Dunning rat prostate adenocarcinomas and alternative splicing reporters: powerful tools to study epithelial plasticity in prostate tumors in vivo
Using alternative splicing reporters we have previously observed mesenchymal epithelial transitions in Dunning AT3 rat prostate tumors. We demonstrate here that the Dunning DT and AT3 cells, which express epithelial and mesenchymal markers, respectively, represent an excellent model to study epithelial transitions since these cells recapitulate gene expression profiles observed during human prostate cancer progression. In this manuscript we also present the development of two new tools to study the epithelial transitions by imaging alternative splicing decisions: a bichromatic fluorescence reporter to evaluate epithelial transitions in culture and in vivo, and a luciferase reporter to visualize the distribution of mesenchymal epithelial transitions in vivo.
KeywordsDT AT3 Dunning FGFR2 Alternative splicing EMT MET Imaging
Examination of the human genome reveals a preponderance of complex genes that produce equally intricate primary transcripts [1, 2, 3]. In humans, a typical pre-messenger RNA (pre-mRNA) contains seven or eight introns, each averaging more than 3,000 nucleotides in length, and eight or nine exons, which average only 300 nucleotides, with the average internal exon measuring only 145 nucleotides [1, 2]. Components of the spliceosome (U snRNPs and general splicing factors) recognize intronic and exonic sequence elements that critically define exon–intron boundaries and mediate the removal of introns and ligation of exons in the process known as pre-mRNA splicing. Alternative RNA splicing is the process by which a primary transcript yields different mature RNAs leading to the production of protein isoforms with diverse and even antagonistic functions [4, 5, 6, 7, 8, 9]. The annotation of mammalian genomes reveals that the bulk of intron-containing transcripts are alternatively spliced [10, 11, 12, 13]. Many different forms of alternative splicing together with differential use of promoters and 3′ cleavage/polyA sites leads to staggering versatility among transcripts driven by RNAP II. Most importantly, alternative splicing is one of the several mechanisms by which broad programs of gene expression are established in different cell types.
An elegant example of regulated alternative splicing is the choice between IIIb and IIIc isoforms of FGFR2. Alternative inclusion of exon 8 (IIIb) or exon 9 (IIIc) leads to the expression of FGFR2(IIIb) and FGFR2(IIIc), respectively. This decision is cell-type specific, most epithelial cells express the FGFR2(IIIb) isoform exclusively, and most mesenchymal cells express solely FGFR2(IIIc). The alternative use of exons IIIb or IIIc alters ligand binding [14, 15] and its regulation is required for normal development in mice and humans [16, 17, 18]. The status of FGFR2 alternative splicing depends on the interplay between several cis-acting elements in the FGFR2 pre-mRNA and trans-acting factors, some of which are cell-type specific. Two cell-type specific cis-acting elements, intronic activating sequence 2 (IAS2) and intronic splicing activator and repressor (ISAR, also known as IAS3), which form a stem, collaborate to activate exon IIIb and repress exon IIIc in an epithelial-specific manner [19, 20, 21, 22, 23]. Additionally, we identified two gcaug sequence elements, located downstream of ISAR and within exon IIIc, and showed that these play a role in epithelial exon IIIc repression . Moreover, we showed that splicing factors of the Fox family of proteins are responsible for this repression of exon IIIc in epithelial cells . The Fox family of proteins, and in particular Fox-2, is a likely candidate for a master regulator of FGFR2 splicing during epithelial plasticity.
The differential splicing of FGFR2 transcripts underscores large differences in gene expression programs in epithelial and mesenchymal cells. Epithelial cells are tightly connected to each other by specialized structures on their surface and form well organized apical-basal polarized layers . Mesenchymal cells, on the other hand, do not form well-organized layers, are motile, and have a front-back polarity. Epithelial cells can become mesenchymal cells and vice versa via phenotypic transitions, known as epithelial–mesenchymal (EMT) and mesenchymal–epithelial transitions (MET). These are critical for vertebrate ontogeny and have been proposed to play important roles in tumorigenesis [25, 26, 27, 28]. Indeed, the evidence that these epithelial transitions are seen in prostate cancer is now overwhelming, and the data that support their importance are compelling [29, 30, 31, 32].
