PKA Signaling Drives Mammary Tumorigenesis Through Src

ORIGINAL ARTICLE PKA signaling drives mammary tumorigenesis through Src

AG Beristain1, SD Molyneux1,2, PA Joshi1,2, NC Pomroy1, MA Di Grappa1, MC Chang3, LS Kirschner4, GG Privé2, MA Pujana5 and R Khokha1,2,3

Protein kinase A (PKA) hyperactivation causes hereditary endocrine neoplasias; however, its role in sporadic epithelial cancers is unknown. Here, we show that heightened PKA activity in the mammary epithelium generates tumors. Mammary-restricted biallelic ablation of Prkar1a, which encodes for the critical type-I PKA regulatory subunit, induced spontaneous breast tumors characterized by enhanced type-II PKA activity. Downstream of this, Src phosphorylation occurs at residues serine-17 and tyrosine-416 and mammary cell transformation is driven through a mechanism involving Src signaling. The phenotypic consequences of these alterations consisted of increased cell proliferation and, accordingly, expansion of both luminal and basal epithelial cell populations. In human breast cancer, low PRKAR1A/high SRC expression deﬁnes basal-like and HER2 breast tumors associated with poor clinical outcome. Together, the results of this study deﬁne a novel molecular mechanism altered in breast carcinogenesis and highlight the potential strategy of inhibiting SRC signaling in treating this cancer subtype in humans.

Oncogene (2015) 34, 1160–1173; doi:10.1038/onc.2014.41; published online 24 March 2014

INTRODUCTION of Prkar1a in mesenchymal lineage cells is sufficient for 18 Cyclic-AMP dependent protein kinase A (PKA) ubiquitously spontaneous osteosarcoma development. However, the role of functions as a signaling hub downstream of G-protein coupled PKA signaling in other cancer types and, particularly, in mammary receptors and cAMP to regulate a spectrum of biological processes carcinogenesis remains unknown. across tissues.1–6 PKA impacts multiple signaling networks in both In this study, we show that altered PKA regulation leading to physiological and pathological conditions by phosphorylating increased PKA activity in mammary tissue promotes carcino- target proteins on serine/threonine residues. The complexity genesis. Prkar1a loss results in heightened PKA activity defined by associated with PKA function stems from its presence as two an increase in type-II PKA isozyme in mammary epithelial cells and distinct heterotetramers, termed type-I and type-II PKA,4,7 with this hyperactivation drives mammary cell transformation through each PKA isozyme varying with respect to protein subunit a mechanism involving Src. We further find that low PRKAR1A/high composition, cellular localization and turnover. Four regulatory SRC marks a tumor subset of poor-prognosis basal-like and HER2 (R) subunits (R1α,R1β,R2α and R2β) and four catalytic (C) subunits breast cancer. (Cα,Cβ,Cγ and Prkx) have been identified, where the presence of R1 or R2 subunits defines the type of PKA isozyme as type-I or RESULTS type-II, respectively.8 The balance between type-I/-II PKA can influence cell cycle entry and terminal differentiation in multiple Prkar1a loss in the mammary gland is sufficient to cause systems.4,7 Dysregulated PKA activity leads to the development of mammary tumors tumors in cAMP-responsive endocrine tissues9,10 and this is To explore the effects of PKA hyperactivation in epithelial cells, thought to stem from imbalances in activities of either type-I or we selected the mammary gland, a tissue outside of classical type-II PKA;7,11 however, its role as a cancer driver in a wider endocrine epithelium harboring an extensive ductal network with spectrum of tissues is less well known. marker-defined lineages. It can develop a diverse family of The discovery of autosomal dominant inactivating mutations of molecular cancer subtypes in humans.19,20 Several mouse models the PRKAR1A gene as the cause of Carney complex syndrome first are available, including those that allow homozygous gene linked PKA dysregulation to carcinogenesis.12,13 PRKAR1A encodes deletion in the majority of mammary ductal epithelium. We the PKA regulatory subunit R1α, and of the four PKA regulatory adopted a genetic strategy (Figure 1a) in which mammary-specific subunits, only PRKAR1A is essential for tissue development and deletion of Prkar1a was created by crossing Prkar1alox/lox mice cAMP-dependent regulation.14 Mutations to this gene in humans (Prkar1a exon 2 flanked by LoxP sites) with transgenic mice and mice induce multiple endocrine tumors as well as myxomas, expressing Cre recombinase under the mammary epithelial- osteoblastic neoplasias and schwannomas.9,15 When combined specific MMTV promoter. Unexpectedly, MMTV-Cre deletion of À À À with Tp53+/ or Rb1+/ backgrounds, Prkar1a+/ mice exhibit a Prkar1a was sufficient to generate mammary tumors (Figure 1b); generally increased incidence of sarcomas, pituitary tumors, Cre-mediated excision of the Prkar1a gene in this cohort of mice is thyroid tumors and chemically induced skin papillomas.10,16,17 In shown in Figure 1c. Prkar1aΔMam mice developed multiple tumors addition, we have found that tissue-specific heterozygous deletion with 100% penetrance and a latency of 9–15 months of age (18/18

1Ontario Cancer Institute, Princess Margaret Hospital, Toronto, Ontario, Canada; 2Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada; 3Department of Laboratory Medicine and Pathobiology, University of Toronto, Toronto, Ontario, Canada; 4Division of Endocrinology, Diabetes and Metabolism, The Ohio State University, Columbus, OH, USA and 5Breast Cancer and Systems Biology Unit, Translational Research Laboratory, Catalan Institute of Oncology, IDIBELL, L’Hospitalet del Llobregat, Barcelona, Spain. Correspondence: Dr R Khokha, Department of Medical Biophysics and Department of Laboratory Medicine and Pathobiology, Ontario Cancer Institute, University of Toronto, Toronto, Ontario, Canada M5G 2M9. E-mail: [email protected] Received 1 August 2013; revised 20 December 2013; accepted 24 December 2013; published online 24 March 2014 PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1161

Figure 1. Conditional loss of Prkar1a is sufficient to generate mammary tumors. (a) Schematic describes mouse-breeding strategy for the generation of Prkar1aΔMam mice. Prkar1aΔMam (MMTV-Cre/Prkar1afl/fl) denotes homozygous deletion by Cre recombinase expressed under the control of the MMTV promoter. (b) Survival plot of Prkar1aΔMam (n = 18; solid black line) and control MMTV-Cre mice (n = 6; dashed line). (c) Cre-mediated genomic excision of Prkar1a in primary mammary epithelial cells derived from Prkar1aΔMam mice assessed by PCR; MMTV-Cre mammary epithelial cells do not exhibit Cre-directed Prkar1a excision shown by lack of 175-bp PCR product. Prkar1afl/+ mouse osteoblast cultures transduced with retroviral Cre-recombinase (p-Cre) or GFP (p-GFP) serve as positive or negative controls.19 ‘L’ indicates DNA ladder. (d) Mammary gland whole-mounts of Prkar1aΔMam from 1.4 to 13 months of age highlights tissue progression to tumors. LN denotes lymph node; red arrows highlight progression to tumors, ‘mo’ indicates age in months. Representative (e) H&E staining of tumors from Prkar1aΔMam mice showing papillary, mixed (mix) and invasive ductal carcinoma (IDC) mammary tumors. Scale bars, 100 μm. (f) Representative immunofluorescent images of Prkar1aΔMam mammary tumors dual-labeled with epithelial lineage markers keratin 14 (basal) and keratin 18 (luminal). Merged images show the combination of keratin 14 (red), keratin 18 (green) and DAPI (blue) positivity. Scale bars, 50 μm. See also Supplementary Figure 1 mice aged o16 months; Figure 1b; Supplementary Figure 1A) mammary tumors showed immunohistochemical positivity for and progression from ductal hyperplasia (4.2 months) to palpable estrogen receptor α (Erα) and progesterone receptor (Pgr), tumors (13 months) was observed by whole-mount analysis whereas less-differentiated tumors harboring an invasive pheno- (Figure 1d). type were immuno-negative for Erα (Supplementary Figure 1B). Histologically, mammary glands of >10 month-old Prkar1aΔMam Next, to examine whether Prkar1a loss cooperates with mice had an abundance of lobular hyperplasia, back-to-back molecular pathways known to be activated in human breast growth and areas of atypia (Figure 1e). The tumors had cancer, Prkar1aΔMam mice were bred into the widely used MMTV- characteristics of papillomas with gradual progression to ductal PyMT (polyoma virus middle T-antigen) model that induces carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) activation of ErbB2, Src, c-Myc and Ras/PI3 kinase signaling (Figure 1e). In the majority of mammary tumors, a profound networks (Figure 2a).21,22 Confirmation of Cre-mediated Prkar1a expansion of keratin 18-positive luminal cells was observed, which gene excision in this cohort was confirmed by PCR (Figure 2b). was accompanied with scattered expression of the myoepithelial Tumor development was faster, with increased tumor burden, in marker keratin 14 (Figure 1f). Additionally, select Prkar1aΔMam Prkar1aΔMam/PyMT mice compared with PyMT controls (Figures 2).

