Application of Proximity Ligation Assay for Multidirectional Studies

To My Family

“We must have perseverance and above all confidence in ourselves. We must believe that we are gifted for something and that this thing must be attained.” ― Maria Skłodowska-Curie

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Zieba A, Pardali K, Söderberg O, Lindbom L, Nyström E, Moustakas A, Heldin C-H and Landegren U. Intercellular variation in signaling through the TGF-β pathway and its relation to cell density and cell cycle phase. Mol Cell Proteomics. 2012 Mar 22. [Epub ahead of print]

II Sundqvist A, Zieba A, Vasilaki E, Herrera Hidalgo C, Söder- berg O, Heldin C-H, Landegren U, ten Dijke P and van Dam H. Specific interactions between Smad proteins and AP1 components determine TGFβ–induced breast cancer cell invasion. Submitted.

III Pinidiyaarachchi A, Zieba A, Allalou A, Pardali K, Wählby C. A detailed analysis of 3D subcellular signal localization. Cy- tometry A. 2009 Apr;75(4):319-28.

IV Zieba A, Wählby C, Hjelm F, Jordan L, Berg J, Landegren U, Pardali K . Bright-field microscopy visualization of proteins and protein complexes by in situ proximity ligation with peroxidase detection. Clin Chem. 2010 Jan;56(1):99-110.

Reprints were made with permission from the respective publishers.

Other publications by the author

V Raja E, Zieba A, Morén A, Voytyuk I, Söderberg O, Heldin C-H and Moustakas A. The tumour suppressor kinase LKB1 negatively regulates bone morphogenetic protein signalling. Submitted. VI Dahl M, Lönn P, Vanlandewijck, M, Zieba A, Hottiger M, Söder- berg O, Heldin C-H and Moustakas A. PARP-2 activation and association with Smads during regulation of TGF TGF-β signaling. Man- uscript. VII Dieterich L, Mellberg S, Langenkamp E, Zhang L, Zieba A, Sa- lomäki H, Teichert M, Huang H, Edqvist P-H, Kraus T, Augustin H, Olofsson T, Larsson E, Söderberg O, Molema G, Pontén P, Georgii- Hemming P, Alafuzoff I, Dimberg A. Transcriptional profiling of tumor vessels reveals a gene expression signature indicative of a key role of VEGF-A and TGF-β2 in vascular abnormalization in human glioblastoma. Submitted. VIII Zieba A, Grannas K, Söderberg O, Nilsson M and Landegren L. Molecular tools for companion diagnostics. Submitted review. IX Leja J, Yu D, Nilsson B, Gedda L, Zieba A, Hakkarainen T, Åker- ström G, Öberg K, Giandomenico V, Essand M. Oncolytic adenovi- rus modified with somatostatin motifs for selective infection of neu- roendocrine tumor cells. Gene Ther. 2011 Nov;18(11):1052-62. X Florczyk U, Golda S, Zieba A, Cisowski J, Jozkowicz A, Dulak J. Overexpression of biliverdin reductase enhances resistance to chemotherapeutics. Cancer Lett. 2011 Jan 1;300(1):40-7. XI Wang X, Hu B, Zieba A, Neumann NG, Kasper-Sonnenberg M, Honsbein A, Hultqvist G, Conze T, Witt W, Limbach C, Geitmann M, Danielson H, Kolarow R, Niemann G, Lessmann V, Kilimann MW. A protein interaction node at the neurotransmitter release site: domains of Aczonin/Piccolo, Bassoon, CAST, and rim converge on the N-terminal domain of Munc13-1. Neurosci. 2009 Oct 7;29(40):12584-96. XII Looman C, Sun T, Yu Y, Zieba A, Ahgren A, Feinstein R, Forsberg H, Hellberg C, Heldin CH, Zhang XQ, Forsberg-Nilsson K, Khoo N, Fundele R, Heuchel R. An activating mutation in the PDGF receptor- beta causes abnormal morphology in the mouse placenta. Int J Dev Biol. 2007;51(5):361-70.

Contents

Introduction ...... 11 TGF-β signaling- the classical frame ...... 12 Down to the nucleus ...... 12 They are not alone ...... 14 Phospho-“diversion” – when one is not enough ...... 15 TGF-β in cancer ...... 18 TGF-β in colon carcinogenesis ...... 19 TGF-β in biomarker discovery ...... 20 Paper I and II - With the tool in my hand ...... 22 Paper I: Intercellular variation in signaling through the TGF-β pathway and its relation to cell density and cell cycle phase ...... 23 Paper II: Specific interactions between Smad proteins and AP1 components determine TGFβ–induced breast cancer cell invasion...... 24 Paper III and IV - Necessity is the mother of invention ...... 26 Paper III: A detailed analysis of 3D subcellular signal localization...... 26 Paper IV: Bright-field microscopy visualization of proteins and protein complexes by in situ proximity ligation with peroxidase detection...... 27 Current investigations and future plans ...... 28 The context-dependent signaling: Perspectives on the single cell analysis of the TGF-β pathway ...... 28 Smad phosphorylation ...... 28 Hippo pathway- The window on the world ...... 29 PLA in cancer diagnostics ...... 32 Identification of a gene expression signature for colorectal cancer outcome ...... 32 TGF-β and tumor vasculature ...... 34 Acknowledgments...... 36 References ...... 39

Abbreviations

ActR-II activin receptor type II ALK1-7 activin receptor-like kinases 1-7 AMHR-II anti-Müllerian hormone receptor type II AP-1 activating protein-1 ATF activating transcription factor BMP bone morphogenetic protein BMPR-II bone morphogenetic protein type II bZIP basic leucine zipper CDK cyclin-dependent kinase cFos FBJ murine osteosarcoma viral oncogene homolog chIP chromatin-immunoprecipitation Co-Smad common mediator Smad CRC colorectal cancer EGF epidermal growth factor EMT epithelial to mesenchymal transition EPSC EMT promoting Smad complexes ERK extracellular signal-regulated kinases FGF fibroblast growth factor Fra1/2 Fos-related antigen 1/2 HRP horseradish peroxidase I-Smad inhibitory Smad isPLA in situ proximity ligation assay JNK c-Jun N-terminal kinase K-ras Kirsten rat sarcoma viral oncogene homolog LNA locked nucleic acid sequences Maf musculoaponeurotic fibrosarcoma oncogene homolog MSI microsatellite instability

Noch neurogenic locus notch homolog protein PI3 kinase phosphatidylinositol 3-kinase PLA proximity ligation assay PTM post-translational modification Ras rat sarcoma RCA rolling circle amplification R-Smad receptor regulated Smad RCP rolling circle product Smad small mothers against decapentaplegic Ski Sloan-Kettering avian retrovirus transforming protein SnoN Ski-related novel protein N Taz transcriptional coactivator with PDZ-binding motif TGF-β transforming growth factor β TGF-β RI TGF-β Receptor Type I TGF-β RII TGF-β Receptor Type II TMA tissue microarrays TS thymidylate synthase Wnt wingless-type MMTV integration site family Yap Yes-associated protein

Introduction

A number of mechanisms can modulate Smad signaling on different levels. To understand all aspects of that signaling it is important to build a “map” of interactions and modifications that are involved. Smads never work alone and there is not only one path of signaling through Smads. To project a com- plete picture of TGF-β signal transduction one needs to change the percep- tion of Smad signaling pathway and see it as a sophisticated network instead of a linear signaling pipeline. The core components of Smad signaling function as a “backbone” of TGF-β signaling, while the most important regulatory elements for the outcome of the cascade activity come from the outside of the canonical pathway. These are the ones we now need to focus on. Analyses of cellular responses are generally done at the population level, ignoring sources of heterogeneity and thereby complicating interpretation of how cells coordinate responses to changes in the external and internal envi- ronments. Outcomes of analysis of in vitro cultured cell lines are often ex- trapolated to biology of cells in their natural environment, the body. Taking into account the variety of cellular responses in tissue of organs or tumors it is unlikely that this is the right path to follow. Therefore analysis of signaling events needs to be performed at the single cell level within the cells normal environment, i.e. within tissue sections.

