Introduction

Cytoskeletal proteins form the backbone of the different types of structural polymers found in every eukaryotic cell. Such polymers include microfilaments (MF), mini-filaments, microtubules (MT) and intermediate filaments (IF). Each polymer has a relatively homogeneous composition. Monomeric cytoskeletal proteins bind in a head-to-tail manner to form long chains of different geometries and biophysical properties. These monomers include actin (which forms MF), myosin (mini-filaments), tubulin (MT), and various families of IF proteins, including keratins, desmins, glial fibrillary acidic protein (GFAP), peripherin, vimentin, internexins, nestins and others (reviewed in [1]). MFs and mini-filaments enable cells to adapt to their surroundings. They exert several roles in cell division and support cell migration in physiological and pathological contexts, for example during invasion and metastasis. MTs are essential as they form the physical scaffold that mediates an even separation of genetic material during cell division, but they play limited roles in cell migration. IFs confer mechanical resistance to the cells.

Like every protein in the eukaryotic proteome, cytoskeletal proteins are substrates of diverse protein kinases. Phosphorylation changes their interactive and dynamic properties with respect to their non-phosphorylated forms. In the specific case of cytoskeletal proteins, phosphorylation controls their assembly and disassembly affinity and dynamics, as well as the biochemical and biophysical properties of the polymers themselves. Phosphorylation also modulates their interactome, which also affects the stability and function of the polymers.

Genetic modifications affecting protein kinases are very frequent in cancer [2]. The most typical are mutations or deletions that cause loss of function or increased catalysis [3]. These two outcomes also emerge when portions of kinases become fused with incorrect pieces of DNA as part of genomic cancer recombination, leading to abnormal activation. One example is the Philadelphia translocation, which is typical of chronic myeloid leukemia (CML) cells. It consists of a reciprocal translocation between parts of human chromosomes 9 and 22 that leads to the production of a constitutively active Abelson kinase (BCR-ABL), which triggers uncontrolled proliferation by phosphorylating multiple substrates [4]. On the other hand, activating mutations may lead to unexpected effects on the different cytoskeletal systems. For example, mutations to the small GTPase RhoA may lead to increased activation of proteins that control mini-filament formation [5]. These events disturb the delicate balance of MFs and mini-filament dynamics that enable cells to maintain their form and function in the context of the host tissue.

These and other examples found in the latter sections of the present work highlight the fact that alterations of the phosphorylation of cytoskeletal proteins caused by cancer-related mutations may change the dynamics, architecture, and function of the different cytoskeletal polymers. These events cause aberrant molecular behaviors that may confer specific properties to cancer cells, e.g., increased cell migration, invasion, division, mechanosensing, etc.

The study of these modifications is complicated by the existence of multiple isoforms of these proteins, some of which have non-overlapping functions. There are many isoforms of actin, tubulin and myosin, and multiple forms and variants of IF. How phosphorylation impacts every isoform of every cytoskeletal protein is not only impossible to describe, but mostly unknown. Because of this, we describe only the current state of the art regarding the phosphorylation of selected isoforms of actin, myosin II, tubulin and vimentin. We focus on the functional effect of these phosphorylations, and what the consequences would be if the extent of these phosphorylations was altered in the context of cancer progression. While phosphorylation has proven crucial for the function of some of these filament-forming polymers, e.g., myosin II, the function of many of the phosphorylations described here is still unexplored. Thus, a major goal of this work is to provide a wide, yet incomplete, perspective of the field, identifying potential hotspots that may be amenable to specific targeting to treat diverse forms of cancer.

Actin

Actin forms MFs by adenosine triphosphate (ATP)-dependent polymerization. Together with myosin II mini-filaments, MFs constitute the contractile apparatus of animal cells. A vast array of nucleators, cross-linkers and other binding partners regulate multiple aspects of its ability to form filaments and the multiple cellular functions they enable [6, 7].

