Paxillin family of focal adhesion adaptor proteins and regulation of cancer cell invasion

4.1. Actin cytoskeleton-mediated effects

4.2. Crosstalk with microtubules

To migrate effectively and persistently, cells must establish front-rear polarization and efficient anterograde vesicle trafficking to enable directed delivery of factors required for migration to the leading edge of the cell and for recycling of focal adhesion-associated integrins (Mellman and Nelson, 2008; Paul et al., 2015; Petrie et al., 2009; Wilson et al., 2018). The microtubule (MT) network is a major mediator of cell polarization through its role in coordinating cell-ECM adhesions, in positioning of the Golgi apparatus, and in facilitating vesicle trafficking (Garcin and Straube, 2019).

Microtubules are 25-nm wide tubes composed of protofilaments of tubulin subunits (Garcin and Straube, 2019). These subunits can be post-translationally modified by acetylation, phosphorylation, and tyrosination, among others, and these modifications affects MT dynamics (Janke and Bulinski, 2011). MT acetylation is commonly associated with longer-lived and more stable microtubules (Janke and Montagnac, 2017). Importantly, acetylated MTs are enriched at the Golgi apparatus and toward the leading edge in cells undergoing both 2D and 3D cell migration, suggesting that this post-translational modification may be important in establishing a polarized phenotype (Doyle et al., 2009; Ryan et al., 2012). One important regulator of MT acetylation and stability is the histone deacetylase 6 (HDAC6), which also deacetylates α-tubulin (Asthana et al., 2013). In fact, HDAC6 inhibitors have been successful in reducing tumor cell migration and invasion in pre-clinical studies and are currently being used in a number of clinical trials (Bian et al., 2018; Li et al., 2018).

It has long been known that paxillin is involved in establishing and maintaining mesenchymal cell front-rear polarity through its LD4 motif. Expression of a paxillin mutant lacking LD4 impairs cells’ ability to orient their Golgi apparatus toward the leading edge and to migrate in a directed manner, but promotes the formation of random, non-directed protrusions, likely as a result of dysregulated Rac1 activity (West et al., 2001). However, recent work has more clearly established another mechanism for paxillin’s crucial role in cell polarity.

Paxillin regulates MT acetylation by binding to and inhibiting HDAC6 (Fig. 4A, MT acetylation, polarity, & trafficking) (Deakin and Turner, 2014). Depletion of paxillin results in a significant decrease in MT acetylation but does not affect overall tubulin expression or distribution. The change in acetylation also reduces cell migration and importantly, can be reversed by treatment with the HDAC6-specific inhibitor tubacin or by HDAC6 knockdown (Deakin and Turner, 2014). Together with data demonstrating HDAC6 enrichment adjacent to focal adhesions and paxillin/HDAC6 binding in cells, these observations suggest that paxillin inhibition of HDAC6 functions to permit normal MT acetylation, thereby promoting cell motility, and is potentially linked to MT-mediated focal adhesion turnover (Efimov et al., 2008; Kaverina et al., 1999). However, since HDAC6 inhibitors can reduce tumor invasion in vivo as mentioned above, it is likely that a delicate balance of HDAC6 activity and microtubule acetylation is required for optimal cell migration and invasion.

Paxillin depletion and resulting MT deacetylation also affect the integrity and positioning of the Golgi apparatus and centrosomes (Deakin and Turner, 2014; Dubois et al., 2017). Paxillin knockdown causes fragmentation of the Golgi and prevents its efficient, polarized orientation toward the leading edge in 1D, 2D, and 3D systems (Dubois et al., 2017). This is likely due to MT-dependent effects, since MTs are known to tether Golgi stacks together and support their cohesion (Lippincott-Schwartz et al., 2000). Interestingly, paxillin is also required for centrosome polarization and cohesion—the maintenance of close proximity between centrioles in non-dividing cells (Fig. 4A) (Dubois et al., 2017). A small amount of paxillin localizes to the centrosome and paxillin depletion leads to significant centriole separation (Dubois et al., 2017). Centrioles are typically closely associated in normal cells during interphase (Gönczy, 2015). However, they are often aberrantly duplicated and separated in cancer cells and it is believed that these changes contribute to genomic instability due to chromosome segregation errors during mitosis (Pease and Tirnauer, 2011). Centriole cohesion also depends on MT integrity, suggesting that paxillin-mediated MT acetylation changes may be responsible for the separation phenotype previously observed following paxillin knockdown in MDA-MB-231 cells (Burakov et al., 2003; Dubois et al., 2017). Once again, these centrosome- and Golgi-related phenotypes can be rescued in paxillin-depleted cells by tubacin treatment or HDAC6 RNAi, confirming that they are mediated through paxillin-dependent effects on MT acetylation (Deakin and Turner, 2014; Dubois et al., 2017). However, the precise mechanism connecting paxillin, MT acetylation, and Golgi/centrosome cohesion and orientation has yet to be clearly elucidated.