Figure 2. Prkar1a ablation in the PyMT mammary tumor model accelerates tumorigenesis. (a) Schematic describes mouse-breeding strategy for the generation of Prkar1aΔMam/PyMT mice. (b) Cre-mediated genomic excision of Prkar1a in primary mammary epithelial cells derived from Prkar1aΔMam/PyMT mice assessed by PCR; MMTV-PyMT mammary epithelial cells (PyMT) do not display Cre-directed Prkar1a excision shown by lack of 175-bp PCR product. Prkar1aﬂ/+ mouse osteoblast cultures transduced with Cre-recombinase (p-Cre) or GFP (p-GFP) serve as positive or negative controls.19 ‘L’ indicates DNA ladder. (c) Combined inguinal mammary gland weights (mg) of Prkar1aΔMam/PyMT (n = 8) and PyMT (n = 12) at 40 and 70 days of age. *Po0.05. (d) Survival plot for Prkar1aΔMam/PyMT (n = 15; red line) and PyMT (n = 14; gray line) mice. Representative gross (e), whole mount (f) and H&E (g) images of 40-day-old inguinal mammary glands of Prkar1aΔMam/PyMT and PyMT mice. Dotted line and red arrows highlight tumor periphery and location. (h) Ki67 immunohistochemistry in mammary tissue from age-matched 40-day-old Prkar1aΔMam/PyMT and PyMT, as well as in 55-day-old PyMT mammary tumor. Bars, 100 μm. See also Supplementary Figure 2.

In these mice, tumors were palpable by 33 days of age, reaching PKA activity in protein lysates from age-matched non-tumor end point earlier compared with PyMT controls (detection median, (MMTV-Cre) and tumor-bearing (Prkar1aΔMam) mammary tissue 38 versus 70 days; end point median, 72 versus 95 days), as shown revealed significantly higher baseline PKA activity in Prkar1aΔMam in Figure 2d and Supplementary Figure 2. Gross tumor images and relative to MMTV-Cre controls (P = 0.047). Treatment with the PKA mammary gland whole mounts displayed extensive hyperplasia as inhibitor, PKI, abolished the effect of cAMP-induced PKA activity early as 40 days of age in Prkar1aΔMam/PyMT mice (Figures 2). This demonstrating kinase specificity of the assay (Figure 3a). In demonstrates that mammary tumorigenesis is accelerated upon addition, baseline and total PKA activity increased four- to fivefold conditional deletion of Prkar1a in an established breast cancer in 40-day Prkar1aΔMam/PyMT tumors compared with age-matched model. Tumors in Prkar1aΔMam/PyMT cohort were classified as PyMT mammary tissue lysates (Figure 3b). Notably, PKA activity in DCIS harboring similar characteristics to those known to develop non-tumor bearing 40-day-old mammary glands of Prkar1aΔMam in the PyMT model (Figure 2g). Forty-day-old lesions in mice or 55-day-old PyMT mammary tumors was not elevated. Prkar1aΔMam/PyMT had far greater Ki67 positivity than in control Accordingly, PyMT/Prkar1aΔMam tumors at 40 and 90 days of age cohorts at 40 or 55 day of age, indicating that mammary deletion exhibited elevated phosphorylated CREB, a canonical downstream of Prkar1a drives cellular proliferation (Figure 2h). target of PKA (Figures 3). Taken together, we demonstrate that mammary-specific We next explored the cellular makeup of Prkar1aΔMam deletion of Prkar1a is sufficient to induce stochastic mammary mammary tumors by flow cytometry to determine the effect of tumorigenesis exhibiting step-wise stages of histological progres- PKA hyperactivation on distinct mammary epithelial lineages. This sion. Further, in the aggressive MMTV-PyMT mouse model of analysis revealed a significant expansion of both the luminal CD24 breast cancer, Prkar1a loss profoundly accelerates mammary +/CD49flo and basal CD24+/CD49fhi epithelial subpopulations tumorigenesis. Our data show that Prkar1a have a tumor (Figure 3e). The basal fraction, known to contain mammary stem suppressor-like function in the mammary epithelium. cells,24 was increased to >13% of total cells compared with 5% in control MMTV-Cre tissue. Luminal-type breast cancers that Mammary epithelial lineage expansion upon PKA hyperactivation develop in the PyMT model are reported to exhibit a profound We and others have demonstrated that loss of the Prkar1a PKA expansion of luminal (CD24+/CD49flo) mammary epithelial cells,21 regulatory subunit results in elevated PKA enzymatic activity in and we observed a more extreme expansion of this population bone and endocrine tissues.18,23 We sought to test whether this in the Prkar1aΔMam/PyMT compared with the PyMT cohort occurs in mammary epithelium, following Prkar1a tissue-specific (Figure 3f). Biochemically, fluorescence-activated cell sorting ablation. Analyzing basal (unstimulated) and total (cAMP-induced) (FACS)-purified CD24+/CD49flo luminal mammary epithelial cells

Figure 3. Mammary-specific deletion of Prkar1a results in elevated PKA activity and mammary epithelial cell expansion. (a) Quantification of total (cAMP-treated) and baseline (unstimulated) PKA activity in adult age-matched (10–13 months of age) littermate Prkar1aΔMam mammary tumors (n = 5) and control MMTV-Cre mammary glands (n = 4). PKI inhibitor co-treatment demonstrates PKA activity specificity. (b) Baseline and total PKA activity of age-matched 40-day-old Prkar1aΔMam/PyMT tumors (n = 4), tumor-free mammary tissues from Prkar1aΔMam (n = 1), MMTV-Cre (Wt; n = 1) and PyMT (n = 4) mice and 55-day-old (55 d) PyMT mammary tumors (n = 3)* Po0.05. (c) Phosphorylated Creb (Ser 133) and total Creb protein levels in mammary protein lysates from 40- and 90-day-old Prkar1aΔMam/PyMT and MMTV-Cre mice. Wt = age-matched MMTV-Cre control. Numbers indicate individual mice from each cohort. (d) Phosphorylated Creb immunohistochemistry in Prkar1aΔMam/PyMT mammary tumor and PyMT 40-day-old mammary gland. Bar, 100 μm. (e) Representative FACS plots segregated by cell surface markers CD24 and CD49f identifying luminal (L), basal (B) and stromal (S) subpopulations in adult (4–6 months) age-matched Prkar1aΔMam and control MMTV-Cre glands; percentage of cells representing each cell population is included. Bar graphs compare total numbers of luminal (linÀ/CD24+/CD49flo) and basal (linÀ/CD24+/CD49fhi) epithelial cells in each cohort. N = 5/cohort; *Po0.05. (f) FACS plots segregated by cell surface markers CD24 and CD49f identifying luminal (L), basal (B) and stromal (S) subpopulations in 40-day-old Prkar1aΔMam/PyMT and PyMT mammary tissues; percentage of cells representing each cell population is included. Bar graphs compare total numbers of luminal (linÀ/CD24+/CD49flo) and basal (linÀ/CD24+/CD49fhi) epithelial cells in each cohort. N = 5/cohort. *Po0.05. (g) Baseline and total PKA activities in FACS-purified luminal mammary cells from PyMT (n = 3) and Prkar1aΔMam/PyMT (n = 3) mice. *P ⩽ 0.05. from Prkar1aΔMam/PyMT showed increased baseline and total PKA Prkar1aΔMam and Prkar1aΔMam/PyMT mice. Gene expression activity (Figure 3g). These data demonstrate that mammary- analysis of PKA regulatory (Prkar1a, Prkar1b, Prkar2a and Prkar2b) specific loss of Prkar1a elevates PKA activity and this coincides and catalytic (Prkaca) subunits was performed using FACS-purified with substantial expansion of the luminal and basal epithelial luminal, basal and stromal mammary cells. The lineage markers subpopulations. keratin 14 and keratin 18 were used to verify that sorted cells were purely basal (keratin 14-positive) or luminal (keratin 18-positive) (Figure 4a). As expected, Prkar1a expression was reduced in both ΔMam Type-II PKA isozyme hyperactivation in Prkar1a tumors the luminal and basal epithelial subpopulations from Prkar1aΔMam We next set out to determine the predominant PKA isozyme mice, while remaining unchanged in the stromal cell subpopula- subtype in mammary tissues and mammary epithelial cells of tion, confirming the specificity of the MMTV-directed conditional

Figure 4. PKA-II isozyme defines Prkar1a mammary-specific loss. (a) qPCR expression analysis of keratin 14, keratin 18, Prkar1a, Prkar1b, Prkar2a, Prkar2b and Prkaca in luminal (L), basal (B) and stromal (S) FACS-purified cell subpopulations derived from adult age-matched Prkar1aΔMam mammary tumors and control MMTV-Cre mammary glands (n = 5 mice per cohort); *Po0.05 among individual cell populations. (b) Expression of Prkar1a, Prkar1b, Prkar2a, Prkar2b and Prkaca PKA subunits in flow-purified luminal mammary epithelial cells derived from 40-day-old age-matched Prkar1aΔMam/PyMT and PyMT mice (n = 5 mice/cohort); *Po0.05. β-actin was used for normalization in all cohorts. (c, d) Measurement of PKA activity in individual FPLC protein fractions and subsequent verification of PKA isozyme by western blotting. PKA activity in Prkar1aΔMam tumors (n = 3) and control MMTV-Cre mammary glands (n = 2) are shown in (c), while PKA activity in age-matched 40-day-old Prkar1aΔMam/PyMT and PyMT mammary tumors (n = 4/cohort) are in shown in (d). Specifically, protein lysates from tissues were fractionated by DEAE column chromatography over a linear salt gradient. Activity peaks correspond to type-I PKA and type-II PKA isozymes, which elute at different salt concentrations. Fraction numbers (14–44) are specified on the x axis, while kinase activities measured by radioactive counts/min are shown on the y axis. Western blots verify the holoenzyme identity of eluted PKA heterotetramers based on the antibody specific for R1α (type-I PKA) or R2β (type-II PKA) PKA regulatory subunits in FPLC-fractionated protein lysates. Non-fractionated protein lysates served as positive (input) controls.

