Immunobiology of human mucin 1 in a preclinical ovarian tumor model.

Continued improvement on the development of more effective therapies for ovarian cancer, including immune-based therapies with biologics and cancer vaccines, depends in part on the availability of adequate preclinical models for in vivo testing of treatment efficacy. Historically, most animal models of ovarian cancer were based on xenografting human ovarian cancer cells into immune-deficient mice, an approach largely ineffective on replicating immune regulation during tumor initiation and progression. 1 Over the past decade, several transgenic (Tg) mouse models tailored to immune-competent hosts have emerged, advancing our understanding of oncogenic pathways during ovarian tumorigenesis. 2 – 8 In 2005, seminal work from Dinulescu et al . 9 reported a genetically engineered, conditional (Cre-loxP) mouse in which Cre-mediated activation of the oncogenic K-rasG12D mutation and homozygous inactivation of Pten in the ovarian surface epithelium (OSE) triggers the formation of ovarian tumors with endometrioid histology. The lox P-Stop- lox P- K-ras G12D /+ Pten loxP/loxP mice (called herein KrasPten, for brevity) represent the first immune-competent mice with histologically defined, orthotopic ovarian tumors, closely resembling the human disease. Although the KrasPten mice provide many opportunities for investigations on disease pathogenesis, 2 , 9 , 10 they are limiting for antigen-based vaccine research due to the paucity of well-defined ovarian cancer-associated murine antigens and overall reduced translational potential of mouse vaccines to human studies. To address this, we generated triple Tg MUC1 +/− Kras G12D/+ Pten loxP/loxP (or briefly MUC1KrasPten) mice that express physiological levels of human mucin 1 (MUC1) as self-antigen and progress to MUC1-overexpressing ovarian epithelial tumors upon intrabursal administration of Cre recombinase-encoding adenovirus (AdCre). MUC1 is a highly glycosylated transmembrane glycoprotein, normally present on the apical surface of polarized epithelial cells in tissues of the human genitourinary, respiratory and digestive tracts. 11 During tumor progression, MUC1 becomes overexpressed and differentially glycosylated. These changes in MUC1 biochemistry and expression pattern on ovarian tumor cells are recognized by the immune system of cancer patients who can spontaneously develop low, yet detectable levels of both antibody and T-cell responses to MUC1 epitopes. 12 – 15 In addition to being a tumor-associated antigen, MUC1 is also an oncoprotein, playing an active role in various cellular pathways, which impact cell growth, proliferation and migration. Most of the evidence supporting the MUC1 roles in transformation and metastasis comes primarily from breast, pancreatic and colon cancers, 16 – 22 with only a few studies to date focused on MUC1 in ovarian cancer where another mucin, CA125 (MUC16) is most frequently studied, especially as a marker for response to therapy. 23 – 26 The advent of our MUC1KrasPten mouse model using conditional mice with MUC1-positive, orthotopic ovarian tumors allowed us to study MUC1 function in ovarian cancer pathogenesis, explore its roles in disease immunobiology and identify its value as a therapeutic cancer vaccine candidate.

