Chapter 1: Introduction
The discovery and cloning of TP63 gene, a homolog of the tumor suppressor TP53, in 1998 has brought enormous attentions to lesser known members of the p53 family (1). The first p63 knockout (TP63-/-) mouse models were generated and characterized with striking defects in limb, craniofacial and epithelial development, which prompted a question on the significance of this p53 family member during normal development and diseases (2, 3). Since its discovery 20 years ago, there have been numerous studies focusing on investigating its structure and functions in many physiological and pathological processes. However, the TP63 locus has a very complex structure with two separate promoters and multiple alternative splicing, resulting in transcription of at least six different isoforms. This complexity has challenged studies that aim to fully understand specific functions of each isoform in biological processes and diseases, including cancers (4).
Initial reports on p63 expression in human cancers led to contradictory findings on whether p63 was a tumor suppressor or oncogene due to structural complexity at the TP63 locus (4). Isoform-specific conditional knockout mouse models for TAp63 and ΔNp63 have elucidated distinct roles of p63 isoforms in different cellular compartments, emphasizing the importance of investigating specific isoforms during the initiation and progression of diseases (5-7).
The TP63 gene locates at chromosome 3q27–29 and encodes an acidic N-terminal transactivation (TA) domain, a central DNA-binding domain (DBD) and a carboxy-oligomerization domain (OD) (1, 8, 9). The DBD shares the highest homology with other members of the family (more than 60% identical with p53 and around 85% identical with p73). Therefore, p63 is able to bind and activate transcription from canonical p53-responsive elements. Importantly, these structural similarities also account for both physical and functional interactions among p53, p63 and p73, indicating an intricate interplay among p53 family members in development and diseases. The TA domain is the least conserved among p53, p63 and p73 with less than 30% homology (Figure 1) (4, 8).
As mentioned before, the TP63 gene contains two different promoters, which results in the transcription of multiple proteins categorized into two main isoforms, TAp63 (with a TA domain) and ΔNp63 (lack a TA domain) (9). In addition, both TAp63 and ΔNp63 genes can be transcribed into proteins with at least three different C-termini termed α, β and γ via alternative splicing. The full-length α isoforms (TAp63α and ΔNp63α) contain a carboxy-terminal sterile α-motif (SAM) domain, a protein-protein interaction domain that has been implicated in many cellular processes. It has been reported that TP53 and TP73 genes can also transcribe into multiple isoforms through separate promoters and alternative splicing, adding more layers to the complexity and functional diversity of the family (8-10).
Due to significant homology to p53 at the DBD, p63 was predicted to have transcriptional activities on p53-responsive genes. Initial in vitro studies showed that TAp63 isoforms can transactivate p53 targets, whereas ∆Np63 isoforms were thought to have dominant-negative effects on p53- and TAp63- mediated transcription due to the lack of the TA domain (1, 11). However, an increasing number of papers have demonstrated that ∆Np63 has its own transactivation domain at the C-terminal end and is able to induce transcription of its own target genes (12-14).
Even though these two isoforms are derived from the same gene locus, they are expressed in different cellular compartments and have distinct functions in many biological processes. While TAp63 knockout (TAp63-/-) mice were viable with shorter lifespan than wild-type mice and developed spontaneous metastatic tumors, ∆Np63 knockout (∆Np63-/-) mice were born with severe defects in epidermal differentiation, limb and craniofacial development and died shortly after birth (5, 7, 15). ∆Np63 is the predominant isoform expressed in the basal layer of the epidermis and plays crucial roles in maintaining proliferation and terminal differentiation of these cells (4, 5, 10). ∆Np63 also localizes in basal cells of the epithelial tissues in mammary glands, trachea, cervix, prostates, tongue and hair follicles (9). TAp63, on the other hand, is found at very low level in the epidermal compartment of the skin and acts as a molecular switch during epithelial stratification program. TAp63 is expressed very early on at day E7.5 in mouse embryos, then get turned off by ∆Np63 to allow maturation of the epidermis during embryogenesis (16). In stratified epithelium, TAp63 isoforms are expressed in a specific subset of dermal stem cells known as skin-derived precursor cells (SKPs) and maintain these cells in quiescence to prevent genomic instability and premature senescence (Figure 2). TAp63-/- mice developed accelerated aging phenotype, spontaneous blisters, skin ulceration and delayed wound healing due to premature depletion of dermal and epidermal precursors (6). Taken together, these findings suggest distinct contributions of TAp63 and ∆Np63 during development.