There are multiple isoforms of actin, including muscle-specific and non-muscle [8]. Due to its abundance, ubiquity and high degree of homology among isoforms (Figure 1), we focus on cytoplasmic, β-actin. Human β-actin (ACTB gene; Uniprot #P60709) is located in chromosome 7p22.1 in humans. β-actin forms MFs in every non-muscle cell lineage, mediating protrusion, assembly of contractile structures and other motility-related structures, e.g., podosomes [9]. Somatic mutations of the ACTB gene associated to cancer have not been reported. However, the filamentous state of actin is a checkpoint for cell proliferation that is deregulated in several types of cancer [10]. Although the effect of actin phosphorylation in MF dynamics has yet to be studied in detail, some phosphorylation events have been described that potentially affect actin filamentation and/or disassembly (see Table 1 for a list). For example, serine (Ser)33 appears phosphorylated in multiple phospho-proteomic analyses, including diverse types of cancer cells that include human epidermal growth factor receptor 2 (HER2)-positive, luminal A and triple negative breast cancer [2, 11] and lung cancer (http://phosphosite.org, search term = ACTB, Ser33). Its phosphorylation lies downstream of polo-like kinase 1 (PLK1) since the specific inhibitor BI 4834 abrogates it [12]. Although it is unclear whether PLK1 directly phosphorylates β-actin in Ser33, the crucial role of this kinase in cancer cell division [13] suggests that PLK1-dependent actin phosphorylation may control actin function (by controlling actin polymerization and/or cross-linking) in cancer cell proliferation.

Display full size

Figure 1. 

Homology between human actin isoforms

Another potentially important residue is Tyr53. It is conserved from Dictyostelium discoideum to humans. It resides in the D-loop of actin [14], which is essential for actin polymerization [15]. Tyr53 phosphorylation decreases the affinity of actin monomers for each other, causing filament shortening [16]. This mechanism is critically important in the central nervous system. During synaptogenesis, Tyr53 phosphorylation increased actin turnover [17]. Although the mitogen-activated protein kinase kinase (MEK) inhibitor U0126 inhibits its phosphorylation (and that of Tyr362, of unknown function), it is unlikely that this kinase, which phosphorylates Ser/threonine 18 (Thr), directly phosphorylates Tyr53 (or Tyr362). However, MEK control, directly or indirectly, some potential Tyr kinases that may phosphorylate Tyr53 (and/or Tyr362). For example, extracellular signal-regulated kinase (ERK), the canonical target of MEK [18], can phosphorylate and activate RhoA [19], which increases focal adhesion maturation [20]. In this manner, ERK would increase Src activation by recruiting it to focal adhesions. However, active Src does not remain in focal adhesions, but propagates rapidly [21]. Based on these data, a possible model emerges in which Src is activated at focal adhesions in a MEK/ERK-dependent manner. Active Src would diffuse from focal adhesions and phosphorylate filamentous actin in Tyr53. This could destabilize MFs, thereby decreasing its assembly in contractile actomyosin bundles associated to focal adhesions. According to the actin treadmilling model [7], monomeric actin would eventually undergo Tyr53 dephosphorylation to be reused by the cell to form other structures, e.g., lamellipodia or podosomes/invadopodia in less contractile/more protrusive regions. In this regard, Src overexpression promotes invadopodia [22] which, similar to podosomes, appear in regions in which actomyosin bundles are scarce. Importantly, Tyr53 also appears nitrated [23], which accelerated filament elongation, promoting the formation of disorganized F-actin aggregates, which may be very important for actin dynamics in highly oxidative contexts, e.g., in lung cancer [24].

Finally, actin phosphorylation in Tyr91 has been observed in multiple types of cancer, including diverse subtypes of breast cancer [2], colorectal carcinoma [25], lung cancer [26] and diverse types of leukemia (http://phosphosite.org, search term = ACTB, Tyr91). Tyr91 phosphorylation was modestly affected by treatment of non-small cell lung cancer cells with the EGFR inhibitor erlotinib [27]. However, the effect of this phosphorylation in the regulation of the actin cytoskeleton in cancer cells has yet to be addressed, although it could be related to its ability to polymerize and/or form filaments. Similar to Tyr53, Tyr91 also appears nitrated in vivo [23], with potential implications in the regulation of actin dynamics by the oxidative state of the cell.

