Abstract
Protein arginine methyltransferases (PRMT) are generally not mutated in diseased states, but they are overexpressed in a number of cancers, including breast cancer. To address the possible roles of PRMT overexpression in mammary gland tumorigenesis, we generated Cre-activated PRMT1, CARM1, and PRMT6 overexpression mouse models. These three enzymes are the primary type I PRMTs and are responsible for the majority of the asymmetric arginine methylation deposited in the cells. Using either a keratin 5-Cre recombinase (K5-Cre) cross or an MMTV-NIC mouse, we investigated the impact of PRMT overexpression alone or in the context of a HER2-driven model of breast cancer, respectively. The overexpression of all three PRMTs induced hyper-branching of the mammary glands and increased Ki-67 staining. When combined with the MMTV-NIC model, these in vivo experiments provided the first genetic evidence implicating elevated levels of these three PRMTs in mammary gland tumorigenesis, albeit with variable degrees of tumor promotion and latency. In addition, these mouse models provided valuable tools for exploring the biological roles and molecular mechanisms of PRMT overexpression in the mammary gland. For example, transcriptome analysis of purified mammary epithelial cells isolated from bigenic NIC-PRMT1Tg and NIC-PRMT6Tg mice revealed a deregulated PI3K–AKT pathway. In the future, these PRMTTg lines can be leveraged to investigate the roles of arginine methylation in other tissues and tumor model systems using different tissue-specific Cre crosses, and they can also be used for testing the in vivo efficacy of small molecule inhibitors that target these PRMT.
These findings establish Cre-activated mouse models of three different arginine methyltransferases, PRMT1, CARM1, and PRMT6, which are overexpressed in human cancers, providing a valuable tool for the study of PRMT function in tumorigenesis.
See related commentary by Watson and Bitler, p. 3
Introduction
In mammals, arginine methylation is catalyzed by a family of nine related protein arginine methyltransferases (PRMT); this posttranslational modification is implicated in a multitude of biological processes, including cell growth, proliferation, differentiation, and transformation. The PRMTs are classed into three types, based on the chemical modification they deposit. Type I enzymes generate asymmetrical di-methylarginine (ADMA) marks (PRMT1, 2, 3, 4, 6, 8); Type II enzymes deposit symmetrical di-methylarginine (SDMA) marks (PRMT5 and 9); Type III enzymes catalyze the mono-methylarginine (MMA) mark (PRMT7; ref. 1). Most PRMTs methylate glycine/arginine-rich (GAR) motifs, but the exceptions are PRMT4 (also called CARM1) that targets an ill-defined proline-rich motif in its substrates, and PRMT7 that recognizes a RXR motif.
Deregulated PRMT expression (particularly overexpression of PRMT1, 4, 5, and 6) has been well documented in a number of solid and hematological malignancies (1–3). PRMT1 is the major type I enzyme that is responsible for >85% of ADMA deposition (4, 5). PRMT1 is overexpressed in breast cancer tumor samples, as compared to the adjacent normal tissue, and the degree of expression correlates with tumor grade (6, 7). PRMT1 is not only upregulated, but also aberrantly spliced in breast cancer, and the overexpression of certain splice variants is tightly correlated with poor disease prognosis (8). CARM1 displays elevated expression levels in grade 3 breast cancer tumors where it likely functions as a coactivator of the E2F1-regulated transcription (9). It also works closely with the AIB1 (amplified in breast cancer 1) in the context of estrogen stimulated cell proliferation (10). Furthermore, elevated CARM1 activity promotes migration and metastasis of breast cancer cells in in vitro and in vivo models (11). Recently, in the context of ovarian cancer, it was found that CARM1 overexpression tumors are sensitive to EZH2 inhibitors (12), thus identifying the first therapeutic vulnerability for CARM1-driven tumors. PRMT5 and its cofactor MEP50 are essential for regulating transcriptional programs that promote invasive phenotypes in breast carcinoma cells and other cancers (13). Finally, there are a number of reports that support an association of high PRMT6 expression levels with poor breast cancer prognosis. Two studies found that high levels of PRMT6 expression correlate with poor patient prognosis (7, 14), and a third study reported a similar correlation when looking at PRMT6-dependent gene expression signatures (15).
The remaining five PRMTs have not been strongly linked to cancer. Indeed, PRMT2 has weak type I activity (16). PRMT3 methylates GAR motifs, like PRMT1 and PRMT6, but its primary substrate is the ribosomal protein rpS2, and not histones or RNA binding proteins (17). PRMT7 is the only type III enzyme, and it is also the only PRMT to be mutated in a disease setting, namely in an intellectual disability syndrome called SBIDDA (18). PRMT8 is very similar in primary sequence to PRMT1, but it is membrane-associated through an N-terminal myristoylation modification and displays neuron-restricted expression (19). PRMT9 is a type II enzyme that seems largely dedicated to methylating the SAP145 splicing factor (20).