gene deletion in the epithelial compartment (Figure 4a). Next, To biochemically profilePKAholoenzymekinaseactivity, while Prkar1b expression did not change in any cell type, the we fractionated total mammary tissue protein, and used a expression of both Prkar2a and Prkar2b regulatory subunit genes combination of kinase assays and immunoblotting to match PKA increased significantly in luminal and basal subpopulations, but activity with type-I and type-II regulatory subunit abundance were unchanged in the stroma (Figure 4a). Expression of the (Figures 4). Control MMTV-Cre and Prkar1aΔMam (Figure 4c) catalytic subunit, Prkaca, was higher in all cellular subpopulations or PyMT and Prkar1aΔMam/PyMT (Figure 4d) protein lysates (Figure 4a). Given the luminal cell expansion observed in the prepared from whole mammary tissues were subjected to anion PyMT model (Figure 3f), the expression of PKA subunits in exchange chromatography and activity assays were performed this subpopulation was further analyzed in Prkar1aΔMam/PyMT on each collected fraction. Immunoblotting against R1α and R2β mice. Consistent with the above observations, significant over- proteins in fractionated samples confirmed the identity of each expression of both regulatory type-II and catalytic subunits was PKA holoenzyme species (type-I or type-II PKA) eluting across the observed, while Prkar1a levels were reduced and Prkar1b levels linear salt gradient.15 In total profiles, MMTV-Cre mammary unchanged in the luminal fraction (Figure 4b). Altogether, these glands exhibited comparable amounts of type-I and -II PKA changes suggest a shift towards the type-II PKA holoenzyme in activity, with PyMT control glands showing more PKA type-I mammary epithelium following the loss of Prkar1a. activity (Figures 4). Notably, Prkar1aΔMam and Prkar1aΔMam/PyMT

Figure 5. Aberrant PKA activity drives mammary cell proliferation. (a) Cell proliferation over 96 h time course of Prkar1aΔMam/PyMT mammary ΔMam epithelial cells stimulated with 1.0 mM cAMP. (b) Images of 3D mammary epithelial cell colonies at day 8 of culture. Prkar1a /PyMT and PyMT cells were cultured in the presence of cAMP (1.0 mM), H89 (10 μM) or vehicle 1% DMSO ( À ). (c) Comparison of colonies formed between Prkar1aΔMam/PyMT and PyMT. Colonies were quantified based on the indicated area and averages from five fields of view were calculated. (d) PKA-R2β immunoprecipitates (IP) from Prkar1aΔMam/PyMT cells treated with control peptide (Control) or AKAP-StHT31 inhibitor (Ht31) immunoblotted (IB) for AKAP1; protein lysates from identical treatments were immunoblotted for AKAP1 or PKA-R2β. Molecular weights (kDa) ΔMam are shown on the left. (e) Cell proliferation over 96 h time course of Prkar1a /PyMT mammary epithelial cells treated with 5 μM AKAP-St- Ht31 (Ht31), control peptide (Control) or left untreated ( À ). (f) Images of Prkar1aΔMam/PyMT 3D mammary epithelial cell colonies at day 8 of culture; cAMP-stimulated (cAMP) or unstimulated ( À ) cells were cultured with control peptide (Control) or AKAP-St-Ht31 inhibitor (Ht31). (g) Quantification of Prkar1aΔMam/PyMT colonies ⩾25mm2 in diameter. Colony diameters were measured and averaged from five fields of view. a = P ⩽ 0.05 when compared with cAMP-treated PyMT sample; b = P ⩽ 0.05 when compared with non-treated PyMT sample. * = P ⩽ 0.05 compared with control peptide. Cell proliferation and colony formation assays were performed in triplicate, and repeated three times. See also Supplementary Figure 3. tumors exhibited dramatic increases in type-II PKA activity, epithelial deletion of Prkar1a results in type-II PKA isozyme and experienced a decrease in type-I PKA (Figures 4), as hyperactivation. previously suggested by our mRNA expression analyses; modest type-I PKA activity observed in Prkar1aΔMam is likely Heightened PKA activity drives mammary epithelial cell from non-epithelial (stromal) contributions from homogenized proliferation mammary tissues. This series of gene/protein expression Having showed that Prkar1a ablation in mammary epithelium is and kinase activity studies demonstrate that mammary sufficient to induce tumorigenesis, we next asked whether the

Figure 6. Mammary gland-speciﬁc loss of Prkar1a induces Src activation. (a) Phosphorylated PKA substrates from 40-day-old age-matched Prkar1aΔMam/PyMT or PyMT mammary tissues were immunoprecipitated (IP) using an antibody-speciﬁc for PKA-phosphorylated serine/ threonine (RRXS*/T*) epitopes, followed by immunoblotting (IB) with anti-Src antibody. Molecular weights (kDa) are shown on the left and numbers represent individual mice. (b) Immunoblots of phosphorylated and total protein levels of Src (Ser17; Tyr416) and Akt (Ser473) in mammary protein lysates. β-actin indicates loading control. Numbers represent individual mice. (c) Immunoprecipitation (IP) of phosphorylated PKA substrates from Prkar1aΔMam tumors and control MMTV-Cre mammary glands followed by Src immunoblotting (IB). Numbers indicate individual mice from respective cohorts. (d) Src Tyr416 phosphorylation indicates active Src in protein lysates from Prkar1aΔMam and control mammary tissue. Protein lysate from Prkar1aΔMam/PyMT tumor are a positive control. See also Supplementary Figure 4.

increases in PKA activity downstream of Prkar1a deletion are control peptide (St-Ht31P) had no effect (Figures 5). AKAP St-Ht31 required for mammary epithelial proliferation. Primary mammary modestly, but significantly inhibited cAMP-induced colony forma- epithelial cells derived from Prkar1aΔMam/PyMT and control PyMT tion (Figures 5). Together, these experiments demonstrate that mice were subjected to proliferation assays stimulated with cAMP type-II PKA isozyme is fundamental in driving Prkar1aΔMam/PyMT and treated with the PKA inhibitors H89 or PKI. To begin with, cell growth. Thus, heightened type-II PKA activation is sufficient untreated Prkar1aΔMam/PyMT cells had higher baseline prolifera- and required to promote mammary epithelial cell hyperprolifera- tion rates than PyMT cells over 96 h of culture (Figure 5a; tion in the context of Prkar1a deletion. Supplementary Figure 3A). The addition of cAMP enhanced proliferation in both groups; however, Prkar1aΔMam/PyMT cells were more sensitive to this stimulation, showing a two-fold PKA activates Src signaling during mammary tumorigenesis induction in proliferation over wild-type PyMT cells (Figure 5a). Given our findings that aberrant PKA activity regulates cell The PKA inhibitors H89 and PKI abolished the effect of cAMP on proliferation, we sought the molecular mechanism through which mammary epithelial cell proliferation across the time course in PKA drives breast carcinogenesis. To this end, we first investigated both genotypes (Supplementary Figure 3). the Wnt and Erα signaling pathways, which are modulated To assess the influence of PKA activity on cellular transforma- downstream of PKA in other cancer types.5,10,29 Expression of Δ tion, Prkar1a Mam/PyMT and PyMT mammary epithelial cells were several Wnt and Erα target genes (Axin2, Tcf4, Greb1 and Pra/b) grown in Matrigel to assess the potential of colony formation. Cells were measured in 40-day-old Prkar1aΔMam/PyMT and control from both models grew significantly more large colonies in the PyMT FACS-purified luminal cells. This analysis failed to reveal Δ presence of cAMP (Figures 5). Importantly, Prkar1a Mam/PyMT cells differences in the expression of these target genes between these generated larger colonies than PyMT alone, even in the absence of cohorts (Supplementary Figure 4), suggesting that type-II PKA cAMP, and this was abrogated upon H89 co-treatment. To directly hyperactivity accelerates PyMT breast cancers by mechanisms examine the importance of elevated type-II PKA in driving cell other than Erα or Wnt pathways. Δ proliferation and transformation, we treated Prkar1a Mam/PyMT PKA can phosphorylate c-Src (herein referred to as Src) on cells with the peptide inhibitor, AKAP St-Ht31, which prevents serine-17 to regulate its activity.30 Given that Src is involved in type-II PKA regulatory subunit interaction with A-kinase anchoring PyMT tumorigenesis,22,31 its activation was evaluated in 40-day- proteins (AKAPs)25,26 important for optimal activity and cellular old PyMT cohort mammary tissues. PKA protein substrates were localization.27,28 AKAP St-Ht31 peptide mediated disruption of immunoprecipitated using an antibody specifically targeting PKA–R2AKAP interaction was verified by PKA–R2α immunopreci- phosphorylated serine/threonine epitopes (RRXS*/T*) of PKA pitation followed by AKAP1 immunoblotting, which showed substrates,32 followed by Src immunoblotting. This showed highly reduced R2βAKAP1 interaction (Figure 5d). Importantly, AKAP elevated PKA-phosphorylated Src across Prkar1aΔMam/PyMT St-Ht31 peptide inhibitor treatment effectively decreased baseline mammary tissues (Figure 6a). Since Ser17 phosphorylation of Src proliferation and 3D colony formation, while treatment with a by PKA directs its activation through auto-phosphorylation of