Non-muscle myosin II

Functionally, non-muscle myosin II (NMII) is a hexameric molecular motor made of different combinations of genes. It always comprises two heavy chains [myosin heavy chain II (MHCII)] and four light chains, two regulatory (RLC) and two structural (essential, ELC). There are three genes that encode MHCII isoforms: MYH9 (Uniprot #P35579, human chromosome 22q12.3), MYH10 (Uniprot #P35580, human chromosome 17p13.1) and MYH14 (Uniprot #Q7Z406, human chromosome 19q13.33); three genes that encode RLC: MYL9 (Uniprot #P24844, human chromosome 20q11.23), MYL12A (Uniprot #P19105, human chromosome 18p11.31) and MYL12B (Uniprot #O14950, human chromosome 18p11.31); and one gene that encodes ELC, MYL6 (Uniprot #P60660, human chromosome 12q13.2). The typical structure of NMII involves MHCII from the same gene forming a central homodimer (they do not heterodimerize). Each heavy chain contains two tandem IQ motifs that bind to ELC and RLC. These binding sites define the “neck” of the hexamer, which is flexible and enables the conformational movement that generates mechanical work upon ATP hydrolysis when the hexamer is bound to actin [28]. Actin binding and ATPase activities lie upstream of the neck in a ≈ 800 amino acid long globular head domain. Downstream of the neck, both heavy chains display a ≈ 1,000 amino acid long coiled-coil domain that supports dimerization. The C-terminus of the heavy chains ends in a non-helical domain of variable length that controls the oligomerization of the hexamer into larger order units termed mini-filaments (the name has a historic connotation based on electron microscopy (EM) visualization of thick and thin bands in muscle sarcomeres, which are made of myosin II and actin, respectively).

Whereas binding to calcium-sensitive proteins controls muscle myosin II function, NMII is largely controlled by phosphorylation. In fact, RLC phosphorylation is essential for the conformational extension that is required for NMII hexamers to form mini-filaments [29]. Likewise, phosphorylations in the coiled coil domain controls dimerization; and those in the non-helical tailpiece (NHT) domain regulate oligomerization [30]. Phosphorylations of the globular domain of the heavy chain are less characterized. In Table 2, it summarizes the phosphorylations affecting the different chains of the NMII hexamer, including phosphorylations of RLC (MYL9/12) that modulate its function as well as that of the entire NMII hexamer; those of ELC (MYL6); as well as of the three genes of MHCII (MYH9/10/14).

RLC is the most important regulatory hotspot of myosin II by phosphorylation. Several residues have been described, including Ser1/2. Their phosphorylation inhibits NMII function, as they decrease ATP catalysis on the NMII hexamer head [31]. Although they control NMII function in response to growth factors downstream of conventional PKCs [32, 33], their mutation to a non-phosphorylatable version does not prevent cell division [34].

Conversely, Ser19 phosphorylation mediates the conversion of folded, assembly incompetent into extended, assembly competent NMII hexamers. Extended hexamers whose RLC is phosphorylated in Ser19 immediately form bipolar filaments that grow by lateral association, as outlined below [29]. Ser19 phosphorylation also increases ATP catalysis in the associated MHCII [35, 36]. On the other hand, Thr18 only appears phosphorylated if Ser19 is also phosphorylated [37]. Based on its in vitro effect boosting ATP catalysis of the bound MHCII, Thr18 is considered a synergy site with Ser19. It also has increases the half-life of NMII mini-filaments, which is essential during cell migration [38]. Several kinases induce these phosphorylations, including ROCK, MLCK, CITK, MRCK and ZIPK/death-associated protein kinase 3 (DAPK3) [30].