Importantly, the Golgi apparatus and centrosomes are both microtubule-organizing centers (MTOCs) in the cell (Wu and Akhmanova, 2017). Therefore, paxillin is required for the cohesion and effective polarization of each of the primary MTOCs in both normal and cancer cells. This mechanism is independent of paxillin localization to focal adhesions, since the observed phenotypes are evident even when cells were plated on poly-l-lysine-coated dishes, which prevents focal adhesion formation (Dubois et al., 2017). Thus, reduced paxillin expression in certain cancers could result in loss of cell polarization to impair effective cell migration/invasion, while a loss of centriole cohesion could promote carcinogenesis due to genomic instability.

Pharmacologic inhibition of FAK, an important signaling partner of paxillin, also prevents paxillin localization to the centrosome and mimics many of the phenotypes which occurred following paxillin depletion, suggesting that these proteins likely work together in regulating MT acetylation (Dubois et al., 2017). FAK stimulates MT acetylation through RhoA activation, so activated FAK could potentially modulate paxillin’s regulation of HDAC6 activity (Palazzo et al., 2004).

Paxillin’s regulation of MT acetylation also indirectly impacts anterograde vesicle trafficking (Fig. 4A) (Dubois et al., 2017). As mentioned previously, effective transport of pro-migratory components to the leading edge is crucial for optimal directed migration, and this transport often occurs along MTs (Naslavsky and Caplan, 2018). In paxillin-depleted cells, trafficking of GFP-tagged vesicular stomatitis virus G (VSVG)—a marker of anterograde trafficking—from the endoplasmic reticulum to the Golgi and from the Golgi to the plasma membrane are both impaired (Dubois et al., 2017). The mechanism by which this occurs remains to be elucidated, but it is possible that paxillin-dependent changes in MT acetylation affect the binding and/or activity of the molecular motors that traffic cargo along MTs or the activity of Rab GTPases, which function in vesicle trafficking and integrin recycling and therefore, focal adhesion dynamics (Garcin and Straube, 2019; Paul et al., 2015). Regardless of the mechanism, this result suggests an interesting focal adhesion-independent mechanism by which paxillin regulates cell migration.

Using a conditional paxillin knockout mouse, it has been shown that paxillin is also critical for the establishment of apical-basal polarity in normal mouse mammary glands in vivo and the formation of 3D polarized acini in a mammary epithelial cell in vitro model system, including the apical enrichment of acetylated MTs (Fig. 4B) (Xu et al., 2019). This newly described role for paxillin in normal epithelial cells could become dysregulated during carcinogenesis, as cancer cells often undergo an EMT and lose their apical-basal polarity prior to invasion (EMT will be discussed in a subsequent section). Indeed, 3D acini of mammary epithelial cells derived from the normal mammary gland of the paxillin knockout mice closely resemble the disorganized morphology and defects in cell-cell junction organization seen in tumor organoids generated using tumor fragments from conditional paxillin knockout MMTV-PyMT breast cancer mice (our unpublished results). Remarkably, pharmacologic inhibition of HDAC6 with tubacin rescued the defects in acini polarization (Xu et al., 2019), suggesting that paxillin regulation of MT acetylation not only affects front-rear polarization in single cells, but can also affect the polarization of multicellular units, potentially through its effects on intracellular trafficking and crosstalk with the actin cytoskeleton (Booth et al., 2014).

4.3. Crosstalk with intermediate filaments

Intermediate filaments (IFs) generally play an important role in determining cell morphology and response to mechanical stress due to their characteristic structural integrity (Leduc and Etienne-Manneville, 2015). IFs are built from several proteins and protein families, including cytokeratins, lamins, neurofilaments, desmin, and vimentin, and their expression varies widely by cell type and between normal and cancerous cells (Lowery et al., 2015). In fact, several IFs have been used clinically as markers for disease grade and tumor invasiveness for decades (Osborn and Weber, 1983). For example, vimentin expression is typically limited to mesenchymal cells, so its expression in epithelial cells may indicate a more malignant phenotype (Korsching et al., 2005). Furthermore, cytokeratins are used as markers for tumors of epithelial origin, and individual cytokeratins may be enriched in certain cell populations. For example, cells leading strands of collectively migrating cells in breast cancer are known to express cytokeratin 14 (Diepenbruck and Christofori, 2016). Different IFs have differing effects on cell migration (Leduc and Etienne-Manneville, 2015).

