The hormone receptor positive breast cancer cell line T47D^M (CLS Cat# 300353/p525_T-47D, RRID : CVCL_0553) was used in all experiments unless otherwise stated. T47D^M cells were routinely grown in RPMI (Supplemented with 10% foetal bovine serum (FBS), penicillin/streptomycin (pen/strep), L-glutamine (L-glut) as previously described (21). For hormone induction experiments, cells were seeded at a concentration of 5 x 10⁶ per 150mm cell culture dish in RPM1 white (15% charcoal stripped FBS, Pen/strep, L-glut) for 48 hours. 16 hours prior to hormone induction for time indicated (10nM R5020), the medium was replaced with RPM1 white (0% FBS, Pen/strep, L-glut). Samples were harvested at the time points indicated.

The total protein content of the samples was calculated prior to antibody array, mass spec or western blotting analysis using BCA assay (Thermo Fisher, catalogue number 23227) according to manufacturer’s instructions.

Changes in phosphorylation sites within individual proteins identified was confirmed by western blotting as previously described (19) using specific antibodies; Progesterone receptor (PGR) phospho-S162 (Abcam Cat# ab58564, RRID : AB_883089), and as a loading control, total PGR (Santa Cruz Biotechnology Cat# sc-7208, RRID : AB_2164331), total CDK2 (Santa Cruz Biotechnology Cat# sc-6248, RRID : AB_627238) or CDK2 phospho-T160 (Abcam Cat# ab47330, RRID : AB_869087).

Phosphorylation antibody array analysis was carried out by Kinexus™ using Kinexus ™ Antibody Microarray (KAM) technology. For each time point 3 biological replicates were prepared independently. For each replicate, 50ug of protein lysate was prepared and samples prepared by Kinexus™ in house (Kinexus Bioinformatics Corporation, RRID : SCR_012553). Signal quantification was performed using ImaGene 8.0 (ImaGene, RRID : SCR_002178) from BioDiscovery (BioDiscovery, RRID : SCR_004557). Background corrected raw intensity data was logarithmically transformed with base 2 and Z scores calculated (29). Any poor-quality spots based on morphology and/or background, were flagged as unreliable and removed from any subsequent analysis.

Samples (triplicates from independent experiments, per time point) were precipitated with 6 volumes of cold acetone and the pellet was dissolved in 6M Urea/200mM ammonium bicarbonate at a concentration of 1 µg/µL. 260 µg of each sample was reduced with dithiothreitol (DTT, 10mM, 37°C, 60 min) and alkylated with iodoacetamide (IAM, 20mM, 25°, 30 min). The resulting protein extract was first diluted to 2M urea with 200 mM ammonium bicarbonate for digestion with endoproteinase LysC (1:10 w:w, 37°C, o/n, Wako, cat # 129-02541), and then diluted 2-fold with 200 mM ammonium bicarbonate for trypsin digestion (1:10 w:w, 37°C, 8h, Promega cat # V5113). Digested peptides were subjected to phosphopeptide enrichment using the “TiO2 Phosphopeptide Enrichment and Clean-up Kit (Pierce)”

45% of each enriched sample was analyzed using an Orbitrap Fusion Lumos with an EASY-Spray nanosource coupled to a nano-UPLC system (EASY-nanoLC 1000 liquid chromatograph) equipped with a 50-cm C18 column (EASY-Spray; 75 µm id, PepMap RSLC C18, 2 µm particles, 45 °C). Chromatographic gradients started at 5% buffer B with a flow rate of 300 nL/min and gradually increased to 22% buffer B in 105 min and to 32% in 10 minutes. After each analysis, the column was washed for 10 min with 95% buffer B (Buffer A: 0.1% formic acid in water. Buffer B: 0.1% formic acid in acetonitrile).

The mass spectrometer was operated in data-dependent acquisition mode, with full MS scans over a mass range of m/z 350–1500 with detection in the Orbitrap (120K resolution) and with auto gain control (AGC) set to 100,000. In each cycle of data-dependent acquisition analysis, following each survey scan, the most intense ions above a threshold ion count of 10,000 were selected for fragmentation with HCD at normalized collision energy of 28%. The number of selected precursor ions for fragmentation was determined by the “Top Speed” acquisition algorithm (maximum cycle time of 3 seconds), and a dynamic exclusion of 60 s was set. Fragment ion spectra were acquired in the ion trap with an AGC of 4000 and a maximum injection time of 300 ms.