In this manuscript we build upon our previous observation that cells in Dunning AT3 rat prostate tumors show heterogeneity in the alternative splicing of FGFR2 transcripts and this is due to MET . We demonstrate here that the Dunning DT and AT3 cells, which express epithelial and mesenchymal markers, respectively, represent an excellent model to study epithelial transitions in prostate cancer. Moreover, we introduce two new tools to study the epithelial transitions by imaging alternative splicing decisions: a bichromatic fluorescence reporter to evaluate epithelial transitions in culture and in vivo, and a luciferase reporter to visualize the distribution of MET in vivo.
Materials and methods
Gene expression analysis
Triplicate cultures of AT3 and DT cells were grown to ~ 60% confluency. Total RNA was isolated using the RNeasy kit (Qiagen, Inc.) and triplicate samples were submitted to the Duke Microarray Facility. Gene expression analysis was performed using the RO27 K rat spotted arrays 3.0 (Operon, Inc.). Bioinformatical analysis of expression differences between AT3 and DT cells was done using the GeneSpring GX software version 7.3.1 (Agilent Technologies, Inc.). The data files (representing signals for 26,986 gene probes in all six data points, three for AT3 and three for DT) were normalized using the feature: per Spot and per Chip—intensity dependent (lowess) normalization. The resulting gene list was used to determine the significantly differentially expressed genes between AT3 and DT using the “Filtering on Volcano plot” feature with the following characteristics: (1) Test type: parametric test, don’t assume variances equal; (2) Multiple testing correction: none; (3) Fold Difference: 2-fold or greater and a P-value cutoff of 0.01.
Gene set enrichment analysis
To infer differential pathway activity from gene expression differences between the AT3 and DT cells, gene set analysis was performed (GSEA) . Briefly, we first generated a gene set from microarray features from the R027 K rat spotted array (v 3.0) with increased expression in the AT-3 cells compared to the DT cells using an a priori established cut off of ≥2-fold increase in average expression. The features identified using the rat array were mapped to Affymetrix U133A accession numbers using annotation files provided by Operon (See Supplemental Table 1). Differential enrichment of this gene set was then tested across a cohort of local (n = 16) and metastatic (n = 32) prostate cancer tumors using previously published data . The statistical significance of differential enrichment in one class (metastatic) vs. the other (local) was determined using a weighted Kolmogorov–Smirnoff metric to determine an enrichment score (ES), and random permutation of phenotypic labels was performed to compare the observed results with reiterative random permutations of class labels to establish a false discovery rate (FDR). Parameters for GSEA included: (1) genes were ranked according differential expression between metastatic (up) and local (down) prostate samples using a Student’s t-test, (2) weighted analysis was performed with the weight assigned a value of 1.0, and (3) 1,000 random permutations of randomized class labels were used to establish the FDR.
Bichromatic fluorescent reporter pRGIIIc
The minigene pRGIIIc was constructed on the backbone of minigene RG6  kindly provided by Dr Thomas Cooper (Baylor College of Medicine). The exon IIIc and flanking intronic regions of the FGFR2 gene were amplified from the pRIIIcI2 minigene  and inserted between the XbaI and AgeI sites of the RG6 minigene. A 21 nucleotide region encoding a nuclear localization signal in the first exon of the RG6 minigene was deleted. Complete sequence of the plasmid will be provided upon request. AT3 and DT rat prostate tumor cells were kindly provided to our laboratory by Dr. Wallace McKeehan (Texas A&M University Health Sciences Center, Houston, TX). Cells were transfected using lipofectamine (Invitrogen Corporation, Inc.) and stable transfectants were selected using 500 μg/ml G418.
The minigenes pFFint and pFFIIIcI2 are identical to pRint and pRIIIcI2 (described elsewhere ) except for having the firefly luciferase ORF instead of the RFP ORF. Complete sequence of the plasmids will be provided upon request. AT3 cells were transfected using lipofectamine (Invitrogen Corporation) and stable transfectants were selected using 15 μg/ml blasticidin.
