Pygopus 2 (Pygo2) is a coactivator of Wnt/β-catenin signaling that can bind bi- or trimethylated lysine 4 of histone-3 (H3K4me2/3) and participate in chromatin reading and writing. It remains unknown whether the Pygo2–H3K4me2/3 association has a functional relevance in breast cancer progression in vivo. To investigate the functional relevance of histone-binding activity of Pygo2 in malignant progression of breast cancer, we generated a knock-in mouse model where binding of Pygo2 to H3K4me2/3 was rendered ineffective. Loss of Pygo2–histone interaction resulted in smaller, differentiated, and less metastatic tumors, due, in part, to decreased canonical Wnt/β-catenin signaling. RNA- and ATAC-sequencing analyses of tumor-derived cell lines revealed downregulation of TGFβ signaling and upregulation of differentiation pathways such as PDGFR signaling. Increased differentiation correlated with a luminal cell fate that could be reversed by inhibition of PDGFR activity. Mechanistically, the Pygo2–histone interaction potentiated Wnt/β-catenin signaling, in part, by repressing the expression of Wnt signaling antagonists. Furthermore, Pygo2 and β-catenin regulated the expression of miR-29 family members, which, in turn, repressed PDGFR expression to promote dedifferentiation of wild-type Pygo2 mammary epithelial tumor cells. Collectively, these results demonstrate that the histone binding function of Pygo2 is important for driving dedifferentiation and malignancy of breast tumors, and loss of this binding activates various differentiation pathways that attenuate primary tumor growth and metastasis formation. Interfering with the Pygo2–H3K4me2/3 interaction may therefore serve as an attractive therapeutic target for metastatic breast cancer.
Pygo2 represents a potential therapeutic target in metastatic breast cancer, as its histone-binding capability promotes β-catenin–mediated Wnt signaling and transcriptional control in breast cancer cell dedifferentiation, EMT, and metastasis.
Breast cancer is the most frequent malignancy in women worldwide. Localized, differentiated breast cancers, as represented by the luminal subtypes, are responsive to therapeutic intervention (1). However, dedifferentiated, aggressive subtypes, characterized as basal-like/triple-negative, in most cases represent therapy refractory primary tumors concomitant with the formation of distant metastasis (1). Inappropriate activation of the Wnt signaling pathway, for example by activating mutations in positively acting components or loss of function of negative regulators of the pathway, has been shown to drive several malignancies including breast cancer (2). The transcriptional activity of β-catenin, the key transcriptional effector of Wnt signaling, is aided by binding to the cotranscriptional activators Pygopus 1 and 2 (Pygo1 and 2) via the Bcl9/9L adaptor proteins (3, 4).
Pygo proteins were first discovered in Drosophila melanogaster (5–7) and in Xenopus (8). Of the two mammalian Pygo homologs, Pygo2 has been shown to be more abundantly expressed and functionally relevant than Pygo1 (9, 10). While deletion of Pygo2 affects the proper development of multiple tissues, additional deletion of Pygo1 does not appear to exacerbate the Pygo2 phenotype (9–12). Furthermore, in mammals, the involvement of Pygo proteins in Wnt/β-catenin signaling has been found to be context-dependent. For instance, Pygo2 functions in a β-catenin–independent manner in lens development (12, 13), in tooth enamel formation (14) and in spermatogenesis (11). However, during the activation of hair follicle stem/progenitor cells and the regeneration of skin (15) and also during the development of mammary glands and expansion of mammary stem/progenitor cells, Pygo2 functions at least in part by regulating Wnt/β-catenin signaling (16).
The best-characterized function of the Pygo proteins is in cell proliferation. At actively transcribing gene loci, Pygo2 binds activating histone marks, such as bi- or trimethylated histones (H3K4me2/3), and recruits histone acetyltransferases (HAT), which in turn promote an open chromatin structure (17–19). Moreover, by acting as an epigenetic accessory protein, Pygo2 drives Myc-dependent activation of mitosis-related genes (20). Befitting its role in proliferation, an increased expression of Pygo2 has been observed in several malignancies, including ovarian cancer (21), glioma (22), lung cancer (23), hepatocellular carcinoma (24), and prostate cancer (25). Importantly, Pygo2 upregulation has also been observed in breast cancer (18, 26). Interestingly, Pygo2 resides in the chromosomal region 1q21-q22, which is found to be amplified in more than half of breast cancer cases (27).
In the breast, Pygo2 plays an important role in mammary lineage differentiation. Pygo2 maintains the fate of mammary stem cells/basal cells by suppressing their luminal/alveolar differentiation, at least in part, through the suppression of Notch signaling (28). Finally, ablation of Pygo2 in the mammary gland was found to delay the onset of tumors in the Wnt signaling–driven MMTV-Wnt1 transgenic mouse model of breast cancer (29).
The Pygo family proteins contain a highly conserved C-terminal plant homology domain (PHD), which mediates the direct binding to H3K4me2/3, a mark of active transcription (16, 19, 30). PHD-containing proteins can act as “code readers,” thus linking chromatin remodeling to changes in gene transcription (31). Moreover, mammalian Pygo2 also participates in “writing” of the histone code by interacting with and recruiting histone acetyltransferases (HAT), such as CBP [CREB (cAMP-responsive-element-binding protein)-binding protein] (17), or histone methyl transferases (HMT), such as mixed lineage leukemia-2 (MLL2; ref. 18), and thus augments Wnt/β-catenin–mediated transcriptional activation. Hence, Pygo2 can function as an efficient chromatin effector participating in both, “reading” and “writing” of the histone code (11, 16).
However, whether, the Pygo2–H3K4me2/3 association has a functional relevance in breast cancer progression in vivo has remained elusive. Here, we report that the interaction of Pygo2 with histones is critical for promoting Wnt/β-catenin-driven dedifferentiated invasive breast cancer and metastasis formation. In mammary tumors of knock-in mice carrying Pygo2 alleles that are unable to bind to H3K4me2/3, differentiation-inducing pathways like PDGFR get activated and promote a luminal cell fate, thereby repressing primary tumor invasiveness and metastasis formation.
Materials and Methods
Antibodies and reagents
Axin-2 (Abcam, ab32197), α-tubulin (Sigma, T-9026), β-catenin (BD Transduction Laboratories, 610154; used for immunofluorescence staining of tumor spheroids), β-catenin (Novus Biologicals, NBP1-32239; used for ChIP experiments), non-phospho (active) β-catenin (Ser33/37/Thr41; D13A1; Cell Signaling Technology, 8814; used for immunoblotting), cleaved caspase-3 (Cell Signaling Technology, 9664), c-Myc (D84C12; Cell Signaling Technology, 5605), cyclin-D1 (EPR2241) (Abcam, ab134175), cytokeratin-14 (Thermo Fisher Scientific, RB-9020-P0), cytokeratin-8/18 (Fitzgerald, 20R-CP004), E-cadherin (Zymed, 13-1900), fibronectin (Sigma-Aldrich, F3648), GAPDH (Abcam, ab9485), Notch 3 (Abcam, ab23426), phospho-Histone3 (Merck Millipore, 06-570), phospho-PDGFRβ (Tyr751; 88H8; Cell Signaling Technology, 3166), total-PDGFRβ Y-92-C-terminal (Abcam, ab32570), Pygopus 2 (Novus Biologicals, NBP1-46171; used for IHC on mouse tumor sections), Pygopus 2 (Abcam, ab155262; used for IHC on human breast tumors and also for ChIP experiments), total-AKT (Cell Signaling Technology, 9272), vimentin (Novus Biologicals, NB300-223), WIF1 (Abcam, ab155101), Alexa-Fluor 488 and 568 (Molecular Probes), secondary horseradish peroxidase-conjugated antibodies against mouse and rabbit (Jackson ImmunoResearch).