Raw files, quantified peptides, proteins and phosphosites (5% FDR) are deposited and publicly available in the UCSD Data depository, MassIVE (Project ID: MSV000089820, ftp://MSV000089820@massive.ucsd.edu).

Acquired data were analyzed using the Proteome Discoverer software suite (v2.0, Thermo Fisher Scientific), and the Mascot search engine (v2.5.1, Matrix Science) was used for peptide identification. Data were searched against a Human protein database derived from SwissProt plus the most common contaminants. A precursor ion mass tolerance of 7 ppm at the MS1 level was used, and up to three missed cleavages for trypsin were allowed. The fragment ion mass tolerance was set to 0.5Da. Oxidation of Methionine, N-terminal protein acetylation and phosphorylation in Serine, Threonine and Tyrosine were defined as variable modification and carbamidomethylation of Cysteines was set as fixed modification. The identified peptides were filtered by 5%FDR. Peptide areas were obtained with the area under from extracted ion chromatogram using the “Precursor Ions Area Detector” node from Proteome Discoverer.

Gene Ontology (GO) GO-Biological process (GO-BP), GO-Molecular Process (GO-MF), KEGG (KEGG, RRID : SCR_012773) and Biocarta (BioCarta Pathways, RRID : SCR_006917) pathway analysis of networks was carried out using GeneMania application (GeneMANIA, RRID : SCR_005709) (30) within Cytoscape (Cytoscape, RRID : SCR_003032). Clustering analysis, similarity analysis was carried out using GeneE. Network analysis was performed using Cytoscape v3.5 (31). The initial prior knowledge network (PKN) was generated based on known protein-protein interactions only validated experimentally. Significantly enriched pathways were analyzed within the network using CytoKEGG application within Cytoscape. The parent network and each of the individual pathway networks are available for visualisation and further analysis (cytoscape session link Supplementary Materials). Comprehensive resource of mammalian protein complexes (Corum analysis) was carried out using online tool http://mips.helmholtz-muenchen.de/corum/CORUM, RRID : SCR_002254 (Giurgiu et al., 2018). Functional classification GO biological process (BP), molecular function (MF) and cellular component (CC) were carried out using molecular signatures database (MSigD) within Gene Set Enrichment (GSEA) tool (Gene Set Enrichment Analysis, RRID : SCR_003199) and terms with a p value of less than 0.001 were considered significantly enriched (32, 33).

Analysis of the overall survival of breast cancer patients using a Kaplan-Meier plot were carried out using KMPlotter (34), https://kmplot.com/analysis/to assess the correlation between the expression of the gene (mRNA) and survival in 4929 breast cancer patient samples. Sources for the databases include GEO, EGA, and TCGA. All patients’ samples were included in the analysis shown, i.e ER, PR status, subtype, lymph node status and grade. Analysis of protein expression levels in breast tumor versus normal samples were representative of those within the Human Protein Atlas database (35) http://www.proteinatlas.org.

Before starting to add quantitative dynamic data to the already existing knowledge of progesterone signalling events in T47D cells, we have generated a “prior knowledge network (PKN)”, based on the published literature (Figures S1A, B, and Supplementary Material; Cytoscape Session 1). Each protein-protein interaction is characterized based on type (interaction, phosphorylation or dissociation) and is displayed as a unique edge. The corresponding literature is given in Figure S1B and within the Cytoscape session. The PKN already shows the key role played by kinases in response to progestins. First, progestins via ERa activate SRC1 that phosphorylates MAPKK1, that activates ERK1, that phosphorylates PGR, resulting in dissociation from the HSP90A and B proteins (36, 37). Activated ERK1 also phosphorylates ERa at S118 (38). ERK1 in association with hormone receptors translocates to the cell nucleus where it phosphorylates MSK1 (39), leading to the formation of an active complex PR-ERK-MSK1 that interacts with chromatin containing accessible PRE. Activated PR also interacts with PLK1 that activates MLL2 (40), with CDK2 that phosphorylates and activates ARTD1 (7), and with JAK2 that activates STAT5 (16). Simultaneously, membrane activated SRC1, also activates RAS and EGFR (41), which feeds back activating the MAPK cascade. Membrane associated ERa also activates PI3K and cAMP, which upon binding with AKT and PKA respectively lead to the activation (via interaction and direct phosphorylation) of GSK3, mTOR (42, 43) and the arginine methyltransferases CARM1 and PRMT1 within the nucleus (44–46). This brief description of the PKN shows that it already encompasses a great degree of complexity and complementary connections that need additional data to be resolved.