p53 is a well-known tumor suppressor for its roles in promoting cell cycle arrest, cellular senescence and apoptosis through its transcriptional activities on p21, retinoblastoma (RB) and Bcl-2 family member (17). Due to high structural similarities between p63 and p53, together with the ability of p63 to activate p53-target genes, p63 was initially believed to function as a tumor suppressor. Interestingly, while TP53 is widely mutated across multiple cancers and TP53 mutation is considered one of the hallmarks of cancers, TP63 is rarely mutated and frequently found to be overexpressed in multiple epithelial cancers. However, these studies often failed to distinguish between TAp63 and ∆Np63 isoforms (18, 19). Early genetic mouse models demonstrated that TP63 heterozygous (TP63+/-) mice had shorter lifespan than their wild-type littermates and developed various types of tumors, including squamous cell carcinomas, thymic lymphoma, lung adenocarcinomas and hemangiosarcomas. Interestingly, double mutant TP53+/-; TP63+/- mice exhibited a more aggressive tumor phenotype with higher metastatic frequency compared to TP53+/-mice, suggesting an important tumor suppressive role of p63 in cancer development and metastasis (20). In another mouse model, TP63+/- mice showed a significant decrease in spontaneous tumors compared to wild type mice. Features of accelerated aging and hyperplasia in the epithelium were often observed in these mice. In this case, TP53+/-; TP63+/- mice had lower tumor incidence than TP53+/- mice, proposing TP63 as an oncogene in cancers (21). These contradictory results arose from the existence of multiple p63 isoforms with potential opposing functions and the lack of isoform specific knockout mouse models.
The first TAp63 knockout (TAp63-/-) mouse model was reported to develop spontaneous metastatic carcinomas and sarcomas. Strikingly, TAp63 heterozygous (TAp63+/-) mice displayed a higher metastatic frequency in sarcomas than TAp63-/- mice, whereas TAp63-/- and TAp63+/- mice had a similar metastatic frequency in carcinomas, indicating a tissue-specific contribution of loss of heterogeneity (LOH) in the aggressiveness of cancers. Moreover, TAp63 was shown to control metastasis through direct transcriptional regulation of Dicer and miR-130b. Overexpressing Dicer and miR-130b in TAp63-deficient mouse embryonic fibroblasts (MEFs) decreased migratory and invasive capacity of these cells (15). In another independent study, TAp63 was also demonstrated as a suppressor of tumorigenesis in vivo through inducing senescence via oncogenic Ras, and this activity was independent of p53 (22). Through several in vitro studies, TAp63 has been shown to not only be induced upon DNA damage but be able to activate transcriptions of genes involved in apoptosis (23, 24). In summary, these data confirmed TAp63 as a master regulator of tumorigenesis and metastasis.
The fact that ∆Np63-/- mice die shortly after birth was a major challenge for studies that aimed to investigate the roles of this isoform in tumorigenesis and metastasis. Due to its antagonistic effects on transcriptional activities of p53, TAp63 and TAp73, ∆Np63 has been implicated as an oncogene in cancers (1, 4). ∆Np63 is highly expressed in lung and head and neck squamous cell carcinomas (12, 18, 25). Moreover, a recent in vivo study utilized a ∆Np63 conditional knockout mouse model to demonstrate that the ablation of ∆Np63 suppresses thymic lymphomas in p53-deficient mice through TAp63-induced transcription of IAPP (26). These accumulating data further emphasize potential oncogenic activities of ∆Np63 in cancers. However, there are evidences that ∆Np63 is a suppressor of epithelial to mesenchymal transition (EMT) and potentially cancer metastasis, indicating that functions of ∆Np63 in cancer development and metastasis are far more complex (27-30).