On the human female genital tract, MUC1 is expressed by the epithelial cells of the uterine lining, endometrial glands, fallopian tubes and cervical epithelium. 27 The OSE cells from healthy human ovaries express low (or no) MUC1. In contrast, more than 90% of all ovarian epithelial tumors overexpress MUC1, regardless of histology 28 – 32 ( Figure 1a ). To generate a preclinical mouse model for MUC1-expressing ovarian tumors, we performed a series of cross-breedings between MUC1 Tg mice that heterozygously express human MUC1 +/− as a transgene driven by the endogenous promoter 33 and mice carrying the conditional K-ras G12D oncoallele (loxP- Stop -loxP- K-ras G12D /+ ) and the floxed Pten gene Pten / loxP/loxP (Dinulescu et al . 9 ) ( Figure 1b ). For selection of triple Tg mice, the genotype of breeders and all littermates was confirmed by PCR analysis of tail DNA ( Figure 1c ). MUC1 presence in tissues throughout the genital tract of healthy, triple Tg female mice resembles the profile normally seen in women, with higher expression in endometrial glands and oviducts than on OSE ( Figures 1d and e ). To induce ovarian tumors, we injected 2.5 × 10 7 p.f.u. (plaque-forming units) per ovary of AdCre under the ovarian bursa, during survival surgery, as previously described. 9 The ovarian tumors were often accompanied by hemorrhagic ascites ( Figure 2a ), occurred in the injected ovary only ( Figures 2b and c ) and often disseminated throughout the abdominal cavity and on the diaphragm ( Figure 2d ), mirroring late-stage human disease. No histological modifications of the ipsilateral oviduct and uterine horn were observed, confirming that intrabursal AdCre injections remain anatomically confined, as previously demonstrated by us and others using reporter gene experiments. 9 , 34 The ovarian MUC1KrasPten tumor cells are positive for estrogen receptor, epithelial cell marker cytokeratin 7 and Ki67 nuclear proliferation marker and, as in KrasPten tumors, 9 overexpress pMEK and pAkt, confirming simultaneous activation of the two oncogenic pathways, downstream of Kras and PI3K, respectively ( Supplementary Figure S1A ). PCR analysis of tumor-extracted DNA further corroborated the Cre-mediated recombination of loxP sites at both Kras and Pten loci ( Supplementary Figure S1B ). The tumors in triple Tg mice appear histomorphologically identical with those arising in double Tg mice ( Figures 2e and f ) and display all the characteristics previously described ( Supplementary Figure S2A ). 9 The contrasting feature is the presence of human MUC1 protein, expressed as expected, in tumors from triple Tg mice only ( Figure 2f ). The staining pattern identifies MUC1 as being abundantly present on the membrane and cytosol of ovarian tumor cells from the primary ( Figure 2f ) as well as metastatic sites ( Figure 2g ). Compared with normal OSE cells, the OSE-derived tumor cells display drastically increased MUC1 both at the protein ( Figure 2h ) and RNA levels ( Figure 2i ) mirroring MUC1 overexpression seen in human ovarian cancer ( Figure 1a ). Profuse MUC1 expression could be detected in earlier, smaller tumors as well as in more advanced, larger tumors ( Figure 2h ). The tumor-bearing MUC1KrasPten mice also have increased accumulation of soluble human MUC1 (CA15-3) in serum and ascites (not shown), another common occurrence in advanced stage ovarian cancer patients. 35 These findings demonstrate the close parallelism between the tumors in mice and patients, with MUC1 presence bringing this preclinical model even closer to the human disease. The triple Tg MUC1KrasPten mice express human MUC1 in addition to mouse Muc1 protein, 33 raising the question whether the co-expression of these two molecules may contribute to the phenotypes observed. To address this, we explored the tissue distribution of both human and mouse mucin 1 proteins throughout the female mouse genital tract of MUC1KrasPten and KrasPten mice, using an antibody (clone MH1, Supplementary Figure S2A ) that recognizes a conserved epitope within amino-acids 239–255 from the cytoplasmic tail of mucin 1 and is cross-reactive to both mouse (Muc1) and human (MUC1) molecule. We detected positive staining in the endometrial glands and oviducts of both KrasPten and MUC1KrasPten mice, suggesting that mouse Muc1 is present in the uterus and oviducts of both strains. In contrast, the OSE cells and the OSE-derived ovarian tumors stained positive in the MUC1KrasPten mice only, indicating that the human MUC1 and murine Muc1 are differentially expressed in healthy ovaries and ovarian tumors. In line with these findings, gene expression analysis by qRT–PCR for both Muc1 and MUC1 genes further demonstrates that although in healthy mice the ovaries