, stromal cells are more easily cultured long term [28]. A Rho kinase inhibitor (ROCK), in combination with fibroblast feeder cells, induces epithelial cells from many tissues to proliferate indefinitely ex vivo, without the need to immortalise them [34,35]. The culture of pEECs with conditional reprogramming by feeder cells and ROCK inhibitor did not improve the culture system in our hands. Despite some minor growth in culture, pEECs displayed strong degenerative changes, such as giant intra-nuclear vacuoles, and did not survive the first passage in culture. Since it was not possible to expand and passage pEECs which retain lineage commitment and normal growth, we tried to generate immortalised cell lines through Trp53 tumour suppressor inactivation or Hras oncogene activation with the use of ecotropic lentiviruses. This approach did not work, possibly due to the fact that ecotropic lentiviruses seem to efficiently infect contaminating stromal cells.
We also discovered that MuLE lentiviruses efficiently infect uterine stromal cells but not endometrial epithelial cells when injected into the uterus of cycling, pseudopregnant or ovarectomized mice. Consistent with this cellular infection spectrum, we show that intra-uterine injection of ecotropic MuLE viruses expressing oncogenic Hras G12V together with knockdown of Cdkn2a induce high-grade endometrial stromal sarcomas. This model is analagous to a similar aggressive uterine soft tissue sarcoma model that was previously generated through a Cre-loxP germline strategy. In this model, uterus-specific deletion of Trp53 was combined with expression of a constitutively active form of Kras [22]. The advantage of the MuLE system is that it is not dependent on the prior generation of germline-modified, Cre-regulatable mice and is therefore better suited to higher throughput testing of the pathological effects of different combinations of genetic alterations.
A theoretical drawback of the use of lentiviral vectors is the possibility of insertional mutagenesis. We refer to our previous detailed discussion of this issue [24] with respect to the fact that lentiviral vectors appear to be much less problematic than retroviral vectors. Data presented in this study also support the fact that insertional mutagenesis is unlikely to be a complicating factor. While MuLE viruses expressing Hras G12V plus shCdkn2a induced tumours at high frequency (4 of 5 mice), we injected a total of 53 mice with numerous vectors designed to manipulate a variety of known cancer-relevant genes and did not observe a single tumour in any mouse. This underlines the specificity of the Hras G12V plus shCdkn2a tumour phenotype and further demonstrates lack of non-specific tumour formation of the MuLE system in general.
Since we have previously shown that ecotropic MuLE vectors can infect a wide variety of epithelial and non-epithelial cell types [24], it remains unclear why endometrial epithelial cells are resistant to infection. It has been shown that mucin, a glycoconjugate, inhibits the entry of lenti-and adenoviruses into epithelial cells and that viral transduction could be improved by pretreating either the cells or the virus with the glycolyse hydrolase neuraminidase before infection [36,37]. It would be worth testing in the future if the pretreatment of either ecotropic lentiviruses or pEECs with neuraminidase increases epithelial cell transduction. In vivo studies have shown that adenoviral vector infection efficiency is influenced by the stage of the estrous cycle at the time of injection, with best results occuring during late metestrous and diestrous stages of the estrous cycle [33]. During the estrous cycle apoptosis is increased ten-fold in the luminar epithelium compared to other cell compartments [38]. Even if epithelial cells were initially infected, the high rates of apoptosis may lead to the rapid loss of these cells during the cycle. By the use of R26-lox-STOP-lox-tdTomato mice [29] as a reporter system to test cellular infection, we discovered that the injection of ecotropic lentiviruses expressing Cre into the uterus of cycling, pseudopregnant and ovarectomized mice induced tdTomato expression in stromal cells but not epithelial cells regardless of stage of the estrous cycle and the viral dose. In rodent cells a cationic amino acid transporter, termed CAT1, serves as the receptor for the envelope glycoprotein gp70 of ecotropic MMLV [39]. In the endometrium CAT1 is expressed on the basal surface of the epithelial cells [40]. Limited accessibility of CAT1 is another possible reason why we were not able to infect epithelial cells in vivo. Additionally, we injected Creliposome complexes into the uterine lumen of pseudopregnant and ovarectomized mice but also with this approach we detected reporter gene activity only in stromal cells.
Only when we injected Cre-expressing adenoviruses in ovarectomized mice, were we able to see not only tdTomato-positive stromal cells but also very few positive epithelial cells. This confirmed the earlier observations that adenoviruses infect only few epithelial cells and many more stromal cells when injected in the uterus of lacZ reporter mice [6]. The injection of an adenovirus expressing Cre into the uterus of Pten fl/fl mice caused Pten deletion in both stromal and epithelial cells, however this treatment caused endometrial carcinoma development, but none of the injected uteri exhibited stromal tumours. This indicates that the deletion of Pten in stromal cells does not offer an advantage to the cells and therefore is not selected for in terms of causing a tumour [6]. Our studies in which injected ecotropic MuLE lentiviruses expressing shPten failed to induce stromal tumours when injected into the uterine lumen of SCID/beige mice are consistent with these data.
In summary, we discovered that ecotropic MuLE lentiviruses are able to inefficiently transduce epithelial cells ex vivo and that infection strongly favors stromal cells. The injection of ecotropic lentiviruses into the uterus of mice infected stromal cells but not epithelial cells regardless of stage of the estrous cycle. The MuLE system appears to be suited for use in modelling endometrial sarcomas but not carcinomas and in general represents a rapid method for performing complex genetic manipulations in uterine stromal cells in vivo.