Oncogene (2015) 1160 – 1173 © 2015 Macmillan Publishers Limited PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1167 Tyr416,33 we examined whether this occurs in Prkar1aΔMam/PyMT somatic copy number and expression (Pearson’s correlation mammary glands. Prkar1aΔMam/PyMT tissues exhibited consis- coefficient, PCC = 0.67; P=2.2 × 10À16; Supplementary Figure 6A). tently elevated Ser17 and Tyr416 Src phosphorylation (Figure 6b). High PRKAR1A copy number/expression was associated with high Additionally, levels of total Src were also higher in Prkar1aΔMam/ expression of genes of the oxidative phosphorylation pathway, PyMT mice. Further, phospho-Akt (Ser473), a signaling effector whereas low copy number/expression associated with high known to be downstream of Src, was probed and found to be expression of genes of the ribosome pathway (Supplementary higher in Prkar1aΔMam/PyMT (Figure 6b). Figure 6B). Subsequent evaluation of PRKAR1A expression To determine whether Src is also phosphorylated by PKA in differences across tumor subtypes as defined using the PAM50 Prkar1aΔMam tumors that develop spontaneously without the classifier indicated significant under-expression in the basal-like PyMT oncogene, we analyzed tissue bearing advanced tumors subtype relative to all others (t-test P-values o0.001, from these mice. Immunoprecipitation of phosphorylated PKA Supplementary Figure 6C). Accordingly, PRKAR1A expression was substrates demonstrated highly elevated phosphorylated Src in 3 positively correlated with ESR1, which encodes for ERα, and of 4 tumors (Figure 6c). Furthermore, Tyr416 Src phosphorylation to a lower degree with ERBB2, which encodes for the human was elevated in two of four Prkar1aΔMam mammary tumors, epidermal growth factor receptor 2 (HER2) (Supplementary suggesting that Src is a frequent mediator in PKA-initiated Figure 6D). mammary tumors (Figure 6d). Together, these results highlight Our above experiments in the mouse had established increased Src as a critical pathway impacted in breast tumors linked with Src signaling to be a main mechanism critical for PKA-directed aberrant PKA activity. mammary cell growth and transformation. To explore this molecular link in human breast cancer, the TCGA series was PKA-triggered Src activation is responsible for mammary cell divided in two tumor sets corresponding to low or high PRKAR1A growth expression values (based on tertile categorization). Analyzing the expression differences between these sets identified an associa- Next, the importance of Src in PKA-driven mammary epithelial cell tion with an oncogenic SRC signature previously derived in the transformation was dissected by evaluating cell proliferation and non-tumorigenic MCF10A mammary cell line:35 genes over- colony growth upon pathway modulation using inhibitors against Δ expressed with SRC activation in MCF10A cells were also found ERα, ErbB2/EGFR, PI3K/Akt and/or Src. Prkar1a Mam/PyMT mam- to be over-expressed in breast tumors with low PRKAR1A mary epithelial cells proliferated at a higher rate than PyMT, and expression (P = 0.023; Figure 8a, top left panel). This association cAMP promoted this phenotype (Figures 7a and b; Supplementary was further confirmed by examining all genes over-expressed with Figure 5A), as previously observed. Addition of the ERα antagonist SRC activation in MCF10A cells (at a false discovery rate o5%); fulvestrant or the ErbB2/EGFR inhibitor lapatinib had incomplete P = 0.003; Figure 8a, top right panel). The genes under-expressed inhibitory effects on baseline and cAMP-induced proliferation in Δ with SRC oncogenic activation did not show significant results both PyMT and Prkar1a Mam/PyMT genotypes (Figure 7b; (Figure 8a, bottom panels). GSEA analysis in the TCGA data set Supplementary Figure 5A). In contrast, inhibition of Src with revealed that low-PRKAR1A/high-SRC tumors are enriched for dasatinib reduced baseline, and profoundly prevented cAMP- Δ pathways regulating GPI anchor biosynthesis and the peroxisome induced proliferation of Prkar1a Mam/PyMT mammary epithelial metabolic pathways (Supplementary Figure 6E), while pathways cells (Figures 7a and b). Wortmannin, which blocks PI3K/Akt, regulating glycosphingolipid biosynthesis were under-expressed suppressed proliferation in a similar manner (Figures 7a and b). (Supplementary Figure 6F). Mammary epithelial cell growth in Matrigel cultures provides a Examining clinical characteristics in the TCGA tumors, those readout of transformation. Here, ERα inhibition by fulvestrant had with concurrent low PRKAR1A and high SRC expression (low no effect, whereas ErbB2/EGFR inhibition with lapatinib did not Δ PRKAR1A/high SRC) were found to be more likely to recur (log-rank impact baseline Prkar1a Mam/PyMT or cAMP-induced colony test P = 0.007; Figure 8b). In addition, recurrence was also higher in formation (Figures 7c and d; Supplementary Figure 5B). In the low-PRKAR1A/high-SRC tumor subset compared with tumors contrast, both dasatinib and wortmannin significantly reduced classified as only low PRKAR1A, high SRC expression, or all others baseline growth of colonies and additionally inhibited cAMP- (log-rank test P-valueso0.05; Figure 8c). Subsequently, examining induced colony formation in both genotypes (Figures 7c and d). Δ the concordance with PAM50-based subtypes, 33% of the basal- To further dissect the importance of Src in Prkar1a Mam/PyMT like and 17% of HER2 subtype had low-PRKAR1A/high-SRC cell transformation, small interfering RNA (siRNA)-directed deple- expression (Figure 8d). Analysis of recurrence in these subtypes tion of endogenous Src was performed. Four siRNAs (Src-si1, -si2, Δ showed similar trends, although only HER2 reached significance -si3 and -si4) targeting Src were transfected into Prkar1a Mam/ despite containing fewer cases (P = 0.021; Supplementary Figure PyMT mammary epithelial cells and Src knockdown was assayed 6G). We repeated these analyses in the independent NKI-295 data 96 h post transfection by immunoblotting; all four siRNAs resulted set36 both for the full data and the HER2 subtype (Figures 8e and f; in reduced Src protein levels, whereas a control siRNA had no Supplementary Figure 6H). This recapitulated our findings from effect (Figure 7e). Strikingly, Src depletion using two of the the TCGA analysis, showing similar trends when examining above siRNA duplexes (Src-si1 or Src-si4) significantly inhibited Δ prognosis among tumors classified as low PRKAR1A/high SRC, baseline and cAMP-stimulated Prkar1a Mam/PyMT cell proliferation low PRKAR1A, high SRC or others. Informed by our studies in mice, (Figure 7f; Supplementary Figure 5C) and 3D colony growth this work points to a distinct subset of human breast cancers with (Figure 7g). Thus, aberrant PKA activity in the mammary gland high rates of recurrence defined by low PRKAR1A/high SRC promotes epithelial cell proliferation and transformation through a expression, and these tumors are found particularly within the mechanism involving the activation of the tyrosine kinase Src. basal-like and HER2 subtypes. Further, Src activation may act via Akt signaling to provide a growth advantage to these mammary epithelial cells. DISCUSSION Poor patient survival in low-PRKAR1A/high-SRC human breast Here we describe a pro-neoplastic role for PKA isozyme cancers hyperactivation in the mouse mammary gland. We show that Our findings that PKA hyperactivation drives mammary tumor- mammary-specific loss of Prkar1a leads to elevated type-II PKA igenesis in the mouse prompted us to examine this in human isozyme activation and this is sufficient to drive breast carcino- breast cancer. We first analyzed data from the TCGA data set34 and genesis. Further, we show that heightened PKA activity associates observed a significant correlation between altered PRKAR1A with Src pathway activation and this is in part responsible

* *

Figure 7. PKA hyperactivity induces proliferation through Src. (a) Mammary epithelial cell proliferation of Prkar1aΔMam/PyMT cells grown over 72 h. Cells were left untreated or stimulated with cAMP in combination with the small molecule inhibitors dasatinib (100 nM) or wortmannin (200 nM). Treatments were performed in triplicate and the experiments were replicated three times. (b) Comparison of cell proliferation at 72 h between Prkar1aΔMam/PyMT and PyMT cells treated with the small-molecule inhibitors fulvestrant, lapatinib, dasatinib or wortmannin. (c) The above cells were also subjected to Matrigel colony formation assays performed over 8 days. The images shown are representative of three independent experiments. (d) Small-molecule inhibitors dasatinib and wortmannin inhibit cAMP-induced colony formation in Prkar1aΔMam/ PyMT mammary epithelial cells, whereas fulvestrant and lapatinib do not. Bar graph represents number of cell colonies ⩾ 28 mm2 counted from indicated genotypes. (e) Immunoblot of Src from Prkar1aΔMam/PyMT mammary epithelial cells transfected with four siRNA duplexes directed against Src (Src-si1, -si2, -si3 and -si4), non-silencing siRNA control (NS) or transfection reagent alone (À ) 96 h post transfection. (f) Cell proliferation over 96 h time course of Prkar1aΔMam/PyMT mammary epithelial cells transfected with NS, Src-si1 or Src-si4 siRNA. (g) Images of cAMP-stimulated (cAMP) or unstimulated ( À ) Prkar1aΔMam/PyMT mammary epithelial cell colonies at 5 days of culture following NS, Src-si1 or Src-si4 siRNA transfection. Quantification of colonies ⩾25mm2 in diameter are shown in the bar graph to the right where averages from five fields of view were calculated. Experiments were performed in triplicate on three independent occasions (n = 3). a = P ⩽ 0.05 compared with untreated control; b = P ⩽ 0.05 compared with cAMP-stimulated control; *P ⩽ 0.05 compared with NS. See also Supplementary Figure 5.