express lower mucin 1 levels (as compared with oviducts and endometrium, Supplementary Figure S2B ), there is marked overexpression of MUC1 (but not Muc1) in mouse ovarian tumors ( Supplementary Figure S2C ), similarly to the human disease ( Supplementary Figure S2D ). These results indicate that MUC1 and Muc1 (which share little homology in the extracellular domain) are differentially expressed in ovarian tumors and that the MUC1KrasPten model is ideally suited for human MUC1 preclinical studies in ovarian cancer. To evaluate the potential role of MUC1 in ovarian tumorigenesis, we compared the survival pattern and tumor biology characteristics of double Tg mice ( n = 13) versus triple Tg mice ( n = 13) with tumors. All mice were killed when moribund, in accordance to institutional guidance. The overall survival of triple Tg MUC1KrasPten mice (median 68.62 days) was not significantly different than the one observed in double Tg KrasPten mice (median 94.62 days) ( P = 0.128) ( Figure 3a ). In order to assess the disease extent throughout the genital tract and peritoneal cavity of tumor-bearing mice, we measured at necropsy the following variables: ascites volume, primary tumor score and number of metastatic sites (diaphragm, peritoneal wall, colon, liver capsule, splenic surface and so on). Our analyses revealed that the primary ovarian tumors had the tendency to be smaller in MUC1KrasPten mice than KrasPten mice (tumor score of 2.1 and 2.7, respectively, P = 0.08) and to be accompanied more often by ascites (90 and 54%, respectively, P = 0.09) ( Table 1 ). Furthermore, the triple Tg mice had an average of 2.6 metastatic sites versus 1.4 in double Tg mice ( P = 0.03), suggesting that the MUC1-expressing tumors migrate and metastasize more widely in the peritoneal cavity. To explore further the MUC1 roles in cell invasion, we transfected the murine IG10 ovarian cancer cell line 36 with a MUC1-encoding plasmid ( Figure 3b ). Our results show that the IG10-MUC1 cells exhibit increased invasive abilities over the parental (wild type, MUC1-negative) controls ( Figure 3c ). Furthermore, the effect was significantly decreased ( P < 0.005) in cells treated with a pool of four different small interfering RNAs (siRNAs) targeting 3’-UTR/ORF of human MUC1 gene ( Figure 3c ). To assess the efficacy of siRNA treatment, MUC1 expression was measured at mRNA and protein level by RT–PCR and flow cytometry, respectively ( Supplementary Figures S3A and B ). Altogether, these results suggest that MUC1 influences the ovarian tumor cells’ invasive behavior both in vitro ( Figure 3c ) and in vivo ( Table 1 ). To identify MUC1-induced molecular changes in loco-regional tumor metastasis, we used triple and double Tg mice with comparable tumor scores (tumor score of 3, n = 4, each) and studied the gene expression profiles in their ovarian tumors using a targeted, preselected array of 84 genes with well-defined roles in cell adhesion and metastasis. Based on a threshold of twofold change and P value ≤ 0.05, we identified significant downregulation for cadherin 11 (Cdh11; fold regulation = 2.57; P = 0.032) in MUC1KrasPten versus KrasPten tumors ( Figure 3d and Supplementary Table 1 ). The Cdh-11 protein expression in KrasPten and MUC1KrasPten primary ovarian tumors was next confirmed by immunohistochemistry. In line with our qRT–PCR results, the MUC1-expressing tumors from triple Tg mice exhibit diminished Cdh11 expression compared with MUC1-negative tumors from double Tg mice ( Figure 3e ). We also confirmed this inverse correlation in vitro , using the mouse IG10 ovarian cancer cell line and MUC1-tranfectants (IG10-MUC1). The IG10-MUC1 cells show significantly less Cdh11 when compared with IG10 cells ( Figure 3f ). Treatment of IG10 cells with siGENOME SMARTpool containing four specific siRNAs targeting the ORF sequence of mouse Cdh11 gene ( Supplementary Figures S3C and E ) significantly increases invasion ( Figure 3c ). However, downregulation of Cdh11 ( Supplementary Figure S3D ) inhibits invasion of IG10-MUC1 cells (which express low but detectable Cdh11 protein levels, Figure 3c ), suggesting that the low levels of Cdh11 present at baseline may contribute to the cells’ invasive potential. The inverse correlation between MUC1 and cadherin 11 was mirrored in three human ovarian cancer cell lines (OvCa432, OvCa420 and TOV112D) with high, intermediate and low MUC1 levels, respectively ( Figures 3g–i ), further confirming our findings in mice. Having identified MUC1-induced epithelial cell changes, we profiled next the immune status of triple versus double Tg mice with ovarian tumors. Our results from multicolor flow cytometry of splenic T cells revealed that although in healthy KrasPten and MUC1KrasPten mice there are no differences in splenic CD4+FOXP3+ regulatory T cells (Treg) before tumor induction ( Figure 4a ), there is a significant increase in the Treg percentages in mice with advanced tumors ( Figure 4a , upper panel), similar to findings in ovarian cancer patients. 