Our plan was to combine antibody microarray technology and shotgun phosphoproteomics in T47D^M breast cancer cells exposed to 10 nM R5020 for different lengths of time, as previously described (7). For each experimental approach and exposure time total protein extracts were harvested in triplicate. For the antibody arrays, data was collected, filtered for quality control and summarized as log₂ ratio over time zero (as described in materials and methods, Figure S2A). This dataset provided 246 unique phosphorylation sites corresponding to 155 proteins (Figure 1A; Supplementary Table S1). The majority of proteins contain 1 phosphorylation site, although for several proteins (Tau, RB1, MAP2K1, PTK2 and the protein kinase RPS6KA1) 7 or more significantly regulated phosphorylation sites were identified (Figure S2B).

Analysis of the number of phosphorylation sites clearly showed a rapid activation already 1-minute following hormone (68 significant phosphorylation events Figure 1B). Signalling persists throughout the time course showing two peaks at 30- and 360-minutes following hormone (Figure 1B). Phosphorylation sites were characterized as up or down-regulated, using a threshold for the log₂ fold change with respect to time 0 of -0.6< or >0.6 respectively (Figure 1C). We see a trend for early phosphorylation sites to be dynamically increased compared to time zero, in contrast to later time points where protein phosphorylation sites as a whole decrease compared to time zero (Figure 1C). The majority of phosphorylation sites identified belong to protein kinases (45%), co-factors (11%), transcription factors (18%) and structural proteins (13%). (Figure 1D). Combining the identified phosphorylation sites over the time course reveals that the majority of sites are regulated at more than one time point (Figure 1E), however the protein function enrichment does not alter significantly over time, with kinases and transcription factors being the main protein groups where the phosphorylation sites are observed (Figure S2C).

Pathway and gene ontology (GO) for biological function (BP) molecular function (MF) analysis revealed a significant increase in Cancer pathways (Figure S2D), signal transduction, biopolymer metabolic process and kinase activity (Figures 2SE and F). Within this dataset we observed a strongly upregulated phosphorylation of the MAPK Signal-Integrating Kinase 1, MNK1 at T250/T255 (Figure 1F) in T47D in response to progesterone stimulation. Phosphorylation of MNK1 at T250/T255 by ERK induces the activity of MNK1 (47). Once activated, MNK1 phosphorylates its targets, including the proto-oncogene Eukaryotic Translation Initiation Factor 4E (EIF4E), for which we also observed a modest phosphorylation which follows a similar pattern to MNK1 (Figure 1F). Activation of MNK1 has been shown to promote cell proliferation thus MNK1 inhibitors appear as an exciting opportunity for cancer therapy. MNK1 signalling play a key role in invasive breast cancer growth (48), MNK1 inhibitors have been shown to block breast cancer proliferation in multiple cell lines (49), and its downstream target EIF4E is overexpressed in tumor versus normal samples from breast cancer patients (Figure 1G) and associated with a poor overall survival (Figure 1H). Our results are the first indication that MNK1 activation may be relevant for progesterone induced breast cancer cell proliferation.