Figure 8. Association between PRKAR1A/SRC expression levels, SRC oncogenic activation and breast cancer recurrence. (a) Gene set enrichment analysis (GSEA) graphical outputs for the association analysis of expression differences between high and low PRKAR1A-expressing TCGA tumors and genes over-/under-expressed with SRC oncogenic activation in MCF10A cells. The top panels correspond to the over-expressed genes in the original SRC oncogenic signature (left panel) or to the over-expressed genes at false discovery rate o5% (that is, SRC-mediated perturbation, right panel). Similarly, the bottom panels correspond to the under-expressed genes with SRC activation or mediated perturbation. The GSEA enrichment score and the nominal P-values are shown. (b) Kaplan–Meier breast cancer recurrence curves for categorization of TCGA tumors based on concurrent low PRKAR1A and high SRC expression versus the rest of cases with recurrence information available. The log-rank test P value and number of cases (n) are indicated. (c) Kaplan–Meier breast cancer recurrence curves for categorization of TCGA tumors based on low and/or high PRKAR1A/SRC expression values. (d) For the identiﬁcation of low-PRKAR1A/high-SRC expression TCGA tumors, the complete series was ordered according to PRKAR1A or SRC expression and, subsequently, the intersection was identiﬁed. The distribution of these tumors (that is, low PRKAR1A plus high SRC) across the PAM50-based subtypes is shown in the right panel, which reveals higher concordance with basal-like type. (e) Kaplan–Meier survival curves for categorization of tumors based on concurrent low PRKAR1A and high SRC expression versus the rest of cases within each PAM50-based subtype in the full NKI-295 data set. (f) Kaplan–Meier survival curves for categorization of tumors based on low and/or high PRKAR1A/SRC expression values in the full NKI-295 data set. See also Supplementary Figure 6.

© 2015 Macmillan Publishers Limited Oncogene (2015) 1160 – 1173 PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1170 for mammary epithelial cell proliferation and transformation. downstream mediator of PKA-directed cell proliferation and Our work additionally exposes a putative low-PRKAR1A/high-SRC colony growth. Thus, mammary-specific PKA hyperactivation in human tumor subset associated with poor patient outcome. this model cooperates with PyMT to augment Src-mediated PKA is the main mediator of the highly conserved cAMP proliferation and cellular transformation. Importantly, we observed signaling pathway and aberrant activation of this kinase is known Src activation even in the absence of the PyMT oncogene, to be causal in hereditary endocrine neoplasias of the Carney highlighting the conserved signaling mechanism attributed to complex.12,13 However, its importance in sporadic tumors Prkar1a genetic ablation. However, our observation that not all originating from epithelial lineage is less understood. Mouse Prkar1aΔmam tumors exhibited elevated phosphorylated Src (Y416) modeling of Prkar1a heterozygosity in the germ line setting combined with our findings that Src inhibition partially blocked recapitulates the spectrum of endocrine tumors seen in Carney PKA-mediated cell proliferation suggests that aberrant PKA activity complex patients harboring autosomal dominant inactivating may interact with other unidentified signaling pathways important PRKAR1A mutations.9 In mesenchymal cells, osteoblast-specific in tumorigenesis. heterozygous Prkar1a deletion leads to spontaneous bone Human breast cancers exhibit substantial complexity and tumorigenesis.18,37 We now demonstrate that mammary epithelial integrative genomic studies are now progressively stratifying cells lacking both copies of Prkar1a demonstrate aberrant type-II tumors into novel molecular subsets building on the PAM50 PKA activation. This mirrors reports of type-II PKA activity increases designations.19,20,34,43 These studies have exposed considerable in Carney complex tumors.15,23 Importantly, we found that heterogeneity within the major types of breast cancer. Patients heightened PKA activation is responsible for highly augmented with basal-like or HER2 breast cancers have adverse prognosis baseline and cAMP-induced proliferation of primary mammary compared to those categorized as luminal A and luminal B. cells in multiple assays in vitro and mammary lineage expansion Aberrant Src activity has been found to correlate with HER2 in vivo, enabling stochastic mammary cancer development by positivity, higher breast tumor grade and inversely correlates with 1 year of age. Although it is uncommon, breast ductal adenomas ERα status.16,17 Informed by the observation that Prkar1a deletion have been reported in women with Carney complex, indicating activates the Src pathway in the mouse, we probed for a putative that this phenomenon is not limited to genetically engineered low-PRKAR1A/high-SRC breast cancer subset and uncovered a mice.38 Our findings generalize the importance of PKA holoen- substantial proportion of the patient population with tumors zyme homeostasis to epithelial tissue that impacts the most harboring this signature. In both the TCGA and NKI-295 breast common form of cancer in women. cancer data sets, low-PRKAR1A/high-SRC tumors exhibited distinctly A major finding of this study highlights the importance of type- poor patient outcome, even when compared with low-PRKAR1A or II PKA as a critical mechanism in directing tumor cell proliferation. high-SRC designations alone. Further, our analyses suggested that This observation is contrary to what has been previously reported low-PRKAR1A/high-SRC tumors make up a notable proportion of about the roles of type-I and type-II PKA. Specifically, type-II PKA basal-like as well as HER2 cancers. Given that SRC inhibitors such as predominantly associates with well-differentiated non-prolifera- dasatinib are in clinical trials for a variety of human cancers tive cell types, while a strong interaction exists between type-I PKA including advanced breast cancers, it will be important to isozyme, cell proliferation and tumorigenesis.8 A possible explana- determine whether SRC inhibitors can provide benefits to this tion for our findings centers on the complexity of PKA holoenzyme putative novel low-PRKAR1A/high-SRC breast cancer subset. homeostasis, where the Prkar1a regulatory subunit exerts dominance over other regulatory subunits in regulating PKA activity. In our tumor model, where Prkar1a is absent in mammary MATERIALS AND METHODS epithelia, the increase in PKA activity can be attributed by either a Reagents compensatory increase in type-II PKA isozyme or by an increase The mouse mammary epithelial cell line (NMuMG) was obtained from in the unbound (unregulated) PKA catalytic subunit. In the MMTV- ATCC (CRL-1636). Ongoing cultures were maintained in DMEM containing PyMT background, hyperactive PKA activity (the result of 25 mM glucose, L-glutamine, antibiotics (100 U/ml penicillin, 100 μg/ml increased type-II PKA or free PKA catalytic subunit) may cooperate streptomycin) and supplemented with 10% FBS. 8-bromoadenosine 3′, with underlying PyMT cancer drivers (that is, Src) leading to 5′-cyclic monophosphate (cAMP), cholera toxin, hydrocortisone, insulin accelerated tumorigenesis. The importance of Prkar1b in con- and fulvestrant were purchased from Sigma Aldrich (St Louis, MO, USA). tributing to type-I PKA isozyme function cannot be overlooked in Phenol red-free, growth factor-reduced Matrigel was purchased from BD Biosciences (San Jose, CA, USA). EGF was purchased from Peprotech (Rock our mammary tumor model, even thought alterations in its fi Hill, NJ, USA). Dasatinib and wortmannin were generously provided by Drs expression were not signi cantly altered. It will be important Shereen Ezzat and Vuk Stambolic, Ontario Cancer Institute, Toronto. to determine whether human breast cancers defined by low Lapatinib was purchased from Selleck Chemicals. H89 was purchased from PRKAR1A exhibit an altered type-I/II PKA isozyme ratio. Cell Signaling Technology (Boston, MA, USA). PKI, InCELLect AKAP-St- PKA has been studied as a mediator of ERα signaling in breast Ht31inhibitor and InCELLect St-Ht31P control peptide were purchased cancer, including ligand-dependent/-independent induction and from Promega (Sunnyvale, CA, USA). For cell signaling experiments, cells tamoxifen resistance.2,5,29,39–41 Using small-molecule inhibitors, we were washed in PBS and starved in DMEM media for 24 h; cells were ruled out the involvement of ERα and EGF receptor signaling in treated with recombinant mouse EGF (10 nM) or insulin (50 nM) for 15 min. our PKA-driven mammary tumors. We also did not observe The following small molecules were used at the given concentrations: H89 (10 μM), PKI (20 μM), cAMP (1.0 mM), InCELLect AKAP-St-Ht31inhibitor and increases in expression of ERα-responsive genes Pgr or Greb1,or control peptide (5μM), fulvestrant (100 nM), lapatinib (100 nM), wortmannin Wnt reporter genes Axin2 and Tcf4. Instead we found PKA-induced (200 nM) and dasatinib (100 nM). Src signaling to be a downstream mechanism in Prkar1aΔmam cohorts. It has been shown that the development of PyMT tumors depends in part on Src activity, where mammary epithelial Mice ablation of Src significantly delays tumor onset.31 In our model, MMTV-PyMT mice and MMTV-Cre mice in the FVB background were Δmam obtained from Dr W Muller (McGill University, Canada). The generation of Prkar1a /PyMT tumors were associated with elevated phos- lox/lox phorylated Src at both serine 17 and tyrosine 416 residues. the Prkar1a mice and genotyping of the WT, loxP and deletion allele have been described.16 To generate the experimental mice, the following PKA-directed regulation of Src has previously been observed in breeding pairs were set: Prkar1alox/lox/MMTV-PYMT+/MMTV-Cre+ males fi 33,42 fi – À broblasts and adrenal cells in vitro; our ndings provide the were bred with Prkar1alox/lox/MMTV-PYMT /MMTV-Cre females. The first evidence of in vivo PKA-Src regulation. Although we did not proper Mendelian ratios were obtained. Mammary tumor onset was Δmam directly examine Src activity in Prkar1a tumors, Src RNAi determined by manual palpation of mammary glands. Tumor endpoints, experiments demonstrated the importance of Src as a determined by Animal Care Committee guidelines, were reached when