11 TGF-β signaling- the classical frame

The TGF-β pathway is one of the most well studied regulatory systems of the cell that is involved in most of major cell activities under physiological and pathological conditions. Deregulation in signaling downstream of TGF-β is associated with many diseases such as cancer and fibrosis [1-4]. TGF-β has a very complex role in tumorigenesis, exhibiting both oncogenic and suppressive properties, depending on tumor stage. At early stages TGF-β induces apoptosis and growth arrest, it also has the ability to suppress te- lomerase activity inducing a replicative senescence, maintaining genomic stability [4]. During tumor development cells become resistant to the suppressive properties of TGF-β signaling and the pro-oncogenic activities dom- inate. This results in epithelial to mesenchymal transition (EMT), augmented cells motility, invasiveness and acquirement of metastatic properties, en- hanced angiogenesis and suppressed immune responses [2-4] .

Down to the nucleus The TGF-β family of ligands includes TGF-βs, bone morphogenetic proteins (BMPs), activins and inhibins. These multifunctional proteins regulate a variety of cellular events including cell differentiation, immune response, cellular senescence, apoptosis and wound healing [5, 6]. Binding of TGF-β to the TGF-β receptor type II (TGF-β RII) recruits the type I receptor (TGF- β RI), inducing the assembly of a heteromeric receptor complex. The formation of such heterocomplex enables TGF-β RII to activate the TGF-β RI serine/threonine kinase domain through phosphorylation of the glycine- serine rich region (GS domain) [7, 8]. Activated TGF-β RI will then phosphorylate receptor-linked R-Smad proteins at their SSXS motif, thereby re- leasing them from the receptor complex. The released R-Smad will bind common Smad4 and the formed R-Smad/Smad4 complex translocates from the membrane to the nucleus, where it will transmit the signal by binding promoters and induce gene expression with a variety of transcriptional co- factors (Fig.1) [9-12]. Besides R-Smads and the common Smad4 there is also a third group of inhibitory Smads (I-Smads), which consist of two members Smad6 and Smad7.

Figure 1. Core signaling in the TGFβ pathway. Smad2 and Smad3 become phosphorylated upon TGFR activation and bind to Smad4. The complexes then translocate to the nucleus were they regulate gene transcription of Smad responsive genes.

R-Smads, that comprise Smad1/2/3/5/8 proteins, are phosphorylated by an active heterocomplex consisting of one type I receptor, a family of seven members which are often denoted as activin receptor-like kinases 1-7 (ALK1-7), and one type II receptor, i.e. TβRII, ActR-II, BMPR-II, AMHR- II, ActR-IIB. The different pairwise combinations of type I and II receptors are activated by different TGF-β ligands. ALK5 is phosphorylated and medi- ate signaling via TGF-β while ALK4 and ALK7 are specific for activins and nodal ligands. Smad2 and 3 are downstream effectors of activated ALK4/5/7 in canonical signaling. ALK1/2/3/6 are activated by ligands such as BMPs and phosphorylate R-Smad1, 5 and 8 [13].

13 They are not alone TGF-β- induced cell responses are regulated by variety of factors and mechanism that tightly control this pathway under physiological conditions. There are several proteins influencing TGF-β signaling at different levels. The major ones are two proteins belonging to the same family, called SnoN and Ski. Both were discovered as proto-oncoproteins, promoting transformation of avian fibroblasts and myogenic differentiation of quail embryo cells, suggesting that in normal cells overexpression of SnoN and Ski results in oncogenic transformation [14-16]. Expression of both proteins was subsequently shown to be elevated in many cancer cell lines and tissues, such as colorectal carcinoma [17, 18], esophageal squamous cell carcinoma [19], leukemia [20] and melanoma [21]. Ski and SnoN were shown to negatively regulate TGF-β signaling by binding to Smad proteins [22-24]. They are by them- selves unable to either bind DNA directly or expose any catalytic behavior, thus it is believed they can only function through interaction with other cellular proteins as negative transcriptional regulators. It has been proposed that the ability of Ski and SnoN to physically interact with Smad proteins and suppress their transcriptional activity is essential for their transforming activity [25]. The level of expression and potency of SnoN and Ski actions is regulated at multiple levels: by regulation of their transcription, TGF-β induced protein degradation, post-translational modifications and subcellular localization. Activation of the TGF-β pathway, by exposure to TGF-β, causes an immediate degradation of SnoN and Ski. After 2 hours of TGF-β treatment the SnoN level will increase significantly, whereas Ski level seems to be uninfluenced, suggesting that TGF-β plays dual role in regulation of this oncoprotein, both at levels of transcriptional regulation and protein stability [23, 26]. Weather the regulation of SnoN and Ski expression is cell cycle dependent, or not, is still a controversy with a lot of discrepancy in published data, some groups have observed elevated levels of SnoN in the G1-phase of the cell cycle [27, 28] whereas others have found higher levels of Ski and SnoN in the G2/M-phase [29-31].

14 Phospho-“diversion” – when one is not enough

R-Smads consist of two highly conserved polypeptide fragments, the MH1 and MH2 domains, joined by a variable serine/threonine-rich linker region. Until recently studies of TGF-β signaling have been primary focused on Smads activation by TGFR induced phosphorylation in their SSXS C- terminal region of MH2 domain. Currently we experience an increasing interest in TGF-β pathway alterations induced by additional, or alternative, phosphorylation events of serine and threonine residues inside the linker region of Smads sequences. The intermolecular interactions between Smads and other regulatory molecules and intramolecular interactions of the MH1 and MH2 domains can be modulated by the linker phosphorylation (Fig.2).

Figure 2. Schematic representation of phosphorylation sites in Smad2 and Smad3. JNK, p38 MAPK, CDK4 and ERK phosphorylate Smad2 and Smad3 at specific residues in linker regions. Activated TGF-β RI phosphorylates COOH-terminal SSXS motif of MH2 domain.

The linkers multiple phosphorylation sites for proline-directed kinases are targeted by mitogen-activated protein kinases (MAP kinases), cell division cyclin-dependent kinases (CDKs) or ERK, all with preferences for specific residues within the linker. CDK2 and CDK4 phosphorylate Smad3 at the Thr 8 in the N-terminus and T179 and S213 in the linker region at the basal cell state [32-34]. TGF-ß and EGF treatment causes rapid phosphorylation of T179, S204 and S208 in linker region that inhibits Smad3 activity similarly to CDKs [34, 35]. As a consequence thereof phosphorylation of Smad by the same kinase can lead to diverse or opposite effects in different experimental

15 settings, which suggests that multiple combinations of phosphorylated sites can occur. In addition, one phosphorylation event can prevent or stimulate other phosphorylation events that further increase the level of complexity (Fig.3).

16 Figure 3. Localization of different Smad2 and Smad3 phosphoisoform types: pSmad2-COOH, pSmad3-COOH; pSmad2-Linker, pSmad3-Linker, pSmad2- Linker/COOH and pSmad3-Linker/COOH. TGF-β RI, JNK and CDK differentially phosphorylate Smad2 and Smad3. Only pSmad2-Linker is localized to the cyto- plasmic, the other phosphoisoforms are restrictedly present in the cell nuclei (based on Carcinogenesis vol.32 no.11 pp.1578–1588, 2011).

It has been shown that linker phosphorylation changes during cancer progression, e.g. phosphorylation at both the C-terminal and linker domains of Smad2 and Smad3 promotes malignant signaling resulting in increased proliferation and invasiveness of the cancer cells [36]. The extent, magnitude, timing, and functional consequences of linker phosphorylation differs between cell types and stage of cancer, substantial increase in Ser208/Ser213 phosphorylation of Smad3 has been shown to positively correlate with late colorectal tumors. These findings suggest that the phosphorylation of Smad3 linker is involved in the TGF-β promoted growth in late stages of tumorigenesis [37]. Phosphorylation and dephosphorylation of several residues of Smad sequences can induce, prevent or terminate active signaling. Neverthe- less the precise mechanism, by which the linker phosphorylation controls Smad signaling, and the influence of linker phosphorylation on C-terminal phosphorylation, still remains to be clarified. Further investigation of the regulatory role of Smad linkers require studies on specific (de)phosphorylation events and physiological conditions under which these events occur.