CDK2 plays an important role in progesterone signalling, activating ARTD1, and phosphorylating histone H1 (7). CDK2 activity is controlled by the formation of an active complex with the cyclin partner; either Cyclin E or A. In addition to binding the cyclin partner, CDKs are also controlled via interactions with Kinase Inhibitory Proteins (KIPs). p27/KIP is rapidly dephosphorylated at T187 in response to hormone, dropping sharply at 1 minute after hormone exposure (Figure 1J), when CDK2 is phosphorylated and activated. Phosphorylation of p27 at T187 results in its ubiquitination and degradation and inhibits the interaction with CDK2 (50), which would result in the release of CDK2 from the inhibitory protein complex resulting in the activation of ARTD1 and subsequent nuclear effects. In addition, we observe the coordinated activation of the upstream kinase of CDK2 at T160; ERK at Y202/204 (Figure 1K) and could validate the phosphorylation of CDK2 T160 and its dependence on ERK activity by western blotting in the presence of ERK inhibition (Figure 1L). CDK2 at T160 is the active phosphorylation site of CDK2 peaking at 1-minute following hormone exposure (Figure 1K) in contrast to the inactive phosphorylation site of CDK2 (T14/Y15) which peaks at 60 minutes following hormone exposure to silence the kinase (Figure 1M). The activity of the phosphatase CDC25C is key for the removal of the inhibitory T14/Y15 phosphorylation sites of CDK2. The phosphatase itself is inactivated by phosphorylation at S216. We observe a peak in CDC25C phosphorylation prior to and following 60 minutes of hormone exposure, which would permit the phosphorylation of the inhibitory phosphorylation site in CDK2 (Figures 1M, N). Going one step further within this pathway; MAPKAPK2, the kinase which phosphorylates CDC25C at S216 is activated following the same time dynamic as its target (Figure 1O). Although the importance of CDK2 in progestin induced cell proliferation has been studied (7, 51) the complex mechanism of CDK2 activation; phosphorylation of active/inactive marks, activation and regulation of upstream phosphatases and kinases was not clear until now (Figure 1P). These examples of the dynamic phosphorylation of MNK1 and CDK2 highlight the insight that can be gained by this type of global signalling datasets.

To complement the microarray dataset, we performed shotgun phosphoproteomic analysis using mass spec. Phospho-peptides from T47D^M cells exposed to 10 nM R5020 for the same duration as in the array experiments, were enriched using TiO2 and phosphorylated peptides identified by LC-MS-MS (Figure 2A; Figure S3A; Supplementary Table S1, S2). We identified 7482 phosphopeptides with a FDR 5% (Table S2) and dynamic, significant changes in 310 unique phosphorylation sites within 264 unique proteins (Figures 2B, C). The majority of proteins exhibited regulation of a single phosphosite, except for the serine/arginine repetitive matrix protein, SRRM1, involved in mRNA processing and the TP53 enhancing protein TP53BP1, that exhibited 8 and 10 regulated phosphorylation sites respectively (Figure S3B). Most phosphorylation sites identified were phosphor-serine consistent with the biological ratio of residue specific phosphorylation (Figure 2D). Over the time course, changes at each time point were identified as either up (log₂FC>0.6) or down (log₂FC<-0.6) regulated (Figure 2E). Up-regulated sites prevailed at early time points and many of these phosphorylation sites were significantly regulated at more than one time point (Figure 2F). Pathway and GO-BP (Biological Process) and MF (Molecular Function) enrichment analysis was consistent with the antibody array enrichment and revealed an increase in pathways in cancer, biopolymer metabolic process and kinase activity (Figures S3C–E).

PR S294 is rapidly phosphorylated in response to hormone resulting in its activation and dissociation from chaperone complexes and increased protein turnover (52). In recent years it has been shown that clinical samples assigned as “PR low” actually have elevated levels of phosphorylated PR S294 and that this phosphorylation is associated with a genetic signature linked to cancer stem cell growth and increased recurrence which may have implications for the treatment of PR low patients with anti-progestins (53). Phosphorylation of PR S162 in the hinge region showed a strong hormone induced increase by mass spec (Figures 2G, H). Phosphorylation within this region of PGR has been previously reported to be mediated by CDK2 (54), which we were able to confirm as the specific phosphorylation of S162 PR in response to progesterone was strongly decreased in the presence of CDK2 inhibition (Figure 2H).

In order to investigate the dynamics of progestin signalling over time and with the aim of avoiding inherent biases generated from either technical approach, we combined the significantly regulated phosphorylation sites from both datasets (Figures 1, 2) resulting in a list of 420 unique phosphorylation sites within 390 proteins (Figure S4A). PCA analysis of the samples reveals a clear separation of the phosphorylation data at 6 hours following the hormone, given the majority of phosphorylation sites are rapid, this separation of the latest time point may reveal changes in protein abundance at this time point. The majority of these proteins showed the regulation of a single phosphorylation event with the exception of several highlighted proteins, including FAK, MAPT and EGFR (Figure S4C). As in the individual analysis, phosphorylation sites were significantly regulated over several time points (Figure S4D) and showed a switch from up-regulated sites early after hormone exposure to down-regulated sites at later time points (Figure S4D). KEGG pathway analysis shows a significant enrichment in MAPK, PI3K-AKT, neurotrophin (TRK) and ERBB signalling pathways (Figure 3A; Supplementary Table S2), in addition to pathways key in the progression of cancer, specifically cancer stem cells, such as focal adhesion (Figure 3A).