Oncogene (2015) 1160 – 1173 © 2015 Macmillan Publishers Limited PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1171 mammary tumor diameter measured 1.5 cm. All animal experiments were mix containing unlabeled PCR primers and FAM labeled TaqMan MGB approved by the Animal Care Committee, Ontario Cancer Institute. probes were used to detect expression of specific genes as listed by the catalog numbers in Supplementary Table 1. All raw data were analyzed using Sequence Detection System software Version 2.1 (Applied Biosys- Whole-mount analysis, immunohistochemistry and tems). The threshold cycle (CT) values were used to calculate relative RNA immunofluorescence expression levels. Expression levels of target genes were normalized to Thoracic mammary glands from 40- or 90-day-old mice were analyzed endogenous β-actin transcripts and compared with control MMTV-Cre or for mammary morphology by carmine-alum whole-mount staining as PyMT luminal population (relative expression = 1). described previously.44 For immunohistochemistry, 4% paraformaldehyde- fixed paraffin-embedded tissue sections were de-paraffinized in xylene, gradually rehydrated in descending concentrations of ethanol and Immunoprecipitation and immunoblot analysis subsequently treated in Borg Decloaker antigen retrieval solution (pH 9) μ for 30 min at 121 °C and 10 seconds at 90 °C using a Decloaking chamber Protein (500 g) from whole mammary gland extracted by RIPA extraction – (Biocare Medical, Concord, CA, USA). Tissue sections were stained using buffer (Tris HCl pH 7.6, 1% Triton X-100, 0.1% SDS, 1% NP-40, 1% sodium ’ deoxycholate, 5 mM EDTA, 50 mM NaCl) supplemented with 200 μM HRP-AEC tissue staining kit according to manufacturer s instructions (R&D μ Systems, Minneapolis, MN, USA). For dual immunofluorescence micro- Na3VO4, 200 M NaF, 2 mM PMSF and an appropriate dilution of Complete scopy, paraffin-imbedded sections were hydrated and subjected to Mini, EDTA-free protease inhibition cocktail tablets (Roche, Indianapolis, IN, antigen retrieval as described above. Tissues were permeabilized with USA) was subjected to immunoprecipitation overnight at 4 °C using (1) an fi 0.2% Tritin X-100 and 0.1% sodium borohydride, and blocked in 5% normal antibody that speci cally recognizes RRXS*/T*-phosphorylated PKA sub- goat serum containing 0.1 mg/ml saponin. Primary antibodies (anti-keratin strate epitopes (Cell Signaling Technology) or (2) a polyclonal antibody recognizing the R2β PKA regulatory subunit (ab38949, Abcam, Cambridge, 14 and anti-kertain 18) diluted in PBS were incubated overnight at 4 °C in a fi humidified chamber. Primary antibodies were labeled with Alexa Fluor 488 UK). Immunoprecipitated protein was puri ed following incubation with goat anti-mouse and Alexa Fluor 594 goat anti-rabbit secondary protein G plus protein A/G agarose beads (Invitrogen) for 1 h. The antibodies. Slides were mounted in DAPI mounting medium (Invitrogen, precipitated protein complexes were washed at 4 °C in RIPA buffer lacking – Carlsbad, CA, USA) and imaged using an Axio Observer inverted sodium deoxycholate and SDS and was then subjected to SDS PAGE microscope (Carl Zeiss, Jena, Germany). Antibodies used were anti- followed by immunoblotting. For immunoblotting, cells were washed in mouse antibodies against keratin 14 (Covance, Princeton, NJ, USA), keratin PBS and incubated in RIPA cell extraction buffer, for 30 min. Protein concentrations were determined using a BCA kit (Pierce Chemicals; 18 (Fitzgerald, Burlington, ON, Canada), Ki67 (Novus Biologicals, Oakville, μ – ON, Canada), PR (hPRa7; generated in-house), ERα (clone M20, Santa Cruz Rockford, IL, USA). Cell protein lysate (30 g) was resolved by SDS PAGE Biotechnology; Dallas, TX, USA) and phosphorylated CREB (p-Creb Ser133) and transferred to nitrocellulose membranes. The membranes were from Cell Signaling Technology (Beverly, MA, USA). probed using the following mouse antibodies: anti-AKAP1, anti- phosphorylated CREB (pCreb, Ser133) anti-Creb, Akt/PKB (p-Akt, Ser473), anti-Akt/PKB, anti-phosphorylated Src (Ser17), anti-phosphorylated Src Mammary epithelial cell preparation, FACS analysis and cell (Tyr416), anti-Src (all from Cell Signaling Technology). The blots were sorting stripped and re-probed with an HRP-conjugated monoclonal antibody directed against mouse β-actin (Santa Cruz Biotechnology). Single mammary cell suspensions were generated from freshly isolated pairs of fourth inguinal mammary glands of individual mice by enzymatic digestion and analyzed by flow cytometry as reported previously44 or cultured on collagen-coated tissue culture plates. Briefly, mammary glands siRNA transfection were digested for 2.5 h at 37 °C in mouse Epicult-B with 5% FBS, 750 U/ml Four ON-TARGETplus siRNAs (Thermo Scientific; LQ-040877-00-0002; 25 nM) collagenase and 250 U/ml hyaluronidase. Organoids obtained after targeting the mouse c-Src mRNA transcript (Src-si1 50-CCAAGGGC 0 0 0 vortexing were subjected to red blood cell lysis in 0.8% NH4Cl, further CUCAACGUGAA-3 ; Src-si2 5 -CCUCAGGCAUGGCGUACGU-3 ; Src-si3 dissociation in 0.25% trypsin for 2 min, 5 mg/ml dispase with 0.1 mg/ml 50-CGUCCAAGCCGCAGACUCA-30;Src-si450-GAGAACCUGGUGUGCAAAG-30) fi μ DNase I for 2 min, and ltered through a 40- m mesh to obtain single cells. were transfected into primary mammary epithelial cells using Lipofecta- All reagents were from StemCell Technologies Inc. (Vancouver, BC, Canada) mine RNAi Max transfection reagent (Life Technologies, Carlsbad, CA, USA) and antibodies were from BD Pharmingen (San Diego, CA, USA) unless according to the manufacturer’s protocol. Cells transfected with ON- otherwise stated. Cells were blocked with Fc receptor antibody, incubated TARGETplus Non-silencing siRNA#1 (NS; Cat# D-001810-01-20) or cultured with biotinylated StemSep mouse/human chimera cocktail and anti-CD31 in the presence of transfection reagent alone, served as negative controls. in order to label hematopoietic CD45+/Ter119+ cells and endothelial cells respectively, which were excluded by secondary conjugation with streptavidin-PE-Cy7 using flow cytometry. Dead cells were excluded from analysis by staining with propidium iodide (Sigma). Anti-CD49f-FITC (clone Mammary cell proliferation assays GoH3) and anti-CD24-R-PE (clone M1/69) were used to identify the Cells were seeded at 2 × 103 cells/well into opaque collagen-coated 96-well mammary epithelial cell populations. FACS analysis was performed using microplates in 100 μl of DMEM/F12 1:1 medium containing 2% charcoal- FACSCalibur (BD) and FlowJo software (Tree Star Inc., Ashland, OR, USA). stripped FBS, EGF (5 ng/ml), insulin (10 ng/ml), LA complex (5 μg/ml), Cell sorting was performed on a FACSAria (BD). The purity of sorted penicillin/streptomycin and anti-microbiotic (Wisent, St Bruno, QC, populations was routinely greater than 96%. Canada). Cells were cultured for 0, 24, 48, 72 or 96 h in the presence of cAMP (1.0 mM) or the small molecules H89, PKI, fulvestrant, lapatinib, wortmannin or dasatinib, after which cell proliferation was measured using RNA isolation and real-time PCR analysis the CellTiter-Glo Luminescent Cell Viability Assay (Promega) following the Total RNA was prepared from FACS-sorted primary mammary cell manufacturer’s instructions on a POLARstar Omega plate reader (BMG; subpopulations using the PicoPure RNA Isolation Kit (Arcturus, Carlsbad, Offenburg, Germany). In vitro Matrigel colony forming assays were CA, USA) as described in Joshi et al.44 The quality and concentration of RNA performed as described45 with modifications. Single mammary cell was determined by visualizing purified RNA samples on SyBr Green II suspensions were prepared from individual mice, and 5000 total mammary (Invitrogen)-stained formaldehyde agarose gels and by analysis with a cells from the inguinal glands of PyMT or Prkar1aΔMam /PyMT were seeded NanoDrop 2000 Spectrometer (260/280 ratio; Thermo Scientific, Waltham, onto 80 μL of pre-coated Matrigel in eight-well chamber slides and MA, USA). Isolated and purified total RNA was reverse transcribed into first cultured in DMEM/F12 1:1 medium containing 2% charcoal-stripped FBS, strand cDNA and amplified using the SMARTer PCR cDNA Synthesis Kit and 2.5% Matrigel, EGF, insulin and cholera toxin (200 ng/ml). After 24 h, culture Advantage2 PCR Kit (Clontech). Amplified cDNA aliquots (5 μl) were media was replaced with media containing the small-molecule inhibitors analyzed on ethidium bromide-stained agarose gels (1.2%) to determine described above. All untreated cells were supplemented with 1.0% DMSO. the optimal number of LD-PCR cycles for cDNA amplification that ranged Cell colonies were fixed at day 8 of culture in 4% paraformaldehyde for 30 from 18–21 cycles. Relative quantification real-time PCR (ΔΔCt) was min at 4 °C. Colonies were imaged using an inverted light microscope performed on 4 ng of cDNA generated from FACS-purified primary using a × 4 objective; colonies were counted in five random fields of view. mammary cells using an ABI PRISM 7900HT Sequence Detection System Colony size (μm2) was determined by measuring the height and width of (Applied Biosystems, Foster City, CA, USA). TaqMan gene expression assay each colony using Image J Software.