17 TGF-β in cancer

TGF-β pathway plays a central role in progression of human cancer; mal- function of this pathway severely influences both cell and an organ condi- tion. At early stages of cancer development TGF-β act as a growth- suppressor, which was demonstrated with several experiments by introducing mutations in the basic components of the TGF-β pathway [38]. When the cancer cells by-passes the suppressive functions of the TGF-β pathway it will progress into a malignant stage that utilizes TGF-β signaling to promote tumor growth, invasion and metastasis to secondary organs. In progressive- stage tumor, cells produce elevated levels of TGF-β; overexpression of TGF- β strongly correlates with presence of metastasis in colon, breast, and prostate cancer [39-41] . The expression pattern in these secondary tumors was considerable higher than in their primary tumors suggesting important role of TGF-β in invasion and metastasis. Even though the TGF-β level is upregulated in certain cancer cells that may remain TGF-β-responsive, they often loose sensitivity for TGF-β-induced growth arrest, which in turn induce unrestricted proliferation. Deactivation of components in the TGF-β pathway, by mutations or loss of heterozygosity, is essential for the tumor to escape the suppressive functions of this pathway. This functional switch of TGF-β from a suppressive to an oncogenic role is a dynamic process that involves multiple factors. Not only genetic changes but also epigenetic events and abnormal alterations within tumor microenvironments are associated with the oncogenic activity by TGF-β. These changes collectively promotes EMT, a first step in tumor cell invasion and metastasis that gives rise to cells with stem-like properties, including capability for self-renewal and a tumor-initiating phenotype, that result in increased resistance to chemothera- py [42]. During TGF-β induced EMT also non-canonical effectors of the pathway appear to have elevated or altered activity. The protein complexes involved in the TGF-β driven EMT is called EMT promoting Smad complexes (EPSC). AP-1 is the one of this EMT-linked family of transcription factors that by interaction with Smads regulates gene expression, promoting expression of mesenchymal genes and repressing epithelial ones. AP-1 transcription factors belong to several different families, the common feature for all of them is the presence of basic leucine zipper (bZIP) domains that are required for proteins dimerization and binding to DNA. The main factors belonging to the AP-1 family are Fos (FosB, c-Fos, Fra1, Fra2) and Jun (JunB, c-Jun, JunD) subgroups. Less studied but also important are activat-

18 ing transcription factor (ATF) proteins and the Maf subfamily. The multiple possible combinations between these proteins, together with a multitude of co-factors, gives rise to a large panel of complexes that ensures affinity and specificity for a large repertoire of genes. The AP-1 binding sequences vary and are determined by interactions with proteins such as, for example, Smads [43]. Both Smad2 and Smad3bind all three Jun family members and TGF-β stimulation has been shown to cause rapid phosphorylation of cJun, promoting its association to Smads and thereby inducing transcriptional activity of AP-1 sites [44]. The activation of mechanisms involved in the in- duction of EMT in tumors, involvement of various signaling pathways and transcription factors and formation of regulatory complexes is of highest interest in cancer research, [45] as this provide the cancer cells the ability to migrate, invade surrounding tissue and to form metastases. By EMT the cells acquire features analogous to stem cells and several stem cells pathways like Noch, Ras or Wnt are likely to be involved in this process. Also members of Hippo pathway, Taz and Yap, were found to promote proliferation, EMT and therefore oncogenesis in cell culture overexpression studies [46]. It has been suggested that Yap could promote stem cell like properties, by directly binding to promoters of the genes inducing stem cell pluripotency [47], a finding that might shed lights on the connection between EMT and the con- cept of cancer stem cells, or cancer initiating niche. Interestingly the Hippo pathway has been recently associated with TGF- β signaling in a microenvironment and context-dependent fashion [48]. The tumor suppressive functions of TGF-β are sustained by the core elements of the pathway. Increase in non-canonical signaling and interactions with other molecules may interfere with these functions and result in imbalance, leading to metastatic progression. The synergistic outcome of interactions among all components involved in TGF-β induced EMT is cell-context dependent.

TGF-β in colon carcinogenesis There are several types of cancer where TGF-β signaling plays a prominent role. Mutations of Smad2 and Smad4 together with TGF-β receptor mutations are predominantly present in a subset of colorectal cancers (CRC) which there is linked to their development and progression. TGF-β inhibits cell proliferation and induces apoptosis in the normal intestinal epithelium, but through mutations in the signaling cascade many colorectal tumors gain resistance to this growth inhibitory signal [49]. At the late stages of colorectal carcinogenesis TGF-β is usually highly expressed, which then is associated with increased probability of recurrence and low survival rate [50]. Ex- tended exposure to high levels of TGF-β results in neoplastic transformation of intestinal cells and both preoperative and the postoperative high levels of

19 TGF-β are predictive factor for metastasis [51]. TGF-β increases the expression of numerous mitogenic growth factors such as FGF, EGF and TGF-α and activates other oncogenic pathways including JNK, Ras/MAPK and PI3 kinase/Akt pathways. In addition TGF-β controls of cell motility, adhesion and the extracellular matrix composition, and hence is involved in cancer cells invasion, and promotes immunosuppression and tumor angiogenesis. Several mutations in TGF-β RI have been identified in human colorectal cancer cell lines, but those mutations are not common [52, 53]. Some studies have established association between CRC and a polymorphic allele of the TGF-β RI named TGF-β RI *6A and indicate that *6A may have a function of a tumor predisposition allele. A collective evaluation of role of *6A in several studies of CRC cases shows that patients carrying TGF-β RI*6A have higher risk of developing CRC [49, 54]. Deletions or mutations in the Smad genes are common events in colon cancerogenesis. The most fre- quently altered TGF-β related gene in CRC is Smad4, with a mutation rate of 16–25% while Smad2 mutations are identified in approximately 6% of the cases. Both Smad2 and Smad4 are localized at chromosome 18q, a region that often is deleted in colon adenocarcinomas. Mutations of Smad3 in CRC are infrequent and they role is still elusive, but it has been proposed that occurrence of specific phosphoisoforms of Smad3 linker region promotes cell invasion.

TGF-β in biomarker discovery CRC is the second most common malignancy in most western countries and almost 50% of diagnosed patients with CRC die from the disease [55]. There are currently no robust methods for determining risk of recurrence after diagnosis and although a number of effective treatments exist today selection of treatment for individual patients are hampered by the lack of biomarker predictive of prognosis and response to therapy [56]. The basis for diagnosis and decisions about treatment is usually the histo- logical evaluation of tissue sections. CRC can be staged according to the Duke system, wherein a patient is classified as Duke A if the tumor is affect- ing the inside layer of the colon or rectum or slightly invading the muscle layer. In Duke B the cancer spreads into the muscle layer of the colon or rectum, without infecting lymph nodes, while spreading into a least one lymphnode in the area of tumor classifies it as Duke C stage. When the cancer gains metastatic properties it is classified as a Duke D. This evaluation system is satisfactory for prediction of the outcomes of therapies of A and D stages, but not for the stages in-between [57-59] . The lack of accuracy in the oncopathologic evaluation results in different clinical outcomes of patients with similar immunohistochemical staining pattern, which highlights the need for new prognostic and predictive mark-

20 ers to improve the assessment and treatment of patients [60]. There are currently several candidate markers under investigation that show promise in determining the risk for progression and responsiveness to adjuvant therapy. Nevertheless it is becoming evident that there are multiple, cross-talking pathways responsible for tumor progression and that using a single marker is not sufficient. Therefore up to date there are no recommendations of predictive markers, to be used in clinical routine diagnostics, for evaluation of Duke B and Duke C stages. Candidate markers in clinical trials, such as 18q and MSI status, p53, K-ras, TS, dihydropirimidyne dehydrogenase or tymi- dine phosphorylase, has so far been shown to provide limited value [61] . It has been shown that the level of expression of Smad4 in Duke C patients can be employed as predictive marker whereas Smad4 mutations or imbalance of 18q21 locus are of less prognostic value [62]. The use of protein expression or genetic alterations as biomarkers can be insufficient, misleading or simply problematic to perform. Some, but not all, genetic anomalies and changes in protein expression will affect the function of a protein, thus may the functional activity of proteins be considered as future key marker for prognosis and therapy assessment. Thus, methods such as in situ PLA may be used to provide functional analysis of pathways promoting cancer development and for detecting disturbances in the interplay among proteins.