Vimentin is a type III IF expressed primarily in fibroblasts and cells that have undergone EMT (Lowery et al., 2015). Recent studies have shown a role for vimentin in promoting directional persistence of migrating cells, and vimentin filaments occasionally terminate at focal adhesions, suggesting some degree of crosstalk between the actin and IF cytoskeletons at these sites (Battaglia et al., 2018), a process that appears to be dependent on Hic-5 (Vohnoutka et al., 2019). Additionally, vimentin expression in fibroblasts protects against nuclear rupture and DNA damage as cells migrate, by forming a cage around the nucleus (Patteson et al., 2019). A similar mechanism could be important for protecting cancer cells migrating through tight spaces in the ECM, especially since DNA damage is an important part of carcinogenesis through its impact on genomic stability (Hanahan and Weinberg, 2011). Further evidence of focal adhesion/vimentin crosstalk and the role of Hic-5 will be discussed below.

4.4. Epithelial-mesenchymal transition

As briefly described above, epithelial-mesenchymal transition (EMT) is a key process in both normal development and in cancer cell invasion and metastasis, although its role in cancer is highly controversial. (Diepenbruck and Christofori, 2016; Ye and Weinberg, 2015). This controversy stems largely from the inability to monitor the transient and reversible EMT that occurs in vivo and evidence that EMT may not be required for metastasis in certain cancers (Fischer et al., 2015). Furthermore, immunohistochemical staining of tumors often show cells migrating collectively, which does not require a complete EMT; in fact, cells which have undergone a total EMT cannot invade collectively (Clark and Vignjevic, 2015). However, new tools have improved our understanding of EMT and its contribution to invasion in certain contexts, and a whole range of intermediate EMT states have been described (Pastushenko et al., 2018).

The process of EMT involves the disassembly of cell-cell junctions, the down-regulation of cell-cell junction components such as E-cadherin, and reorganization of the actin cytoskeleton to promote cell-matrix adhesion and increased migration (Brabletz et al., 2018). EMT also results in a switch from epithelial markers (including several cytokeratins) to mesenchymal markers (including vimentin) (Ye and Weinberg, 2015). Mesenchymal cells are much more motile than epithelial cells in both 2D and 3D culture systems (Aiello et al., 2018; Brabletz et al., 2018). TGF-β, a versatile cytokine which can, depending on the context, act as both an oncogene and a tumor suppressor, is a potent inducer of EMT and stimulates upregulation of Hic-5 expression which, as noted, is typically very low in epithelial cells (Fig. 3A) (Tumbarello and Turner, 2007). As mentioned above, Hic-5 is also known as transforming growth factor beta 1 induced transcript 1 (TGFB1i1) and participates in the non-canonical arm of TGF-β signaling, which primarily impacts RhoGTPase signaling and the actin cytoskeleton (Varney et al., 2016).

Following TGF-β stimulation, Hic-5 facilitates EMT through a downstream pathway involving RhoA and ROCK (Tumbarello and Turner, 2007). Hic-5 RNAi after TGF-β stimulation results in reduced cell migration and the retention of epithelial morphology. Additionally, Hic-5 RNAi prevents TGF-β-induced RhoA activation, while ectopic Hic-5 overexpression promotes ROCK kinase activity and the formation of ROCK-dependent stress fibers (Tumbarello and Turner, 2007). Interestingly, inhibiting RhoA/ROCK activity by expressing dominant negative RhoA or by using a ROCK inhibitor prevents an increase in Hic-5 expression following TGF-β-induced EMT, suggesting that Hic-5 and RhoA/ROCK work in a positive feedback loop to promote EMT and maintain a mesenchymal phenotype. As discussed earlier, Hic-5 upregulation in response to TGF-β also promotes the formation of invadopodia that are important for the resulting mesenchymal invasion (Pignatelli et al., 2012a) (Fig. 3B).