GO cellular component analysis reveals a dynamic pattern of specific cellular compartments over time (Figure 3B; Supplementary Table S3). As expected, over the whole-time course, proteins are mainly found within the cytosol and nucleoplasm. However, prior to hormone exposure, phosphorylated proteins are enriched in RNA transcription repression complex and nuclear chromatin. The addition of hormone rapidly induces the phosphorylation of the membrane rafts, components of focal adhesion and protein kinases consistent with published works whereby signalling initiates from the plasma membrane. This transient phosphorylation of the membrane rafts diminish after 1 minute and is followed by the phosphorylation of transcription factors and proteins within the cytoskeleton (Figure 3B). Interestingly, in line with our findings showing the generation of nuclear ATP synthesis independent of mitochondrial supplementation at 30 minutes after hormone we observe an enrichment in phosphorylated proteins located within the mitochondrial membrane at 15 minutes (Figure 3B). The dynamic regulation of these proteins; CYB5B (cytochrome b5), the transcriptional activator ATF2 (Cyclic AMP-dependent transcription factor ATF-2), RPS6KB1 (Ribosomal protein S6 kinase beta-1) and PI4KB (Phosphatidylinositol 4-kinase beta) (Figure S4F) may suggest an as yet undiscovered crosstalk between the nuclear and mitochondrial ATP synthesis pathways. Mitochondrial PR (PR-M) is a truncated isoform of the nuclear progesterone receptors PRB and PRA, which lacks the N-terminal DNA binding domain present in PRA and PRB but does contain the hinge region responsible for dimerization and the ligand binding domain (55). PR-M has been shown to increase cellular respiration hence cell energy levels in response to ligand in various physiological situations and animal models (56). Therefore, the coordinated phosphorylation of proteins within the mitochondria in response to ligand (Figure S4F) may provide an interesting insight into a possible crosstalk between mitochondrial PR-M and the nuclear receptors PRA and PRB.

At 60 minutes following hormone exposure the main localization of phosphorylation changes and shifts again to nuclear matrix proteins and proteins found within distinct regions of the nucleus, such as PML bodies (Figure 3B group IV), which may be involved in the reorganization of chromatin in response to progestins (22). At 6 hours following hormone exposure cells enter the early stages of the cell cycle and cellular movement is increased. This is also evident by the enrichment of phosphorylation sites in proteins within cell-cell junctions, the cytoskeleton and microtubules (Figure 3B group V).

The majority of identified phosphorylated proteins (60%) were assigned to one class, however due to the promiscuous nature of enzymes nearly 40% were assigned to more than one class (Figure S4G). Taking only the parent class into account, we observed 5 distinct functions; 1) nucleic acid binding, 2) enzymes, 3) structural proteins, 4) protein modulators, and 5) proteins involved in signalling, membrane and cell-cell contacts (Figure S4H). Each function class consists of sub-groups (Figures S5A–F). The Nucleic Acid binding class includes DNA binding proteins, helicases, nucleases and RNA binding protein subgroups (Figure S5A). The enzyme class is dominated by kinases but also includes histone modifying enzymes, hydrolases, ligases and oxidoreductases (Figure S5D). The structural class is dominated by cytoskeleton proteins (Figure S5F). The protein modulator class includes chaperones and various kinases and G proteins regulators (Figure S5C). The cell signalling and the membrane/cell-cell contact classes are more complex and include many specialized proteins such as signalling molecules, receptors and transporters (Figures S5B, E).