21 Paper I and II - With the tool in my hand

TGF-β pathway is widely and extensively studied for its multidirectional roles and importance in retaining normal functioning of healthy cells; mutations in this pathway are one of the most common alterations in cancer. In paper I and II we performed in situ analysis of TGF-β pathway signaling by functional characterization of Smad proteins and related, accompanying interaction partners. Our aim was to investigate the signal transduction in the TGF-β pathway at a single cell resolution, to enable analysis on cell to cell variations and to follow the kinetics of Smad complexes formation, and we thus utilized in situ PLA for visualization of complexes between two or more proteins as well as post-translational modifications of proteins in cultured cells and tissue sections. In situ PLA is based on antibody recognition of proteins in a complex, converting a binding event into a DNA-based report- er. To achieve this, the antibodies are converted into PLA probes by conju- gating short DNA strands to them. Only when a pair of proximity probes binds adjacent epitopes will they template hybridization and subsequent ligation of two circularization probes, leading to the formation of a circular DNA molecule. This DNA circle can then be amplified by rolling circle amplification (RCA) primed by oligonucleotides of the proximity probes, generating a long single stranded DNA molecule that still is attached to the proximity probe that bind its antigen. The concatameric repetitions of the complementary sequence of the circle enable hybridization of multiple detection oligonucleotides, generating a bright florescent dot. Each rolling circle amplification RCA product can easily be visualized in florescence microscope and quantified. Such an approach of double recognition provides high selectivity and sensitivity of analysis, providing single-molecule resolution using conventional microscopes [63-66], as binding of single proximity probe is insufficient to create an amplifiable DNA circle (Fig.4).

Figure 4. Schematic representation of proteins complex detection with in situ PLA. PLA-probes bind to proteins in a complex and thus are brought in close proximity. Subsequently added circularization oligonucleotides hybridize to the PLA-probes and will upon ligation form a circlular DNA molecule. This DNA circle is multi- plied by rolling circle amplification, generating a concatemeric amplification product of the circle that is detected by hybridization of fluorescence-labeled detection oligonucleotides.

Paper I: Intercellular variation in signaling through the TGF-β pathway and its relation to cell density and cell cycle phase The members of transforming growth factor-β superfamily initiate signaling through complex ligand–receptor interactions followed by formation of Smad complexes that are transduced to the nucleus, resulting in a nuclear accumulation of transcriptionally active Smad complexes that initiates a wide range of transcriptional programs. The quantitative aspects of Smad signaling as the duration and strength of signaling affect those responses and are regulated on many levels. In my first paper I have used the in situ PLA technique to characterize how stimulation via TGF-β affects Smad proteins in genetically unmodified cells. Herein we focus on the assembly of functionally active complexes between endogenous Smad proteins, during TGF-β signal transduction from cytoplasm to nucleus by visualizing complexes of Smad proteins in individual cells. The single-molecule resolution of in situ PLA enabled us to investigate sub-cellular localization of the interacting proteins, to determine were interactions occur and to monitor trafficking. We show that in presence of TGF-β Smad-2/4 and Smad-3/4 complexes are mainly localized in the nucleus and their accumulation is dependent on con- centration of TGF-β and time of exposure to the cytokine. We also demonstrated that complex assembly involving Smad proteins is influenced by the

23 cellular environment, and that it correlates with the cell cycle phase, resulting in striking cell to cell variation of complex formation. Our data clearly show that Smad3 activation, i.e. Smad3 phosphorylation, does not depend on cell density. However, Smad complexes were found to be significantly re- duced in numbers in dense cell cultures, suggesting that activation by Smad phosphorylation need not always result in productive signaling. These findings are consistent with recently published data [48]. We speculate that activated Smad2/3 might be sequestered in cytoplasm due to interactions with Taz or Yap in a density-dependent manner. In addition, we demonstrate a possible role for SnoN and Ski, potent transcriptional repressors of the signaling pathway, in which they sequester Smads proteins in a cell cycle- dependent manner. Earlier studies of Smad interactions have mostly relied on engineered systems of transfected, over-expressing cells, with measurements made across populations of cells. Due to the limitations of such methods, important questions remain to be addressed: regarding mechanisms and kinetics of endogenous cell signaling, the localization of complexes within different cells and compartments of the cell, and about the quantitative nature of these processes. Our data supports and extends earlier findings about TGF-β signaling, and demonstrates the potential of the in situ PLA method to reveal new mechanisms of regulation of cell signaling in genetically unmodified cells. We further demonstrate that analysis can be performed in formalin-fixed patient tissue sections, providing a basis for characterization of tumors with respect to the activity status of signaling pathways. Thus, analysis of Smad proteins via in situ PLA might provide new biomarkers for developmental and disease-related studies of critical signaling pathways.

Paper II: Specific interactions between Smad proteins and AP1 components determine TGFβ–induced breast cancer cell invasion. The role of Smad proteins as transcriptional effectors of TGF-β signaling is well studied and defined. However, in addition to their role as DNA binding transcription factors Smads also function as transcriptional coactivators, transducing signal from TGF-β to promote AP-1 regulated genes activation. This is achieved by direct physical interaction between Smads and members of the AP-1 family. The composition of AP-1 complexes depends on expression of the AP-1 members, which differs between different cell types. The affinity of Smad proteins to AP-1 family members varies and certain complexes, like Smad- Jun, prevent Smad association with other factors. The Smad-AP-1 complexes exhibit different affinities to DNA-elements and will thereby modulate the

24 capability to activate responsive promoters. In paper II, I describe our recent findings on the transcriptional crosstalk between TGF-β and AP-1 family members in context of tumor progression and metastasis. The aim with this study was to investigate how AP-1 signaling is influenced by TGF-β. As a 3D model of breast cancer invasion we used spheroids of MCF10A-MII (pre-malignant human breast cancer cells) imbedded in collagen. This model was previously used in studies on TGF-β-induced invasion, showing that TGF-β-RI and Smad proteins play a critical role in cancer breast invasion by activating the secreted proteases mmp2 and mmp9 [67] . With this 3D model we determined that AP-1 transcription factor such JunB, cJun, cFos and Fra1 are essential for tumor cell invasion induced by TGF-β. We also established that these factors are responsible for regulation of numerous TGF-β induced genes involved in EMT and tumor invasion. To determine protein complexes formation between AP-1 transcription factor and Smad proteins we used in situ PLA in MCF10A-MI cells, and were able to detect Smad proteins interacting with several AP-1 family members. We showed that signaling from TGF-β promotes complex formation in the nucleus between Smad3 and Smad4, but also complexes between Smad2/3 and Fra1 proteins. Interesting- ly Smad3 complexes with cJun and JunB were found prior to TGFβ- stimulation. Using chromatin-immunoprecipitations (chIP) we could determine the functional consequence of these complexes, showing that association of Smad2/3 to the mmp10 and PAI-1 promoters depends on binding of cJun, JunB and Fra1, in a TGF-β-dependent manner. We conclude that specific types of complexes formed between Smad2/3 and Fra1 may trigger the activation of the TGF-β-induced invasion program.

25 Paper III and IV - Necessity is the mother of invention

In the course of our work we often find ourselves limited by a lack of proper methods to find answers to important questions we raise. In two of my pro- jects presented in this thesis I was faced with such problems to overcome.

Paper III: A detailed analysis of 3D subcellular signal localization The function of cellular components often depends on their localization within the cell. Accordingly it is of great importance to identify the subcellular localization of proteins and protein complexes in many biological studies. Most of current methods for image analysis are restricted to 2D signal detection and therefore the possibilities of detailed analysis of localization are limited. To overcome this obstacle, in paper III we developed an approach for 3D analysis of subnuclear localization of RCA products obtained with in situ PLA. Our aim with this paper was to develop an analysis tool for detection of 3D point-like fluorescence signals, providing subnuclear positioning of such signals to enable a more detailed analysis for localization of complexes between Smad proteins. In situ PLA signals were detected using a stable wave detector, and analysis of signals was then confined to a restricted area in relation to the nuclear membrane. This tool was used to explore Smad complex localization within nuclei of mouse embryonic fibroblasts, treated with TGF-β for different times. We found that the Smad proteins exist as complexes close to the inner nuclear membrane, which implies temporal regulation of transcription. The quantitative data agreed with the generally proposed model of Smad kinetics.