Here, we generated three independent transgenic mouse lines to address the effects of type I PRMT overexpression in different in vivo settings. The forced overexpression of human PRMTs (PRMT1, CARM1, and PRMT6) was achieved by Cre-mediated removal of a floxed STOP cassette. Two independent Cre drivers were used. First, we crossed the three PRMT transgenic mice to a K5-Cre line, which induced PRMT expression broadly in epithelial tissues including the basal cells of the mammary epithelium. Second, the three lines were crossed with the MMTV-NIC (Neu-IRES-Cre) mouse, which harbors Neu/ErbB2/HER2 under the control of a MMTV promoter, has Cre recombinase coexpressed with the oncogene, and provides a tight system for the analysis of factors that synergize with HER2 (21).
K5-driven conditional overexpression of Prmt1, Carm1, and Prmt6 led to the hyperbranching of mammary ducts and epithelial hyperplasia at different time points with varied degree. Additionally, spontaneous mammary tumors were observed in the aged Prmt1- and Carm1-overexpressing mammary glands. In the Neu-induced oncogenic context, overexpression of Prmt1 and Prmt6 significantly accelerated mammary tumor onset, whereas Carm1 augmented the tumor progression exclusively upon tumor initiation. These data demonstrate that all three type I PRMTs possess oncogenic activity that predisposed the mouse mammary gland to tumor development, and support the therapeutic targeting of these PRMTs for the treatment of patients with breast cancer.
Materials and Methods
Mice
The full-length cDNA sequences of open reading frame (ORF) for human PRMT1 (variant 1) and CARM1 were inserted in-frame into the multiple cloning site (MCS) of pCAG-floxed STOP-3XFlag-MCS-IRES-GFP plasmid backbone. Human PRMT6 ORF was inserted in-frame into the pCAG-floxed STOP-3XFlag-MCS vector without GFP tagging. The constructed plasmids encoding all three PRMT ORFs were sequenced to ascertain mutation-free, and PRMT protein overexpression was verified through transient transfection into 293T cells with the PRMT plasmids and Cre plasmid, followed by Western blot analysis using Flag antibody. To generate the PRMT overexpression transgenic mice, the respective plasmids were introduced into the pronucleus of the zygotes (FVB/Nhsd substrain background) through microinjection. Founder pups were genotyped through PCR using the tail clip to examine the germline transmission. Primers are listed in Supplementary Table S1. All mouse experiments were approved by the Institutional Animal Care and Use Committee.
The inducible overexpression of Flag-tagged PRMTs in the epithelial cells of skin and mammary glands was initially confirmed by crossing with keratin 5 promoter–driven Cre (K5-Cre) mouse and Western blot analysis of the bigenic K5-Prmt mouse models (male K5-Cre mice were crossed with female PrmtTg mice). The K5 promoter is activated in utero (E13.5) in epithelial precursor cells (22), and removal of the STOP cassette at this early stage results in ectopic expression (driven by the CAGGS promoter) of the PRMTs in epithelial cells of the skin and mammary glands in adults.
Transgenic NIC mice [MMTV-NEU(NDL2-5)-IRES-CRE] were generated and characterized as previously described (21). To eliminate the high variability of tumorigenic onset due to genetic background, NIC transgenic mice used in this study were maintained in the inbred FVB/N background (23, 24). Tumor onset and progression was monitored by manual palpation bi-weekly starting from 10 weeks of age. Following the tumor onset by palpation, the mice were killed 6 weeks thereafter, except when the tumor size reached the maximal burden in length (1.5 cm) allowed by our protocol. Primary tumor volume burden was monitored by caliper measurements on live sedated mice twice a week for six consecutive weeks after tumor onset, and the tumor volume was estimated with the formula: volume = L × S2 × π/6 (mm3), where L was the longest dimension whereas S represents the shortest dimension. Tumor-free survival was analyzed using the Kaplan–Meier approach, and the statistical significance was determined by a log-rank (Mantel–Cox) test using Prism software.
All mice were housed through 12-hour light/12-hour dark cycles with free access to food and water. All animal studies were approved by the Institutional Animal Care and Use Committee at UT MD Anderson Cancer Center.
Purification of mammary epithelial cells
Mouse mammary epithelial cells (MEC) were purified following the procedures published as previously described with minor modifications (25). Briefly, #3, #4, and #5 mammary glands were removed from the mice, free of muscle tissue contamination. The lymph lode from the #4 inguinal mammary gland was carefully excised under a dissection microscope. The mammary gland tissues were then manually minced to pieces (1 mm × 1 mm), followed by the digestion in Collagenase IV (1 μg/μL Collagenase IV+DNase I in DMEM/F12) medium at 37°C for 2 hours. After washing three times in DMEM/F12 medium, enriched MECs were finally collected by pulse centrifuge.