The skin is the primary barrier to protect the body from dehydration, environmental insults and pathogens. It is comprised of two compartments, the dermis and the epidermis that function cooperatively to regulate epithelial stratification and homeostasis. The epidermis is separated from the dermis by a basement membrane and continuously regenerated by proliferative keratinocytes in the basal layer. The epidermis also provides a variety of appendages throughout the whole body, including mammary glands, sweat glands and hair follicles (31-33).
The stratification program starts as early as day E8.5 during embryogenesis when cells in the ectoderm receive signals to switch to an epidermal fate (31). After this commitment, basal keratinocytes in the single-layer epidermis begin parallel divisions to expand the epidermal area. Then, a large portion of these cells start to divide asymmetrically perpendicular to the basement membrane to generate one basal and one intermediate suprabasal daughter cell. Later, these intermediate cells mature into spinous cells which further undergo terminal differentiation into granular and cornified cells (31, 34) (Figure 3). These epidermal cells not only have different morphology but express distinct sets of cytokeratin markers. Basal keratinocytes express keratin 5 (K5) and keratin 14 (K14), whereas suprabasal cells express keratin 1 (K1) and keratin 10 (K10) (31). Skin stratification is a very complex process orchestrated by various signaling factors and coordination between multiple cell populations, especially the tissue-resident stem cells with self-renewal capacity, in order to maintain structural integrity and homeostasis of the skin (34).
∆Np63-/- mice were born with severe developmental defects in epithelial stratification and died shortly after birth due to dehydration, which resembled phenotype of TP63-/- mice (5, 7). While TAp63 is the first isoform to be expressed in the single-layer epidermis and required for initiation of a stratification program through activation of K14 expression, ∆Np63 is the most abundant isoform found in the basal layers of stratified epidermis and essential for both maintenance of resident stem cell population in the epidermis and terminal differentiation of epidermal cells (16, 31-33). Moreover, only crossing ∆Np63α-overexpressing transgenic mice, but not TAp63α, to the TP63-/- mice was able to partially rescue the epithelialization of the epidermal layer (35, 36).
In one study, ∆Np63-null keratinocytes exhibited unbalanced and accelerated differentiation, resulting in a failure to form a mature stratified epithelium. There were keratinocytes expressing K5 in a patchy fashion in the ∆Np63-null skin, indicating that a subset of ∆Np63-deficientkeratinocytes were still able to commit to an epidermal fate. Interestingly, clusters of keratinocytes were found not only to express K5 but to retain expression of K8 as well. Co-expression of K5, K1 and K10 was also found in the epidermal layer of E15.5 ∆Np63-null embryos. These findings suggest an altered differentiation program associated with premature expression of terminal differentiation markers in ∆Np63-null keratinocytes. Furthermore, this study demonstrated that ∆Np63-null skin showed a significant decrease in cell proliferation, loss of extracellular matrix and alterations in Notch signaling pathway (5). Another study utilizing a ∆Np63 conditional knockout mouse model found that ∆Np63-null epidermal cells simultaneously expressed multiple differentiation markers, including K8, K18, K14 and K10, indicating defects in terminal differentiation in these cells. However, these ∆Np63-null epidermal cells displayed hyperproliferation with increased BrdU incorporation and self-renewing capacity with high expression of induced-pluripotency associated factors such as Oct4, Sox2 and Nanog. ∆Np63 is believed to regulate multipotency in keratinocytes through transcriptional activation of DGCR8, a component in microRNA biogenesis machinery (7). Taken together, although there are conflicting data on how ∆Np63 regulates proliferation in keratinocytes, it is evident that ∆Np63 is a master regulator of epithelial development.
One of the important molecular switches that has been shown to control ∆Np63 level between epidermal proliferation and differentiation is microRNA-203 (miR-203). In situ hybridization of miR-203 in the skin showed that miR-203 was mainly expressed in the suprabasal layer, but not in the basal layer. This expression pattern was mutually exclusive with ∆Np63 (37). Similarly, human keratinocytes induced differentiation by calcium treatment in vitro had higher level of miR-203 and lower level of ∆Np63 than untreated cells. ∆Np63 was shown to be a direct target of miR-203 during epidermal terminal differentiation (37, 38).