4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich, D9542), recombinant human TGFβ1 (R&D Systems, 240-B), Wnt3a (PeproTech, 315-20), CHIR99021 (Abcam, ab120890), PDGF-BB (PeproTech, 315-18), Tyrphostin AG1296 (Selleckchem, S8024), and DAPT (Selleckchem, S2215).
Generation and validation of Pygo2 knock-in mouse strain in C57BL/6 background has been described previously (32). These mice were backcrossed into an FVB/N background using speed congenics (Taconic Inc.). Thereafter, the heterozygous-mutant mice bearing the A342E point mutation (Pygo2A342E/+) were crossed with MMTV-PyMT (33) transgenic males. MMTV-PyMT-Pygo2+/+ (WT), MMTV-PyMT-Pygo2A342E/+, and MMTV-PyMT-Pygo2A342E/A342E female littermates were used for subsequent analysis. Animal experiments have been approved by the Cantonal Veterinary Office of Basel-Stadt (license numbers 1878 and 1907). In animal experiments, all efforts were made to minimize suffering, and mice were kept and bred under specific pathogen-free (SPF) conditions.
Primary tumor–derived cell line generation and treatments
Pygo2+/+ and Pygo2AE/AE cell lines were isolated from mammary gland tumors (mammary gland 2/3) of littermate MMTV-PyMT transgenic female mice (FVB/N background). In brief, a small piece of a tumor was minced and subjected to enzymatic digestion with 0.1 mg/mL DNaseI (Roche, 11284932001) and 1 mg/mL Collagenase D (Roche, 11088858001) supplemented with 50 μg/mL gentamycin (Sigma, G1397) and 1× antibiotic–antimycotic (Thermo Fisher Scientific, 15240062) for 30 minutes in sterile conditions. The cell mixture was passed through a 70-μm cell strainer (BD Falcon, 352350) and the single cells obtained were plated as a polyclonal population in a 10-cm dish in DMEM supplemented with 10% FBS (Sigma), 10% horse serum (Amimed), 100 U penicillin (Sigma-Aldrich), and 0.1 mg/mL streptomycin (Sigma-Aldrich). Medium was changed regularly and any fibroblasts in culture were removed by several passages of differential trypsinization until only epithelial cells remained. The cells were thereof cultured in DMEM (Sigma-Aldrich, D5671) supplemented with 10% FBS (Sigma-Aldrich, F7524), 2 mmol/L glutamine (Sigma-Aldrich, G7513), 100 U penicillin (Sigma-Aldrich), and 0.1 mg/mL streptomycin (Sigma-Aldrich). All cell lines were grown at 37°C, 5% CO2, 95% humidity. Isolation of these cell lines was done with approval, and according to the rules and guidelines of the Swiss Federal Veterinary Office (SFVO) and the local ethics committee (Cantonal Veterinary Office, Basel-Stadt, Switzerland; license 1878). All the cell lines in this study were regularly confirmed for the absence of Mycoplasma contamination, especially after thawing, and prior to use in any experiment. Cells were used up to 15 passages, after which, fresh cells of an early passage were thawed and used. Py2T cells (34) was used as a second WT cell line (indicated in the figures as +/+ #2) in some experiments. For EMT experiments, cells were treated with 2 ng/mL TGFβ1 for the time points indicated. Cells were treated with 100 ng/mL Wnt3a for time points specified in appropriate experiments. Cells were treated with 3 μmol/L CHIR99021 for three days.
Tumor spheroid generation
Following mechanical and enzymatic digestion of primary tumors (described above under cell line generation), primary cells were washed and resuspended in serum-free DMEM-F12 medium containing 10 ng/mL hEGF, 1 mg/mL hydrocortisone, 10 mg/mL insulin, 20 ng/mL bFGF, 4 ng/mL heparin (Sigma Aldrich), B27 (Invitrogen) supplemented with 100 U penicillin (Sigma-Aldrich), and 0.1 mg/mL streptomycin (Sigma-Aldrich). A total of 2 × 105 primary tumor cells/1.5 mL were then seeded in 6-well ultra-low attachment plates (Corning, CLS3471). Two-hundred–microliter medium was supplemented every third day and spheroids were harvested 1 week after culture. Similar seeding methodology was followed for generating tumor spheroids from primary tumor–derived cell lines.
EdU/PI cell-cycle analysis
For proliferation/cell-cycle analysis, primary tumor–derived cell lines were incubated with 10 μmol/L EdU reagent for 2 hours in the incubator. The cells were then harvested and subjected to fixation, permeabilization, and click reaction using Base Click EdU flow cytometry kit (Base Click, BCK-FC647-100) as per manufacturer's instructions. After the completion of the staining protocol, the cells were resuspended in propidium iodide (PI; 50 μg/mL; Sigma, P4170) + RNAse (10 μg/mL; Roche, 11119915001) for 2 hours at 37°C and then taken to BD FACSCanto II Analyzer for flow cytometric analysis of proliferation and cell-cycle stages. Analysis was performed using FlowJo software.
Transwell migration assay
50,000 untreated or long-term (>20 days) TGFβ-treated cells were suspended in 500 μL of DMEM +0.2% FBS and seeded into 24 transwell migration inserts (Corning, 353097) in duplicates. The bottom chambers were filled with 700 μL of DMEM +20% FBS to create a chemoattractant gradient. The cells were incubated in a tissue culture incubator at 37°C with 5% CO2. After 18 hours, inserts were fixed with 4% paraformaldehyde for 10 minutes. Cells that had not crossed the membrane were removed with a cotton swab, and cells on the bottom of the membrane were stained with DAPI. Images of five fields per insert were taken with a Leica DMI 4000 microscope and stained cells were counted using an ImageJ software plugin developed in-house.
Promoter reporter assay
For TOP/FOP-Flash promoter reporter assays, cells were seeded in triplicates in 24-well plates and next day were cotransfected with 500 ng of TOP- or FOP-Flash Firefly luciferase construct and 10 ng of Renilla luciferase construct using Lipofectamine 3000 (Thermo Fisher Scientific, L3000015). Twenty-four hours posttransfection, medium was changed and 100 ng/mL Wnt3a was added for 48 hours. Cells were then prepared for passive lysis and reading using the Dual luciferase Reporter Assay system (Promega, E1960) according to manufacturer's instructions. Samples were read using a luminometer (Berthold Technologies; Centro LB 960). Firefly luciferase activity was normalized to the Renilla luciferase activity and expressed as relative light units.
For Smad4 promoter reporter assays, cells were cotransfected with 500 ng of Smad4 Firefly reporter and 100 ng of Renilla luciferase construct using Lipofectamine 3000. pGL3-empty vector was used as a control. Twenty-four hours posttransfection, cells were treated with 2 ng/mL TGFβ for another 48 hours, following which, they were harvested and measured as above.