Western blot analysis
Mice were anesthetized by CO2, and tissues were freshly harvested. Primary tissues were freshly collected and flash-frozen in liquid nitrogen and stored at −80°C. When ready, tissue lysates were prepared on ice by an electronic homogenizer in the RIPA lysis buffer (50 mmol/L Tris-HCl, pH 8.0, 150 mmol/L NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, 2 mmol/L EDTA plus 1Xprotease inhibitor; Roche). Protein concentrations of the cleared tissue lysates were determined by Bradford assay (Thermal Fisher). Protein samples were separated in 10% to 12% fresh SDS-PAGE gels and semi-dry transferred to a 0.45 μm PVDF membrane. After blocking in 5% skim milk for 1 hour at room temperature, the blots were incubated with the primary antibodies as indicated at 4°C overnight. The next day, the blots were incubated for 1 hour with the HRP-conjugated secondary antibodies (IgG, H+L; GE Healthcare), before being subjected to visualization using the enhanced chemiluminescence (Amersham). HRP-conjugated anti-mouse secondary antibody (light chain-specific; Jackson ImmunoResearch) was used in order to eliminate the nonspecific bands for mouse Flag antibody. The primary antibodies used were: PRMT1 (CST:E5A8F, 1:1,000 dilution), CARM1 (Bethyl:A300-421A, 1:1,000 dilution), PRMT6 (Bethyl: A300-929A, 1:1,000 dilution), mouse monoclonal Flag M2 antibody (Sigma, F3165: 1:5,000 dilution), rabbit Flag antibody (Sigma: F7425, 1:1,000 dilution), phospho-AKT (pAKTS473; D9E; CST: 1:1,000 dilution), AKT (C67E7; CST: 1:1,000 dilution), CCND1 (ab134175; Abcam: 1:2,000 dilution), CCND2 (D52F9; CST: 1:500 dilution), PIK3R3 (GTX107779; GenTex: 1:500 dilution), MAT2A (A304-279A; Bethyl: 1:500 dilution).
Mammary gland whole mount preparation
Mice were anesthetized by CO2, followed by cervical dislocation. The inguinal mammary glands (#4 pair) were carefully dissected out, mounted on slides, and fixed in 10% formalin solution for 24 hours. Mammary glands were stained in 0.2% Carmine alum solution overnight at room temperature, followed by a series of washing in gradient ethanol solution (70%, 95%, and 100%) for 1 hour each. Fat tissue was cleared in Xylene overnight, and whole mount was stored in methyl salicylate buffer until imaging.
Quantification of mammary intraepithelial neoplastic lesions
The number four inguinal tumor-free mammary glands from the NIC and NIC-Prmt bitransgenic animals were freshly harvested and fixed in 10% formalin solution. The designation and evaluation of mammary intraepithelial neoplasias (MIN) and adenocarcinomas is in accordance with Annapolis guidelines (26, 27). MINs are defined as the focal epithelial lesions harboring cells with atypical cytology and/or organization. Those cells are potentially capable of undergoing malignant transformation, and thus are linked to breast cancer. To identify MINs, hematoxylin and eosin stained sections (4 μm thickness) of one inguinal mammary gland from each of five mice in each of the four genotypes were analyzed, and total number of MINs were counted from the entire section. The total number of MINs from inguinal mammary glands of animal cohorts were compared among four genotypes. Statistics were performed with unpaired two-tailed Student t test.
RT-qPCR
MECs were collected as described above. Total RNA was extracted using TRIzol reagent following the manufacturer's manual. Two micrograms of total RNA was employed for reverse-transcription using the first-strand cDNA Synthesis Kit (Invitrogen) for each sample in a total of 20 μL reaction volume. Quantitative PCR was performed using 10 ng cDNA template in biological triplicates using a SYBR Green-based system (Bio-Rad). Data were normalized by the expression levels of actin and were analyzed using the ΔCt method. Primers are listed in Supplementary Table S1.
Hematoxylin and eosin staining
The dorsal skin or mammary glands were dissected carefully and fixed in 10% formalin solution for 48 hours at 4°C. Hematoxylin and eosin (H&E) staining on paraffin-embedded sections was performed following the standard protocol as described previously (28).