Gene Ontology of Biological Processes (GO-BP) and Molecular Function (GO-MF) showed an enrichment in signal transduction and general biological processes across the entire time course (Figures S6A, B; Supplementary Tables S4, S5). However, several interesting dynamic functions were identified. For instance, transcription cofactors, transcriptional repressors and transcription factor binding were already enriched 1 minute after hormone exposure (Figure S6B) consistent with our previous observations of rapid transcription factor recruitment following hormone exposure (9, 19). We observed enrichment in ATP binding and Adenyl-ribonucleotide binding after 5 and 60 minutes of hormone exposure (Figure S6B), which may represent regulation of the two cycles of ATP dependent chromatin modifiers in response to progesterone (21, 57). KEGG pathway analysis reveals a significant enrichment in signalling and in many cancer pathways, including Prostate, Glioma, CML, lung, AML, endometrial and pancreatic cancer, as well as focal adhesion and tight junctions (Figure S6C). Annotated signalling cascades were significantly enriched at all time points in response to progestin, including MAPK, neurotrophin (TRK), ERBB, FC-receptor and insulin signalling (Figure S6C).

K Means clustering analysis revealed six patterns of regulation over the time course (Figure 3C). Similarity analysis of all phosphorylation sites within all clusters shows several interesting dynamics. First, “Early-risers” cluster 1 and 4, are positively correlated on the similarity matrix and show their initial increase in phosphorylation early at 1 and 5 minutes, respectively (Figure 3D). GO-BP analysis of the proteins contained within the “early riser” clusters shows an enrichment in signal regulation, and signalling cascades including Hippo, NFk-B and MAPK pathways (Figure 3E; Supplementary Table S9). Second, clusters 3 and 6 show an opposing nature (negative correlation Figure 3D). This antagonistic behavior of the two clusters is clearly shown averaging the signal of all phosphorylation within each cluster (Figure 3F). Corum (comprehensive resource of mammalian protein complexes) analysis of the significantly enriched protein complexes contained within clusters 3 and 6 (Supplementary Table S6) showed that most protein complexes were enriched in one cluster or the other (Figure 3G), likely representing crosstalk. Ten protein complexes were found to be enriched in both cluster 3 and 6, having phosphorylation sites within the same protein complex regulated in an opposite manner (Figure 3G).

One such complex was the cMyc-ATPase-Helicase complex, which contains 5 proteins; cMyc, the chromatin remodelling component BAF53, the ATP-dependent helicases RUVBL1 and RUVBL2 (also known as TIP48 and 49) and the histone acetyltransferase, TRRAP. This complex is involved in chromatin organization, histone acetylation and transcriptional regulation (58). Analysis of the phosphorylation sites showed that two sites (S373, T58) within Myc were increased early after hormone exposure, and decreased after 60 minutes, whereas one site of BAF53 (S233) shows the opposite dynamic (Figure 3H). Database analysis also reveals a strong overexpression of BAF53 in tumor versus normal samples in multiple cancer types (Figure 3I). Myc has an important role in breast cancer growth via the activation of AMPK (59).

The AMP-activated protein kinase (AMPK), exhibited a decrease in T183 phosphorylation in response to hormone. This site is phosphorylated by CAMKK1 or 2 (60). AMPK is a master sensor, and its activation inhibits several kinase pathways including mTOR, NfkB, JAK/STAT, insulin and Hippo (61–64). In addition, active AMPK inhibits the phosphorylation of PR S294, PR recruitment to chromatin and the activation of progesterone regulated genes (65). Activation of the kinase, specifically requires the phosphorylation of AMPK at T183 by CAMKKs, and de-phosphorylation of this site has been shown to be induced by estrogens and androgens in adipocytes (66, 67). Previous published results and the data presented here suggests a model where AMPK must be silenced in order for regulatory pathways described and PR itself to be active, which is what we observe within 1 minute of progesterone stimulation (Figures 3J, K).

Further in-depth analysis of the complexes which are regulated by phosphorylation in response to progestin revealed a full list of complexes with at least 2 proteins phosphorylated in response to hormone. The Sam68-p85 P13K-IRS-1-IR signalling complex, which encompasses the insulin receptor (INSR), the insulin receptor substrate 1 (IRS1), the KH domain containing transduction-associated protein 1 (Sam68) and the phosphatidylinositol 3-kinase regulatory subunit alpha (GRB1). This protein complex is involved in insulin signalling and has been proposed to provide a link between the PI3K pathway and other signalling cascades of insulin or p21/RAS (68). We observed a dynamic phosphorylation of several sites within the complex (Figure 4A), including 4 distinct phosphorylation events within IRS1, two of which peak at 1 minute (S312, S639) and two sites where the peak in phosphorylation is observed at 60 minutes (Y1179, and Y612). S312 has been shown to be directly phosphorylated by c-Jun N-terminal kinase (JNK1) in breast cancer signalling (69) and this phosphorylation inhibits its interaction with IKKA. S639 is phosphorylated by mTOR has been linked to PI3K/Akt/mTOR signalling in breast cancer (70, 71) and effects the intracellular localization of IRS1 (72). Y1179 has been reported to be phosphorylated by IGF1R or INSR itself (73) and Y612 phosphorylation activates the interaction with PIK3R1 (74).