Since its first description in the early 1980s by Elizabeth Hay, the concept of EMT has emerged as an integral mechanism in many physiological and pathological processes, such as gastrulation, morphogenesis, wound repair, cell adhesion and movement, fibrosis and especially cancers (39, 40). Epithelial cells are single layer (or multilayer) cells that are closely in contact with each other through various structures on cellular membranes, including adherens junctions, tight junctions, gap junctions and desmosomes. In addition, epithelial cells show apical-basolateral polarization based on sequential arrangement of cadherins and integrins on the basement membrane, and communicate with each other through specialized junctional complexes. Mesenchymal or stromal cells, on the other hand, neither organize into cell layers nor exhibit apical-basal cell polarity, and only focally contact with each other. Mesenchymal cells in culture display spindle shapes with high motility (39, 41-43).
EMT is a series of events in which epithelial cells undergo complex changes in their cell morphology and behaviors to acquire characteristics of mesenchymal cells in order to enable cell motility and invasion. Hallmarks of EMT include loss of cell-cell contact, decreased expression of epithelial cell junction markers such as E-cadherin and occludins, together with increase in mesenchymal markers, including N-cadherin and Vimentin (43, 44). The reverse process, mesenchymal to epithelial transition (MET) in which mesenchymal cells change their characteristics to become epithelial cells, was also reported. These findings suggest a dynamic nature of EMT to allow cells to transition between two states of epithelial and mesenchymal in a highly plastic fashion (39, 45) (Figure 4). There are three distinct types of EMT based on different biological contexts that EMT is involved. Type 1 EMT happens during embryogenesis, gastrulation, organ morphogenesis and development. Type 2 EMT is associated with tissue regeneration, wound repair and organ fibrosis. Type 3 EMT is involved in cancer and metastasis where tumor cells gain mesenchymal phenotype, allowing them to invade local stroma and metastasize to distant organs (43).
EMT is orchestrated by a very complex regulatory network of signaling molecules, transcription factors and microRNAs. Several transcription factors have been demonstrated to function as master drivers of EMT, including Twist1/2, Snail1/2 and Zeb1/2. These transcription factors trigger an EMT program by repression of epithelial markers and activation of mesenchymal ones through various mechanisms (45-47). Twist1 binds to E-boxes in the promoter of E-cadherin to repress its transcription (47). Twist1 interaction with NuRD complex or polycomb repressor complexes PRC1 and PRC2 is required for the repression of E-cadherin (48-50). Likewise, Twist1 induces N-cadherin expression through direct transcriptional activation (51). Similarly, Snail1 (Snail) and Snail2 (Slug) inhibit E-cadherin transcription by recruiting polycomb repressor complex 2 (PRC2) to its promoter (52-54). In addition, Snail and Slug suppress expression of several other epithelial markers such as cytokeratins, occludins, claudins and desmoplakin, whereas they increase expression of pro-invasive genes such as vimentin and fibronectin (47, 55). Snail also binds to Smad3/4 proteins to promote EMT in the presence of TGFβ (56). In certain contexts, Snail and Twist factors can cooperate to regulate a same set of genes involved in EMT. Interestingly, Zeb1 and Zeb2 are downstream targets of both Snail and Twist (46, 47, 57). Several signaling pathways, such as Wnt, TGFβ, Notch, FGF, TNFα and tyrosine kinase receptors, can directly upregulate Twist, Snail and Zeb factors. These findings suggest an intricate cross-regulation network among EMT-inducers to promote cell migratory and invasive abilities (39, 47).
Recently, microRNAs have been intensively demonstrated as integral post-transcriptional regulations of EMT and MET. A complex network of microRNAs (miRs) has been involved in controlling translation of multiple EMT-associated factors. Conversely, Twist, Snail and Zeb can also regulate expression of EMT-related microRNAs, establishing a negative feedback loop to maintain epithelial homeostasis (58, 59). Among the microRNAs governing EMT, miR-200 and miR-205 families stand out. miR-200 and miR-205 members have been shown to directly bind to 3’UTR of Zeb1 and Zeb2 to repress their expression (60). In turn, miR-200 family is inhibited by not only Zeb1/2 but Snail1 to form a double negative feedback loop (61, 62). Similarly, while Snail1 is a direct target of miR-34 and miR-203, it also binds to E-boxes in the promoters of miR-34 and miR-203 to regulate their expression (63, 64). Fewer microRNAs have been connected to Twist. Twist1 has been shown to be repressed by let-7d, miR-29b and miR-214 (65-67). miR-10b, an onco-microRNA highly expressed in metastatic breast cancer, is a direct target of Twist1 (68). Furthermore, there are other aspects of EMT regulated by microRNAs. For example, miR-9 directly targets E-cadherin in mammary epithelial cells to induce EMT and increase invasiveness (69).
