Carcinoma Initiation via Rb Tumor Suppressor Inactivation: A Versatile Approach to Epithelial Subtype-Dependent Cancer Initiation in Diverse Tissues

Carcinomas arise in a complex microenvironment consisting of multiple distinct epithelial lineages surrounded by a variety of stromal cell types. Understanding cancer etiologies requires evaluating the relationship among cell types during disease initiation and through progression. Genetically engineered mouse (GEM) models facilitate the prospective examination of early oncogenic events, which is not possible in humans. Since most solid tumors harbor aberrations in the RB network, we developed an inducible GEM approach for the establishment and assessment of carcinoma initiation in a diverse range of epithelial tissues and subtypes upon inactivation of RB-mediated tumor suppression (RB-TS). The system allows independent assessment of epithelial subtypes that express either cytokeratins (K) 18 or 19. By Cre-dependent expression of a protein that dominantly inactivates RB and functionally redundant proteins p107 and p130, neoplasia could be initiated in either K18 or K19 expressing cells of numerous tissues. By design, because only a single pathway aberration was engineered, carcinomas developed stochastically only after long latency. Hence, this system, which allows for directed cell type-specific carcinoma initiation, facilitates further definition of events that can progress neoplasms to aggressive cancers via engineered, carcinogen-induced and/or spontaneous evolution.


Introduction
Malignant cancers evolve through mechanisms of selective growth, survival, and invasive properties in a diverse microenvironment. Thus, cancers are heterogeneous with a complex constituency of multiple cell types. Genetically engineered mouse (GEM) models provide a powerful approach to assess both cause/ effect relationships and cell type susceptibility. Over the past few decades, progress has been made in identifying molecular aberrations that can contribute to tumorigenesis, thus offering insight into potential causal mechanisms and therapeutic targets. However, since many events have been engineered together and often don't initiate disease, relatively few studies have explored the etiology of tumor evolution from initiation through progression to advanced disease. Here, we sought to develop ''GEM cancerinitiator'' models that could be induced to initiate tumorigenesis in a wide variety of epithelial tissues and in distinct subtype compartments for further use in defining tissue and cell-specific mechanisms for carcinoma progression.
Cell subtypes of an epithelium can be characterized by specifically paired cytokeratin (K) expression profiles [1][2][3]. Cytokeratins are cytoskeleton protein intermediate filaments assembled from heterodimeric subunits of acidic type I (K9-K28) and basic type II (K1-K8, and K71-K80) proteins [4]. K18, usually paired with K8, is the first keratin expressed at the eightcell stage of mouse development [5,6], followed by K19 and K7 [7]. However, K19, the smallest known acidic keratin, has no known basic type II keratin partner. Although both K18 and K19 are expressed in simple epithelia, they are differentially expressed in subtypes of some epithelia (e.g. K18 in umbrella cells and K19 in basal and intermediate cells of urothelium [2,3,8]). Therefore, distinct keratin expression patterns can be utilized as the basis for targeting tumor initiation events to specific epithelial subtypes.
Retinoblastoma 1 (RB) is a negative regulator of proliferation in the eukaryotic cell cycle, particularly during cellular differentiation [9], and the RB pathway plays a critical role in tumorigenesis. Aberrant RB pathway activity, resulting from defects in RB itself, CDKN2A, CCND1, or CDK4, is observed in most solid human cancers [10]. In the few cancer types wherein early human tissue is routinely accessible, such aberrations are often present, suggesting a role in initiation [11][12][13][14][15]. However, in some cases, the first association of RB pathway aberration is apparent in more advanced cancer, arguing a role in progression [16]. Indeed, cause and effect relationships likely depend on multiple variables, including the cell and tissue of origin and the stochastic order of causal events, emphasizing the need for model systems to determine plausible roles in disease etiology. Because of functional redundancy among RB and its family members p107 and p130 in most murine cell types [17][18][19][20], we have used a dominant inactivating protein, T 121 , to inactivate RB tumor suppression (RB-TS) in the mouse [21][22][23][24][25]. T 121 is derived from the Nterminal 121 amino acids of Simian Virus 40 (SV40) large T antigen, which evolved to inactivate the RB-mediated cell cycle brake. When directed by tissue specific promoters in transgenic mice, T 121 is sufficient to initiate tumorigenesis in prostate [26], mammary [27], ovarian [28] and choroid plexus [29] epithelial cells, as well as in central nervous system astrocytes [30,31]. The initiation phenotype is dependent on the T 121 RB/p107/p130 binding site as demonstrated by point mutagenesis [32]. In each case examined thus far, initiation is associated with aberrant proliferation accompanied by apoptosis, and tumor progression has been driven by engineered and/or stochastic aberration of specific cancer-associated molecular networks [28,31,33].