The phosphorylation of several components of the TNFa/NFkB signalling complex were also identified (Figure 4B). This complex is involved in I-kB kinase/NF-kB signalling in tumor progression. Indeed, complex components IKKα, RelB and p52 are associated with decreased cancer-specific survival in ERa-positive breast cancer (75). This may be linked to the cancer stem cell niche, which we showed recently was present in T47D cells grown in 3D cultures (24). NFkB regulates self-renewal in breast cancer stem cell (BCSC) models and deletion of IKKα in mammary-gland epithelial cells affects progestin-driven breast cancer (76, 77). Indeed, the upstream activator RANK ligand (RANKL) and hence the RANK pathway promotes mammary tumor formation, (78), (76). Another example is the P130Cas-ER-cSrc-PIK3 kinase complex (Figure 4C). Which has been shown to induce transcriptional changes in response to oestrogen and mammary proliferation in breast cancer. The authors showed that estradiol triggers the association of ERa, c-Src, the p85 subunit of PI 3-kinase (PI3K) and p130Cas in a macromolecular complex and activates the c-Src kinase leading to p130Cas-dependent Erk1/2 phosphorylation (79, 80). Given the similarity of the phosphorylation dynamics, peaking early at 5 and 15 minutes across Src, PIK3 and ESR1 within the complex induced by progestin (Figure 4C right panel) this may (similarly to the induction by oestrogen shown by others) present a novel ERa-ERK-cSrc activation mechanism in response to progestin in breast cancer cells.

Recently, it has been shown that the majority of PARylation events on eukaryotic nuclear proteins take place on serine residues rather than acidic residues as previously accepted (81–86). Given the enrichment of serine in the phosphorylation dataset (Figure 2D) and the importance of both PTMs in progestin gene regulation (21), we investigated the overlap between PARylation sites and phosphorylation sites within protein complexes. We identified 52 proteins, which were both phosphorylated and PARylated in response to progestins in T47D breast cancer cells (Figure 4D). Cellular component analysis of this set of 52 proteins indicates a significant enrichment in nuclear, cytoskeleton and chromatin associated proteins (Figure 4E; Supplementary Table S7), in line with the well described role of PAR in the nucleus, nuclear organization, chromatin organization, relaxation and transcriptional regulation (87–90). This finding may indicate a crosstalk between PARylation and phosphorylation with regards to nuclear structure and chromatin organization. Analysis of complexes significantly enriched within this group of proteins revealed 12 protein complexes (Supplementary Table S8), which contained proteins both PARylated and phosphorylated. One of them is the KSR1-RAF1-MEK complex composed of MEK1 and 2, both PARylated and phosphorylated, and RAF1 which is phosphorylated (Figure 4F). This complex is involved in the MAKPKKK cascade, and in response to EGF it activates BRAF mediated phosphorylation of MEK1, at 3 sites, and MEK2, which activate MAPK1 and 3 (91). In our dataset we observe a clear change in phosphorylation of all members of the complex in response to progestin (Figure 4F right panel).

We also observed the phosphorylation and PARylation of the Emerin complex (Figure 4G). This complex is involved in DNA replication, transcription and structural integrity of the nucleus, specifically of the inner nuclear membrane (92). Depletion of Emerin results in changes in the organization and dynamics of the nucleus, increased chromatin mobility and a mis-localization of chromosome territories (93). Within this complex we find proteins phosphorylated, PARylated, or phosphorylated and PARylated (Figure 4G). Given the role of PAR in the structure of the nucleus, this complex may present an interesting example for studying the PARylation, phosphorylation crosstalk.