Chromatin immunoprecipitation (ChIP) experiments were performed as described previously (35). Briefly, crosslinked protein-bound DNA of WT or AE/AE–mutant cells was sonicated (Bioruptor, Diagenode) to achieve chromatin fragments of sizes between 100 and 500 bps. For ChIP of endogenous Pygopus 2 or β-catenin, at least 150 μg of chromatin was incubated with 10 μg of Pygopus 2 antibody (Abcam, ab155262) or β-catenin antibody (Novus Biologicals, NBP-32239), and immunocomplexes were precipitated with 40 μL of preblocked Sepharose Protein A beads (Affi-Prep Protein A Support, Bio-Rad; 1560006). Immunocomplexes were eluted from the beads, decrosslinked, and genomic DNA was purified by phenol/chloroform extraction and precipitated with sodium acetate. One out of 40 of the ChIP sample and 1% of input DNA were used for quantitative RT-PCR. Fold enrichments for specific miR-29a/b1, PDGFRβ, or WIF1 promoter regions were calculated by IP over input samples and normalized to isotype-specific IgG as negative control. Primers targeting different genomic regions of the miR-29a/b1, PDGFRβ, or WIF1 promoters are listed in Supplementary Table SI.
Cells from a 15-cm dish each were trypsinized, washed with FACS buffer (2 % FCS, 2 mmol/L EDTA/PBS) followed by blocking with anti-CD16/CD32 Fcg III/II receptor antibody (BioLegend, 101302) for 5 minutes at room temperature. Following washing, cells were divided for staining with specific antibodies (CD24-PE, CD29-APC, CD61-BV510) at 1:100 dilution for 20 minutes on ice. After washing, cells were resuspended in FACS buffer and filtered with a 40-μm mesh before applying to flow cytometry. DRAQ7 was added just prior to putting cells on the FACS to mark dead cells.
For flow cytometry–based stem cell analysis in mouse tumor tissues, primary tumors from WT or AE/AE-mutant mice were collected, minced, and subjected to enzymatic digestion first with 1 mg/mL Collagenase D (Roche, 11088858001) and then with 0.1 mg/mL DNaseI (Roche, 11284932001) under sterile conditions. After blocking with anti-CD16/CD32 antibody, lineage-positive cells were removed by staining with biotinylated antibodies against CD45, CD31, and Ter119. Dead cells were then removed using EasySep Dead Cell Removal (Annexin V) Kit (StemCell Technologies, #17899) on a magnet. The resulting cell suspension was washed with FACS buffer and subjected to staining for stem cell markers using specific antibodies (CD24-PE, CD29-APC, CD61-BV510) as described above. Cells were analyzed using Cytoflex (Beckman Coulter) and data analysis was performed by Flowjo 10.6.2.
Statistical analysis was performed using GraphPad Prism 7.0 software. All data are presented as mean ± SEM. P values <0.05 were considered statistically significant. All experiments were repeated thrice, unless otherwise stated.
Data and material availability
The datasets generated and analyzed within this study are deposited at Gene Expression Omnibus (GEO; RNA-sequencing data: accession number GSE130330; ATAC-sequencing data: accession number GSE133066). RNA- and ATAC-sequencing data can also be accessed together under super series GSE133068.
The cell lines generated within this work will be made available upon contacting the corresponding authors.
Pygo2 binding to histones is required for primary tumor growth and distant metastasis formation
To assess the biological relevance of Pygo2 binding to H3K4me2/3 histone marks in mammary gland tumor formation, we crossed knock-in mice bearing an Ala-to-Glu point mutation at position 342 (A342E) in the PHD domain of Pygo2 (previously described in ref. 32) to MMTV-PyMT mice, a well-established mouse model of metastatic breast cancer (33). It has previously been shown that the A342E mutation effectively abolishes the binding of Pygo2 to histones (32). Interestingly, compared with Pygo2+/+ wild-type (WT) mice, Pygo2A342E/+ (AE/+) heterozygous, and Pygo2A342E/A342E (AE/AE) homozygous–mutant mice had significantly smaller tumors in the mammary glands at 12 weeks of age (Fig. 1A, i and ii). This was due to a decrease in the number of proliferating cells in the mutants as assessed by immunofluorescence for phospho-histone 3 (Fig. 1B, i and ii). Cell death by apoptosis was unaffected in the mutants as assessed by immunofluorescence for cleaved caspase-3, an apoptosis marker (Supplementary Fig. S1A, i and ii). IHC analysis of tumor sections confirmed that Pygo2 expression levels and localization remained unchanged between Pygo2 wild-type and mutant mice (Supplementary Fig. S1B), indicating that the inability of Pygo2 to bind histones did not affect its nuclear localization.
To assess whether the tumors of mutant Pygo2 mice exhibit a proliferation block or reduced cell-cycle progression, tumor-bearing AE/+ heterozygous and AE/AE homozygous–mutant mice were sacrificed at 13, 14, and 15 weeks of age before they reached the legally imposed termination endpoint. The tumor size consistently increased in both hetero- and homozygous mutants at 13, 14, and 15 weeks compared with 12 weeks of age (Supplementary Fig. S1C), suggesting that the mutant tumor cells still proliferated, albeit slower than WT. This proliferation defect in Pygo2 AE/+ and AE/AE mutants was also corroborated in primary tumor-derived WT, AE/+ and AE/AE cell lines by EdU/PI labeling and flow cytometry–based cell-cycle analysis. Endpoint PCR using specific primers on genomic DNA isolated from the tumor-derived cell lines confirmed their genotype (Supplementary Fig. S1D). While 70%–80% WT cells were found in S phase, only 10%–20% AE-mutant cells were found in S phase and 60%–70% in G0–G1 phase (Supplementary Fig. S1E).
To determine whether Pygo1 could compensate for the loss of Pygo2 histone binding, we analyzed the effect of the Pygo2 AE mutation on tumor growth in PyMT mice carrying a genetic deletion of Pygo1 in mammary tumor cells (Supplementary Fig. S1F, i). Knockout of Pygo1 did not affect tumor growth, neither in WT mice nor in Pygo2 AE/+ and AE/AE–mutant mice (Supplementary Fig. S1F, ii), suggesting that Pygo1 does not play an additive or compensatory role to Pygo2 in PyMT mammary tumors.
To assess the differentiation status of tumors between Pygo2 WT and AE-mutant mice, we analyzed the expression of several differentiation markers such as ERα, ERβ, Pgr, Gata3, Wap, and Fabp4. Increased mRNA expression of these markers indicated that loss of Pygo2-histone binding leads to increased differentiation of tumors (Fig. 1C). Tumor staging by quantifying the area fractions of hyperplasia, adenoma, and carcinoma on hematoxylin and eosin–stained histologic sections also confirmed the differentiated status of AE-mutant tumors (Fig. 1D, i). Interestingly, WT mice developed significantly more carcinoma and less hyperplasia as compared with AE-mutant mice (Fig. 1D, ii). This change in malignant phenotype was further corroborated by the observation that WT mice formed significantly higher numbers of metastases in the lungs as compared with AE-mutant mice (Fig. 1E, i). The metastasis index (number of metastases in relation to primary tumor mass) and the incidence of metastasis was also significantly higher in the WT mice as compared with the AE-mutant mice (Fig. 1E, ii and iii).