Here, we describe transgenic mouse lines in which RB-TS inactivation can be induced in either K19 or K18-expressing epithelial subtypes in a Cre recombinase-dependent fashion. Transgenic lines were established in which Cre-conditional expression of T 121 was driven under cytokeratin18 or 19 transcriptional control using bacterial artificial chromosome (BAC) transgenes that retain endogenous expression patterns in diverse tissues. Thus, a single strain for each subtype can be used to drive cancer initiation tissue-specifically via organ-specific Cre expression after germline or somatic introduction of a Cre transgene. By selective inactivation of RB-TS in K18-or K19expressing cells, we demonstrate the broad utility of these ''cancerinitiator'' mice in mechanistic studies of cancer development and show roles for RB-TS inactivation and epithelial subtype specificity in neoplastic initiation within numerous epithelial organs.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by Animal Care and Use Committee (ACUC) of UNC-Chapel Hill (Permit Number: 04-232.0), and National Cancer Institute (NCI)-Frederick (Permit Number: 11-030). All animals in this study were euthanized per the ''Guidelines For The Euthanasia of Mouse and Rat Fetuses and Neonates'' as defined by the ACUC of UNC-Chapel Hill and NCI-Frederick to minimize pain and suffering.

Generation of Subtype-specific Keratin-directed BAC Transgenic Mice
A floxed eGFP stop T 121 cassette (transgene cassette, Figure 1A) was derived from the MFT 121 construct [34]. T 121 is comprised of the first 121 amino acids (aa) of the SV40 large T antigen followed by 11 missense amino acids resulting from a 31 bp deletion. The transgene cannot express small t antigen due to a second deletion that removes its splice acceptor site [35]. T 121 inactivates RB and family members p107 and p130 and does not contain the p53 and p300 inactivating domain [36]. The RPCI-22 mouse genomic DNA library was screened using specific probes designed for K18 (Krt1-18 on chromosome 15) and K19 (Krt1-19 on chromosome 11) via the RPCI BAC screening service (Roswell Park Cancer Institute, NY). Positive clones were used for recombineering- directed insertion [37,38] of floxed eGFP-stop-T 121 cassettes into exon 1 start codons of K18 and K19. BAC DNA was purified (MTR Scientific, Ijamsville, MD) and injected into fertilized eggs harvested from a B6D2F1 (JAX Laboratory, Bar Harbor, Maine) cross at a concentration of 4 ng/ul without linearization as described [39]. Resulting and subsequent generations of transgenic mice were identified by PCR amplification of a 215-bp fragment using primers 59 GAATCTTTGCAGCTAATGGACC 39 and 59 GCATCCCAGAAGCTCCAAAG 39 and digit-or ear-derived genomic DNA as template. The cycling profile was: 94uC, 2 minutes; followed by 35 cycles of 94uC, 20 seconds; 62uC, 45 seconds; 72uC, 45 seconds; and a final incubation of 72uC, 2 minutes. Mouse lines were maintained by mating with wild type B6D2F1 mice and therefore are designated as B6;D2-Tg(K18GT 121 )Tvd (TgK18GT 121 ) and B6;D2-Tg(K19GT 121 )Tvd (TgK19GT 121 ).