26 Paper IV: Bright-field microscopy visualization of proteins and protein complexes by in situ proximity ligation with peroxidase detection. Bright-field microscopy is by far the most commonly used type of microscopy. An advantage, compared to fluorescence microscopy, is that bright-field microscopy provides a better view of the histology than fluorescence microscopy. In addition, bright-field microscopy is not affected by the auto- fluorescence of a sample, something that hampers detection of weakly fluorescent objects. The level of intrinsic or fixation induced, unspecific back- ground fluorescence is tissue type specific and can severely influence sample analysis and data acquisition. In paper IV we aimed to replace the fluorescent detection probes with horseradish peroxidase (HRP) labeled detection probes, as this would allow colorimetric detection of RCA products and enable quantification of RCA products using bright-field microscopy. To test the new protocol we applied HRP-labeled detection probes to vis- ualize in situ PLA signals, representing individual proteins or protein complexes, in different cells and tissue sections. To specifically enumerate the in situ PLA signals we developed software to separate black and red color from hematoxyline-eosine staining, and with this we were able to quantify in situ PLA signals. This approach provides better visualization of tissue structures, without compromising the detection of in situ PLA signals. Equivalent results were obtained when comparing the enzyme-linked with fluorescent detection. We have herein shown that in situ PLA can be used together with bright- field microscopy, providing an approach that readily can be adopted in routine clinical histopathology.

27 Current investigations and future plans

The context-dependent signaling: Perspectives on the single cell analysis of the TGF-β pathway

Smad phosphorylation A wide range of kinases and phosphatases control activity of TGF-β signaling by generating multiple Smad phosphoisoform, which exhibit various biological functions. The diversity of phosphorylation events controls function of Smad2 and Smad3 in a temporal- and context-dependent manner (Fig.3). Therefor I will focus on the involvement of cyclin dependent kinases in phospho-control of TGF-β activity in relation to cell cycle phases. Intermittent activation and inactivation of CDKs and their complex formation with cyclins governs cell cycle progression and the processes thus need to be tightly regulated. CDKs undergo cell cycle dependent changes of their activity, which influences their subcellular localization, i.e. nucleocyto- plasmic translocation that is regulated by phosphorylation at the beginning of mitosis [68]. CDKs play a fundamental role in regulation of mitosis and it has been shown in a large-scale analysis of the phosphoproteome that most of identified by mass spectrometry phosphorylation sites in CDKs are mainly modulated in mitotic cell. Almost full phosphorylation was observed in the majority of phospho-residues regulated by CDK1 and CDK2, suggesting phospho- inactivation of these proteins in mitotic cells [69]. It has also been shown that CDK4 is activated by cyclinD, whose expression is induced by the Ras/ERK pathway during mitosis. This also significantly increases Smad3 phosphorylation, although the cell cycle-dependent phosphorylation of Smad3 peaks at the G1/S transition [34]. However, the phosphorylation of Smad3 during G1/S phases does not prevent Smad3 and Smad4 complex formation upon TGF-β stimulation, while in mitotic cells these complexes are not present in the nuclei (Paper I). Interestingly, in mitotic cells we observed complexes between Smad4 and a form of Smad3 that is phosphorylated at the T179 residue in the linker region. This complex is localized only to the cytoplasm (Zieba A. and Blokzijl A. unpublished data, Fig.5).

Figure 5. The variation in localization of complexes formed between Smad4 and Smad3 phosphorylated at Linker (T179) or COOH-terminal region (SXSS). The nuclei are stained with Hoechst (blue) and cytoplasm is detected with Alexa488- labeled phallodin (green). The complex formation is visualized using probes labeled with Alexa555 (orange). Complexes between Smad4 and pSmad3 (T179) are localized mostly in cytoplasm of mitotic cells. Smad4 and pSmad3-COOH complexes are not present in mitotic cells.

Further studies are needed to explain the effect of linker phosphorylations of Smads during cell cycle progression. A synergistic effect has been proposed for the Smad phosphorylation by CDK and ERK. This might explain the observed differences in phosphorylation and Smads complex formation in studies using different biological systems. One may speculate that it is the variable stoichiometry of phosphorylation events in time and site-specific manner, which accounts for variable effects on signaling. The impact of each single phosphorylation event on Smad signaling regulation might be determined by using specific antibodies against individual phosphorylation sides. Phosphorylation of the same Smad residues may have diverse effects on pathway activity in different cellular compartments, therefore identification of phosphorylation events and analysis of how they influence signal progression will be crucial for understanding context-dependent Smad signaling.

Hippo pathway- The window on the world The cell can respond to specific clues and stress signals, and the outcome of the stimuli relies on integration and responses to those stimuli. The cues cells receive via cell-to-cell contacts or from the extracellular matrix can act in a coordinated manner and be additionally modified downstream the pathways [70-73]. Uncontrolled mutation-driven or stimuli-independent cellular responses can result in significant alterations in growth and proliferation pro-

29 cesses of the cell, leading to disease. Unrestrained growth of organs has been linked to mutations in the Hippo pathway that senses cell density and controls organ size [74, 75], although there may be discrepancies between organs [76]. The same mechanism that controls global organ size signals, when altered in cancer cells, may promote proliferation or inhibition of stress- induced apoptosis [75]. The pathway transduces multiple external signals from the cell membrane to the nucleus and controls expression of numerous genes in a cell-type and tissue-type dependent manner. Hippo pathway is regulated by variety of upstream inputs but many of controlling events still remain to be explored. The outcome of the Hippo cascade signaling is therefore diverse and depends on cell context. Multiple binding partners of Hippo pathway downstream effectors, such as Taz and Yap, play a significant role in modulating the signaling and negative feedbacks, e.g. by regulating phosphorylation and ubiquitination. The Hippo pathway has recently been linked to TGF-β signaling, in which the main factors of the Hippo regulatory system, Taz and Yap, transduce cell density information. In sparsely growing cells Taz and Yap are localized to the nucleus, where they bind to other transcription factors and promote transcription, while in dense cells both factors become phosphorylated and are retained in the cytoplasm, preventing signaling from other pathways. It has been shown that Taz and Yap regulate TGF-β-Smad signaling in response to cell density by dictating the localization of Smad complexes. Taz and Yap have a dominant control over the subcellular localization of phosphorylated Smads. In high-density cell cultures Taz and Yap sequesters Smad complexes in cytoplasm, suppressing the TGF-β induced signaling. Consequently TGF-β responsiveness is coupled to cell density, utilizing the Hippo pathway through Taz and Yap to transduce external stress stimuli into TGF-β pathway that moderates Smad-dependent gene expression [48]. Due to the integrating properties of the Hippo pathway in regulating cell proliferation, apoptosis and organ growth, together with its involvement in unrestrained proliferation in various types of tumors, there is an increasing interest in alterations and mutations of the Hippo pathway with respect to cancer development and progression. It is well known how the core elements of the Hippo pathway interact, but the molecular and mechanistic character- istics of regulatory components remain unclear. During cancer development the core components of the pathway along with their upstream regulators function as tumor suppressors, but Taz, Yap and other associated downstream factors are considered to be involved in tumor progression. Therefore it is of great importance to identify not only direct downstream target genes but also other signaling molecules or transcription factors that cross-talk with the Hippo pathway [77]. I have therefore initiated a project, to analyze of interactions between Taz, Yap and Smads in the context of cell density-dependent modulation of TGF-

30 β regulated genes. Preliminary data on complex formation in normal and cancer cell lines have been very promising (Fig.6) and we will hence contin- ue analyses also in human cancer tissue sections.

31 Figure 6. Cell density dependent Taz/Yap interactions with Smad. (A) In cells growing at low or moderate densities Taz and Yap translocate to the nucleus and act as transcription co-factors by interaction with Smad molecules. (B) In dense cell cultures Taz and Yap are phosphorylated and remain in the cytoplasms where they sequester Smad complexes, blocking downstream genes activation. (C) Cells growing at different cell densities were analyzed by IF for endogenous Smad2/3 and Taz expression, and we observed a correlation in localization between Smad2/3 and Taz in sparse cell cultures. (D) Using in situ PLA we could detect Smad2/3 and Taz complexes in sparse cell cultures, while in high density cultures the cells were empty of complexes. The complexes, visualized by in situ PLA, were detected with hybridization probes labeled with Alexa555 (orange). The cytoplasm was stained with Alexa488-labeled phallodin (green) (D).