5.1. Extracellular matrix deposition

During wound healing, resident tissue fibroblasts differentiate to become myofibroblasts, which participate in wound contraction and fibrotic changes (Bochaton-Piallat et al., 2016; Dabiri et al., 2008). In a similar manner, resident tumor fibroblasts are activated by ECM rigidity- and signaling-mediated changes to become smooth muscle actin-positive, highly contractile cancer-associated fibroblasts (CAFs), which generate and remodel the cancer ECM and play a significant role in tumor progression (Kalluri, 2016). Crosstalk between CAFs and tumor cells can produce a feed-forward loop resulting in increased tumor cell invasion and metastasis through changes in the rigidity and molecular composition of the ECM and secretion of cytokines. For instance, more rigid and aligned matrices rich in fibrillar collagens and fibronectin generally promote cell invasion and, clinically, increasing tumor rigidity is associated with cancer progression, metastasis, and worse prognoses in breast and colorectal cancer, among others (Emon et al., 2018; Levental et al., 2009; Wei and Yang, 2016). Increased ECM rigidity can result from a number of factors, including increased expression of the collagen cross-linking protein lysyl oxidase (Omoto et al., 2018; Wei et al., 2017) or increased deposition of ECM fibers such as collagen and hyaluronan (Gkretsi and Stylianopoulos, 2018). The relative abundance of ECM proteins is also important for tumor stroma crosstalk. For example, increased stromal expression levels of fibronectin and tenascin C correlate with decreased overall patient survival in breast cancer (Ioachim et al., 2002). Increased collagen density and alignment, quantified by a “tumor-associated collagen signature” score, are also associated with negative patient outcomes (Provenzano et al., 2008).

Information about ECM rigidity and molecular composition is at least partially transmitted through focal adhesions in both tumor cells and CAFs. Multiple focal adhesion proteins can act as mechanosensors, including paxillin and Hic-5, as well as vinculin, talin, and integrins (Jansen et al., 2015). Tumor cells can then respond to these changes through complex signaling pathways involving activation of proteins such as Rho GTPases, MAPKs (mitogen-activated protein kinases), YAP/TAZ (Yes-associated protein/transcriptional coactivator with PDZ-binding motif), FAK, and MRTF (myocardin-related transcription factor) (Maller et al., 2013; Provenzano and Keely, 2011). Together, these pathways induce changes in cytoskeletal organization and gene expression that will be discussed in more detail below.

Hic-5 expression regulates myofibroblast function in non-cancerous wound healing and is also important in regulating CAF function in numerous cancers and pre-cancerous lesions (Fig. 5, CAFs & ECM deposition) (Dabiri et al., 2008; Du et al., 2019; Goreczny et al., 2017; Omoto et al., 2018; Varney et al., 2016). For example, Hic-5 deficiency in pancreatic fibroblasts (known as pancreatic stellate cells or PSCs) protects from cerulein-induced pancreatitis in mice (Chen et al., 2020). In humans, chronic pancreatitis predisposes patients to pancreatic cancer, and the PSCs of patients with pancreatitis exhibit increased Hic-5 expression. If pancreatic cancer develops, Hic-5 expression in PSCs then promotes proliferation, survival, and invasion of pancreatic cancer cells in vitro when these cells are cultured together. Additionally, Hic-5 depletion in PSCs results in decreased MMP-9 expression in the fibroblasts (Qian et al., 2020). Furthermore, higher Hic-5 expression correlates with decreased post-operative survival. These data suggest that Hic-5-mediated fibrosis may play an important role in tissue stiffening and other pre-cancerous changes that may ultimately lead to cancer.

A recent study using patient-derived CAFs shows that Hic-5 expression is important in colorectal cancer (Omoto et al., 2018). Relative to normal fibroblasts from tumor-adjacent healthy tissue, colorectal CAFs have significantly higher Hic-5 expression. Hic-5 expression was induced in the normal fibroblasts by treatment with culture media that had been conditioned by CAFs or with cytokines including TGF-β, IL-1 β, and SDF-1/CXCL12. Colon cancer cell growth was shown to be inhibited in the presence of CAFs with low Hic-5 levels in in vitro assays, compared with cells cultured with wild-type CAFs (Omoto et al., 2018). Most interestingly, Hic-5 knockout mice did not develop any tumors in an azoxymethane-induced colorectal cancer model, in contrast to a >50% incidence in wild-type mice (Omoto et al., 2018). This was suggested to be partially due to Hic-5 stimulation of lysyl oxidase and collagen I expression. Lysyl oxidase is a secreted protein responsible for cross-linking collagen in vivo, and its expression correlates with increasing tumor stiffness, a negative prognostic marker in many cancers (Levental et al., 2009). Therefore, Hic-5 expression in human CAFs promotes a pro-tumorigenic stroma in colorectal cancer. Similar results were observed in oesophageal cancer CAFs (Du et al., 2019).