37 – 39 However, we observed a more robust increase in the splenic ( P = 0.0068) ( Figures 4b and c ) Tregs of triple Tg MUC1KrasPten mice, pointing out to MUC1 as modulator of loco-regional immunity, potentially via increased Treg accumulation. Overall, these findings demonstrate that despite having a largely similar histology, the MUC1-expressing tumors are more metastatic and develop in a more suppressive environment than their MUC1-negative counterparts. Based on our findings that MUC1 expression is elevated in tumors ( Figure 1a and Figures 2g and h ) where it has an active role in loco-regional spread ( Figure 3c and Table 1 ) and immune suppression ( Figure 4 ), we asked next whether MUC1-targeted therapy can prolong survival and reverse immune dysregulation in MUC1KrasPten mice. To address this, we used a vaccine based on a synthetic 100mer MUC1 peptide (comprising five tandem repeats from the MUC1 extracellular domain), loaded onto dendritic cells (DC), 40 matured under type 1 polarizing conditions (DC1), using a slight modification of our previously described protocol 41 The DC1-inducing maturation cocktail (combining granulocyte-macrophage colony-stimulating factor (GM-CSF), Interleukin-4 (IL-4), lypo-polysaccharide (LPS), Interferon-gamma (IFNγ) and polyinosinic:polycytidylic acid (Poly I:C)) is superior to other DC maturation stimuli and has been shown to enhance the DC ability to secrete high levels of interleukin-12 (IL-12) during subsequent interactions with T cells. 41 The type 1 polarized, mature DC1 upregulate all emblematic co-stimulatory and maturation markers (CD40, CD80, CD86 and CD83) and display only slight phenotypic differences with classical (IL-4 and LPS-matured) DC ( Figure 5a ). However, despite a downregulation in IL-12 production following MUC1 peptide load, compared with unloaded cells, only the DC1 (and not the LPS-matured DC) released detectable level of IL-12p70 ( Figure 5b ), supporting our rationale for DC1-based vaccination. The DC1–MUC1 vaccine was administered subcutaneously (SC) in the right flank, at weeks 4, 6 and 8 after AdCre injection (according to the vaccination diagram in Figure 5c ). A total of 3000 MUC1-loaded DC1 cells were administered in each vaccine. This low number represents the mouse mass-adjusted equivalent of an adult human vaccine of ~ 6.5 million DC. Our results show that the DC1–MUC1 vaccine significantly prolongs survival in vaccinated MUC1KrasPten mice ( n = 10, P = 0.033) compared with MUC1KrasPten non-vaccinated mice ( n = 13, Figure 6a ). No mice died before the scheduled time for the first vaccine dose. In order to identify biomarkers of vaccination-induced immune changes in the host, we used a combination of ELISA (for MUC1-specific, serum IgM and IgG antibody detection), flow cytometry (for splenic Treg analysis and effector T-cell cytokine detection) and real-time PCR array (for T-cell gene induction). MUC1-specific antibodies were detectable in the serum and ascites of both vaccinated and non-vaccinated mice, although the frequency and amplitude of systemic antibody responses were not increased by DC1–MUC1 vaccination (not shown). These findings confirmed our expectation that DC1 vaccination does not favor humoral responses and are in line with our previous results showing that in order for antibody responses to be triggered, soluble peptide antigen needs to be administered alongside the preloaded DCs. 40 , 42 Although IFNγ was not significantly upregulated in the spleen ( Figure 6b ) of DC1–MUC1 recipients, flow-cytometry profiling revealed decreases in CD4+CD25+Foxp3+ Treg accumulation ( P = 0.00001, Figure 6c ) and significant improvement in the CD8 to Treg ratio in the spleen of vaccinated mice ( P = 0.007, Figure 6d ). In addition, multiplex profiling with a targeted, quantitative real-time PCR immune array revealed significant upregulation of CD80, CD40, CD45 and Jak2 genes in the spleens of vaccinated mice ( Figure 6e ). Of note, as our experiments had overall survival as the primary end point, all our immune measurements (for antibodies, T cell, cytokines and so on) were performed at necropsy, which occurred later in mice subjected to vaccination and may account for some of the variability in the observed immune responses. Altogether, our results demonstrate that type 1 polarized DC1–MUC1 can circumvent tumor-mediated immune suppression in the host, activate multiple immune effector genes and effectively prolong survival.