Animals and tumor cells implantation
AT3 cells were trypsinized, resuspended in PBS to a final concentration of 106/ml and kept on ice. 3 × 105 cells were injected subcutis (sc) in the flanks of Copenhagen 2331 rats (Harlan, Inc.) (75-90 g and ~2 months in age). Animals were continuously monitored for tumor growth. All animal procedures were performed according to the Duke University Institutional and Animal Care and Use guidelines.
For cell culture studies d-luciferin was added to the culture medium (final concentration 50 μg/ml) and 10 min later imaging was performed using the IVIS imaging system (Xenogen). For animal in vivo imaging, following anaesthesia, rats were injected intraperitonealy with d-luciferin (50 mg/kg) and 10 min later sequential imaging was performed as described.
Results and discussion
The Dunning DT and AT3 cell lines as a model to study epithelial transitions in prostate cancer
Epithelial plasticity related transcripts overexpressed in DT relative to AT3 cells
Common gene name
CLI; APOJ; SGP2; SGP-2; SP-40; Trpm2; Trpmb; TRPM-2; TRPM2B; RATTRPM2B
Insulin-like growth factor binding protein 7
Notch gene homolog 3
Cx26; CXN-26; MGC93804
Gap junction membrane channel protein beta 2
Lipolysis stimulated lipoprotein receptor
Ly6/Plaur domain containing 3
Cell adhesion molecule JCAM
Cadherin 1 (E-cadherin)
Growth factor receptor bound protein 7
Matrix metallopeptidase 9
Integrin, beta 6
Epithelial plasticity related transcripts overexpressed in AT3 relative to DT cells
Common gene name
Bone morphogenetic protein 2
Insulin-like growth factor binding protein 4
Twist gene homolog 1
Sema domain, immunoglobulin domain (Ig), short basic domain, secreted, (semaphorin) 3A
Snail homolog 1
In order to evaluate the relevance of Dunning model to human prostate cancer, we compared the gene expression profiles in these cells with that in human prostate tumors and metastases. In order to do this, we used Gene Set Enrichment Analysis (GSEA)  and compared our data to that from patient specimens . GSEA is a powerful way of matching a defined gene set to the rank order of another en masse gene expression experiment. The effectiveness of GSEA hinges on the ability to quantify and visualize the distribution of the defined gene set within the data of another microarray evaluation. By relying on the distribution, GSEA is not affected by varying fold change between cell types. The software determines if genes in a set (in this case the genes with >2-fold average expression in AT3 when compared to DT) occur more frequently at the top or bottom of a list of genes ranked according to their differential expression across the phenotype of interest (metastatic vs. local prostate cancer). The program provides an enrichment score (ES) based on a weighted Kolmogorov–Smirnov statistic and provides a false discovery rate (FDR) based upon the value of the observed ES compared to the distribution of ES across 1,000 randomly permutations of phenotypic labels .
A bichromatic reporter can image splicing differences in epithelial and mesenchymal cells
Alternatively spliced bioluminescence reporters can image mesenchymal–epithelial transitions in lung metastases
To test this potential, we decided to examine the behavior of AT3 tumors bearing these bioluminescence reporters. Whereas AT3 cells in culture and the great majority of these cells in tumors display a mesenchymal phenotype, we previously noted that clusters of cells in tumors and lung metastasis undergo mesenchymal–epithelial transitions (MET) . We injected one Copenhagen 2331 rat sc in both flanks with AT3 cells stably transfected with pFFint and three animals were injected sc in both flanks with AT3 cells stably transfected with pFFIIIcI2. Tumors were left to grow and appearance of metastasis in lungs was monitored by in vivo luminescence imaging every other day. In all four animals, luminescence-reporters AT3 cells were mixed with AT3 cells stably expressing EGFP, so metastasis observed with luminescence could be confirmed at microscopic level.