Transgene Copy Number Determination
Tail genomic DNA was extracted using Qiagen DNeasy Blood and Tissue kit, and quantitative real time PCR was used to detect transgene copy numbers. PCR amplifications were performed in 96-well plates in the ABI 7500 sequence detector. 30 ul samples included 10 ul (50 ng) genomic DNA sample and 20 ul PCR reaction mixture. Amplification was with 10 minutes at 94uC followed by 40 cycles of 15 seconds at 94uC and 1 minute at 60uC.    Table 1. Summary of histopathological findings in TgK19GT 121 ; b-actin Cre and TgK18GT 121 ; b-actin Cre mice.

Statistical Analyses
Student t or Mann-Whitney rank sum tests were performed to evaluate the statistical significance among genotypes and tissues. P,0.05 was considered statistically significant.

Transgene Expression Accurately Reflects Epithelial Subtype Specificity
To target RB-TS inactivation to epithelial subtypes, we employed keratin regulation within BACs to drive conditional transgene expression. A Cre-conditional loxP-eGFP-stop-loxP (LSL) T 121 cassette ( Figure 1A) was introduced into the first exon of either the K18 or the K19 gene within respective BACs to generate TgK18GT 121 and TgK19GT 121 transgenic mice ( Figure 1B). Thus, eGFP was expressed under K18 or K19 regulation until expression/activation of Cre, which mediated eGFP deletion and induced T 121 expression. Two transmitting founder lines for each construct were generated and characterized. As assessed by quantitative real time PCR, mice harbored one transgene copy in lines TgK19GT 121 -34 and -43, and TgK18GT 121 -36, and 3 transgene copies in line TgK18GT 121 -34 ( Figure 1C).
To detect endogenous K19 and K18 expression, and thus the patterns for accurate transgene regulation, wild type epithelial tissues were examined by IHC or IF. Consistent with reported results [2,3,8], both K19 and K18 were widely expressed in a variety of mouse epithelial tissues ( Figure S1 and Table S1).
TgK18GT 121 and TgK19GT 121 eGFP transgene expression was visualized in whole tissues using a stereo fluorescence microscope (SteREO Discovery.V20, Carl Zeiss Meditec, Inc). Strong green fluorescence was observed in some tissues of both lines of TgK19GT 121 mice (e.g. intestine, stomach, urinary bladder, and gallbladder), but weak or no green fluorescence in others (e.g. mammary gland, ovary, and pancreas, Figure S2). Many tissues showed strong green fluorescence in both lines of TgK18GT 121 mice (e.g. intestine, stomach, thymus, mammary gland, ovary, salivary gland, urinary bladder and vas deferens, epididymis, and gallbladder), but not in pancreas, lung, and liver ( Figure S3). In TgK19GT 121 mice, immunostaining of tissue sections consistently showed strong and widespread eGFP expression in intestine, stomach, urinary bladder, gallbladder, and renal pelvis epithelium. Readily detectable staining was limited to a smaller percentage of cells in lung, mammary gland, and prostate ( Figure S4 and Table  S1). Other tissues had very weak or no detectable eGFP expression (Table S1). In TgK18GT 121 mice, eGFP expression was widespread in most K18-positive epithelial tissues ( Figure S5 and Table S1). Importantly, based on double IF, eGFP was co-expressed with respective endogenous keratins in TgK19GT 121 ( Figure S4) and TgK18GT 121 mice ( Figure S5), but excluded from K14 cells (Figure Tissues from TgK18GT 121 -34 mice showed more intense eGFP staining compared to those from TgK18GT 121 -36 mice (data not shown), possibly due to the higher transgene copy number in line 34 ( Figure 1C).