As discussed, prior to hormone exposure PR is present in an inactive complex with the HSP70 and 90 proteins as part of the Kinase Maturation Complex. We know that progestins promote the phosphorylation and dimerization of the receptor and we found that phosphorylation of the HSP90 and 70, along with other members of the complex, is initiated within 1 minute of hormone exposure (Figure 4H), again showing a rapid and concerted phosphorylation of several members of the complex (Figure 4H). In addition, the HSPs are also PARylated as compared to other components where only phosphorylation (MARK2, MAP2K5) or PARylation (14-3-3 components) are present (Figures 4I, J). Further investigation regarding the crosstalk between PARylation and phosphorylation within protein complexes will be the focus of future studies.

In order to understand the crosstalk between the signalling pathways activated by Progesterone in breast cancer cells, a Protein-Protein Interaction (PPI) network was generated using all identified phosphorylation sites (Supplementary File: Network session 2), based on known PPI (evidence based). The resulting network consists of 427 nodes (proteins) and 4309 unique interactions (edges) (Figure S7A). Pathway analysis was carried out on this network, using Genemania™, and 23 statistically significant (p<0.01) pathways were identified (Supplementary Tables S9 and S10). The proteins and interactions (nodes and edges) associated with each pathway were selected and new networks generated (Supplementary File: Network session 2). Several of which are shown in Figures S7B–G and discussed briefly below.

One such pathway, the Fc receptor signalling pathway was enriched (Figure S6C) and the PPI network is shown in Figure S7B. Fc receptors are cell surface proteins that recognize the FC fragment of antibodies, mainly on immune cells. However, recent studies have shown that different subsets of Fc receptors may play a role in tumor cells (94). In particular, it was shown that T47D cells express the FcyRI (CD64). These FC-receptor expressing breast cancer cells can activate the tyrosine kinase signal transduction pathway. Indeed, T47D cells treated with selective tyrosine kinase inhibitors do not proliferate in a FC receptor- tyrosine kinase signalling dependent manner (94).

As mentioned before (Figure S6C), another pathway identified as activated in response to progestin is the ERBB-EGF network (Figures S7D, H). ERBB2 (HER2) is overexpressed in 15-20% of breast cancer in response to EGF activation, and plays a major role in EMT (95). PR interacts with ERBBs and induces the translocation of ERBB2-PR-STAT3 complex to the nucleus. ERBB2 acts as a co-activator of STAT3 and drives the activation of progestin regulated genes, especially genes such as Cyclin D1 that do not contain HREs (96, 97). Blocking PR signalling in PR-ERBB2 positive breast cancer patients has been suggested as a treatment (98). The ERBB pathway may represent a new mechanism for further study to understand the activation of these “non-classical” PR dependent genes in response to progestin. In addition, to the role of ERBB2, the role of ER activation in response to progesterone in breast cancer cells is also critical, as shown in Figure 4C we observe the coordinated activation of the ESR1-Src-PIK3 complex peaking at 15 minutes following hormone exposure. The phosphorylation site of ESR1 which increases is S104. ER S104 phosphorylation is essential for ER activity (99) and it has been suggested that hyperphosphorylation of ER at these sites may contribute to resistance to tamoxifen in hormone receptor positive breast cancer (100–102). ER S104 phosphorylation by ERK has been shown previously in response to oestrogen and EGF but not progesterone exposure. In addition, ER S104 has been implicated in mTOR signalling (103). Given the phosphorylation of ER, the dynamic activation of the ER membrane complex and the role of mTOR in AMPK and insulin signalling described earlier (Figures 3J and 4C) this phosphorylation site may present a key step in the cellular response to progesterone in breast cancer.

A pathway exhibiting strong activation by progestins is the Insulin signalling (Figure S6C; Figure S7F). Insulin-like growth factors (IGFs) and progestins both play a major role in normal mammary gland development and R5020 has been shown to induce the expression of insulin receptor substrate-2 in MCF7 cells (104, 105). Moreover, IGF signalling via IRS2 is known to be essential for breast cancer cell migration. It has also been shown that R5020 pretreatment followed by IGF stimulation increases binding of IRS to PI3K-p85 regulatory complex, which in turn activates ERK and AKT signalling (106). Interestingly, not only do we observe the activation of the insulin pathway in network analysis (Figure S7F, but also the coordinated phosphorylation of all members of the IRS-PIK3 complex (Figure 4A), indicating that indeed progesterone stimulation of T47D cells activates not only the insulin pathway in general terms but also triggers the coordinated regulation of complexes within it.