Collectively, these results suggest that the Pygo2–histone interaction is important for tumor growth, dedifferentiation, malignant tumor progression, and for metastasis formation. Notably, the presence of one AE-mutant allele was sufficient to induce the mutant phenotypes, indicating a dominant-negative function of the AE-mutant protein. Interestingly, in support of previous findings (26), analysis of the The Cancer Genome Atlas (TCGA) dataset for patients with human breast cancer revealed that PYGO2 expression was significantly higher in primary tumors and metastatic lesions as compared with normal breast tissue in both, matched as well as unmatched samples (Supplementary Fig. S1G), suggesting that PYGO2 might also be important in malignant progression of breast tumors in humans. A significant difference in PYGO2 levels across breast cancer subtypes (Supplementary Fig. S1H, i), TNM stages 1–3 (Supplementary Fig. S1H, ii), BRE grades 1–3 (Supplementary Fig. S1H, iii), or in-patient samples with lymph node metastasis positivity (Supplementary Fig. S1H, iv) was, however, not observed.
Pygo2 binding to histones is required for β-catenin–dependent Wnt signaling
Because Pygo2 is a component of the nuclear β-catenin transcriptional complex, we asked whether the inability of Pygo2 to bind histones would affect the output of Wnt/β-catenin signaling. Tumor lysates from primary tumors of WT and AE/+ and AE/AE-mutant mice were analyzed by immunoblotting for the expression of several known Wnt target genes. Interestingly, the expression of well-established Wnt targets, such as Axin-2, c-Myc, and cyclin-D1, was found significantly reduced in AE/AE-mutant tumors as compared with WT tumors (Fig. 2A, i and ii). Furthermore, nonphosphorylated, active β-catenin levels were also found to be significantly higher in the WT than in the AE/AE-mutant tumors (Fig. 2A, i and ii). IHC analysis of histologic sections for c-Myc confirmed its higher expression in WT tumors as compared with AE/AE-mutant tumors (Fig. 2B). These results suggest that Pygo2 binding to histones contributes to the signaling output of Wnt/β-catenin in mammary tumors.
To corroborate these findings, primary tumor-derived cell lines were treated with Wnt3a to activate canonical Wnt signaling. qRT-PCR analysis revealed a significantly higher upregulation of Wnt target genes, such as Axin2 and cyclin-D1 (Ccnd1), in the WT cells as compared to the AE-mutant cells (Fig. 2C). Similar results were observed for Wnt3a-treated cells by immunoblotting analysis for c-Myc, cyclin-D1, and active β-catenin (Supplementary Fig. S2A). Treatment with the Wnt agonist, CHIR99021, which activates the Wnt pathway by selectively inhibiting GSK3β, also led to an upregulation of the Wnt target genes Axin2 and Ccnd1 at a much higher extent in WT cells as compared with the mutant cells (Supplementary Fig. S2B). In addition, TOP/FOP-Flash reporter assays also revealed an increased TCF/LEF reporter activity upon Wnt3a treatment in WT cells compared with mutant cells (Fig. 2D). Finally, active β-catenin could be detected only in nuclei of tumor spheroids formed by the WT primary tumor cells as compared with mutant tumor spheroids (Supplementary Fig. S2C; Supplementary Videos S1 and S2). Likewise, Axin-2 was expressed only in tumor spheroids formed by WT primary cells as compared with spheroids formed by mutant primary cells (Supplementary Fig. S2D; Supplementary Videos S3 and S4).
RNA sequencing of the primary tumor-derived cell lines and subsequent gene ontology (GO) analysis revealed a significant upregulation of pathways involved in the negative regulation of canonical Wnt signaling in the mutant cell line compared with the WT cell line (Supplementary Fig. S2E). This is due to a significant upregulation of several Wnt signaling antagonists, including WIF1, sFRP-2, Dkk-3, Kremen-1, Wnt-5a, and APCDD1, upon loss of Pygo2-histone binding in the mutant cells (Supplementary Table SII). Analysis of ATAC sequencing further corroborated the above findings, where open chromatin was observed near the transcription start sites of several genes encoding for Wnt signaling antagonists, including Wif1, Sfrp2, Wnt5a, and Apcdd1 (Supplementary Fig. S2F).
These findings could also be extended to human cell lines with different driver mutations, where an siRNA-mediated knockdown of Pygo2 in BT-474, BT-549, and MDA-MB-231 cells prevented a Wnt3a-mediated increase in Wnt target genes AXIN-2 and CYCLIN-D1 (Fig. 2E). Furthermore, analysis of the expression of several Wnt pathway genes in the TCGA database of patients with breast cancer revealed an elevated expression of PORCN, NOTUM, LEF1, AXIN-1, and CYCLIN-D1 in patients with high PYGO2 expression compared with patients with low PYGO2 expression. Interestingly, patients with low levels of PYGO2, which implicitly would also have less interaction with histones, were found to have elevated levels of several WNT antagonist genes, such as APCDD1, DKK-2, DKK-3, DKK-4, sFRP1, sFRP4, sFRP,5 and WIF1 corroborating our observations in the Pygo2 AE/AE-mutant mice (Supplementary Fig. S2G).
Collectively, these results suggest that the Pygo2–histone interaction contributes to the signaling output of β-catenin in mammary tumors of MMTV-PyMT mice. Furthermore, the changes in phenotypes observed upon loss of Pygo2′s histone binding can be explained, at least in part, by a decrease in Wnt/β-catenin signaling in tumors.
Pygo2 binding to histones is required for TGFβ signaling and EMT
Pathway analysis of the RNA sequencing results also suggested a significant downregulation of TGFβ signaling pathway in the AE/AE-mutant cells as compared with the WT cells at the basal level (Supplementary Fig. S2E). Interestingly, the transcriptomic analysis revealed a decrease in the expression of TGFβ receptor-1 (TGFβR1), Smad1, 2, 3, 4, and 5 and an upregulation of inhibitory Smad6 in the mutant cells as compared with WT cells (Supplementary Table SII). Moreover, pathway analysis of RNA-sequencing results performed on TGFβ-treated WT or AE/AE-mutant cells revealed a significant upregulation of the TGFβ signaling pathway only in the WT cells (Supplementary Fig. S3A). To validate the functional defect in TGFβ signaling upon loss of Pygo2-histone binding, we performed a promoter reporter assay for Smad4 activity. TGFβ treatment led to a significant increase in Smad4-induced promoter reporter activity only in WT cells as compared with AE/AE-mutant cells (Fig. 3A).
To assess whether the defect in TGFβ signaling in Pygo2-mutant cells could affect the induction of an epithelial-to-mesenchymal transition (EMT), we compared TGFβ-treated WT, AE/+, and AE/AE-mutant cells. While TGFβ treatment of Pygo2 WT cells induced a complete EMT, as observed by a clear change in cell morphology from epithelial to fully mesenchymal over 20 days, such a morphologic change was not appreciable in the AE/AE-mutant cells (Supplementary Fig. S3B). Immunocytochemical analysis revealed the loss of E-cadherin expression, a prototype epithelial marker, and an upregulation of vimentin, a mesenchymal marker, only in WT cells and not in AE-mutant cells (Fig. 3B). Interestingly, the level of vimentin expression in the absence of TGFβ is higher in AE-mutant cells, as compared with WT cells, which could not be further increased by TGFβ treatment. RNA expression analysis by qRT-PCR also revealed the increased expression of the mesenchymal markers N-cadherin, NCAM-1, fibronectin, vimentin, Zeb1, and Zeb2 in 4 and 20 days TGFβ-treated WT cells (Fig. 3C; Supplementary Fig. S3C). Furthermore, compared with the AE/AE-mutant cells, the fold-increase in the expression of EMT markers was significantly higher in the WT cells (Supplementary Fig. S3C). These results were also supported by performing a Transwell Boyden chamber migration assay with WT and AE-mutant cells treated with TGFβ for 20 days. While the WT cells increased their migratory ability by approximately 7-fold compared with untreated (UT) cells, this increase in migration was strongly reduced in TGFβ-treated AE/+ and AE/AE-mutant cells (Fig. 3D).