Immunohistochemistry
Freshly harvested tissues were immediately fixed in 10% formalin solution for 24 to 48 hours at room temperature. Paraffin-embedded sections with 10-μm thickness were deparaffinized in xylene and rehydrated through graded alcohols. Antigen retrieval was performed in 10 mmol/L citrate buffer (pH 6.0). Endogenous peroxidase activity was blocked with 3% H2O2 for 10 minutes. Standard IHC staining was applied using the primary antibodies against K5 (1:500; Covance), K8 (1:200; Developmental Studies Hybridoma Bank), K14 (1:500; Covance), ER (1:500; Santa Cruz), PR (1:400; Santa Cruz), HER2 (1:500; Santa Cruz), Flag (1:5,000; Sigma), phospho-AKT (pAKTS473; D9E; 1:1,000; CST), and Ki67 (1:250; Bethyl Lab). Following extensive washing, HRP-conjugated secondary antibodies were incubated with the sections for 30 minutes at room temperature. Signals were visualized using Tablet DAB Monitoring Staining Kit (Sigma). For quantification of Ki-67 staining, slides were scanned at ×20 using an Aperio ScanScope (Aperio). Algorithms within Spectrum software (Aperio) were used at the default settings to score staining, and the Ki-67 index was determined as the percentage of cells with Ki67-positive nuclear immunostaining among the total number of cells. The slides with CK5, CK8, CK14 staining were scored for the staining intensity defined ranging among 0 to 3, for a score as “0” (no staining), “1” (weakly stained), “2” (moderately stained), and “3” (strongly stained). The slides with ER, PR, and HER2 staining were scored for intensity according to the criteria as described before (29).
RNA-seq
Total RNAs purified from the MECs were treated with DNase I from three biological replicates for each genotype. 1.5 μg of total RNAs was exploited to generate RNA-seq libraries using the TruSeq RNA Library Prep Kit v2 (Illumina) following manufacturer's protocols, and were sequenced using 2 × 75 bases paired-end protocol on a HiSeq 2000 instrument (Illumina). Note that one NIC sample and one NIC-Prmt6Tg sample were subsequently removed in the later analysis due to a quality problem. For other samples that passed our quality control, 28 to 39 million pairs of reads were generated per sample, and were mapped to the mouse genome (mm10) by TopHat (version 2.0.10). The overall mapping rates were between 95% and 96%. Ninety-three percent of fragments had both ends mapped to the mouse genome. To call the differentially expressed genes (DEG), the number of fragments for each known gene from RefSeq database (UCSC Genome Browser on July 17, 2015) was enumerated using htseq-count from HTSeq package (version 0.6.0). Genes with less than 10 fragments in all samples were removed before differential expression analysis. The differential expression between conditions was statistically assessed by R/Bioconductor package DESeq (version 1.18.0). Genes with FDR <0.05, fold change >2 and length >200 bp were called as DEGs. To generate heatmaps, hierarchical clustering was performed on the union of DEGs from any of three comparisons (NIC-Prmt1Tg vs. NIC, NIC-Carm1Tg vs. NIC and NIC-Prmt6Tg vs. NIC) using the log2 ratio values of each sample to the average of NIC samples by hclust function in R package. Euclidean distance and ward clustering parameters were adopted to construct the dendrograms against genes and samples. The heatmaps were plotted by heatmap.2 function in R. RNA-seq data were deposited in the gene expression omnibus database in NCBI (GEO number: GSE111624).
Statistical analysis
All experiments were biologically repeated at least three to six times and data were illustrated as average ± SD unless otherwise stated. Kaplan–Meier tumor-free curves were calculated using the Prism software (v7) with pairwise comparisons using the log-rank test. For other statistical analyses, P values were calculated using unpaired Student t test or by two-way ANOVA test, unless otherwise stated. Statistic differences were deemed as significant if P < 0.05 (*) and very significant if P < 0.01 (**).
Results
Establishment of conditional PRMT-overexpressing mouse models
Numerous studies have implicated PRMT1 (6–8), CARM1 (9–11), and PRMT6 (7, 14, 15) overexpression in the promotion of breast cancer cell migration and metastasis, and in poor disease prognosis. To specifically interrogate the roles of these three PRMTs in mammary gland tumorigenesis, we generated three conditional overexpression alleles, in which Flag-tagged ORFs of three human PRMTs were inserted following a floxed STOP cassette driven by the CAGGS promoter, which facilitates robust and ubiquitous expression (Fig. 1A). In the case of PRMT1, the v1 splice variant was used. All three conditional targeting constructs were injected into the pronucleus of zygotes (FVB background). PCR-mediated genotyping identified two founder lines for PRMT1 (PRMT1Tg), three founder lines for CARM1 (CARM1Tg), and two founder lines for PRMT6 (PRMT6Tg). These lines were further crossed with WT mice (FVB background) to generate the F1 offspring in order to ensure the stable inheritance.