Very recently, we have identified Tyr155 phosphorylation downstream of EGFR. However, this phosphorylation only occurs when RLC is not bound to NMII. Tyr155 phosphorylation prevents the association of RLC with NMII, thus de facto decreasing the amount of NMII available to form filaments [37].

On the other hand, the role of ELC phosphorylation in cellular physiology or NMII function remains practically unexplored. Tyr29 appears more phosphorylated in many types of cancer, but the kinase remains unidentified [39]. Conversely, Tyr89 phosphorylation is dependent of EGFR III [40] and its phosphorylation is inhibited in cells treated with the EGFR inhibitor gefitinib [41]. However, whether this phosphorylation decreases NMII assembly is unknown, and currently under investigation in our lab.

Regarding the heavy chains, there are isoform-specific differences regarding heavy chain phosphorylation that have potential effects on diverse types of filaments depending on their molecular composition. In general, head domain phosphorylations have the potential to control actin binding and ATPase activity of myosin II. However, this has been poorly explored. Conversely, phosphorylations of the coiled coil domains of MHCII-A/B/C are better characterized. These regions are important for dimerization, and phosphorylation has been shown to decrease the stability of the hexamers and hinders their lateral association with other hexamers to form filaments. This revealed that lateral interactions are highly dependent on the net charge of the interacting regions [42]. Phosphorylation of Thr1800, Ser1803 and Ser1808 in MHCII-A, Ser1810 and Thr1815 in MHCII-B and Thr1832/Ser1838 in MHCII-C decrease the formation of filaments of the corresponding isoform [43]. The kinases that regulate most of these phosphorylations are α-kinases, for example TRPM6/7 [43]. However, how theses kinases distinguish between isoforms is unclear.

Finally, phosphorylations at the end of the coiled-coil or into the NHT domain of MHCII decrease mini-filament formation. One such residue is Ser1916 in MHCII-A. Its phosphorylation by PKCβ reduces filament stability [44, 45] as it increases NMII-A interaction with Mts1/S100A4, which forces NMII-A to stay in an assembly-incompetent conformation [46]. Phosphorylation of Ser1943 in MHCII-A has this effect. Casein kinase (CK)-II phosphorylates MHCII-A in Ser1943, promoting NMII-A filament disassembly [47]. In MHCII-B, Ser1935 and Ser1937 (MHCII-B) are phosphorylated by PKCζ [48, 49], but they haveasimilareffect. Importantly, and because NMII-B filaments are more stable than those made of NMII-A, these phosphorylations impair cell polarity and migration [49].

Tubulin

There are three isoforms of tubulin, each one including several variants. For the sake of brevity, and due to the grouped homology among them (Figure 2 and Table S1), we focus on α1-tubulin, encoded by the gene TUBA1A (Uniprot #Q71U36, human chromosome 12q13.12), β1-tubulin, encoded by the gene TUBB1 (Uniprot #Q9H4B7, human chromosome 20q13.32) and γ1-tubulin, encoded by the gene TUBG1 (Uniprot #P23258, human chromosome 17q21.2). α- and β-tubulins are the major components of polymeric MTs. γ-tubulin is mainly present at the centrosome [also known as MT-organizing center, (MTOC)], nucleating polymerization by forming the γ-tubulin ring complex (γTuRC). This complex acts as a template for α/β monomer incorporation [50]. Phosphorylations affecting these subunits are summarized in Table 3, but their effects in tubulin dynamics are very poorly characterized, particularly in cancer. A few significant residues include Ser48 (and Ser75 of β-tubulin). These are likely Aurora kinase sites, as two independent Aurora kinase (AURK) inhibitors reduce their phosphorylation levels [51]. Due to the key role of AURK in cancer [52], it will be extremely interesting to study whether these sites are involved in the potential regulation of MT dynamics. In this regard, Aurora kinase A (AURKA) inhibition inhibits osteosarcoma cell division by preventing MT stabilization to form the mitotic spindle [53].

Display full size

Figure 2. 