In the MMTV-PyMT mouse model of breast cancer, Hic-5 is surprisingly not expressed in the invasive tumor cells but its expression in the CAFs promotes tumor growth, decreases tumor latency, and enhances lung metastasis (Goreczny et al., 2017). Immunostaining of tumors from Hic-5 heterozygous and knockout mice show that collagen and fibronectin staining in the tumor stroma is decreased in the absence of Hic-5. This may be partially due to Hic-5’s role in promoting fibrillar adhesion formation and maturation through its phosphorylation-dependent interaction with tensin-1 (Fig. 5) (Goreczny et al., 2018). These adhesions are important for assembly of fibronectin fibrils, which also affects collagen deposition and organization in vivo. Isolated CAFs from Hic-5 knockout mouse tumors are also less contractile than their Hic-5-expressing counterparts as a result of significantly reduced RhoA activity. In addition, the Hic-5 knockout CAFs produce thinner 3D CDMs with poorly aligned fibronectin and collagen fibers in vitro (Goreczny et al., 2017). This is a potentially clinically relevant phenotype since, as mentioned above, increasing density and alignment of tumor ECMs are inversely correlated with patient outcomes, including overall survival (Provenzano et al., 2008), Interestingly, MDA-MB-231 breast cancer cells seeded in CDMs generated by Hic-5-expressing CAFs are less likely to migrate with an amoeboid phenotype and have decreased directional persistence compared to those seeded in CDMs from Hic-5 knockout CAFs (Goreczny et al., 2017). These changes in migration phenotype are likely one of the contributing factors to the decrease in lung metastases observed in Hic-5 knockout mice.

As noted above, there is growing evidence for crosstalk between the actin and IF cytoskeletons, which likely plays an important role in controling cell shape and motility (Burgstaller et al., 2010; Huber et al., 2015; Osmanagic-Myers et al., 2015).This, in turn, is likely important in myofibroblasts as they organize the surrounding stromal matrix. Evidence of a key role for Hic-5 in connecting vimentin organization with focal adhesions and, therefore, the actin cytoskeleton, was recently described. A collapse of the vimentin cytoskeleton was observed in Hic-5 knockout CAFs and normal lung fibroblasts (Vohnoutka et al., 2019). In contrast, depletion of paxillin did not cause a similar perturbation of the vimentin cytoskeleton. Hic-5 knockout CAFs also exhibit a reduction in centrally localized F-actin stress fibers (an actin “hole”), which could be directly linked to this vimentin phenotype as both are rescued in knockout CAFs by inhibition of the Rho GTPase Cdc42, and phenocopied by pharmacological inhibition of formins in Hic-5 expressing CAFs (Vohnoutka et al., 2019). Together, Hic-5-mediated regulation of Rho GTPase and formin activity may co-ordinately affect both the actin and vimentin cytoskeletons to allow disease-relevant changes in fibroblast phenotype. For instance, low Hic-5 and a collapsed vimentin cytoskeleton may be necessary for a pro-migratory fibroblast phenotype, while high Hic-5 and robust actin and vimentin cytoskeletons are necessary for a highly contractile phenotype, where cells deposit and remodel the surrounding ECM (Fig. 5). Although these observations were made in CAFs, it is interesting to speculate whether a similar phenotype may occur in cancer cells depending on the level of Hic-5 expression and, if so, whether a collapsed vimentin network can still protect against nuclear rupture or facilitate cancer cell amoeboid motility.

5.2. Tumor neovascularization

The tumor stroma also includes blood vessels which support the growing tumor by providing oxygen and nutrients and removing carbon dioxide and waste (Nishida et al., 2006). From early in tumorigenesis, tumor cells promote neovascularization of the tumor stroma by secreting factors which activate an “angiogenic switch,” including vascular endothelial growth factor-A (VEGF-A) and fibroblast growth factor (FGF) (Hanahan and Weinberg, 2011). As tumor progression proceeds, this neovascularization is modulated by both tumor and stromal cells. The process of angiogenesis involves activation of endothelial cells by pro-angiogenic factors, production of proteases to degrade the matrix, cell migration/invasion and proliferation, tube formation, synthesis of new basement membrane, and recruitment of supporting cells including smooth muscle cells and pericytes (Rajabi and Mousa, 2017). The degree of neovascularization varies by cancer type, but most solid tumors are highly vascularized with aberrant, convoluted vessels that support the tumors’ high metabolic requirements. The success of anti-angiogenic drugs such as bevacizumab and cabozantinib in treating a variety of cancers highlights the importance of neovascularization for tumor growth and survival (Rajabi and Mousa, 2017).