© 2015 Macmillan Publishers Limited Oncogene (2015) 1160 – 1173 PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1172 FPLC chromatography ACKNOWLEDGEMENTS All procedures were performed at 4 °C. Samples (1.5 mg protein in 500 μl We thank Megan K Barker and Paul Waterhouse for critical reading of the manuscript. PKA protein buffer) were separated using a 5 ml Bio-Scale Mini Macro-Prep We would also like to acknowledge Shareen Ezzat for sharing the small molecule DEAE cartridge (Bio-Rad, Mississauga, ON, Canada) on an Akta FPLC System Dasatinib, Vuk Stambolic for offering insight into biochemistry experiments and the (GE Healthcare, Buckinghamshire, UK) at a flow rate of 1.0 ml/min. Samples IDIBELL’s Biostatistics Unit for help in analyzing breast cancer data sets. This work was were loaded onto the column in low-salt buffer (10 mM Tris–HCl, 1 mM supported by a Canadian Breast Cancer Foundation (CBCF) grant to RK and the EDTA, 1 mM dithiothreitol, pH 7.1). The column was washed with 25 ml of Spanish Ministry of Health grant FIS-PI12/01528 and RD12/0036/0008 to MAP. AGB low-salt buffer (five column volumes), collecting 5-ml fractions. The column holds a CBCF fellowship. was then eluted with a linear NaCl gradient (0.0–0.8 M) from low-salt buffer to high-salt buffer (10 mM Tris-HCl, 1 mM EDTA, 1 mM dithiothreitol, 1.0 M NaCl, pH 7.1) over 25 ml (five column volumes), collecting 0.5-ml fractions. The column was cleaned with 25 ml (five column volumes) of high-salt REFERENCES buffer before re-equilibrating to low-salt buffer for the following run. A total of 100 μl of each fraction was precipitated with 200 μl 1 Nadella KS, Jones GN, Trimboli A, Stratakis CA, Leone G, Kirschner LS. Targeted trichloroacetic acid (50%) at 4 °C for 20 min. The precipitated fractions deletion of Prkar1a reveals a role for protein kinase A in mesenchymal-to- 68 – were centrifuged at 14 000 × g, 15 min, 4 °C. Supernatants were discarded, epithelial transition. Cancer Res 2008; : 2671 2677. the pellets were washed with 500 μl cold acetone and centrifuged again. 2 Tharakan R, Lepont P, Singleton D, Kumar R, Khan S. Phosphorylation of estrogen The acetone wash was removed, the pellets were air-dried and receptor alpha, serine residue 305 enhances activity. Mol Cell Endocrinol 2008; 295 – resuspended in Laemmli buffer (Tris–HCl 375 mM, pH 6.8; SDS 9%; glycerol :70 78. 50%; betamercaptoethanol 9%; bromophenol blue 0.03%) where samples 3 Yin Z, Jones GN, Towns WH, Zhang X, Abel ED, Binkley PF et al. Heart-specific were resolved by PAGE followed by immunoblotting using antibodies ablation of Prkar1a causes failure of heart development and myxomagenesis. directed against R1α (BD Transduction Labs, Lexington, KY, USA) and R2β Circulation 2008; 117: 1414–1422. (BD Transduction Labs). The remaining 400 μl of each fraction was 4 Almeida MQ, Stratakis CA. How does cAMP/protein kinase A signaling lead to concentrated down to 40 μl using 10 -kDa-cuttoff Microcon Centrifugal tumors in the adrenal cortex and other tissues? Mol Cell Endocrinol 2011; 336: filters (Millipore, Billerica, MA, USA). These concentrated protein samples 162–168. where then used to determine PKA activities (10 μg/reaction). 5 Kok M, Zwart W, Holm C, Fles R, Hauptmann M, Van’t Veer LJ et al. PKA-induced phosphorylation of ERα at serine 305 and high PAK1 levels is associated with sensitivity to tamoxifen in ER-positive breast cancer. Breast Cancer Res Treat 2011; PKA activity assay 125:1–12. The SignaTech protein kinase (PKA) assay system (Promega) was used to 6 Moujalled D, Weston R, Anderton H, Ninnis R, Goel P, Coley A et al. Cyclic-AMP- measure basal (untreated) and total (cAMP-treated) PKA activities from dependent protein kinase A regulates apoptosis by stabilizing the BH3-only total mammary gland protein lysates, FACS-purified mammary epithelial protein Bim. EMBO Rep 2011; 12:77–83. cells and concentrated FPLC protein fractions. Briefly, 10 μg of protein 7 Neary CL, Nesterova M, Cho YS, Cheadle C, Becker KG, Cho-Chung YS. Protein lysate containing 32P-labeled phosphate was incubated with biotinylated kinase A isozyme switching: eliciting differential cAMP signaling and tumor Kemptide, a PKA-specific substrate, in the absence or presence of 0.01 mM reversion. Oncogene 2004; 23: 8847–8856. cAMP. PKA activity specificity was determined using the PKA inhibitor PKI 8 Miller WR. Regulatory subunits of PKA and breast cancer. Ann NY Acad Sci 2002; 32 (20 μM). The P-labeled biotinylated substrate is recovered from the 968:37–48. reaction mix using perforated SAM2 Biotin Capture Membranes that bind 9 Kirschner LS, Kusewitt DF, Matyakhina L, Towns WH, Carney JA, Westphal H et al. to and immobilize the substrate. PKA activity is then measured using A mouse model for the Carney complex tumor syndrome develops neoplasia in standard scintillation to determine radioactive counts/min, reflecting the cyclic AMP-responsive tissues. Cancer Res 2005; 65:4506–4514. amount of phosphorylated substrate. All determinations of PKA activity 10 Almeida MQ, Muchow M, Boikos S, Bauer AJ, Griffin KJ, Tsang KM et al. Mouse were performed in duplicate, corrected for protein concentration, and an Prkar1a haploinsufficiency leads to an increase in tumors in the Trp53+/À or average value was calculated for each experiment. Rb1+/À backgrounds and chemically induced skin papillomas by dysregulation of the cell cycle and Wnt signaling. Hum Mol Genet 2010; 19: 1387–1398. 11 Bossis I, Stratakis CA. Minireview: PRKAR1A: normal and abnormal functions. Bioinformatics Endocrinology 2004; 145: 5452–5458. The publicly available TCGA data were downloaded from the correspond- 12 Kirschner LS, Carney JA, Pack SD, Taymans SE, Giatzakis C, Cho YS et al. Mutations ing repository (tcga-data.nci.nih.gov/tcga/tcgaDownload) on July 2012. of the gene encoding the protein kinase A type I-alpha regulatory subunit in This set contained expression data for 509 primary tumors, of which 241 patients with the Carney complex. Nat Genet 2000; 26:89–92. have also recurrence information available. The TCGA-normalized and pre- 13 Kirschner LS, Sandrini F, Monbo J, Lin JP, Carney JA, Stratakis CA. Genetic het- processed data were used in subsequent analyses. The PAM50 predictor erogeneity and spectrum of mutations of the PRKAR1A gene in patients with the 46 was applied as previously described. Of the 241 tumors set, 65 were carney complex. Hum Mol Genet 2000; 9: 3037–3046. fi classi ed as basal-like, 30 as HER2, 73 as luminal A, 57 as luminal B and 16 14 Amieux PS, Mcknight GS. The essential role of RI alpha in the maintenance of as normal-like. The raw MCF10A expression data were downloaded from regulated PKA activity. Ann NY Acad Sci 2002; 968:75–95. the Gene Expression Omnibus reference GSE3151. Data were normalized 15 Griffin KJ, Kirschner LS, Matyakhina L, Stergiopoulos SG, Robinson-White A, fi 47 and the signi cance analysis of microarrays (SAM) algorithm was applied Lenherr SM et al. A transgenic mouse bearing an antisense construct of regulatory ⩽ to identify differentially expressed probes at a 5% false discovery rate subunit type 1A of protein kinase A develops endocrine and other tumours: (over-expressed, n = 1600; under-expressed, n = 261). Gene expression and comparison with Carney complex and other PRKAR1A induced lesions. JMed clinical data for the NKI data set were downloaded from the Stanford Genet 2004; 41:923–931. Microrray Database. The PRKAR1A and SRC genes were represented by a 16 Wilson GR, Cramer A, Welman A, Knox F, Swindell R, Kawakatsu H et al. Activated single probe each in the NKI data set (NM_002734 and NM_005417, c-SRC in ductal carcinoma in situ correlates with high tumour grade, high pro- respectively). Correlation, clustering and survival analyses were carried out 95 – 48 liferation and HER2 positivity. Br J Cancer 2006; :1410 1414. using R/Bioconductor software (Seattle, WA, USA). The GSEA tool was run 17 Elsberger B, Tan BA, Mitchell TJ, Brown SBF, Mallon EA, Tovey SM et al. Is using default values for all parameters. expression or activation of Src kinase associated with cancer-specific survival in ER-, PR- and HER2-negative breast cancer patients? Am J Pathol 2009; 175: Statistics 1389–1397. 18 Molyneux SD, Di Grappa MA, Beristain AG, Mckee TD, Wai DH, Paderova J et al. Data are reported as mean ± standard error of the mean. All calculations Prkar1a is an osteosarcoma tumor suppressor that defines a molecular subclass were carried out using GraphPad Prism software (La Jolla, CA, USA). in mice. J Clin Invest 2010; 120: 3310–3325. Comparisons were made by two-tailed Student’s t-test and ANOVA. The 19 Gatza ML, Lucas JE, Barry WT, Kim JW, Wang Q, Crawford MD et al. A pathway- differences were accepted as significant at Po0.05. based classification of human breast cancer. Proc Natl Acad Sci USA 2010; 107: 6994–6999. 20 Curtis C, Shah SP, Chin S-F, Turashvili G, Rueda OM, Dunning MJ et al. The CONFLICT OF INTEREST genomic and transcriptomic architecture of 2,000 breast tumours reveals novel The authors declare no conflict of interest. subgroups. Nature 2012; 486:346–352.