In a comprehensive ranking of 75 tumor antigens based on predefined and pre-weighted criteria (including oncogenicity, therapeutic function, immunogenicity and specificity), MUC1 received the second highest priority score (after WT1), emphasizing its potential for future translational studies and vaccine development. 43 Several studies have focused on the role of MUC1 in ovarian cancer progression and explored its potential as a target for immune therapy, but these efforts have been challenging due in part to a paucity of adequate animal models. The mouse mucin 1 (designated Muc1 to distinguish it from human MUC1) shares little homology with the extracellular domain of human MUC1 44 and, as shown here, does not undergo the same changes before and after disease induction, raising the need for studies in MUC1 Tg mice. The only two existing preclinical cancer models for de novo tumors expressing MUC1 as self are for colon and pancreatic adenocarcinomas. 22 , 45 , 46 We describe here the first preclinical ovarian tumor model, which expresses human MUC1 antigen as a self molecule, using triple Tg MUC1KrasPten mice that carry the previously reported, loxP-engineered Kras and Pten loci, 9 , 34 while co-expressing MUC1 antigen, driven by the endogenous mucin 1 promoter. 33 The tumors develop at the primary site, metastasize throughout the peritoneal cavity and on the diaphragm and are accompanied by increased circulating MUC1 levels, replicating more closely the human disease. Ovarian cancer is a highly metastatic disease during which cells undergoing epithelial–mesenchymal transition lose their epithelial morphology, reorganize their cytoskeleton and acquire a motile phenotype through the downregulation of adherent junctions proteins (like cadherins) and upregulation of mesenchymal markers (Snail, Slug and Vimentin). 47 , 48 Although MUC1 has been recently linked to epithelial–mesenchymal transition in pancreatic cancer, 49 no evidence on MUC1 roles in ovarian cancer epithelial–mesenchymal transition currently exist. Our results show that MUC1+ ovarian tumors have augmented capacity for loco-regional spread in vivo and increased invasion in vitro . Furthermore, the tumors with higher invasive potential express less cadherin 11, implicating this molecule for the first time in ovarian cancer biology. Our novel in vivo and in vitro findings from mice, further confirmed in a small number of human ovarian cancer cell lines, stimulate further studies on MUC1 interactions with Cdh 11, β-catenin/Wnt target genes and their mechanistic implications for ovarian cancer metastasis. In addition to Cdh11, we also identified in MUC1KrasPten tumors a significant upregulation of matrix metalloproteinase Mmp7, a well-recognized marker of increased ovarian cancer cell adhesion and invasion ( Supplementary Table 1 ). The interaction between MUC1 and Mmp7 in ovarian cancer has not been yet studied and our results offer the rationale for future investigations. Late-stage ovarian cancer is accompanied by considerable immune suppression 37 – 39 and the conditional tumor model described here also mirrors this characteristic. Furthermore, the tumor-bearing triple Tg MUC1KrasPten mice exhibit drastically increased Treg accumulation in the spleens, implicating MUC1 as modulator of loco-regional immunity. As tumors progress, MUC1 and other mucins are secreted by the tumor cells and accumulate in the systemic circulation as well as the local environment where they can act as a negative regulators of DC biology and inducers of suppressive phenotype in tumor-associated macrophages. 34 , 50 How MUC1 contributes to Treg induction in the ovarian tumor microenvironment has not been explored and ongoing studies in this animal model may provide further insight on mechanisms of tumor-induced and MUC1-mediated immune dysregulation. It is currently accepted that ovarian tumor-associated Tregs are negative prognostic indicators and that overcoming immune suppression in the tumor-bearing host is critical for anti-tumor immunotherapy. 51 – 53 In addition, vaccination with tumor-associated antigens (including MUC1) needs to effectively surmount tolerance in the host. 52 , 54 – 56 We used here a type 1 polarizing maturation cocktail that induced high IL-12 producing DCs in vitro (DC1), was able to reduce CD4+FoxP3+ Tregs in vaccinated mice and to increase survival, providing a strong rationale for further applications of DC1-based vaccines in ovarian cancer and beyond. Furthermore, endometrioid ovarian tumors like the ones studied here are often endometriosis related, as previously demonstrated in both clinical 57 and preclinical 9 , 34 mouse studies. Given the fact that endometriotic lesions are also MUC1-positive, 34 , 58 future examinations of the potential for a MUC1-based vaccine for prevention of endometriosis-associated ovarian tumors merit further investigation. In summary, our preclinical studies demonstrate novel MUC1 functions in ovarian cancer biology and support its suitability as a target for immune-based therapies. The MUC1KrasPten mice described here represent the first orthotopic ovarian tumor model expressing the human MUC1 antigen as self and are ideally suited for further mechanistic studies on MUC1 roles in ovarian cancer initiation, loco-regional progression and tumor-associated inflammation, providing excellent tools for preclinical testing of new and improved MUC1 vaccines.