To direct the targeted overexpression of PRMTs in the murine mammary gland, we first used the widely studied bovine K5-cre model, in which Cre faithfully recapitulates the in vivo keratin 5 expression pattern in the basal cell layer of stratified epithelium of the mammary gland, thymus and skin (30). To verify the conditional expression of transgenic PRMTs in the epithelium, we interbred male K5-cre mice with the F1 PRMTTg females to generate bigenic K5-cre/PRMT+/Tg F2 offspring (referred to as K5-PRMTTg hereafter). To confirm the induced PRMT overexpression of the transgenic vector, we used the αFlag antibody to perform Western blot analysis on skin and thymus tissues, and IHC and H&E staining on skin, from F2 females at 4 weeks of age. Using these approaches, we successfully validated the generation of Founder lines for PRMT1 (line #4: Supplementary Fig. S1A–S1C), CARM1 (lines #6 & #9: Supplementary Fig. S2A–S2C), and PRMT6 (line #4: Supplementary Fig. S3A–S3C), all of which exhibit the expected overexpression of the Flag-tagged PRMTs in the epithelial tissues of the bigenic K5- PRMTTg females, whereas Flag signal was not detectable in the single PRMTTg females. Consistent with Western blot results, the IHC staining revealed robust transgenic tagged-PRMT signal in the basal layer of skin epidermis exclusively in the bigenic K5-PRMTTg females, but not in single PRMTTg females, for PRMT1 (Supplementary Fig. S1B), CARM1 (Supplementary Fig. S2B), and PRMT6 (Supplementary Fig. S3B), indicating that we have successfully generated conditional overexpression transgenic mouse models for PRMT1 (#4), CARM1 (#6 and #9), and PRMT6 (#4). The line #9 exhibits higher expression levels of CARM1 than line #6, and we thus chose line #9 for the subsequent CARM1-related studies.
Overexpression of PRMTs in the mammary gland leads to hyperproliferative phenotypes
To determine the extent to which the three PRMTs are overexpressed in comparison to their respective endogenous gene expression, we performed qPCR analyses on RNA isolated from epidermis tissue of the skin (generated by dorsal skin scraping) and purified MECs. These samples were purified from female mice at 6 weeks of age. The mRNA levels for three PRMTs were nearly three- to five-fold higher in mammary epithelium and skin of bigenic K5-PRMTTg females as compared with those in the single PRMTTg females (Fig. 1B). To specifically explore the protein overexpression levels of transgenic PRMTs relative to endogenous PRMTs, we carried out the Western analysis using specific PRMT antibodies on protein samples extracted from MECs (Fig. 1C). PRMT1, CARM1, and PRMT6 protein levels were clearly elevated in the respective bigenic K5-PRMTTg mouse models (Fig. 1C).
All three models of virgin bigenic K5-PRMTTg female pups grow and develop normally, and are indistinguishable from the single PRMTTg littermates when they are young. However, we observed over side-branching of the mammary glands from the nulliparous K5-PRMT1Tg females starting at 4 months of age (Fig. 2A, left). H&E staining revealed the hyperplasia of the mammary gland epithelium (Fig. 2A, right). This phenomenon was also observed for the K5-PRMT6Tg females when they reached 6 months old (Fig. 2A, bottom). In contrast, hyperbranching was not noticed in the mammary glands of K5-CARM1Tg females until they reached ∼18 months (Fig. 2A, middle), a time point that is significantly delayed when considering the hyperbranching onset in the K5-PRMT1Tg and K5-PRMT6Tg females. The rapidly proliferating epithelial cells were evidenced by the Ki-67 staining, and overexpression eventually promoted the development of MIN in the aged bigenic K5-PRMTTg females of all three models (Fig. 2B). Quantification indicated that MIN lesions in K5-Prmt1Tg, K5-Carm1Tg and K5-Prmt6Tg displayed a Ki-67 index of 63, 47, and 54, respectively. Moreover, we observed spontaneous mammary tumors that occur at higher rates in the aged bigenic K5-PRMT1Tg (in 6 of 28 mice) and K5-CARM1Tg (in 8 of 22 mice) females (with a median onset for both >20 months), when compared with the sporadic background incidence in the single PRMTTg female littermates (Supplementary Fig. S4A). An increased incidence of spontaneous mammary gland tumors was not observed in the aged K5-PRMT6Tg bigenic line. To identify the histological subtypes of those spontaneous tumors, we performed IHC using a panel of markers on three independent mammary gland tumors (only two are shown in Supplementary Fig. S4B) from the two lines that produced an increased incidence of spontaneous mammary tumors. Based on IHC scoring, tumors in the K5-CARM1Tg females display carcinoma feature of luminal cell origin (K8+/K5−), whereas K5-PRMT1Tg tumors exhibit both the basal and luminal features (Supplementary Fig. S4C). In addition to the mammary gland tumors listed above, the K5-CARM1Tg aged females also presented with a high incidence of spontaneous skin tumors (in 9/22 mice). These data suggest that the three PRMTs play distinct roles in predisposing epithelial tissues to transformation.