Oncogene (2015) 1160 – 1173 © 2015 Macmillan Publishers Limited PKA-induced Src drives mammary tumorigenesis AG Beristain et al 1173 21 Lin EY, Jones JG, Li P, Zhu L, Whitney KD, Muller WJ et al. Progression to malig- 35 Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D et al. Oncogenic pathway nancy in the polyoma middle T oncoprotein mouse breast cancer model provides signatures in human cancers as a guide to targeted therapies. Nature 2006; 439: a reliable model for human diseases. Am J Pathol 2003; 163: 2113–2126. 353–357. 22 Guy CT, Webster MA, Schaller M, Parsons TJ, Cardiff RD, Muller WJ. Expression of 36 van de Vijver MJ, He YD, Van’t Veer LJ, Dai H, Hart AAM, Voskuil DW et al. A gene- the neu protooncogene in the mammary epithelium of transgenic mice induces expression signature as a predictor of survival in breast cancer. New Engl J Med metastatic disease. Proc Natl Acad Sci USA 1992; 89: 10578–10582. 2002; 347: 1999–2009. 23 Griffin KJ, Kirschner LS, Matyakhina L, Stergiopoulos S, Robinson-White A, 37 Pavel E, Nadella K, Towns WH, Kirschner LS. Mutation of Prkar1a causes osteoblast Lenherr S et al. Down-regulation of regulatory subunit type 1A of protein kinase A neoplasia driven by dysregulation of protein kinase A. Mol Endocrinol 2008; 22: leads to endocrine and other tumors. Cancer Res 2004; 64: 8811–8815. 430–440. 24 Smalley MJ, Kendrick H, Sheridan JM, Regan JL, Prater MD, Lindeman GJ et al. 38 Carney JA, Toorkey BC. Ductal adenoma of the breast with tubular features. Isolation of mouse mammary epithelial subpopulations: a comparison of leading A probable component of the complex of myxomas, spotty pigmentation, methods. J Mammary Gland Biol Neoplasia 2012; 17:91–97. endocrine overactivity, and schwannomas. Am J Surg Pathol 1991; 15: 25 Fink MA, Zakhary DR, Mackey JA, Desnoyer RW, Apperson-Hansen C, Damron DS 722–731. et al. AKAP-mediated targeting of protein kinase a regulates contractility in 39 Michalides R, Griekspoor A, Balkenende A, Verwoerd D, Janssen L, Jalink K et al. cardiac myocytes. Circulation Res 2001; 88:291–297. Tamoxifen resistance by a conformational arrest of the estrogen receptor alpha 26 Burton KA, Johnson BD, Hausken ZE, Westenbroek RE, Idzerda RL, Scheuer T et al. after PKA activation in breast cancer. Cancer Cell 2004; 5: 597–605. Type II regulatory subunits are not required for the anchoring-dependent mod- 40 Subbaramaiah K, Morris PG, Zhou XK, Morrow M, Du B, Giri D et al. Increased ulation of Ca2+ channel activity by cAMP-dependent protein kinase. Proc Natl levels of COX-2 and prostaglandin E2 contribute to elevated aromatase Acad Sci USA 1997; 94: 11067–11072. expression in inflamed breast tissue of obese women. Cancer Discov 2012; 2: 27 Lim CJ, Han J, Yousefi N, Ma Y, Amieux PS, Mcknight GS et al. Alpha4 integrins are 356–365. type I cAMP-dependent protein kinase-anchoring proteins. Nat Cell Biol 2007; 9: 41 Subbaramaiah K, Howe LR, Bhardwaj P, Du B, Gravaghi C, Yantiss RK et al. Obesity 415–421. is associated with inflammation and elevated aromatase expression in the mouse 28 Gold MG, Fowler DM, Means CK, Pawson CT, Stephany JJ, Langeberg LK et al. mammary gland. Cancer Prev Res 2011; 4:329–346. Engineering A-kinase anchoring protein (AKAP)-selective regulatory subunits of 42 Schmitt JM, Stork PJS. PKA phosphorylation of Src mediates cAMP's inhibition of protein kinase A (PKA) through structure-based phage selection. J Biol Chem 2013; cell growth via Rap1. Mol Cell 2002; 9:85–94. 288: 17111–17121. 43 Prat A, Parker JS, Fan C, Perou CM. PAM50 assay and the three-gene model for 29 Subbaramaiah K, Hudis C, Chang S-H, Hla T, Dannenberg AJ. EP2 and EP4 identifying the major and clinically relevant molecular subtypes of breast cancer. receptors regulate aromatase expression in human adipocytes and breast cancer Breast Cancer Res Treat 2012; 135:301–306. cells. Evidence of a BRCA1 and p300 exchange. J Biol Chem 2008; 283: 3433–3444. 44 Joshi PA, Jackson HW, Beristain AG, Di Grappa MA, Mote PA, Clarke CL et al. 30 Baker MA, Hetherington L, Aitken RJ. Identification of SRC as a key PKA-stimulated Progesterone induces adult mammary stem cell expansion. Nature 2010; 465: tyrosine kinase involved in the capacitation-associated hyperactivation of murine 803–807. spermatozoa. J Cell Sci 2006; 119: 3182–3192. 45 Muthuswamy SK, Li D, Lelievre S, Bissell MJ, Brugge JS. ErbB2 but not ErbB1, 31 Marcotte R, Smith HW, Sanguin-Gendreau V, McDonough RV, Muller WJ. reinitiates proliferation and induces luminal repopulation in epithelial acini. Nat Mammary epithelial-specific disruption of c-Src impairs cell cycle progression and Cell Biol 2001; 3:785–792. tumorigenesis. Proc Natl Acad Sci USA 2012; 109:2808–2813. 46 Parker JS, Mullins M, Cheang MCU, Leung S, Voduc D, Vickery T et al. Supervised 32 Soulard A, Cremonesi A, Moes S, Schutz F, Jenö P, Hall MN. The rapamycin- risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol 2009; 27: sensitive phosphoproteome reveals that TOR controls protein kinase A toward 1160–1167. some but not all substrates. Mol Biol Cell 2010; 21: 3475–3486. 47 Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays 33 Obara Y, Labudda K, Dillon TJ, Stork PJS. PKA phosphorylation of Src mediates applied to the ionizing radiation response. Proc Natl Acad Sci USA 2001; 98: Rap1 activation in NGF and cAMP signaling in PC12 cells. J Cell Sci 2004; 117: 5116–5121. 6085–6094. 48 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA et al. 34 Cancer Genome Atlas Network, Comprehensive molecular portraits of human Gene set enrichment analysis: a knowledge-based approach for interpreting breast tumours. Nature 2012; 490:61–70. genome-wide expression profiles. Proc Natl Acad Sci USA 2005; 102: 15545–15550.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)