All animal experiments were performed according to the protocol approved by University of Pittsburgh International Animal Care and Use Committee. The breeding diagram of MUC1 +/− 33 and KrasPten mice 9 (provided by Dr D Dinulescu) is shown in Figure 1b . Genotyping was performed as previously described. 34 , 59 The triple Tg mice were generated by crossing MUC1 +/− Tg mice, which express human MUC1 under the endogenous promoter 33 with the previously described Kras G12D/+ Pten loxP/loxP mice. 9 The MUC1 +/− mice were first backcrossed with the Pten loxP/loxP mice to ensure loxP/loxP homozygosity at the Pten locus. The newly generated MUC1 +/− Pten loxP/loxP mice were then bred to Kras G12D/+ Pten loxP/loxP mice and each litter checked for the presence of female mice with the triple genotype. Cre recombinase (Ad5CMVCre (AdCre)) adenovirus (University of Iowa Gene Transfer Vector Core) was administered to synchronized animals as previously described by us and others. 9 , 34 Briefly, 7–9–week-old Kras G12D/+ Pten loxP/loxP or MUC1 +/− Kras G12D/+ Pten loxP/loxP virgin females were injected i.p. with 5U of pregnant mare serum gonadotropin (PMSG) followed 48 h later by 5 U human chorionic gonadotropin (hCG). Thirty-six hours later, mice received 5 µl of 2.5 × 10 7 p.f.u. Ad5CMVCre delivered to the OSE of left ovary only, via intrabursal injection, during survival surgery. The contra-lateral ovary served as control. Mice were killed when the signs of disease were clinically visible (tumor mass on the injected side and/or ascites accumulation) and mice were moribund (hunched appearance, ruffled fur and unable to reach for food or water). Internal organs (reproductive tract, spleen, peritoneal tumor masses and diaphragm, and regional lymph nodes) were harvested and blood and ascites collected for further investigations. Age-paired, healthy, uninjected females were used as negative controls as indicated. Primary ovarian tumors were assigned a score of 1 (ovaries appear macroscopically normal, tumors are histologically detected); 2 for visibly enlarged ovaries (intermediate size tumors, easily dissected from the surrounding tissues) and 3 for tumors invading the adjacent structures and abdominal wall. Metastases were assessed according to the number of anatomical locations where tumors were found (diaphragm, peritoneal wall, liver, spleen, small intestine and so on). DC were generated ex vivo from syngeneic mouse bone-marrow precursors, as per our established protocols. 40 , 42 , 60 Following isolation, the precursors were incubated with 1000U/ml granulocyte-macrophage colony-stimulating factor for 7 days. CD11c-purified immature DCs were loaded with 10 µg/ml MUC1 100mer synthetic peptide (HGVTSAPDTRPAPGSTAPPA) X5 (University of Pittsburgh Peptide Synthesis Facility) and then matured for 24 h with a cocktail combining 1000U/ml granulocyte-macrophage colony-stimulating factor, 10 ng/ml IL-4 (Miltenyi Biotec, Cambridge, MA, USA), 250 ng/ml LPS (Sigma-Aldrich, St Louis, MO, USA), 100U/ml IFNγ (PeproTech, Rocky Hill, NJ, USA) and 20 µg/ml poly I:C (Sigma-Aldrich, Cambridge, MA, USA). A total 3 × 10 3 DC1 pulsed with 10 µg/ml 100mer MUC1 peptide was used for each subcutaneous (SC) immunization (right flank), at weeks 4, 6 and 8 after AdCre injection, according to the diagram in Figure 5c . Immunohistochemical studies were performed in formalin-fixed paraffin-embedded tissue with the following antibodies: anti-human MUC1 (clone HMPV, BD Pharmingen, San Diego, CA, USA), anti-human/mouse mucin1 cytosolic tail (clone MH1, NeoMarkers, Fremont, CA, USA), anti-mouse pAkt ser473, pMEK1/2 ser221 rabbit polyclonal, PI3Kp110alpha—all from Cell Signaling, Danvers, MA, USA, anti-cytokeratin 7 (Abcam, Cambridge, MA, USA), anti-ERα (Santa Cruz, Santa Cruz, CA, USA), anti-mouse FOXP3 (eBioscience, San Diego, CA, USA), anti-mouse Ki67 (Abcam) and anti-mouse Cdh11 (Invitrogen, Grand Island, NY, USA). Antigen retrieval was conducted in citrate buffer pH6 (for MUC1, Muc1, pAkt, pMEK, PI3K, Ki67 and Cdh11) or 0.1 m Tris buffer pH9 (for cytokeratin 7, estrogen receptor and FOXP3). Secondary antibodies were: biotinylated goat anti-mouse Ig (1:50; BD Pharmingen), biotinylated goat anti-rat Ig (1:200, BD Pharmingen) or labeled polymer-horseradish peroxidase, rabbit anti-Armenian Hamster IgG (1:250, Abcam) and horseradish peroxidase conjugated anti-rabbit (ready to use, Dakocytomation, Carpinteria, CA USA). Images were acquired with a Canon PowerShot A640 digital camera attached to a Zeiss microscope, using the AxioVision Rel. 4.6 imaging software. Cells from spleens, inguinal and para-aortic lymph nodes collected at necropsy were stained with fluorescent antibodies for CD3 (PerCP), CD4 (Pacific Blue), and CD8 (APC-Cy7) all from BD Biosciences (San Jose, CA, USA) followed by intracellular staining for Foxp3 (eBioscience). The matured DCs were phenotyped with CD40-PE, CD80- APC, CD83-FITC and CD86- FITC all from eBioscience. The IG10 and IG10-MUC1 cell lines were grown in complete Dulbecco’s Modified Eagle Medium (cDMEM) and stained for MUC1 (BD Biosciences) and Cdh11 (Invitrogen) markers. Stained cells were analyzed on a LSR II flow cytometer using the FACSDiva data analysis software (BD Biosciences) or FlowJo software (Tree Star Inc, Ashland, OR, USA). DNA was isolated from snap-frozen ovarian tumor mass at the time of necropsy and from tails of normal, uninjected mice, using Puregene DNA purification system (Gentra Systems, Minneapolis, MN, USA), according to manufacturer’s instructions. The DNA was then subjected to PCR analysis for Lox-Stop-Lox cassette. The sequences of the primers are: Kras forward 5′ CCATGGCTTGAGTAAGTCTGC 3′ ; Kras reverse 5′ CGC AGACTGTAGAGCAGCG 3′ and Pten forward 5′ ACTCAAGGCAGGGATGA GC 3′; Pten reverse 5′ AATCTAGGGCCTCTTGTGCC 3′. PCR conditions were as previously described. 34 , 59 RNA was extracted from either frozen ovarian tumor tissue or mouse splenocytes, as indicated. Total RNA was isolated with TRIzol reagent (Invitrogen) and then purified using an RNeasy Mini kit (Qiagen, Valencia, CA, USA), according to the manufacturer’s protocol. Two 82 gene real-time PCR profilers array were used: mouse Tumor Metastasis and mouse Th1-Th2-Th3 (SABiosciences). The results were analyzed using PCR Array data Analysis Web Portal (SABiosciences, Valencia, CA, USA), with a boundary of twofold change and a statistical significance of P ≤ 0.05. CA15-3 (soluble MUC1) levels were measured in the serum and ascites fluid using a kit (Prolias Technologies, New York, NY, USA), according to the manufacturer’s protocol. Mouse IL-12 in the cell culture supernatant was detected with ELISArray kit from SABiosciences, following manufacturer’s instructions. Anti-MUC1-specific antibodies (IgG and IgM) were detected by ELISA in the blood and ascites samples from vaccinated and non-vaccinated mice according to previously published protocols. 34 , 35 Data were represented using the average of duplicate wells, after subtracting the background readings from control wells. Acquisition and analysis were performed with Ascent Software for Multiskan (Thermo Scientific, Waltham, MA, USA). Five million IG10 cells (kindly provided by Dr K Roby, University of Kansas, Medical Center) were subjected to electroporation (280V, 960 µF) with 10 µl of a MUC1-encoding plasmid pcDNA3-MUC1 (1.6 µg/µl stock), provided by Dr Olivera J Finn (University of Pittsburgh). This plasmid encodes for 22 tandem-repeat MUC1, driven by the CMV promoter. The electroporated cells were rested in culture media (cDMEM-10) overnight and 24 h later, the selection agent (G418 1 mg/ml, Sigma-Aldrich, St Louis, MO, USA) was added. MUC1 expression was checked by flow cytometry (clone HMPV, BD