Wound healing is a physiological response to injury in order to re-establish normal architecture of the epithelium. The process of wound repair consists of three partially overlapping but distinct phases: inflammation, new tissue formation and tissue remodeling (70, 71). During wound healing, keratinocytes at the migratory front of the wound undergo morphological and behavioral changes to acquire a “metastable” phenotype, and move toward damaged area in a process called re-epithelialization. This intermediate state allows the cells to undergo a “partial” EMT and gain motility to heal the wound while maintaining loose cell-cell contact. At the end of the re-epithelialization process, these metastable cells are able to reverse to their original epithelial state in order to fully cover the wound with a new epithelial sheet due to plastic nature of the “partial” EMT (41, 45, 72). Snail2 has been shown to be involved in triggering the “partial” EMT during wounding. Upon EGF treatment in keratinocytes, Erk5 is phosphorylated and induce expression of Snail2, driving the re-epithelialization process in vitro (73). As such, Snail2-deficient mice displayed defects in wound healing (74). In addition, in a N-Acetylglucosaminyltransferase-V (GnT-V) transgenic mouse model, the upregulation of Twist and Snail accelerated cutaneous wound healing in vivo (75).
Metastasis is a rare occurrence of tumor cells in which they lose contact with surrounding cells due to the loss of cell-cell adhesions, enter into circulation, and successfully infiltrate and form metastases at distant organs. EMT has been shown to strongly associate with progression and metastasis of multiple cancers (19, 41, 44). As a key player in metastasis, the EMT program confers malignant traits associated with invasiveness onto epithelial tumor cells to initiate metastatic dissemination (76, 77). In addition, the acquisition of stem cell characteristics in tumor cells after induction of EMT has been observed in multiple cancers, further confirming the necessity of EMT in cancer stem cell formation and tumor dissemination (78-83). More importantly, the MET is believed to be essential for circulating tumor cells to proliferate and seed new tumors at distant organs, revealing the importance of a “reversible” EMT in the multistep process of metastatic cascade (84, 85). Two independent studies have shown that while EMT plays a role in the dissemination of tumor cells, MET facilitates colonization and establishment of new tumors in vivo through two different EMT-associated transcription factors. Tsai et al. demonstrated in a spontaneous squamous cell carcinomas mouse model that the spatiotemporal regulation of EMT and MET through Twist1 has a significant role in increasing number of circulating tumor cells in the circulation and subsequently metastatic frequency (85). Ocana et al., on the other hand, identified the homeobox factor Prrx1 as a novel EMT-inducer. In this study, Prrx1 has been shown to induce EMT in both embryos and tumor cells, whereas the loss of Prrx1 reverses EMT to MET and increases metastasis. While showing Prrx1 cooperates with Twist1 in inducing EMT and invasion, the study also pointed out that knocking down Prrx1 alone or both Prrx1 and Twist, but not Twist1 alone, was able to enhance lung metastasis in a tail vein injection experiment. Strikingly, Prrx1-depleted non-tumorigenic cells acquired stem cell properties, suggesting evidence for the uncoupling of EMT and acquisition of stemness during cancer metastasis (84). However, controversies on whether EMT is a prerequisite for metastasis still exist due to the lack of direct evidence tracing natural occurrence of EMT in spontaneous cancer metastasis in vivo.