PLA in cancer diagnostics The studies on cancer proteomics involve large scale analysis of the com- plete set of proteins, their modifications and interactions to define character- istics of particular pathologic states of cell physiology. However, biomarker research still lacks established technologies for routine analysis of clinical samples. Although methods for genetic analysis of tumor biopsies are commonly used they disregard intra-tumor heterogeneity, which might have tre- mendous consequences for diagnosis and the outcome of the treatment. Therefore the single cell analysis of tumor specimens may instead take advantage on the heterogeneity of tumors, allowing identification of cancer sub-clones that is a prerequisite for a more advanced and personalized thera- peutic approaches. In situ PLA is one of few methods that would be applica- ble in such a context, although the clinical value of the method still needs to be proven, as a viable option for future routine clinical diagnostics.

Identification of a gene expression signature for colorectal cancer outcome Most colorectal cancers are resistant to suppressing effects of TGF-β [2, 3], primarily due to inactivating mutations in the TGF-β signaling pathway, and almost all human colon cell lines have impaired TGF-β signaling [78]. Alt- hough mutations in TGF- β RII and Smad4 are detected in almost 30% of all colorectal tumors [3, 79, 80], whereas mutations in Smad2 are present in only 4% of the cases [81], deeper analysis of TGF-β signal transduction are required to better understand cancer development and to improve diagnosis and treatment of this disease. Together with the Ludwig Institute of Cancer Research (LICR) in Melbourne we will apply the recently established protocol for detection of complexes involving Smad2, Smad3 and Smad4 to analyze the functional state of the TGF-β pathway in genetically characterized human colorectal tumor samples. The samples to be analyzed were taken

32 from primary cancers from patients without recurrence and from patients that have relapsed. The detection of mutations was performed for the Smads and TGFRs genes, known to be mutated in this disease [79, 81], and they were examined individually, and in combination, as markers of prognosis (submitted work by LICR, Melbourne). Our goal with this project is to characterize mutations influencing TGF-β signaling, e.g. mutations that prevent complexes formation and repress the pathway or promote hyper-activation of signaling, by analyzing the functional properties, i.e. complex formation and PTMs, of these at a single cell level. We are also interesting in how signal convergence between Smad and other pathways is affected by Smad mutations and if the mutations in Smad4 can induce compensatory mechanisms e.g. of AP-1 and Smad3 signaling to surmount the pro-oncogenic effects of Smad4 mutations. To accomplish this we plan to use a combination of two methods: padlock probes for detection of individual mRNA molecules and in situ PLA for detection of protein complexes (Fig.7) [82]. Padlock probes can be used to determine expression levels and for genotyping of single mRNA molecules in situ. In this method the mRNA molecules are converted to cDNA molecules using reverse transcriptase and specific sets of primers containing the locked nucleic acid sequences (LNA). The mRNA is then degraded with RNase H, to generate free cDNA to which the padlock probes can hybridize. However, the presence of LNA bases in the cDNA primers blocks degradation at the 5´end of mRNA, thereby anchoring the cDNA to the targeted mRNA. Hybridizing padlock probes to the single stranded cDNA molecules, and subsequent ligation, will result in cDNA templated circularization of the padlock probes. These can then be amplified by RCA, primed from the cDNA, to generate a localized RCA product that can be visualized by fluorescence labeled detection oligonucleotides. In the same cells we will then analyze complex formation between Smad proteins and subunits of TGF-β receptor using in situ PLA. By combining both methods we hope to correlate the mutation status of individual cells with their ability to form complexes of mutated proteins (Fig.7).

Figure 7. A schematic presentation of mRNA detection using padlock probes. mRNA is converted to cDNA by reverse transcriptase, template by an LNA- modified primer, and subsequently degraded by RNase H treatment. Padlock probes will then be added and hybridized to, the newly formed cDNA. Upon ligation and amplification the RCA product is visualized through hybridization of fluorescent detection probes.

By extending genetic analysis of patient tissue samples to analysis at the single cell level of both mRNA and proteins a more precise picture will emerge, taking into account how the individual cancer cells interact with its tumor microenvironment.

TGF-β and tumor vasculature In the process of cancer development the tumor microenvironment is a crucial factor that influences the fate of cancer cells. The interplay between

34 cancer cells and non-malignant cells is a reciprocal process where the non- malignant cells affect the cancer cells, and vice versa, creating an environment that promotes tumor growth. This process includes tumor vasculariza- tion, to support the growing cells with oxygen and nutrition. The vascular system of tumors depends on their type and grade and differs from the normal vasculature with the morphology and function. As a result of extensive and aberrant angiogenesis vessels at later stages increase in size and become permeable. The abnormal vasculature is a characteristic feature of glioblastoma multiforme, which is one of the most aggressive types of brain cancer. However, very little is known about regulations and mechanism that control those events. Anna Dimberg at Uppsala University has recently identified changes in expression of genes in abnormal vessels of glioblastoma multiforme, by analysis of microdissected vessels. Based on evidence from mRNA studies that showed increased level of expression of Smad controlled genes in high grade tumor vessels, we analyzed of complex formation between Smad2/3 and Smad4. The vessels of grade IV tumors showed elevated signaling through Smad2 and Smad3, indicating the involvement of the TGF-β pathway in vasculature aberrations, while healthy and grade II tumor vessels had none or minor TGF-β activity respectively. These results further highlight the need for in situ analysis of tumor tissue sections to better understand the mechanisms underlying the growth of cancer cells, as this is a consequence of the communication between all the cell types present within a tumor. The analysis may lead to high confidence prognostic/predictive assays for personalized medicine.

35 Acknowledgments

Completing a PhD is a challenge and I would not have been able to do it without the help and support of many people over the past years. Therefore I would like to express my sincere gratitude towards:

Ola Söderberg, my PhD supervisor, for your leadership, support and con- stant enthusiasm about our work. Thank you for you kindness, tolerance and understanding. And for creating a great team. It is a great pleasure to work with you.

Ulf Landegren, my PhD co-supervisor, who gave me an opportunity to work in a great scientific environment. The time I spent in your group gave me a lot of inspiration and shaped me as a scientist. Thank you also for your proofreading. I think you might have learned some Polish grammar reading my English texts over the past years.

My collaborators from Ludwig Institute of Cancer research in Uppsala Aris Moustakas and Carl-Henrik Heldin for your kind help and scientific discussions. I look forward to a continuing collaboration with you in the future.

Katerina Pardali for introducing me to the TGF-β field and “babysitting” me at a very beginning of my PhD studies.

Anders Sundqiust and Hans Van Dam for very interesting and fruitful collaboration and your passion for TGF-β.

Mats Nilsson and Masood Kamali for creating a great scientific atmosphere and interesting discussions during our group meetings. Masood for our pas- sionate “non-scientific” discussions.

The Olink people for making a the PLA kit . Erik, Itxaso and Mats for our discussions on PLA and lots of technical help.

Andries for your interest in cell signaling. I would feel very lonely here without you.

Carolina for all the support with blob counting.

I would also like to thank my fellow PhD students. They each made the time during my PhD studies more interesting and entertaining.

Special thanks to our in situ PLA group. We have turned out to be quite sporty team. We have Linda who encourage us to , Karin who promotes and make us to and Carl-Magus who does a lot of . Thanks guys for motivation. I guess I am not a person anymore. Only at 7 a.m still does not work for me.

People sleep at 7 a.m. . Johan thanks for our chats about your and for showing us how to make first steps in a . Thanks to our new- comer (Björn) or Dr.Phil for all the fun chats we have with you in the lab.

Marco thank you for help with and Spyros for help with .

Rachel, thank you for sharing my interest in , supplied by ICA

CLAUSSON. Save some Smarties for me. Johan S, thanks for all the help with .

And (as requested) thanks to Rasel for being a part of my every single conversation with Andries.