Hic-5 is highly expressed in endothelial cells and vascular smooth muscle cells (Kim-Kim-Kaneyama et al., 2011, 2012). Depletion of Hic-5 reduces endothelial sprouting and lumen formation in a 3D in vitro assay, suggesting Hic-5 may play a role in regulating vascularization (Dave et al., 2016). Additionally, inducing sprouting with pro-angiogenic factors including VEGF, FGF, and sphingosine 1-phosphate also increases protein complex formation between Hic-5, FAK, and the matrix metalloproteinase MT1-MMP, which is a membrane-anchored protease critical for endothelial cell invasion during sprouting (Dave et al., 2016). The interaction with MT1-MMP is mediated by Hic-5’s LIM2 and LIM3 domains. While Hic-5 knockdown does not alter MT1-MMP activity, it does reduce its localization at the membrane during endothelial sprouting, which results in impaired invasion.

Hic-5 also plays a role in vascular remodeling after injury due to its mechanosensitive properties (Kim-Kaneyama et al., 2008, 2011). When undisturbed, the femoral arteries of Hic-5-null mice appear similar to those of wild-type mice (Kim-Kaneyama et al., 2008). However, after surgical injury by a wire, the arteries in Hic-5-null mice have delayed recovery and higher numbers of apoptotic cells in the vascular walls. When mechanically stretched in vitro, smooth muscle cells derived from the Hic-5-null mice exhibit an increased cytoplasmic localization of vinculin, another mechanosensing focal adhesion protein that also binds actin (Atherton et al., 2020; Humphrey et al., 2014; Kim-Kaneyama et al., 2005). This provides further evidence that stable focal adhesions help protect against stretch-induced apoptosis and that Hic-5 is required for stable integration of vinculin and other focal adhesion proteins, as previously discussed in the context of mesenchymal tumor cell morphology. This idea is well-supported by other studies and is an important consideration in cells in the vascular wall, which are constantly experiencing mechanical changes (Katsumi et al., 2005; Kim-Kaneyama et al., 2005). Although these studies don’t examine vascular injury in a cancer-related context, the processes of normal vascular wound healing and angiogenesis bear many similarities, suggesting that Hic-5 may be an interesting target to pursue when investigating tumor vascularization (Dabiri et al., 2008; Kareva et al., 2016; Kim-Kaneyama et al., 2012).

Although less well-studied, paxillin also plays a role in remodeling the vasculature. The angiogenic cytokine angiopoietin-1 requires paxillin expression for effective promotion of endothelial cell polarization, migration, and sprouting (Boscher et al., 2019). Angiopoietin-1 binding induces PAK2-mediated paxillin phosphorylation which is necessary for activation of the Rho GTPase Cdc42 at the leading edge and recruitment of the polarity protein Par3. Paxillin also mediates endothelial permeability through control of Rho GTPase signaling, which may be important in allowing tumor cell intravasation and extravasation (Birukova et al., 2009; García-Román and Zentella-Dehesa, 2013; Gawlak et al., 2014).

5.3. Immune cell function

A healthy and intact immune system is a critical defence measure against cancer development (Gonzalez et al., 2018). Accordingly, tumors often demonstrate highly dysregulated immune environments with suppression of normal immune pathways that would destroy the tumor, and activation of inflammatory conditions that promote tumor progression (Pandya et al., 2016). While an in-depth analysis of immune dysregulation is beyond the scope of this review, it is important to note that paxillin has been shown to modulate immune cell signaling, and that a healthy and properly-regulated immune response can impair tumor cell migration and invasion (Gauthier et al., 2017;Herreros et al., 2000; Robertson and Ostergaard, 2011).

In a mouse orthotopic colon cancer model, down-regulation of paxillin expression in macrophages indirectly inhibited tumor growth through effects on the immune system (Zhang et al., 2018). Macrophages are phagocytic immune cells that play roles in both the innate and adaptive immune response, and the M2 class of tumor-associated macrophages (TAMs) are anti-inflammatory, promote tumor growth and progression, and correlate with negative patient prognoses (Aras and Zaidi, 2017). Paxillin expression is upregulated during M2 macrophage activation (Zhang et al., 2018). Furthermore, paxillin depletion inhibits M2 polarization and reduces macrophage invasion and proliferation in vivo, consistent with its role in epithelial cell polarization (Dubois et al., 2017; Xu et al., 2019). Co-injection of colon cancer cells with either wild-type or paxillin-depleted macrophages in a nude mouse model results in significantly decreased tumor volume in mice with paxillin-depleted macrophages, suggesting that paxillin can modulate the immune response in this circumstance to promote tumor growth (Zhang et al., 2018).