Recently, the contributions of EMT in cancer progression and metastasis have been greatly challenged by in vivo data showing EMT is not required for metastasis in several separate mouse models for pancreatic and breast cancers (86, 87). Utilizing a lineage tracing system for EMT in vivo, Fisher et al. provided evidence for a rare occurrence of EMT events in tumor progression in two spontaneous metastatic breast cancer models. In this study, a Cre-switchable fluorescence marker was used to follow the transition of tumor cells from an epithelial (RFP-positive) to a mesenchymal phenotype (GFP-positive) in vivo. Yet, the irreversible nature of this system cannot determine whether the GFP-positive cells eventually returned to epithelial state. This system revealed that tumor cells found in lung metastases remained RFP-positive, indicating that neither Fsp1 nor Vimentin activated-EMT occurred in these metastatic cells. The authors further showed that the inhibition of EMT by overexpressing miR-200 did not impair the formation of metastases in the lungs (86). In another study, Zheng et al. established genetically modified pancreatic cancer mouse models to delete two EMT-drivers, Twist1 and Snail1, and demonstrated that in vivo depletion of neither Twist1 nor Snail1 had significant impact on initiation, progression and metastasis formation in pancreatic cancer (87). Interestingly, both of the studies emphasized synergistic effects of EMT inhibition and chemotherapy in treatment of metastatic cancers (86, 87).
Given the heterogeneous and context-dependent characteristics of cancers, underlying mechanisms of metastasis may vary across different types and stages of cancers. The opposing observations on the roles of EMT in metastatic process reflect the complexity of an open question where further investigations are necessary to provide new insights into the issue. Perhaps, instead of a “black and white” description of EMT in which cells are either in a full epithelial or mesenchymal state, a spectrum of intermediate states between the two poles may provide tumor cells greater flexibility and plasticity in order to survive, outgrowth and establish metastatic colonies (45, 77). An increasing number of reports have described the existence of a “partial” EMT in which cells linger in intermediary phases between epithelial and mesenchymal states and manifest characteristics from both phenotypes (88-90) (Figure 5). Cells with this intermediate hybrid phenotype exhibit a great dynamic potential to spatiotemporally induce or reverse EMT, which ultimately favors tumor progression and metastasis (45, 91).
A knockdown of p63 in human mammary epithelial cells and keratinocytes disrupted cellular adhesion program and increases anoikis, leading to detachment of cells from matrix and ultimately cell death. Overexpression of either ∆Np63α or TAp63γ rescued the phenotype, suggesting an involvement of ∆Np63 in regulation of cell-cell contact, an important aspect of EMT (92). In another study, the loss of p63 in non-transformed and transform squamous cells upregulated genes associated with mesenchymal phenotype such as CDH2 (encodes for N-cadherin), WNT5 and ITGA4, and increased migration and invasion. Reasoning that ∆Np63α is the most abundant isoform expressed in these cells, the authors proposed ∆Np63 as a potential suppressor of invasion and metastasis (93). In line with these findings, ∆Np63 has been demonstrated to inhibit EMT and cell migration through repressing Zeb1 expression by direct transcriptional activity on miR-205 in metastatic prostate and cervical cancer (29, 94). Interestingly, while depletion of ΔNp63α and β isoforms resulted in EMT phenotype in MCF10A cells, ΔNp63γ exhibited ability to upregulate TGFβ and induce EMT (95). More importantly, the loss of ΔNp63 in epithelial-like bladder cancer cells promotes EMT and enhance invasion, while the induction of ΔNp63 in mesenchymal-like bladder cancer cells triggers MET and suppresses invasion (30). This mechanism of action resembles the aforementioned spatiotemporal regulation of Twist1 on EMT during metastasis (85), suggesting a dynamic regulation of ΔNp63 in cancer metastasis. In a different physiological setting, downregulation of TAp63, but not ΔNp63, in MDCK cell lines induced EMT with increased expression of Twist1, Snail1 and Slug. TAp63 is required for tubulogenesis and cell cyst formation (96). However, a mouse model to address the contribution of ΔNp63 in EMT in vivo is still lacking.
The metastatic cascade includes a series of events that enable epithelial primary tumor cells to invade and colonize distant tissues. This process consists of local invasion into surrounding extracellular matrix, intravasation into the blood stream, survival in circulation, arrest and extravasation into distant tissues, adaptation to new environments to form microscopic lesions, and subsequent formation of detectable overt metastatic colonies (Figure 6). This multistep process is orchestrated by a complex network of signaling pathways not only within tumor cells but between tumor cells and surrounding stromal cells (19, 97).