Elin, Christina and Delal, thank you for running our lab, the cell-lab, the PLA SciLife lab…

37 I would like to give many thanks to many graduate and undergraduate students and postdocks I have worked with in our group: Caroline,

Erik, Joakim, Anna, Anne-Li, Wu Di, Elin, Lotte, Gucci, Da- vid, Maria, Rongqin, Anja, Lei, Annika, MONICA, Liza,

Tomek, Tonge, Malte, Camilla, Magda, Junhong, Elin.

Lots of thanks to the former lab members: Kalle, Sara, Olle, Tim, Ida,

Irene, Mikaela, Yuki, Jenny, Malin, Carla, Chatharina…

Thank you all for creating a great workplace and wonderful atmosphere in the lab and for nice during and .

I would also like to thank my and non-polish friends in Uppsala for making me feel here like at and for a great time spend together on , , and .

I wish to thank my husband, Grzegorz. His comments and insights on how the science world works were very valuable and beneficial over the past years. Finally, I thank my parents for implanting in me confidence and determination for pursuing my work.

38 References

1. Blobe, G.C., W.P. Schiemann, and H.F. Lodish, Role of transforming growth factor beta in human disease. N Engl J Med, 2000. 342(18): p. 1350-8. 2. Akhurst, R.J. and R. Derynck, TGF-beta signaling in cancer--a double-edged sword. Trends Cell Biol, 2001. 11(11): p. S44-51. 3. Derynck, R., R.J. Akhurst, and A. Balmain, TGF-beta signaling in tumor suppression and cancer progression. Nat Genet, 2001. 29(2): p. 117-29. 4. Wakefield, L.M. and A.B. Roberts, TGF-beta signaling: positive and negative effects on tumorigenesis. Curr Opin Genet Dev, 2002. 12(1): p. 22-9. 5. Feng, X.H. and R. Derynck, Specificity and versatility in tgf-beta signaling through Smads. Annu Rev Cell Dev Biol, 2005. 21: p. 659-93. 6. Massague, J., TGF-beta signal transduction. Annu Rev Biochem, 1998. 67: p. 753-91. 7. Wrana, J.L., et al., TGF beta signals through a heteromeric protein kinase receptor complex. Cell, 1992. 71(6): p. 1003-14. 8. Franzen, P., et al., Cloning of a TGF beta type I receptor that forms a heteromeric complex with the TGF beta type II receptor. Cell, 1993. 75(4): p. 681-92. 9 Abdollah, S., et al., TbetaRI phosphorylation of Smad2 on Ser465 and Ser467 is required for Smad2-Smad4 complex formation and signaling. J Biol Chem, 1997. 272(44): p. 27678-85. 10. Souchelnytskyi, S., et al., Phosphorylation of Ser465 and Ser467 in the C terminus of Smad2 mediates interaction with Smad4 and is required for transforming growth factor-beta signaling. J Biol Chem, 1997. 272(44): p. 28107-15. 11. Heldin, C.H., K. Miyazono, and P. ten Dijke, TGF-beta signalling from cell membrane to nucleus through SMAD proteins. Nature, 1997. 390(6659): p. 465- 71. 12. Attisano, L. and J.L. Wrana, Mads and Smads in TGF beta signalling. Curr Opin Cell Biol, 1998. 10(2): p. 188-94. 13. Shi, Y. and J. Massague, Mechanisms of TGF-beta signaling from cell membrane to the nucleus. Cell, 2003. 113(6): p. 685-700. 14. Colmenares, C., et al., Activation of the c-ski oncogene by overexpression. J Virol, 1991. 65(9): p. 4929-35. 15. Pearson-White, S. and M. McDuffie, Defective T-cell activation is associated with augmented transforming growth factor Beta sensitivity in mice with mutations in the Sno gene. Mol Cell Biol, 2003. 23(15): p. 5446-59. 16. Boyer, P.L., et al., Sequence and biological activity of chicken snoN cDNA clones. Oncogene, 1993. 8(2): p. 457-66. 17. Buess, M., et al., Amplification of SKI is a prognostic marker in early colorectal cancer. Neoplasia, 2004. 6(3): p. 207-12. 18. Bravou, V., et al., TGF-beta repressors SnoN and Ski are implicated in human colorectal carcinogenesis. Cell Oncol, 2009. 31(1): p. 41-51.

39 19. Imoto, I., et al., SNO is a probable target for gene amplification at 3q26 in squamous-cell carcinomas of the esophagus. Biochem Biophys Res Commun, 2001. 286(3): p. 559-65. 20. Pearson-White, S., et al., The ski/sno protooncogene family in hematopoietic development. Blood, 1995. 86(6): p. 2146-55. 21. Poser, I., et al., Characterization of Sno expression in malignant melanoma. Int J Oncol, 2005. 26(5): p. 1411-7. 22. Luo, K., et al., The Ski oncoprotein interacts with the Smad proteins to repress TGFbeta signaling. Genes Dev, 1999. 13(17): p. 2196-206. 23. Stroschein, S.L., et al., Negative feedback regulation of TGF-beta signaling by the SnoN oncoprotein. Science, 1999. 286(5440): p. 771-4. 24. Luo, K., Ski and SnoN: negative regulators of TGF-beta signaling. Curr Opin Genet Dev, 2004. 14(1): p. 65-70. 25. He, J., et al., The transforming activity of Ski and SnoN is dependent on their ability to repress the activity of Smad proteins. J Biol Chem, 2003. 278(33): p. 30540-7. 26. Sun, Y., et al., SnoN and Ski protooncoproteins are rapidly degraded in response to transforming growth factor beta signaling. Proc Natl Acad Sci U S A, 1999. 96(22): p. 12442-7. 27. Pearson-White, S. and R. Crittenden, Proto-oncogene Sno expression, alternative isoforms and immediate early serum response. Nucleic Acids Res, 1997. 25(14): p. 2930-7. 28. Mimura, N., et al., A transient increase of snoN transcript by growth arrest upon serum deprivation and cell-to-cell contact. FEBS Lett, 1996. 397(2-3): p. 253-9. 29. Wan, Y., X. Liu, and M.W. Kirschner, The anaphase-promoting complex mediates TGF-beta signaling by targeting SnoN for destruction. Mol Cell, 2001. 8(5): p. 1027-39. 30. Macdonald, M., et al., Control of cell cycle-dependent degradation of c-Ski proto-oncoprotein by Cdc34. Oncogene, 2004. 23(33): p. 5643-53. 31. Marcelain, K. and M.J. Hayman, The Ski oncoprotein is upregulated and localized at the centrosomes and mitotic spindle during mitosis. Oncogene, 2005. 24(27): p. 4321-9. 32. Moustakas, A. and C.H. Heldin, Non-Smad TGF-beta signals. J Cell Sci, 2005. 118(Pt 16): p. 3573-84. 33. Matsuura, I., et al., Identification and characterization of ERK MAP kinase phosphorylation sites in Smad3. Biochemistry, 2005. 44(37): p. 12546-53. 34. Matsuura, I., et al., Cyclin-dependent kinases regulate the antiproliferative function of Smads. Nature, 2004. 430(6996): p. 226-31. 35. Wang, G., et al., Transforming growth factor-{beta}-inducible phosphorylation of Smad3. J Biol Chem, 2009. 284(15): p. 9663-73. 36. Matsuzaki, K., et al., Smad2 and Smad3 phosphorylated at both linker and COOH-terminal regions transmit malignant TGF-beta signal in later stages of human colorectal cancer. Cancer Res, 2009. 69(13): p. 5321-30. 37. Yamagata, H., et al., Acceleration of Smad2 and Smad3 phosphorylation via c- Jun NH(2)-terminal kinase during human colorectal carcinogenesis. Cancer Res, 2005. 65(1): p. 157-65. 38. Ikushima, H. and K. Miyazono, TGFbeta signalling: a complex web in cancer progression. Nat Rev Cancer, 2010. 10(6): p. 415-24. 39. Walker, R.A. and S.J. Dearing, Transforming growth factor beta 1 in ductal carcinoma in situ and invasive carcinomas of the breast. Eur J Cancer, 1992. 28(2-3): p. 641-4.