RB-TS Inactivation is Induced in Specific Epithelial Subtypes
To determine whether T 121 expression could be induced in a broad range of epithelial tissues with subtype specificity and to evaluate the effect of RB-TS inactivation, we crossed hemizygous transgenic mice from all lines to transgenic b-actin Cre homozygous mice. Likely due to broad RB-TS inactivation in K19 expressing embryonic cells (see below), some embryonic lethality was observed in TgK19GT 121 ; b-actin Cre mice. Live-born bi-transgenic mice were under-represented relative to littermate controls (34% vs. 66% for TgK19GT 121 -34, 45% vs. 55% for TgK19GT 121 -43, respectively; Table S2), suggesting that RB function in K19 cells may be critical for development. In the surviving mice, consistent with the breadth of endogenous K19 expression ( Figure S1), T 121 expression was readily observed in a variety of tissues first assessed at 2 months of age (Figure 2), even though eGFP expression had been low or below the IHC detection level in some of these tissues (e.g. ovary, mammary gland and pancreas, Figures S2 and S4, and Table S1). Importantly, as eGFP was faithfully expressed in K19 cells, T 121 was also expressed in K19-positive cells (not shown), but excluded from K14 cells ( Figure S6). T 121 expression levels were comparable between TgK19GT 121 -34; b-actin Cre and TgK19GT 121 -43; b-actin Cre mice (not shown).
Contrary to TgK19GT 121 ; b-actin Cre mice, both lines of TgK18GT 121 ; b-actin Cre mice developed with normal Mendelian frequencies (not shown). T 121 was widely expressed in K18 expressing epithelium of some tissues of 2 month old TgK18GT 121 ; b-actin Cre mice (e.g. thymus, mammary gland, and ovary), but showed limited induction in others, albeit in the appropriate cells (e.g. intestine, prostate, and pancreas; Figure 3), even though eGFP expression had been more widespread in some cases ( Figures S3 and S5, and Table S1). Importantly, T 121 was expressed in K18 positive cells ( Figure 3B), but excluded from K14 cells ( Figure S6). The intensity of T 121 expression was comparable between TgK18GT 121 -34; b-actin Cre and TgK18GT 121 -36; b-actin Cre lines (not shown) although eGFP expression was higher in TgK18GT 121 -34 than TgK18GT 121 -36 mice (not shown). This observation may reflect somatic Cre-mediated copy number reduction from a tandem gene array.

RB-TS Inactivation Causes Neoplasia
Compared to wildtype mice ( Figure S7A), all TgK19GT 121 -43; b-actin Cre mice developed multifocal hyperplastic lesions in most K19 positive epithelial tissues [e.g. pancreas (pancreatic ductal epithelium), mammary gland, ovary (ovarian surface epithelial cells, OSECs), lung bronchiole, liver bile ductule, gallbladder, kidney (renal pelvis epithelial cells), urinary bladder, prostate, small intestine, colon, stomach glandular foveolar cells; Figure 4 and Table 1]. This phenotype was concurrent with high proliferation rates in the affected cells ( Figures 5 and 6) compared to wildtype tissues (Figures 6 and S8). As in previous reports of limited tissues [26,27,30], increased proliferation in tissues of TgK19GT 121 ; b-actin Cre mice was accompanied by a significant elevation of apoptosis ( Figure S9) compared to that of wildtype mice ( Figure S10). These results indicate that RB is important for suppressing hyperplasia of K19 positive cells. Some TgK19GT 121 ; b-actin Cre mice developed life-threatening bilateral or unilateral hydronephrosis starting as early as 2 months of age ( Figure S11A). The hydronephrosis was interpreted to be secondary to marked hyperplasia of transitional cell epithelium at the junction of the renal pelvis and ureter resulting in obstruction of urine outflow from the kidney ( Figure S11A). Other mice from line 34 had to be euthanized due to enlarged thymuses, which led to lifethreatening thoracic pressure, at 7 months of age ( Figure S11B). Similar to TgK19GT 121 ; b-actin Cre mice, inactivation of RB-TS in K18-positive cells predominantly led to hyperplasia (Figure 7 and Table 1), which was correlated with T 121 expression in these tissues (Figure 3). With broad induction by the b-actin Cre allele, both TgK18GT 121 -34 and TgK18GT 121 -36 mice developed severe thymic epithelial hyperplasia and subsequent lymphoid hyperplasia (Table 1 and Figure 7). Immunostaining indicated high T 121 expression in K18 cortical thymic epithelium ( Figure S12B) associated with high proliferation (based on Ki-67 IHC) and apoptosis (TUNEL) rates ( Figure S12C). Because these mice were compromised due to life-threatening thoracic pressure from enlarged thymuses ( Figure S12A) and required euthanasia by 6 months of age, TgK18GT 121 ; b-actin Cre phenotypes in other tissues could not be examined past this age. At 1-6 months of age, mild focal hyperplasia was observed in mammary gland, ovarian surface epithelium, intestine, and prostate ( Figure 7 and Table 1), which correlated with elevated proliferation rates and increased apoptosis (Figures 6, 8, and S13). For the most part, when compared to TgK19GT 121 ; b-actin Cre mice, the extent of hyperplasia was less widespread in TgK18GT 121 ; b-actin Cre mice, which correlated with the limited extent of T 121 expression in these tissues. High proliferation rates were observed in both K19 and K18 positive mammary epithelia and ovarian surface epithelia (38.6% vs. 49.2% in mammary, and 51.2% vs. 34.2% in ovary, respectively, Figure 6), which could be explained by co-expression of K19 and K18 in these tissues ( Figure S1). However, this is not the case for other sites of co-expression (e.g. prostate, Figure 6), suggesting that tissue-specificity may play a role in Cre recombination efficiencies.