Tumor cells initiate metastatic dissemination process by locally invading surrounding stromal. During this first step, cancerous cells undergo morphological and genetic changes, and acquire malignant properties to increase mobility and remodel extracellular matrix (ECM) to move toward the bloodstream (77, 97). There are three modes which cancer cells are reported to utilize to leave primary tumors: single cell migration, multicellular streaming and collective migration (98, 99). In the single cell migration mode, tumor cells tend to lose cell-cell adhesions, become more motile and leave primary tumors as individual cells. These single epithelial cells can transform into either mesenchymal (invading cells have spindle shape) or amoeboid cell types (invading cells have round shape) to invade through local microenvironment. In the multicellular streaming mode, individual cells move in a multicellular stream or a small strand of cells toward a chemokine gradient. The movement of these cells through adjacent stromal is driven by changes in cytoskeleton structure of individual cells while the cells still maintain a weak transient contact to each other. Collective migration, on the other hand, involves cohesive groups or clusters of tumor cells at the invasive front moving together through the surrounding stroma. These cells retain high expression of E-cadherin to maintain cell-cell adhesion within the invading clusters (77, 98-100). EMT has been shown to strongly involve in local invasion of cancer cells, especially in single cell migration program. Previously, EMT and collective migration-invasion have been proposed to be mutually exclusive due to opposing phenotype observed in invading tumor cells. However, recent evidence has raised a question on whether collective invasion is actually an alternative program to EMT or function cooperatively with EMT to promote cell invasion (77). In fact, breast cancer metastatic colonies have been shown to comprise of multiple different clones, indicating that these metastases might arise from clusters of tumor cells instead of one single cell during metastatic dissemination (101). These groups of invading cells are led by leader cells at the invasive front which exhibit mesenchymal traits and secret proteases to degrade the extracellular matrix in order to pave the way for the remaining cells in the clusters move through the surrounding stroma (77, 102-104). These findings prompt an important question on an intricate interplay between EMT and collective migration-invasion in the initial step of metastatic cascade. Are EMT and collective migration mutually exclusive? Can a complete block of EMT prevent collective migration in tumor cells (77)?
Intravasation describes a series of events in which invading tumor cells utilize various mechanisms to enter the bloodstream or lymphatic vessels. The mechanisms involve transendothelial migration and vessel disruption, which are facilitated by molecular changes in invasive tumor cells (97). Various signaling pathways have been shown to contribute to disruption of the endothelium by cancer cells. Transforming growth factor β (TGFβ) has been shown to function as a molecular switch between single cell and collective migration. Local and transient TGFβ signaling in the primary tumors enhances single cell migration and intravasation. However, prolonged TGFβ signaling reduces lung colonization, suggesting that a spatiotemporal regulation of TGFβ during metastatic cascade is critical for establishment of overt metastases in distant lungs (105). Interestingly, increasing evidence has supported the concept of early dissemination of tumor cells (106, 107). In a recent study, early disseminated breast cancer cells have been demonstrated to account for 80% of metastases in a Her2-driven breast cancer mouse model (107).
Circulating tumor cells (CTCs)
When invading tumor cells successfully enter the circulation, they have to survive a variety of stresses in order to travel to distant tissues and seed new metastatic colonies. Even though aggressive primary tumors release thousands of cells into the circulation a day, only a few of them eventually develop into overt metastases at distant organs, indicating that metastatic dissemination is a very inefficient process. Half-life of CTCs in the bloodstream is rather short, ranging from a few hours to 2 days (108). Given extreme conditions in the circulation, these CTCs have to undergo dramatic molecular changes to adapt and thrive in the new environment. They have been shown to develop resistance to anchorage-dependent growth and anoikis, and escaping mechanisms from innate immune response (97). In addition, a mixed phenotype of epithelial and mesenchymal traits was often found in the CTCs, together with increased stem-like properties, further confirming the plastic nature of these cells which contributes greatly in the subsequent dissemination to distant sites (83, 102, 109). These CTCs can travel in the bloodstream as either individual cells or clusters of cells which are referred as circulating tumor microemboli (CTMs). These CTMs can derive either from intravasation of invading cell clusters, or less likely from proliferation or aggregation of single CTCs. However, while traveling as clusters might give the CTCs sufficient protection from various attacks, which type of CTCs is responsible for metastatic colonization remains unclear (77, 110).