Distinct oncogenic roles of PRMT1, PRMT6, and CARM1 in the context of the NIC mouse model
The long latency of tumor onset in the K5-PRMT1Tg and K5-CARM1Tg females and the incomplete penetrance of tumor formation implicate that overexpression of PRMT1 or CARM1 itself might not be a major breast cancer-initiating event, which is consistent with the notion that mammary tumorigenesis involves cooperation of oncogenic events (31, 32). Recent IHC studies on almost 250 breast cancer samples revealed that CARM1 is overexpressed in both the HER2 subtype (69%) and TNBC subtype (57%; ref. 33), and the positive correlation between CARM1 and HER2 expression was also observed in a second large study (34). Furthermore, detailed examination of TCGA dataset demonstrated that both CARM1 and PRMT1 genes exhibit higher mRNA expression levels in the HER2 high-expressing breast tumor samples as compared with the HER2 low-expressing tumors (Supplementary Fig. S5). We thus chose to investigate the effects of modulating PRMT levels in a HER2-driven model of breast cancer (HER2 is also known as ERBB2/Neu; Fig. 3A). This model expresses the activated form of Neu under the control of a MMTV promoter, and has Cre recombinase coexpressed with the oncogene due to an internal ribosome entry sequence, thus the name MMTV-NIC (Neu-IRES-Cre; ref. 21). This setup provides a tight system for the analysis of factors that cooperate or synergize with HER2. Importantly, by crossing the MMTV-NIC mouse with the different PRMTTg mice, the same cells will overexpress both HER2 and the PRMT of interest.
PRMTTg female mice were crossed with male MMTV-NIC mice, and cohorts of monogenic and bigenic females were retained and aged (Fig. 3A). As consistently reported, nulliparous MMTV-NIC females developed multi-focal, signature nodular adenocarcinomas in the mammary glands with 100% penetrance. Compared with the average tumor latency of MMTV-NIC females, bigenic NIC-PRMT1Tg and NIC-PRMT6Tg virgin females displayed accelerated tumorigenic onset, with T50 being decreased by ∼2 weeks and ∼three weeks, respectively (Fig. 3B). In contrast, bigenic NIC-CARM1Tg females exhibited delayed tumorigenesis with T50 being increased by ∼2 weeks, suggesting that these three PRMTs play distinct roles in the HER2-induced tumorigenic context (Fig. 3B). Histologic analyses revealed that the average number of the MIN lesions is elevated in the mammary glands of NIC-PRMT1Tg and NIC-PRMT6Tg females, but not in NIC-Carm1Tg females, as compared to their MMTV-NIC female littermates, respectively (Fig. 3C). All three models of nulliparous bigenic females eventually developed multifocal, nodular carcinoma (Fig. 3D, left). Interestingly, whereas ∼100% of the tumorigenic cells displayed the robust Flag-tag staining in the NIC-CARM1Tg and NIC-PRMT6Tg females, ∼45% of tumor cells showed Flag staining in the NIC-PRMT1Tg females (Fig. 3D, right). The underlying mechanism that contributed to this phenomenon is not clear, but could be due to epigenetic silencing of the PRMT1 transgenic locus after or during tumor development.
In addition, we followed up the earlier tumorigenic events by counting the average number of focal lesions in the whole-mount mammary glands in the bigenic females prior to the formation of palpable mammary tumors. The NIC-PRMT1Tg and NIC-PRMT6Tg females displayed a higher than average numbers of MINs, as compared with that of NIC females at ∼12 and ∼15 weeks of age, unlike NIC-CARM1Tg females (Supplementary Fig. S6A). Unexpectedly, the average number of focal tumor nodules is higher in the NIC-CARM1Tg females, compared with the NIC-PRMT1Tg and NIC-PRMT6Tg females at the experimental end-point (Supplementary Fig. S6B). Detailed recording of the cumulative tumor volume showed that tumors from NIC-CARM1Tg females are smaller than those from NIC females at the onset of tumorigenesis, but rapidly caught up and exceeded the volume of tumors from NIC females after ∼24 weeks (Supplementary Fig. S6C). To explore whether the variability of tumor growth behaviors was due to the varied protein overexpression levels of the respective PRMTs, we performed Western analysis with a Flag antibody that detected the ectopic expression level of all three PRMTs simultaneously. As shown in Supplementary Fig. S6D, ectopic PRMT expression was observed in all the tumor tested (two independent tumors for each line), and the levels of expression varied between tumor samples. Ectopic PRMT6 expression was consistently lower than the other two PRMTs. However, NIC-PRMT1Tg and NIC-PRMT6Tg females displayed similar dynamics of MIN and focal tumor development, as compared with NIC-CARM1Tg females. Thus, it is unlikely that the different tumor characteristics were due to the varied amounts of PRMTs overexpressed. Together, these data demonstrated that in the HER2-induced tumorigenic context, both PRMT1 and PRMT6 were capable of accelerating the mammary tumorigenesis, whereas CARM1 initially delayed tumorigenesis, but promoted tumor growth once the tumor initiated.