These findings were validated in human cell lines where siRNA-mediated ablation of PYGO2 expression prevented the TGFβ-induced increase in the EMT-related markers fibronectin and vimentin in BT-474 and BT-549 cell lines (Fig. 3E, i and ii). Similar results were observed in Hs-578T and MDA-MB-231 cells (Supplementary Fig. S3D, i and ii). Notably, knockdown of PYGO2 also led to a decrease in the basal expression of EMT-related markers in BT-474, BT-549, and MDA-MB-231 (Fig. 3E, i and ii; Supplementary Fig. S3D, ii). Furthermore, we performed Gene Set Enrichment Analysis (GSEA) on TCGA gene expression data from human patients with breast cancer with low or high levels of PYGO2 expression. As expected, the signature for genes that are downregulated during EMT, such as epithelial genes, was enriched in PYGO2 high-expressing patient tumors (Supplementary Fig. S3E, i). However, the genes that are typically upregulated in EMT, such as the mesenchymal genes, were found enriched in the PYGO2-low–expressing group (Supplementary Fig. S3Eii).
Collectively, these results suggest that Pygo2–histone interaction is critical for the activation of TGFβ signaling, which in turn induces EMT and drives malignant tumor progression. Hence, a defect in the TGFβ signaling pathway can, at least in part, explain the observation of reduced tumor malignancy and fewer metastases in mice carrying the Pygo2–histone binding mutation.
Pygo2 binding to histones is required to suppress differentiation pathways
Pathway analysis of the RNA- and ATAC-sequencing data described above suggested that the Pygo2–histone interaction was required to suppress pathways involved in cellular differentiation, such as PDGF receptor signaling (Supplementary Figs. S2E and S4A). Because histopathologic analysis had revealed a more differentiated phenotype in AE/AE-mutant tumors (Fig. 1C and D, ii), we investigated the activation of differentiation pathways in tumors with defective Pygo2–histone interaction. Notably, RNA and ATAC sequencing revealed increased mRNA expression (Supplementary Fig. S4B) and open chromatin at the transcriptional start sites or gene bodies (Supplementary Fig. S4C) of genes encoding for most PDGF ligands and their receptors PDGFRα and PDGFRβ in AE/AE-mutant cells. Furthermore, qRT-PCR analysis confirmed the significant upregulation of PDGF ligands and PDGF receptors in tumors of AE-mutant compared with WT mice (Fig. 4A). IHC analysis also revealed that PDGFRβ expression was specifically upregulated in the differentiated cells of AE-mutant tumors, while it was expressed in the stromal cells of both AE-mutant and WT tumors (Fig. 4B). Stimulation with PDGF-BB ligand revealed a phosphorylation and hence activation of PDGFRβ only in AE/AE-mutant cells (Fig. 4C; Supplementary Fig. S4D). This phosphorylation decreased in a dose-dependent manner upon treatment with the PDGFR inhibitor AG1296 (Fig. 4C).
PDGFR signaling has previously been shown to drive fibrotic disease by increasing the production of collagens, glycosaminoglycans, and by ECM remodeling, and its activity has been shown to be promoted by binding to heparan sulfate (36). In fact, among the top significant pathways regulated in AE-mutant cells, pathways involved in ECM organization and metabolism of carbohydrates were enriched (Supplementary Figs. S2E and S4A). Indeed, RNA sequencing revealed that the expression of genes involved in ECM and collagen formation, such as Col2a1, Col4a1, Col4a2, Col9a2, Itga8, 6, 11, and Itgb3, Laminins (Lama4, Lamb3) and fibulins (Fbln2, Fbln5), and genes involved in glycosaminoglycan metabolism, such as Hpse2, Hexa, Hs6st1, B4galt4, Hs3st3a1, Hs3st3b1, Gpc4, and Gpc6, was upregulated in AE-mutant cells as compared with WT cells (Supplementary Table S2). This observation was also confirmed by an increased production of fibronectin, N-cadherin (Supplementary Fig. S4E, i and ii), and collagen deposition (Supplementary Figs. S4F, i and ii) in the stromal compartment of AE/AE-mutant tumors, suggesting that they were indeed more fibrotic. In fact, many mutant tumors at the time of resection were found to have fluid-filled cysts, a phenotype reminiscent of benign fibrocystic disease of the breast in human patients (37).
RNA and ATAC sequencing also revealed a significant upregulation of the Notch pathway in the AE-mutant cells (Supplementary Figs. S2E and S4A), in support of an earlier study, where the genetic deletion of Pygo2 in mammary glands was found to activate Notch signaling (28). In keeping with these findings, qRT-PCR analysis confirmed a higher expression of Notch receptors and their ligands in tumor lysates of mutant mice as compared with WT mice (Supplementary Fig. S4G). In contrast, expression of the canonical Wnt target gene Axin2 was found reduced in AE-mutant tumors as compared with WT tumors (Supplementary Fig. S4G). IHC analysis confirmed the higher expression of Notch 3 in the homozygous AE/AE-mutant tumors compared with WT tumors (Supplementary Fig. S4H). Notably, while Notch 3 was localized at the cell membranes in tumor spheroids formed by WT tumor cells, it was cytoplasmic or nuclear in the tumor spheroids formed by primary cells derived from the AE-mutant tumors (Supplementary Fig. S4I), suggesting an activation of Notch signaling in the Pygo2-mutant tumor cells.
Analysis of the expression of PDGF receptors and ligands in the TCGA database of breast cancer revealed their elevated expression in patients with low PYGO2 expression compared with patients with high PYGO2 expression (Fig. 4D). These results were also corroborated by performing a GSEA for PDGFR pathways in human patients with breast cancer, where it was found elevated in the group with low PYGO2 expression (Supplementary Fig. S4J). Further supporting these results, PYGO2 expression was found to be inversely correlated to the expression of PDGFRβ and NOTCH 3 across 825 patients with breast cancer of the TCGA Network using cBioPortal (Supplementary Fig. S4K; ref. 38). Collectively, these results strongly suggest that Pygo2 interaction with histones is required to suppress differentiation drivers like PDGFR and Notch signaling pathways.