Multiple studies have addressed a role for paxillin in T-cells, an important class of cytotoxic cells in the adaptive immune system that are responsible for killing epithelial tumor cells (Thommen and Schumacher, 2018). Cytotoxic T-cells interface with their target cells at immunologic synapses, which are supramolecular clusters of several transmembrane and membrane-associated proteins. When one of these immunologic synapse proteins, CD103 (the α subunit of integrin α_Eβ₇), binds to a tumor-specific antigen, paxillin is phosphorylated and binds to the CD103 cytoplasmic tail at the immunologic synapse (Gauthier et al., 2017). If paxillin expression is knocked down by shRNA or its phosphorylation is inhibited by the Src kinase inhibitor saracatinib, T-cell spreading and effector activities are impaired in vitro, including their abilities to lyse tumor cells and release the anti-tumor cytokine INFγ (Gauthier et al., 2017).

Interestingly, another role for paxillin in T-cells is related to its ability to regulate the positioning of the MTOC, as discussed earlier in a different context. When a T-cell binds to a target cell, paxillin is recruited to both the immunologic synapse and the MTOC (Herreros et al., 2000; Robertson and Ostergaard, 2011). Importantly, paxillin contributes to reorientation of the MTOC toward the target cell. This is a crucial step in T-cell activation because it allows for directed delivery of cytokines to the plasma membrane for secretion to enhance target cell killing (Martín-Cófreces et al., 2008). These and other studies suggest that paxillin expression in T-cells is important for effective anti-tumor immunity, potentially through HDAC6-mediated changes in MT acetylation and MTOC polarization.

While paxillin is expressed in most circulating and tissue-resident immune cells, Hic-5 exhibits only slightly enhanced RNA expression in basophils and minimal expression in any other immune cells, suggesting it has little direct impact on immune function (paxillin and Hic-5 expression data from proteinatlas.org, v19.3) (Thul et al., 2017). Similarly, although leupaxin expression was first identified in leukocytes, there a lack of subsequent research indicating an important role for this protein in cancer-related immunity (Lipsky et al., 1998).

6.1. Regulation of paxillin expression by miRNA

Paxillin expression can be post-transcriptionally regulated by miRNAs during cancer invasion and metastasis. miRNAs are a class of 19–22 nt single strand non-coding RNAs that typically act by binding to target mRNAs, thus inducing the degradation of the mRNAs or inhibition of their translation (Bartel, 2009; Kusenda et al., 2006). miRNAs are highly involved in tumor progression processes such as tumor cell adhesion, migration, invasion, and proliferation (Hayes et al., 2014; Peng and Croce, 2016). miR-218 negatively regulates paxillin translation (Wu et al., 2010). Accordingly, low miR-218 levels in non-small cell lung cancer correlate with high paxillin expression and, furthermore, with decreased patient survival and increased relapse rates (Wu et al., 2010). A similar miR-218-dependent, post-transcriptional regulation of paxillin has been observed in oral cavity squamous cell carcinoma (Wu et al., 2014b). miR-137 inhibits paxillin translation in colorectal cancer, so low miR-137 levels result in high paxillin expression and reduced survival, (Chen et al., 2013). However, miR-137-mediated paxillin inhibition in lung cancer cells results in reduced proliferation and migration, suggesting that the result of paxillin down-regulation is again, context-specific (Bi et al., 2014). Several other studies in cancer cell lines have shown that paxillin expression can be directly inhibited by other miRNAs, thus impairing the ability of cancer cells to migrate and invade (Li et al., 2015; Matsuyama et al., 2016; Qin et al., 2015; Tao et al., 2016). For example, miR-145 interacts with paxillin mRNA in human colorectal cancer cells to suppress its translation (Qin et al., 2015), and miR-27b inhibits paxillin expression and integrin-mediated cell adhesion (Matsuyama et al., 2016).