Inactivation of RB-TS in Adult-restricted K19+ Cells is Sufficient for Tumor Initiation
To circumvent the developmental defects observed in TgK19GT 121 ; b-actin Cre mice and to demonstrate the utility of these models for studying cancer development in adult organs, we crossed TgK19GT 121 -34 to tamoxifen-inducible K19CreER mice [40]. In contrast to TgK19GT 121 ; b-actin Cre mice, TgK19GT 121 ;K19CreER mice were born with expected Mendelian frequencies (not shown), further supporting the critical role of RB-TS in K19 positive cells during mouse development. Mice were induced with tamoxifen at 6-8 weeks of age, and T 121 expression was assessed by IHC. T 121 was first readily detectable in gallbladder at 2 weeks post induction (not shown) and was Figure 7. Neoplastic lesions in TgK18GT 121 ; b-actin Cre mice. Hyperplasia was observed in some tissues (ovary and prostate: indicated by arrows). All mice were euthanized at 1-7 months of age due to life-threatening thoracic pressure caused by enlarged thymuses. All samples were stained with H&E. doi:10.1371/journal.pone.0080459.g007  widespread in various tissues by 4-6 weeks post induction ( Figure  S14A). Moreover, T 121 was co-expressed with K19 ( Figure S14B). Consistent with expression of eGFP prior to Cre expression, and, typical of transgene expression, T 121 was expressed in a subset of K19 positive cells within a given tissue.
To determine whether RB-TS inactivation predisposed adult K19-expressing epithelial cells to tumorigenesis, tamoxifen-induced TgK19GT 121 ;K19CreER mice were aged. Consistent with the general hyperplastic phenotype observed in TgK19GT 121 ; b-actin Cre mice, a high frequency of atypical hyperplasia was observed in most epithelial tissues (e.g. stomach glandular epithelium 85.7% in males and 90% in females, and lung bronchiole epithelium 92.9% in males and 90% in females; Table 2), consistent with increased proliferation rates (not shown). While a low frequency of malignant neoplasia did occur in some tissues (e.g. stomach adenocarcinoma 13.3% in males and 14.3% in females; Table 3 and Figure 9), low grade lesions present in most organs indicated that RB-TS inactivation was sufficient to initiate but not to progress disease. This is consistent with our previous results in mammary [27], prostate [26], choroid plexus [29] and ovarian epithelium [28], and brain astrocytes [30,31]. Similar to TgK19GT 121 ; b-actin Cre mice, renal hydronephrosis was also observed in 75% of males and 40% of females of induced TgK19GT 121 ;K19CreER mice at 9-16 months post induction (Table 2). However, unlike TgK19GT 121 ; b-actin Cre mice, the cause of death was bladder obstruction due to high expression of K19 in intermediate and basal cells of bladder epithelium ( Figure  S7B) and a high frequency of adenoma development in the bladder (Table 3 and Figure 9). Interestingly, we observed earlier onset of urinary bladder adenoma in males compared to females (9  months vs. 14 months post induction; Table 3). The underlying mechanism is unknown.