Biosciences) and MUC1-positive cells sorted with a FACSArea sorter (BD Biosciences). Commercially available siRNA were scrambled (Silencer Negative Control no. 1 siRNA from Applied Biosystems (Carlsbad, CA USA) or directed against human MUC1 and mouse Cdh11. SMARTpool technology consisting of four mRNA regions directed against human MUC1 and mouse Cdh11 (Thermo Scientific Dharmacon RNAi Technologies, Lafayette, CO, USA) were used. Transfection was performed with DharmaFECT 1 Transfection Reagent (Thermo Scientific Dharmacon RNAi Technologies) and 25 n m siRNA oligonucleotides, according to the manufacturer’s protocol. Following transfection, the cells were incubated for 72 h and downregulation of target proteins was confirmed by RT–PCR and flow cytometry. Forty-eight hours after siRNA treatment, IG10 and IG10-MUC1 cells were subjected to RNA extraction with TRIzol reagent (Invitrogen). The mouse (IG10 and IG10-MUC1) and human (OvCa432, OvCa420 and TOV11D) ovarian cancer cell lines and human T-cell leukemia—Jurkat—cell line were cultured in complete DMEM (10% fetal bovine serum and 1%Pen-Strep) until confluence reached 80–90%; then the RNA was extracted using TRIzol reagent. The normal ovaries, oviducts and uterus were collected from one healthy KrasPten and one healthy MUC1KrasPten females and RNA was extracted using a tissue homogenizer (Polytron, Paterson, NJ, USA) and TRIzol reagent. The RNA was purified using an RNeasy Mini kit (Qiagen), according to the manufacturer’s protocol. Five hundred nanogram of total RNA were reverse-transcribed into cDNA using RT 2 First Strand Kit (SABiosciences) and then qRT–PCR was performed using RT 2 SYBR Green/ ROX master mix (SABiosciences); specific primers for human MUC1, mouse Muc1, human CDH11, mouse Cdh11, mouse IFNγ, all from SABiosciences. Our cohort of human formalin-fixed paraffin-embedded specimens comprised normal human endometrium ( n = 21), normal OSE primary cell lines ( n = 2) and endometrioid ovarian cancer ( n = 15). RNA was extracted using RNeasy FFPE kit (Qiagen) and 100 ng of total RNA was then subjected to hybridization and scanning according to the manufacturer’s protocol. MUC1 copy numbers were detected and quantified after background subtraction and normalization to a pool of n = 9 housekeeping genes (Nanostring nCounter GX Human Immunology Kit). IG10 and IG10-MUC1 (1 × 10 5 ) cells exposed to siRNA treatment for 48 h were placed in the upper chambers of 8 µm polycarbonate membrane coated with a thin layer of ECMMatrix (Millipore, Billerica, MA, USA). One hundred and fifty microliter of serum free media, or supplemented with 10% fetal bovine serum, were added to the wells of the feeder tray (lower chamber). After 24 h, the media from the top side of the insert, containing the cells that were not invaded, was discarded by flipping out the remaining cell suspension. The inserts were then rinsed with 150 µl of phosphate-buffered saline for 1 min, at room temperature. The invaded cells were detached and subsequently lysed in Lysis Buffer containing CyQuant GR Dye (75:1) (Millipore). The results were read with Spectra Max M2 (Molecular Devices, Sunnyvale, CA, USA), using 480/520nm filter set. The statistical analyses were conducted using univariate comparisons. We used χ 2 -tests for categorical variable (presence of ascites), or their nonparametric equivalents such as Fisher’s Exact test, if appropriate. For continuous variables (ascites volume, number of metastases and tumor scores), we used t -test or their nonparametric equivalents, such as Wilcoxon test. The mortality was assessed with Kaplan–Meier plots using log-rank test. All analyses were performed with SAS 9.0 (SAS Institute, Cary, NC, USA) assuming statistical significance at P < 0.05. Comparisons between ELISA readings and flow-cytometry measurements were performed with the Student’s t -test.