Pygo2 binding to histones is required to repress luminal cell fate
It has previously been demonstrated that differentiation drivers like Notch and PDGFR signaling promote the luminal cell fate associated with mammary gland differentiation (28, 39). Because these pathways were activated upon loss of Pygo2 interaction with histones, we asked whether this loss-of-function mutation also promoted breast tumor cells to assume a luminal cell fate. Indeed, immunofluorescence analysis of WT and AE/AE tumors revealed a significant upregulation in the expression of the luminal cell markers cytokeratin-8 and 18 (CK-8/18) in the AE/AE-mutant tumors as compared with the WT tumors, while the expression of the basal cell marker cytokeratin-14 (CK-14) was comparable between the different genotypes (Fig. 5A). Interestingly, while tumor spheroids formed by WT primary cells expressed both CK-14 and CK-8/18, spheroids formed by AE/AE-mutant primary cells consisted almost exclusively of CK-8/18-expressing luminal cells (Fig. 5B). Moreover, tumor spheroids formed by AE/AE-mutant cells were primarily hollow (empty lumen) in contrast to solid spheroids formed by WT primary cells (Fig. 5C; Supplementary Fig. S5A, i). Yet, no differences were observed in the sphere-forming efficiency between spheroids formed by the WT and the AE/AE mutants over three passages (Supplementary Fig. S5A, ii). GSEA further revealed an upregulation of a luminal gene expression signature in mutant cell lines compared with WT tumor-derived cell lines (Supplementary Fig. S5B), lending further support to a luminal fate of cells upon loss of Pygo2–histone interaction. Interestingly, RNA sequencing also revealed a significant decrease in the expression of several stemness markers, such as CD24a, Prom1 (CD133), CD1d, and Klf4 in AE/AE-mutant cells compared with WT cells (Supplementary Table SII). In contrast, expression of Itgb3 (CD61), a marker for luminal progenitors was found significantly higher in mutant as compared with the WT cells (Supplementary Table S2). To further assess the effect of loss of Pygo2-histone binding on cell stemness, we performed flow cytometry analysis for the expression of basal (CD24+CD29hi), luminal progenitor (CD24+CD29loCD61+), and mature luminal (CD24+CD29lo) cell markers (40). While all cells were CD24+ in the WT cell line, a small CD24− population existed in the Pygo2-mutant cell line (Fig. 5D, i). Furthermore, WT cells were enriched in CD29hiCD61lo expression, while 92.9% of the AE/AE-mutant cells were CD29+CD61hi, suggesting a strong enrichment in luminal progenitors (Fig. 5D, ii and iii). In addition, flow cytometry analysis for the stem cell markers CD24, CD29, and CD61 in the lineage-negative (Lin-) population of primary tumors revealed approximately two times larger CD24loCD29+ population in AE/AE-mutant tumors as compared with WT tumors (Supplementary Fig. S5C), again suggesting an enrichment for luminal cells. In contrast, no difference in the expression of CD61was apparent between mutant and WT primary tumors. Collectively, and in keeping with previous studies (28), our results suggest that Pygo2-histone binding is important for expansion of basal mammary stem cells, and a loss in this interaction promotes the expansion of luminal cells.
In line with a role of PDGFR signaling in luminal cell differentiation, the specific inhibition of PDGFR signaling with AG1296 led to a decrease in the number of hollow spheroids formed by AE/AE-mutant cell lines, while it did not affect the phenotype of the solid spheroids formed by WT tumor cell lines (Fig. 5E, i and ii; Supplementary Fig. S5D, i and ii). In fact, PDGFR signaling inhibition also led to a significant decrease in the expression of the differentiation markers Gata3 and Fabp4 in tumor spheroids generated by AE/AE-mutant cells (Supplementary Fig. S5E). Inhibition of Notch signaling with the γ-secretase inhibitor DAPT, also resulted in the formation of solid spheroids by AE/AE-mutant cells; however, this was observed only in one of the two AE/AE-mutant cell lines (Supplementary Fig. S5F). Finally, siRNA-mediated ablation of PDGFRβ expression in AE/AE-mutant cells also led to a significant decrease in the expression of genes involved in ECM organization, such as Col2a1, Col4a1, Col9a2, Bcan, Sdc3, and Chst11, indicating that the desmoplastic/fibrotic phenotype observed upon loss of Pygo2–histone binding was, at least in part, regulated by PDGFRβ signaling (Supplementary Fig. S5G).
Collectively, these results suggest that in mammary tumor cells, Pygo2–histone interaction suppresses differentiation-inducing pathways driven by PDGFR and Notch signaling. Upon activation, they drive breast cancer cells to assume a differentiated luminal, less aggressive, and less metastatic phenotype.
Pygo2 suppresses cell differentiation through miR-29 expression
To mechanistically explain how Pygo2–histone interaction suppressed pathways, which induce differentiation, we performed Integrated Motif Activity Response Analysis (ISMARA) on the RNA-sequencing transcriptomic data generated with the WT and AE-mutant cell lines. Although motifs for Snai1 and Tfdp1 transcription factors were found to be highly enriched (Supplementary Fig. S6A), their knockdown was found not to affect the expression of PDGFRα or PDGFRβ in AE-mutant cells (Supplementary Fig. S6B, i and ii). Yet, ISMARA analysis also revealed an enrichment for motifs of miR-29-3p family members (Supplementary Fig. S6A). Notably, in addition to PDGFRβ, miR-29 family members were predicted to target several genes regulated in our dataset, including many collagens (involved in ECM remodeling), Wisp1 (involved in fibrosis), and Atp1b1 (inhibits breast cancer invasion; Supplementary Fig. S6C). Hence, we probed the functional involvement of miR-29 in defining the phenotypic changes observed in the Pygo2-mutant cells. Indeed, all three members of the miR-29 family (miR-29a-3p, miR-29b-3p, miR-29c-3p) were found to be significantly downregulated in two independent AE/AE-mutant cell lines as compared with WT cells (Fig. 6A). Interestingly, the forced transient expression of all the three miR-29-3p family members in AE-mutant cells led to a significant decrease in the levels of PDGFRα and PDGFRβ (Fig. 6B; Supplementary Fig. S6D, i). Moreover, the forced expression of miR-29 family members also led to a significant decrease in the expression of the Wnt antagonist WIF1 (Supplementary Fig. S6D, ii), yet likely due to an indirect effect, because WIF1 does not have a seed sequence for miR-29 family members in its 3′UTR. In contrast, expression of another Wnt antagonist, sFRP-2, was unaffected by the expression of any of the miR-29 family members (Supplementary Fig. S6D, ii). Immunoblotting analysis for PDGFRβ and WIF1 confirmed their decrease in the mutant cells upon overexpression of miR-29 family members (Supplementary Fig. S6E). The above findings were also confirmed in an independently generated mutant cell line (Supplementary Fig. S6F). Transient overexpression of miR-29 family members in long-term TGFβ-treated AE/AE-mutant cells was able to significantly increase their migratory potential (Fig. 6C), further highlighting the differentiation-repressing function of miR-29 family members. In contrast, cell proliferation was increased only marginally by miR-29 expression (Supplementary Fig. S7A).
To further assess the effect of miR-29 overexpression on tumor growth, differentiation, and metastasis formation, we generated AE/AE-mutant cell lines stably overexpressing either control miRNA or miR-29 a/b/c family members (Supplementary Fig. S7B, i). As with the transient overexpression, stable overexpression of miR-29 family also led to a decrease in PDGFRβ and WIF1 levels, though the decrease in WIF1 was not statistically significant (Supplementary Fig. S7B, ii). AE/AE-mutant cells stably expressing either miR-Ctrl or miR-29 family members were injected orthotopically into mammary gland 9 of female NSG mice and assessed for tumor growth weekly. Interestingly, the mice overexpressing miR-29a-3p and miR-29b-3p exhibited a significantly faster tumor growth rate (Fig. 6D). The tumor weights at the time of termination were also higher in the mice overexpressing miR-29a and b (Supplementary Fig. S7C). Histopathologic analysis of the lungs in these mice indicated a trend to higher numbers of metastatic lesions in the group overexpressing miR-29a, however, without statistical significance (Supplementary Fig. S7D). Interestingly, the tumors of all the 3 groups overexpressing miR-29 family members formed dedifferentiated tumors with cells invading into the surrounding stroma, while the control group developed highly differentiated tumors (Fig. 6E).