40 40. Friedman, E., et al., High levels of transforming growth factor beta 1 correlate with disease progression in human colon cancer. Cancer Epidemiol Biomarkers Prev, 1995. 4(5): p. 549-54. 41. Wikstrom, P., et al., Transforming growth factor beta1 is associated with angiogenesis, metastasis, and poor clinical outcome in prostate cancer. Prostate, 1998. 37(1): p. 19-29. 42. Heldin, C.H., M. Landstrom, and A. Moustakas, Mechanism of TGF-beta signaling to growth arrest, apoptosis, and epithelial-mesenchymal transition. Curr Opin Cell Biol, 2009. 21(2): p. 166-76. 43. Jochum, W., E. Passegue, and E.F. Wagner, AP-1 in mouse development and tumorigenesis. Oncogene, 2001. 20(19): p. 2401-12. 44. Liberati, N.T., et al., Smads bind directly to the Jun family of AP-1 transcription factors. Proc Natl Acad Sci U S A, 1999. 96(9): p. 4844-9. 45. Fuxe, J., T. Vincent, and A. Garcia de Herreros, Transcriptional crosstalk between TGF-beta and stem cell pathways in tumor cell invasion: role of EMT promoting Smad complexes. Cell Cycle, 2010. 9(12): p. 2363-74. 46. Overholtzer, M., et al., Transforming properties of YAP, a candidate oncogene on the chromosome 11q22 amplicon. Proc Natl Acad Sci U S A, 2006. 103(33): p. 12405-10. 47. Lian, I., et al., The role of YAP transcription coactivator in regulating stem cell self-renewal and differentiation. Genes Dev, 2010. 24(11): p. 1106-18. 48. Varelas, X., et al., The Crumbs complex couples cell density sensing to Hippo- dependent control of the TGF-beta-SMAD pathway. Dev Cell, 2010. 19(6): p. 831-44. 49. Xu, Y. and B. Pasche, TGF-beta signaling alterations and susceptibility to colorectal cancer. Hum Mol Genet, 2007. 16 Spec No 1: p. R14-20. 50. Xu, W.Q., et al., Expression of TGF-beta1, TbetaRII and Smad4 in colorectal carcinoma. Exp Mol Pathol, 2007. 82(3): p. 284-91. 51. Jakowlew, S.B., Transforming growth factor-beta in cancer and metastasis. Cancer Metastasis Rev, 2006. 25(3): p. 435-57. 52. Parsons, R., et al., Microsatellite instability and mutations of the transforming growth factor beta type II receptor gene in colorectal cancer. Cancer Res, 1995. 55(23): p. 5548-50. 53. Markowitz, S., et al., Inactivation of the type II TGF-beta receptor in colon cancer cells with microsatellite instability. Science, 1995. 268(5215): p. 1336-8. 54. Bian, Y., et al., Somatic acquisition of TGFBR1*6A by epithelial and stromal cells during head and neck and colon cancer development. Hum Mol Genet, 2007. 16(24): p. 3128-35. 55. Jemal, A., et al., Cancer statistics, 2007. CA Cancer J Clin, 2007. 57(1): p. 43- 66. 56. Lurje, G., W. Zhang, and H.J. Lenz, Molecular prognostic markers in locally advanced colon cancer. Clin Colorectal Cancer, 2007. 6(10): p. 683-90. 57. Le Voyer, T.E., et al., Colon cancer survival is associated with increasing number of lymph nodes analyzed: a secondary survey of intergroup trial INT- 0089. J Clin Oncol, 2003. 21(15): p. 2912-9. 58. Laurent-Puig, P., et al., Survival and acquired genetic alterations in colorectal cancer. Gastroenterology, 1992. 102(4 Pt 1): p. 1136-41. 59. Cohn, K.H., et al., The significance of allelic deletions and aneuploidy in colorectal carcinoma. Results of a 5-year follow-up study. Cancer, 1997. 79(2): p. 233-44. 60. Bendardaf, R., et al., CD44 variant 6 expression predicts response to treatment in advanced colorectal cancer. Oncol Rep, 2004. 11(1): p. 41-5.

41 61. Locker, G.Y., et al., ASCO 2006 update of recommendations for the use of tumor markers in gastrointestinal cancer. J Clin Oncol, 2006. 24(33): p. 5313- 27. 62. Alazzouzi, H., et al., SMAD4 as a prognostic marker in colorectal cancer. Clin Cancer Res, 2005. 11(7): p. 2606-11. 63. Jarvius, M., et al., In situ detection of phosphorylated platelet-derived growth factor receptor beta using a generalized proximity ligation method. Mol Cell Proteomics, 2007. 6(9): p. 1500-9. 64. Soderberg, O., et al., Direct observation of individual endogenous protein complexes in situ by proximity ligation. Nat Methods, 2006. 3(12): p. 995-1000. 65. Soderberg, O., et al., Characterizing proteins and their interactions in cells and tissues using the in situ proximity ligation assay. Methods, 2008. 45(3): p. 227- 32. 66. Soderberg, O., et al., Proximity ligation: a specific and versatile tool for the proteomic era. Genet Eng (N Y), 2007. 28: p. 85-93. 67. Wiercinska, E., et al., The TGF-beta/Smad pathway induces breast cancer cell invasion through the up-regulation of matrix metalloproteinase 2 and 9 in a spheroid invasion model system. Breast Cancer Res Treat, 2011. 128(3): p. 657- 66. 68. Yang, J. and S. Kornbluth, All aboard the cyclin train: subcellular trafficking of cyclins and their CDK partners. Trends Cell Biol, 1999. 9(6): p. 207-10. 69. Olsen, J.V., et al., Quantitative phosphoproteomics reveals widespread full phosphorylation site occupancy during mitosis. Sci Signal, 2010. 3(104): p. ra3. 70. Nishioka, N., et al., The Hippo signaling pathway components Lats and Yap pattern Tead4 activity to distinguish mouse trophectoderm from inner cell mass. Dev Cell, 2009. 16(3): p. 398-410. 71. Oka, T. and M. Sudol, Nuclear localization and pro-apoptotic signaling of YAP2 require intact PDZ-binding motif. Genes Cells, 2009. 14(5): p. 607-15. 72. Rogulja, D., C. Rauskolb, and K.D. Irvine, Morphogen control of wing growth through the Fat signaling pathway. Dev Cell, 2008. 15(2): p. 309-21. 73. Zecca, M. and G. Struhl, A feed-forward circuit linking wingless, fat-dachsous signaling, and the warts-hippo pathway to Drosophila wing growth. PLoS Biol, 2010. 8(6): p. e1000386. 74. Camargo, F.D., et al., YAP1 increases organ size and expands undifferentiated progenitor cells. Curr Biol, 2007. 17(23): p. 2054-60. 75. Dong, J., et al., Elucidation of a universal size-control mechanism in Drosophila and mammals. Cell, 2007. 130(6): p. 1120-33. 76. Song, H., et al., Mammalian Mst1 and Mst2 kinases play essential roles in organ size control and tumor suppression. Proc Natl Acad Sci U S A, 2010. 107(4): p. 1431-6. 77. McNeill, H. and J.R. Woodgett, When pathways collide: collaboration and connivance among signalling proteins in development. Nat Rev Mol Cell Biol, 2010. 11(6): p. 404-13. 78. Ijichi, H., et al., Systematic analysis of the TGF-beta-Smad signaling pathway in gastrointestinal cancer cells. Biochem Biophys Res Commun, 2001. 289(2): p. 350-7. 79. MacGrogan, D., et al., Comparative mutational analysis of DPC4 (Smad4) in prostatic and colorectal carcinomas. Oncogene, 1997. 15(9): p. 1111-4. 80. Takagi, Y., et al., Somatic alterations of the DPC4 gene in human colorectal cancers in vivo. Gastroenterology, 1996. 111(5): p. 1369-72.

42 81. Eppert, K., et al., MADR2 maps to 18q21 and encodes a TGFbeta-regulated MAD-related protein that is functionally mutated in colorectal carcinoma. Cell, 1996. 86(4): p. 543-52. 82. Weibrecht, I., et al., Simultaneous visualization of both signaling cascade activity and end-point gene expression in single cells. PLoS One, 2011. 6(5): p. e20148.