Extravasation starts with the arrest of CTCs or CTMs in blood vessels at distant organs. Then, tumor cells migrate across the endothelium via a process called transendothelial migration similar to intravasation. After successfully nesting in the microvasculature of distant tissues, extravasated cancer cells initiate proliferation and form micro-metastases (77, 110). There are ongoing debates on whether the arrest of CTCs at distant sites is organ-specific based on evidence that several primary cancers favor metastatic dissemination to specific organs, for example breast cancer patients tend to develop metastasis to the bone, lungs and brain. A hypothesis on establishment of a premetastatic niche at distant tissues has been proposed as a mechanism in which primary tumors release various signals to create a more hospitable home for disseminated cancer cells in order to facilitate colonization at specific organs (111).
Metastatic colonization represents the final and most deadly step of tumor progression. In fact, majority of disseminated cells fail to form overt metastatic colonies at distant organs. These successfully extravasated cells may enter a dormancy program in which they retain viability without increase in tumor cell mass. The dormant tumor cells either are quiescent or maintain an equal rate between cell proliferation and apoptosis (110, 112). The dormancy of tumor cells may explain tumor recurrence in cancer patients after removal of primary tumors. An increasing number of studies have identified regulators of tumor dormancy in order to propose potential therapeutic approaches to treat metastatic cancers (113-115). Recently, a study has utilized an in vivo shRNA screening strategy to identify MSK1 as a regulator of metastatic dormancy in ER+ breast cancer (116).
As briefly mentioned in section 1.1.2, given its dominant-negative regulation on p53, TAp63 and TAp73, and high expression in multiple squamous cell carcinomas and metastases, ΔNp63 has been argued to function as an oncogene in human cancers (1, 25, 117). Recent studies have also demonstrated oncogenic activities of ΔNp63 in primary tumors of thymic lymphomas and cutaneous squamous cell carcinomas (cSCC) in vitro and in vivo. While Venkatanarayan et al. identified ΔNp63 as an oncogene in thymic lymphomas in vivo and proposed targeting ΔNp63/TAp63/IAPP axis as a potential therapeutic approach to treat p53-deficient tumors (26), Napoli et al. further explored oncogenic activities of ΔNp63 through ΔNp63/DCGR8/microRNAs axis and utilized HDAC inhibitors to induce tumor regression in both thymic lymphomas and cSCC (118). These findings indicate a tumor-promoting effect of ΔNp63 during tumor development. Similar results were also observed in breast cancer where overexpression of ΔNp63 in patient tumor-derived breast cancer sphere cells promotes tumor formation and metastases to the lungs in xenograft models. The positive regulation of ΔNp63 on metastasis in this study was shown through PI3K/CD44v6 axis (119). In another study, ΔNp63 was shown to promote breast cancer cell migration and invasion via transcriptionally regulating oncogene MTSS1 (120). In keratinocytes expressing HPV16 E6/E7 proteins, p63 was able to drive cell migration and invasion via modulating Src-FAK signaling (121). Adding to the complexity of the pleiotropic roles of ΔNp63 in tumor progression and metastasis, numerous studies have demonstrated ΔNp63 as a suppressor of EMT and potentially metastasis as discussed in section 1.3.4. Several studies also pointed out suppressive activities of ΔNp63 in cancer metastasis through other mechanisms. Bergholz et al. demonstrated that ΔNp63α inhibits cell invasion and metastasis in breast cancer by negatively regulating Erk signaling via direct transcriptional activation of MKP3 (28). These controversial findings arose from the fact that majority of studies on the roles of ΔNp63 in cancer metastasis were performed in vitro with either the lack of using siRNAs or shRNAs specific for ΔNp63, or overexpression of ΔNp63α/β/γ isoforms separately. These approaches did not take into account the existence of TAp63 isoforms and the possible cooperative functions among the isoforms of ΔNp63 in vivo. In addition, the basal level of ΔNp63 in cancer cell lines should be assessed before manipulating expression of ΔNp63 in these cells. Metastatic breast cancer cells with undetectable expression of ΔNp63 may harbor genetic alterations that help them to bypass the requirement of ΔNp63 for metastasis formation.