To assess whether miR-29 expression could also affect metastatic outgrowth, Pygo2 AE/AE cells stably expressing either miR-Ctrl or miR-29 family members were injected into the tail veins of NSG mice. After 4 weeks, mice were sacrificed, and histologic sections of their lungs were analyzed for metastatic lesions. This analysis revealed a significantly higher number of metastases in the mice expressing miR-29b and c, as compared with the miR-Ctrl group (Fig. 6F). Furthermore, we assessed the expression of miR-29 family members in human patients with breast cancer using the TCGA dataset. Low expression of miR-29a and miR-29b seemed to correlate with luminal/low BRE grade of patients with breast cancer, however, this association was not statistically significant (Supplementary Fig. S7E). On the contrary, low expression of miR-29c was found to correlate significantly with basal/high BRE–grade of patients with breast cancer (Supplementary Fig. S7E).
Collectively, these results suggest that Pygo2–histone interaction suppresses differentiation, inducing pathways, such as PDGFR signaling, at least in part, through induction of miR-29 expression. As a result, Pygo2-expressing tumors are proliferative, dedifferentiated, and metastatic.
Pygo2 and β-catenin directly induce miR-29 expression
To assess whether Pygo2 and β-catenin could directly regulate the expression of miR-29, PDGFRβ, and Wnt antagonists like WIF1, we assessed their differential binding at the TCF/LEF motifs in the promoter regions of miR-29a/b1, PDGFRβ, and WIF1 in WT and AE/AE-mutant tumor-derived cell lines using chromatin immunoprecipitation (ChIP) analysis. Interestingly, a significant decrease in binding of β-catenin was observed at 3 different regions of the miR-29a/b1 promoter (region 1, 2, and 7) in AE/AE–mutant cells as compared with the WT cells (Fig. 6G, i). A similar trend was also observed with Pygo2 binding at 2 of these 3 sites (region 2 and 7), although the decrease was not statistically significant. These results suggest, that Pygo2 and β-catenin can directly bind specific regions of the miR-29a/b1 promoter, thereby regulating its expression (Fig. 6G, i). In a comparable manner, both Pygo2 and β-catenin occupancy decreased at regions 3 and 4 of the PDGFRβ gene promoter in AE/AE–mutant cells as compared with WT cells, yet with statistical significance only at region 4 (Fig. 6G, ii). A reduced binding of Pygo2 occupancy (but not β-catenin) was also observed at region 1 of the PDGFRβ gene promoter.
While Pygo2 occupancy was higher at region 2 of the WIF1 gene promoter, binding of β-catenin was significantly decreased at region 3 of the WIF1 promoter, suggesting that some regions are regulated independently between Pygo2 or β-catenin (Supplementary Fig. S7F). An increase in Pygo2 occupancy at region 2 of the WIF1 promoter suggests a mechanism of how Wnt/β-catenin signaling output is decreased in a feedback loop on loss of Pygo2–histone interaction. The decreased occupancy of both Pygo2 and β-catenin at the Notch3 gene promoter in AE/AE-mutant cells as compared with WT cells (28) was used as a positive control for the ChIP experiment (Supplementary Fig. S7F). Collectively, these results imply, that Pygo2 and β-catenin can directly regulate the expression of PDGFRβ, WIF1, and possibly other relevant genes whose expression is also affected on loss of Pygo2-histone binding, while miR-29 can in part regulate the expression of PDGFRβ and other target genes.
Pygopus proteins, initially discovered as cotranscriptional activators for β-catenin in canonical Wnt signaling, have recently been shown to also function as chromatin effectors in a Wnt-dependent and independent fashion (11–16). Yet, whether Pygo2 interaction with chromatin, by binding histone marks, plays only an auxiliary role in potentiating Wnt/β-catenin signaling or has a more direct functional relevance in aiding malignant tumor progression has remained elusive. In our study, using a knock-in mouse model, where a point mutation in the PHD domain prevents Pygo2 to effectively bind histones (32), we demonstrate that the interaction of Pygo2 with histones is critical for breast primary tumor growth, malignant progression and metastasis formation by activating canonical Wnt/β-catenin signaling and by repressing tumor cell differentiation.
Consistent with a role of Pygo2 in proliferation (20), our results demonstrate that loss of Pygo2-histone binding results in smaller tumors due to a proliferation defect rather than an increase in apoptosis. Interestingly, the phenotype was not accentuated by an additional loss of Pygo1 confirming previous observations that Pygo1 is functionally not as relevant as Pygo2 (9–12). Importantly, loss of Pygo2-histone binding decreased Wnt/β-catenin signaling output, as observed by a decrease in Wnt target gene expression in mutant tumors, the inability of mutant tumor cell lines to increase Wnt target gene expression or by TCF/LEF promoter-reporter activity upon stimulation with Wnt3a. Hence, the results suggest that the reduced tumor cell proliferation, malignant progression and metastasis formation observed in Pygo2-mutant tumors are due to a decrease in Wnt/β-catenin signaling output. The results also indicate that the loss of Pygo2–histone binding phenocopies the reported knockout of Pygo2 to a large extent, suggesting that the observations of decreased proliferation and differentiation to a luminal breast cancer subtype were primarily due to a loss of Pygo2 binding to histones.
We find that Pygo2-histone binding is important for malignant tumor progression, that is, the generation of dedifferentiated, aggressive carcinomas and formation of distant lung metastasis. Furthermore, Pygo2-histone binding also seems to be important for cells to undergo a complete EMT, an important pathologic response in cancer cells that enables them to eventually invade and metastasize (41, 42). RNA-sequencing data and functional analysis reveals that AE/AE-mutant tumor cells exert decreased TGFβ/Smad signaling. This defect is probably caused by diminished expression of TGFβR-1 and several regulatory Smads and an increase in the expression of inhibitory Smad6 in the mutant cells. The lack of efficient TGFβ signaling may thus explain the reduced malignant tumor progression and metastasis formation of the Pygo2-mutant tumors. These results are in support of a recent study where overexpression of Pygo2 in prostate cancer cell lines was found to increase regional lymph node invasion in transplantation experiments (25). Collectively, the data suggest that Pygo2 in addition to proliferation, can also affect the malignant phenotype of a cancer cell.
The complete ablation of Pygo2 expression in the normal mammary gland has been previously shown to promote luminal differentiation of the mammary cells, in part, due to activation of the Notch 3 pathway (28). In corroboration of these findings, we find that the binding of Pygo2 to histones is required to suppress the Notch pathway also in the MMTV-PyMT mouse model of breast cancer. Yet, in addition, our results reveal that Pygo2–histone binding is also critical to suppress other pathways known to induce cellular differentiation, importantly here, the PDGFRβ signaling pathway. These findings are strongly supported by the observation that in human patients with breast cancer, Pygo2 expression correlates inversely with the expression of Notch 3 or several PDGF receptors and ligands. Functionally important, the inhibition of PDGFRβ could repress the luminal and differentiated cell fate of the Pygo2-mutant tumor cells. Also, the effect of PDGFRβ inhibition is stronger than Notch inhibition in our experimental systems, as the luminal fate of only one mutant cell line was repressed by Notch inhibition, while PDGFR inhibition was effective in two independent Pygo2-mutant cell lines. In support of these results, PDGFR signaling has previously been shown to promote proliferation of luminal breast cancer cells in the absence of estrogens (39). In another study, a significant association between high PDGF signature score and poor survival has been observed in LN+ (Lymph node +), ER+ (Estrogen Receptor +), luminal A, and grade 1 and 2 breast cancers (43). While autocrine PDGFR signaling has been implicated in glioblastomas and sarcomas (44), PDGFR signaling seems to require the cooperation of other genetic alterations to induce a fully malignant phenotype (45). For instance, autocrine PDGFR signaling, in conjunction with oncogenic Ras, appears to hyperactivate PI3K and drive malignant tumor progression of breast cancer in an MMTV-Neu mouse model (46). All these studies support our findings that PDGFR signaling promotes the growth of luminal breast cells and that additional pathways, such as activated Wnt signaling or additional genetic alterations, are required to suppress cancer cell differentiation and promote the development of dedifferentiated, aggressive breast cancers.