Thirteen days following implantation, strong luminescence signal was seen in the two primary tumors of the FFint animal and much weaker luminescence was detected in the lungs (Fig. 4c; two images of the same animal (FFint) are shown in the two panels). Only three of the six FFIIIcI2 tumors had detectable luminescence (on both flanks of animal 02-FFIIIcI2 and on the left flank of animal 03-FFIIIcI2). This is consistent with our previous ex vivo observations using fluorescence reporters  and we interpret the (low level) luminescence in the FFIIIcI2 primary tumors as originating in MET foci (Fig. 4c). Interestingly, the intensity of the signal from the reporters observed in the lung metastases relative to the signal from the primary tumors was higher for FFIIIcI2 tumors (Fig. 4c). Indeed in the case of the one animal (01-FFIIIcI2—Fig. 4c), only cells metastatic to the lung expressed luciferase, and thus had undergone MET. These observations suggested that MET was more frequent in metastases as we had predicted based on our ex vivo analysis . Moreover, these data show that the luminescence reporters can be used to observe alternative splicing changes in vivo during tumor progression and metastasis.
EMT and MET transitions have been described before and are thought to play an important role during development, when cells activate or repress different gene expression programs in order to migrate at a certain point in time and space and later become stationary [25, 49, 50]. There is an increasing number of reports showing that cells at the edge of carcinomas or metastatic cells are able to reactivate EMT and MET programs in order to successfully complete the metastasis program [28, 49]. Nonetheless, skepticism still exists in the field about the existence of these transitions during tumor invasion and metastasis, especially because of the lack of evidence in vivo . In previous work, we have shown that the AT3 cells, which have been previously described as mesenchymal-like [38, 52], display surprising plasticity during tumor growth and metastasis . Using alternative splicing reporters we observed MET occurring both in tumors (especially near stroma) and in lung metastasis around blood vessels. The previous observations were done ex vivo, by excising the tumors and metastatic lungs and analyzing sections under epifluorescence microscopy. In this work, to our knowledge, we are the first to show that MET associated with tumor growth and metastasis can be visualized in a living animal. Using alternatively spliced luciferase reporters that have a very low signal when cells display a mesenchymal-like phenotype and high signal when cells turn to a more epithelial-phenotype we show that it is possible to follow MET in vivo in living animals. Although we have not ruled out that a very small population of cells could have transitioned in the culture dish, and expanded only after implantation, we believe that the data shown here and our previously published data are much more likely due to epithelial transitions occurring post-implantation. Regardless, we have developed tools that can be used to study the contribution of MET to metastasis in vivo in this particular system. Additionally, here we show the first generation of bichromatic reporters that can sense both EMT and MET.
We propose that Dunning tumor cells (both DT and AT3) might be in metastable phenotypic states. Indeed the expression of Mmp9 and other mesenchymal gene products at higher levels in DT cells than AT3 cells and vice versa for Syndecan 2 in AT3 cells may reflect this metastability. Indeed the expression of EGFP by a small number of DT cells transfected with the pRGIIIc bichromatic reporter is consistent with this idea. We posit that the ability of these cells to toggle between phenotypic states contributes to malignancy. Whereas we make our proposal based on a rat model system for prostate cancer, our comparison of the Dunning cells with human specimens strongly suggests that our hypothesis has relevance to human cancers.
M. A. G-B acknowledges support from NIH (NCI and NIGMS) grants R33 CA97502 and RO1 GM063090. P.G.F. is a Damon Runyon-Lily Clinical Investigator (CI-29-05) and acknowledges support from the DOD (PC060879). We thank Dr. Robert Brazas and Mr. Todd Albrecht for construction of FFint and FFIIIcI2.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- 31.Gravdal K, Halvorsen OJ, Haukaas SA et al (2005) Expression of bFGF/FGFR-1 and vascular proliferation related to clinicopathologic features and tumor progress in localized prostate cancer. Virchows Archiv 448:1–7Google Scholar
- 48.Seth P, Miller HB, Lasda EL et al (2008) Identification of an intronic splicing enhancer essential for the inclusion of FGFR2 exon IIIc. J Biol Chem 283:10058–10067Google Scholar
- 52.Isaacs JT (1987) Development and characteristics of the available animal model systems for the study of prostatic cancer. In: Coffey DS, Bruchovsky N, Gardner WA, Resnick MI, Karr JP (eds) Current concepts and approaches to the study of the prostate. Alan R Liss, Inc., New York, pp 513–576Google Scholar