Discussion
Studies utilizing tissue-specific gene targeting in genetically engineered mouse models have contributed extensively to our understanding of cancer complexity and the molecular and cellular basis of tumorigenesis. Here we describe the generation of transgenic mouse models (TgK18GT 121 and TgK19GT 121 ) using keratin gene transcriptional regulators to drive conditional epithelial subtype-specific inactivation of RB-TS. High fidelity of respective keratin-specific expression patterns was achieved using BAC transgenes that were also engineered for assessing tissues of interest via reporter GFP expression. We demonstrate that these models can be used in combination with specific Cre drivers to initiate tumorigenesis in a wide range of epithelial tissues in either K19 or K18 expressing subtypes. We further demonstrate that RB-TS inactivation can indeed elicit pre-cancerous lesions cell autonomously in all tested epithelial tissues, including upon induction strictly in adult organs, and that early neoplasia is uniformly associated with aberrant proliferation and apoptosis.
Using subtype-specific targeting, a broad range of K18 or K19 positive subtypes within epithelial tissues were susceptible to neoplastic induction by RB-TS inactivation. Neither TgK18GT 121 ; b-actin Cre mice (up to 6 months) nor TgK19GT 121 ; b-actin Cre mice (up to 8 months) developed adenomas or carcinomas. These results confirm that RB-TS inactivation in affected epithelia only initiates a neoplastic phenotype. Based on previous studies of tumor progression in T 121 -initiated cells, we expect that additional lesions in specific cancer genes will be required to drive this early neoplastic phenotype into respective cancers. Indeed, we have demonstrated this case using the K18-driven system in ovarian surface epithelium, wherein lesions disrupting p53 function facilitate progression to metastatic stage IV serous ovarian cancer [28], and in mammary epithelium, where Trp53 and BRCA1 deletion progresses hyperplasia to adenocarcinoma (Van Dyke et al., unpublished). Using the K19-driven system, we have also observed that mammary hyperplasia can progress with additional PTEN deletion (Song and Van Dyke et al., unpublished).
In comparing results from K18 and K19-driven models, most tissues showed a potential subtype-dependency with respect to the extent of hyperplasia/dysplasia. This observation was independent of transgene insertion site, since the same results were obtained regardless of the transgenic line within each genotype. However, the interpretation that K18 and K19 expressing epithelial subtypes are differentially responsive to RB-TS inactivation is confounded by the differential transgene induction rate by b-actin Cre in K18 and K19-driven models. Indeed, the breadth of oncogene expression, and possibly cellular levels, was directly correlated with the extent of neoplasia induced. TgK19GT 121 ; b-actin Cre epithelial tissues showed wide spread T 121 induction and neoplasia, while T 121 was induced only in a subset of cells in most TgK18GT 121 ; b-actin Cre tissues and resulted in minimal neoplasia. This outcome may be due to tissue-dependent Cre recombination efficiencies and/or to temporal differences in K18 and K19 expression onset, since not every TgK18GT 121 ; b-actin Cre tissue showed low T 121 induction. Whatever the explanation, this study could not distinguish susceptibility between K19 and K18 subtypes in most tissues. Mammary epithelium and ovarian surface epithelium are exceptions to this outcome in that RB-TS inactivation under both K18 and K19 regulation led to overt hyperplasia due to co-expression of K18 and K19 in these tissues, an outcome reflecting the expected endogenous keratin expression patterns. Of note, the K18-driven model, in which limited expression within specific cell populations can be induced, will be valuable for cancer initiation studies, since this situation mimics the limited stochastic initiation that is thought to occur in human cancers. Furthermore, focal induction via viral Cre injection in specific tissue might be the best approach to assess the adult induction of RB-TS inactivation in K18 cells in tissues other than thymic epithelium. Overall, these ''GEM cancer-initiator'' models provide useful tools with which to further define tissue and cellspecific mechanisms for carcinoma progression via engineered, carcinogen-induced and/or spontaneous evolution.