In addition to direct binding to paxillin mRNA, miRNAs can also regulate paxillin through indirect mechanisms. For example, the miR-200 family/ZEB1 axis potentiates signaling through integrin β1 in response to binding extracellular collagen I, which promotes FAK-dependent paxillin phosphorylation and thereby increases lung cancer cell invasion and metastasis (Ungewiss et al., 2016). In ovarian cancer, miR-708 inhibits the GTPase activating protein Rap1B, which is a regulator of integrin-based adhesion signaling (Hattori and Minato, 2003). This inhibition results in reduced paxillin phosphorylation at focal adhesions and decreased lung metastases (Lin et al., 2015).

6.2. Regulation of gene expression by Hic-5 and paxillin

As mentioned above, Hic-5 is also known as androgen receptor coactivator 55kDa protein (ARA55) (Fujimoto et al., 1999). In addition to its localization at focal adhesions, Hic-5 can also localize to the nucleus as a coactivator for the androgen receptor transcription factor (Heitzer and DeFranco, 2006a; Shibanuma et al., 2003), and studies have shown that it can regulate expression of numerous genes by promoting the formation of a transcriptional complex with the transcriptional coactivator p300 and transcription factors Sp1 and SMAD3 in a variety of cell types (Heitzer and DeFranco, 2006b; Shibanuma et al., 2012; Yang et al., 2000). One particularly relevant study examined Hic-5 responsive gene expression changes in response to glucocorticoid treatment in U2OS osteosarcoma cells (Chodankar et al., 2014). Four of the top 15 gene ontology categories represented in the Hic-5-responsive group of genes were related to cell migration and adhesion, suggesting that Hic-5 may regulate migration indirectly through transcriptional changes.

Hic-5 also plays an important role in transducing mechanical and actin cytoskeleton-related signals to regulate gene expression through pathways including the actin-MRTF-SRF pathway (Varney et al., 2016). MRTFs bind to, and are sequestered by monomeric G-actin, but incorporation of G-actin monomers into F-actin fibers release MRTFs, allowing them to translocate to the nucleus and interact with the transcription factor serum response factor (SRF) (Olson and Nordheim, 2010). Therefore, actin polymerization into stress fibers promotes MRTF-SRF-dependent transcription of genes including several genes involved in contractility, such as smooth muscle actin and calponin (Crider et al., 2011). In fibroblasts, MRTF-SRF signaling promotes Hic-5 expression (Wang et al., 2011), but Hic-5 is also required for TGF-β-dependent nuclear localization of MRTF-A and promotion of smooth muscle actin expression (Varney et al., 2016). These data indicate that Hic-5 promotes fibroblast differentiation via a TGF-β-dependent feed-forward loop that regulates expression of actin cytoskeleton- and contractility-related genes.

Hic-5 may also play a role in other mechanosensitive transcription pathways involving the YAP/TAZ-TEAD (transcriptional enhancer factor-domain) axis. YAP/TAZ nuclear localization and activation correlate with several Hic-5-related phenotypes, including EMT and ECM rigidity-dependent fibroblast activation, although a direct role for Hic-5 in YAP/TAZ signaling has not been reported (Calvo et al., 2013; Shao et al., 2014). Furthermore, YAP/TAZ activity is modulated by several pathways, including a mechanosensitive regulation similar to MRTF-SRF signaling: high ECM rigidity and/or cellular contractility promotes release of YAP/TAZ from cytoplasmic sequestration and allows nuclear translocation (Totaro et al., 2018). YAP/TAZ activity is elevated in many cancers, and results in increased aggressiveness and cell proliferation, as well as induction of expression of several actin cytoskeleton- and focal adhesion-related genes, including formins, integrins, and RhoGEFs like the Dock proteins (Pocaterra et al., 2020; Totaro et al., 2018). Therefore, a role for Hic-5 in mediating YAP/TAZ transcription of proteins relevant to the actin cytoskeleton and cancer cell migration would not be surprising.

Although paxillin contains a nuclear export sequence and localizes most strongly to focal adhesions, several studies have demonstrated that it can also be shuttled into the nucleus to regulate androgen- and MAPK signaling pathway gene transcription (Kasai et al., 2003; Ma and Hammes, 2018). For example, RNAseq analysis in prostate cancer cells produced a panel of over 1000 paxillin-dependent, androgen-responsive genes (Ma et al., 2019). Paxillin expression in these cells increased transcription of pro-proliferative genes, including the CyclinD/Rb/E2F pathway, while it decreased transcription of pro-apoptotic genes, including CASP1 and TNSF10.

Although much of the signaling through paxillin family members occurs directly at focal adhesions, their respective roles in modulating gene expression is clearly also important to consider when studying their impact on tumor progression.