Among the various phenotypic changes, loss of Pygo2–histone binding also resulted in the formation of fibrotic/desmoplastic tumors as observed by a significant increase in fibronectin, N-cadherin, and collagen deposition in the stromal compartment. RNA- and ATAC-sequencing analysis revealed an enrichment of biological processes involving ECM reorganization and metabolism of carbohydrates, including glycosaminoglycans, heparan sulfate, chondroitin sulfate, and keratan sulfate. In fact, significant upregulation of several collagens (Col2a1, Col4a1, Col4a2, Col9a2), integrins (Itgb3, Itga11), laminins (Lama4, Lamb3, Lamc1), Adamts2 metalloproteinase, heparanase Hpse2, proteoglycans (Sdc3, Bcan), and proteoglycan sulfotransferases (Chst11, Chst12, Hs3st3a1, Hs3st3b1, Hs6st1) suggest active ECM remodeling contributing to the fibrotic phenotype in the mutants (Supplementary Tables S2 and S3). Moreover, Wnt1-inducible signaling pathway protein 1 (WISP1), a protein involved in fibrosis, is also found to be highly upregulated in Pygo2-mutant cells (Supplementary Table S2). Activation of autocrine and paracrine PDGFR signaling, due to an elevated level of PDGFRβ in both, tumor and stromal cells might be responsible for the desmoplastic reaction by recruiting stromal fibroblasts and promoting stromal depositions and ECM remodeling in the mutant tumors as has been described previously (47). However, while tumor fibrosis has often been shown to foster tumor development by itself, it is not a driver of carcinogenesis (48). Fibrocystic tumors of benign nature are in fact often found in patients with breast cancer (49). Interestingly, a study using pancreatic ductal adenocarcinoma (PDAC) has reported that at both early and late stages of pancreatic cancer, fibrosis-associated tumor stroma constitutes a protective response from the host rather than exerting an oncogenic supportive role (50). In our study, the Pygo2-mutant tumors are desmoplastic in their stroma, but the tumor cells themselves are not dedifferentiated. Notably, treatment of Pygo2 AE/AE-mutant cells with Wnt3a significantly downregulated the expression of several ECM remodeling genes (Supplementary Table SIII). In addition, knockdown of PDGFRβ in AE/AE-mutant cells caused a significant decrease in the expression of genes involved in ECM reorganization, indicating that, upon loss of Pygo2–histone interaction, decreased Wnt activity and increased PDGFR signaling are, at least in part, responsible for the fibrotic stroma phenotype.
Finally, we found that the upregulation of PDGFR signaling and a decrease in tumor growth and metastases formation caused by the loss of Pygo2–histone binding is, at least in part, due to a decrease in the levels of miR-29-3p family members. Notably, the expression of miR-29 family members is directly regulated by canonical Pygo2-dependent Wnt signaling. On the other hand, miR-29 family members directly repress the expression of PDGFRβ, thus explaining the upregulated expression of PDGFRβ and a luminal tumor cell fate in addition to desmoplasia upon loss of Pygo2 function. Interestingly, miR-29 is classically known as an “antifibrotic miRNA”, because it has been shown to repress genes involved in ECM remodeling and collagen biogenesis. Moreover, Col4a2 and Atp1b1, known targets of miR-29 and previously shown to inhibit breast cancer invasion, are found to be highly upregulated in the Pygo2 AE-mutant cells (Supplementary Table S2). Importantly, miR-29 has previously been shown to be transcriptionally activated by β-catenin and to target Wnt signaling antagonists, such as Dkk-1, sFRP-2, and Kremen-2, thereby augmenting Wnt/β-catenin signaling (51). Therefore, we propose that Pygo2 binding to histones potentiates Wnt/β-catenin signaling and drives the expression of miR-29, which, in turn, suppresses differentiation-inducing pathways (Fig. 7). Furthermore, the loss of Pygo2–histone binding also facilitates an increase in the expression of several Wnt antagonist genes including WIF1 and potentially others. While we show this could be, in part, due to differential binding of Pygo2 and β-catenin at their promoter regions, we speculate that this could also be due to relieving the repression of an activator for Wnt antagonist genes explaining how this inhibitory feedback loop further decreases the Wnt/β-catenin pathway output. This mechanistic hierarchy also explains how the AE knock-in mutation of Pygo2 regulates Wnt/β-catenin signaling (Fig. 7). Hence, we envisage an attractive therapeutic opportunity by specifically targeting the binding site in the PHD of Pygo2 to prevent Pygo2–histone interaction. Our results predict that such therapeutic treatment will repress primary tumor growth, malignant tumor progression, and metastasis formation. Moreover, the remaining smaller, differentiated tumors may be more readily and successfully treated with conventional chemotherapy or targeted therapy, possibly by PDGFR inhibition.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
M. Saxena: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. R.K.R. Kalathur: Data curation, software, formal analysis, investigation, methodology. N. Rubinstein: Conceptualization, formal analysis, validation, methodology. A. Vettiger: Formal analysis, investigation, methodology. N. Sugiyama: Formal analysis, validation, investigation, methodology. M. Neutzner: Formal analysis, validation, investigation. M. Coto-Llerena: Resources, data curation, validation. V. Kancherla: Resources, formal analysis, validation. C. Ercan: Resources, formal analysis, validation. S. Piscuoglio: Resources, supervision. J. Fischer: Formal analysis, validation. E. Fagiani: Resources, methodology. C. Cantù: Resources, formal analysis, investigation, writing-original draft. K. Basler: Conceptualization, resources, supervision, funding acquisition, writing-original draft. G. Christofori: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, writing-original draft, writing-review and editing.
We thank P. Lorentz and the DBM microscopy facility (DBM, University of Basel) for support with microscopy, E. Panoussis, the DBM Animal Facility and D. Büchel for technical support in animal experimentation, U. Schmieder for mice genotyping, I. Galm for technical support with sectioning and staining of lung sections, and T. Bürglin for help with FACS-related experiments. We are grateful to C. Beisel, K. Eschbach, and the Genomics Facility Basel for library preparation and next-generation RNA and ATAC sequencing. We acknowledge funding for this work by the SystemsX.ch MTD project MetastasiX (2014/268), a project grant by the Swiss National Science Foundation (310030B_163471), a Sinergia Grant by the Swiss National Science Foundation (CRSII3_136274), the Swiss Cancer League (KFS-3479-08-2014), and the Krebsliga Beider Basel (KlbB-4469-03-2018).
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.