PRMT1 and PRMT6 overexpression promoted the activity of the PI3K–AKT pathway
To begin to decipher the molecular mechanisms underlying the accelerated mammary tumorigenic onset in the NIC-PRMT1Tg and NIC-PRMT6Tg females described above, we carried out the RNA-seq analyses using purified MECs. MECs were isolated from virgin female glands at 12 weeks for NIC-PRMT1/6Tg and 15 weeks for NIC-CARM1Tg, when the mammary glands are still tumor-free and pre-neoplastic (Supplementary Fig. S6A). Over 1,000 genes displayed differential gene expression (DEG) in both NIC-PRMT1Tg and NIC-PRMT6Tg female mammary epithelia, as compared with NIC epithelia (Fig. 4A). In contrast, only ∼400 DEGs were uncovered in the MECs isolated from NIC-CARM1Tg female mammary glands (Fig. 4A).
At the global level, hierarchical clustering heatmap demonstrates that the DEGs pattern is very similar between NIC-PRMT1Tg and NIC-PRMT6Tg, which clearly grouped together, whereas the NIC-CARM1Tg DEGs pattern segregates (Fig. 4B). Kyoto Encyclopedia of Genes and Genomes pathway analysis of the DEGs from NIC-CARM1Tg mice did not identify an oncogenic pathway. However, pathway analysis of the PRMT1/6 signature revealed robust activation of the PI3K–AKT signaling pathway, and also the ECM-receptor interaction and focal adhesion pathways (Fig. 4C). The upregulated DEGs in these three pathways overlap substantially. Furthermore, these genes were largely upregulated in the NIC-PRMT1Tg and NIC-PRMT6Tg females, as compared to the NIC-CARM1Tg females (Fig. 4D, left). We validated a number of the DEGs using qPCR, including Itga2b, Itga5, Itga7, Pik3r3, Ccnd2, and Creb5 (Fig. 4D, right). These data suggest that overexpression of PRMT1 and PRMT6 likely augments NIC-induced mammary tumorigenesis by promoting the PI3K–AKT pathway. Indeed, further Western analysis revealed that AKT phosphorylation levels were elevated in these tumors (Fig. 5A). Also, protein level changes were seen for PIK3R3 and CCND2 (Fig. 5A), which were deregulated at the RNA level (Fig. 4D). Furthermore, in the bigenic NIC-PRMT1/6Tg mice we observed elevated levels of CCND1 protein (Fig. 5A), which is stabilized by active AKT signaling (35). IHC staining of mammary tumors demonstrates elevated AKT phosphorylation levels in bigenic NIC-PRMT1/6Tg mice (Fig. 5B). Recently, PRMT1 overexpression was shown to amplify the activation of AKT pathway in colorectal cancer cell lines (36), but this is the first in vivo data supporting this link.
Discussion
Transgenic overexpression mouse models have long been utilized to elucidate the biological roles and molecular mechanisms of proto-oncogenes and tumor suppressors during the mammary gland tumorigenesis (37–42). PRMT1, CARM1, and PRMT6 are overexpressed in a number of cancers and their elevated levels often correlate with poor disease prognosis. TCGA dataset analyses revealed that PRMT1 and CARM1 are markedly amplified and overexpressed in human breast cancer samples, with PRMT6 being upregulated to a less degree (Supplementary Fig. S7A and S7B). Detailed analysis revealed that PRMT1 and CARM1 are prominently upregulated in the basal-like subtype of human breast cancers (Supplementary Fig. S7C). These data suggest that these PRMTs may have oncogenic activity on their own, or that they cooperate with oncogenic drives to promote tumorigenesis. In this study, we investigated the oncogenic potential of PRMT1, CARM1, and PRMT6 by the generation of independent inducible mouse models for each of these PRMTs, and activating them in epithelial tissue using K5-Cre and MMTV-NIC driver lines. We provide in vivo genetic evidence that PRMT1, CARM1, and PRMT6 all promote cell proliferation and transformation, albeit to varying degrees.
K5-driven targeted overexpression of PRMT1, CARM1, and PRMT6 elicited increased side-branching and hyperplasia of the mammary gland (Fig. 2A). This phenotype was most obvious in fairly young (4–6 month) K5-PRMT1Tg and K5-PRMT6Tg virgin females with >85% penetrance, but also became evident in older (18 months) K5-CARM1Tg females with >80% penetrance. Interestingly, overexpression of PRMT6 did not result in an increase of sporadic tumor incidence in either the mammary gland or skin in aged female mice (Supplementary Fig. S4A). However, overexpression of PRMT1 and especially CARM1 did induce a significant increase in sporadic tumors in the bigenic K5-PRMTTg, as opposed to the uninduced PRMTTg littermates. Note that these sporadic tumors required long latency before development (>18 months).
PRMTs may collaborate with well-established oncogenic drivers to promote the development of tumors. To investigate this possibility, we crossed the three different PRMTTg mouse lines to the MMTV-NIC mouse model. The MMTV-NIC mouse model has been effectively used to study the interplay between HER2 and ShcA (21), PTEN (43), β-catenin (44), and noncoding RNAs (45), as well as in preclinical in vivo studies (46). Here we find that in the context of HER2 oncogene-induced mammary tumors, PRMT1 and PRMT6 significantly accelerated tumor growth. Surprisingly, in the same context CARM1 retards the timing of tumor initiation, but once a tumor is initiated, it displays accelerated growth (Supplementary Fig. S6C). Thus, CARM1 seems to function differently to PRMT1/6 in this setting. Transcriptome analysis has shed some light on these differences. Indeed, RNA-seq experiments performed on purified MECs isolated from bigenic NIC-PRMTTg mice revealed that PRMT1 and PRMT6 overexpression induced similar deregulated gene expression patterns, that were distinct from the altered transcriptome pattern seen in MECs from isolated NIC-CARM1Tg. This clustering is perhaps not surprising considering that PRMT1 and PRMT6 recognize similar substrates, which are not shared by CARM1. Pathway analysis of the DEGs from NIC-PRMT1/6Tg mice identified enhanced activation of PI3K–AKT signaling. Further, Western analysis revealed that AKT is phosphorylated in NIC-PRMT1/6Tg mammary tumor samples, and that selected proteins in the PI3K–AKT pathway displayed elevated protein levels (Fig. 5). This observation is consistent with previous findings that PRMT1-mediated ERα methylation at arginine 260 plays a key role in nongenomic signaling mediated by the estrogen receptor, in which ERα methylation facilitates the assembly of ERα/p85/PI3K/SRC/FAK complex in the cytoplasm and activates the downstream PI3K–AKT pathway (47, 48). Also, PRMT1 overexpression was shown to stimulate AKT phosphorylation in colorectal cancer cell lines, and PRMT1 knockdown or small molecule inhibition attenuates this pathway (36).
It should be noted that we did not observe a dramatic increase in arginine methylation levels in the different PRMT overexpressing tumors. The methyl-specific antibodies that we used may miss substrates that display increased methylation. Thus, we cannot rule out the possibility that there may be nonenzymatic scaffolding roles for the PRMTs, which could contribute to tumorigenesis.
Taken together, the three different transgenic PRMT mouse models (PRMT1, CARM1, and PRMT6) that we have established and described here, provide compelling in vivo evidence for oncogenic activities of type I PRMTs. These data also provide important insights into the molecular determinants of mammary gland tumorigenesis driven by PRMT overexpression, and could offer a valuable resource for testing the efficacy of therapeutic small molecules that target these PRMTs.
Disclosure of Potential Conflicts of Interest
M.T. Bedford is a cofounder of EpiCypher. No potential conflicts of interest were disclosed by the other authors.
Database Depositions
Deep sequencing data have been submitted to the NCBI: Geo #GSE111624.
Authors' Contributions
Conception and design: J. Bao, W.J. Muller, M.T. Bedford
Development of methodology: J. Bao, A. Di Lorenzo, Y. Yang, M.T. Bedford
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Bao, A. Di Lorenzo, M.M. Sebastian, Y. Yang
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Bao, K. Lin, Y. Lu, Y. Zhong, M.M. Sebastian, Y. Yang, M.T. Bedford
Writing, review, and/or revision of the manuscript: J. Bao, M.M. Sebastian, M.T. Bedford
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.M. Sebastian, Y. Yang
Study supervision: M.T. Bedford
Acknowledgments
This work was supported by a NIH grant (GM126421) and a CPRIT grant (RP110471) to M.T. Bedford. W.J. Muller is supported by CRC Chair in Molecular Oncology and CIHR Foundation grant.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.