Grouped homology among human tubulin isoforms. Range shown in the homology of an isoform with itself refers to the minimal and maximal homology among sub-isoforms (see Table S1 for full details)

On the other hand, tubulin phosphorylation in Ser165 by PKCα increases MT dynamics, cell motility and acquisition of a mesenchymal phenotype characterized by expression of neural (N)-cadherin [54]. This strongly suggests that this phosphorylation could be a key event in the acquisition of mesenchymal phenotypes, which is a typical event during the transition of tumors from non-invasive to invasive states.

Finally, phosphorylation of centrosomal γ-tubulin in Ser131/385 is mediated by SadB. This phosphorylation is involved in centrosome duplication during mitosis, a key event during cell division. A phospho-mimetic form induces spontaneous centrosome duplication, whereas a non-phosphorylatable mutant impairs centrosome duplication [55]. SadB also phosphorylates γ-tubulin in Ser385, regulating S-phase progression by moderating the activities of E2 promoter-binding factors. When phosphorylated by SadB in this amino acid, γ-tubulin increases its nuclear localization [56].

Vimentin

Vimentin (UniProt # P08670) is located on human chromosome 10p13. Vimentin forms very stable IF. Unlike actin, myosin and tubulin, IFs are not very dynamic, but they endow cells with structural stability. Importantly, vimentin is an IF typical of mesenchymal and hematopoietic cells, and a signature gene of the epithelial-mesenchymal transition (EMT) [57]. As such, it is a marker of cells that evolve into mesenchymal phenotypes, acquiring migratory capability as tumors become invasive. The development of cancer affects this protein, including epigenomic alterations [58, 59] and somatic mutations in squamous lung cancer [60], gastric adenocarcinoma [61], and other types of cancer.

Vimentin undergoes extensive phosphorylation, which potentially controls its cellular function. The best characterized phospho-residues are grouped in the non-helical N-terminus domain [62] (see Table 4). Some crucial residues include Ser5, Ser7, Ser8, Ser9 and Ser10, which are phosphorylated by PKCα [63]. Since phosphorylation of these residues control leukocyte transmigration downstream of phosphatidyl inositol 3-kinase, isoform γ (PI3Kγ) [64], it is possible that they also control CTC extravasation, which tends to mimic leukocyte diapedesis [65]. In this regard, vimentin localizes to the trailing edge of migrating leukocytes and controls cortex rigidity [66]. This may be important to reduce cancer cell attrition in the bloodstream [67]. The ras-related C3 botulinum toxin substrate 1 (Rac1)/p21-activated kinase 1 (PAK1) pathway controls the phosphorylation of Ser26, Ser51 and Ser66, impairing IF assembly. Whether this is related to the weak oncogenic ability of Rac [68] is currently unknown.

Key residues include Ser39, Ser56, Ser72 and Ser83. In cancer cells, Ser39 phosphorylation by protein kinase B (AKT/PKB) protects vimentin from proteolysis and enhances tumor growth and metastasis [69] by altering filament assembly [63], which regulates cortex plasticity and could underlie the fact that cancer cells are overall softer than non-cancer cells [70].

On the other hand, Ser56 is phosphorylated by the cyclin dependent kinase (CDK) 1. This phosphorylation recruits PLK1, which enhances phosphorylation in Ser82 [71]. Consequently, cell arrest in G2/M induced by taxanes lead to an accumulation of phospho-Ser56 vimentin in a CDK1-dependent manner. Ser56 is also phosphorylated downstream of ROCK and PAK1 in hypoxia [72]. Ser56 phosphorylation promotes disassembly of perinuclear vimentin, controlling filament stability [73] and promoting cancer cell invasiveness [74]. Likewise, two independent studies indicated that Ser72 phosphorylation downstream of ROCK1 is important for cancer cell migration. A phospho-mimetic mutation of Ser72 increased cancer cell speed, whereas a non-phosphorylatable form impaired sphingolipid-triggered cell migration, respectively [75, 76]. Finally, Ser83 phosphorylation by calcium/CaMKII or PLK1 controls β1 integrin expression at the plasma membrane, controlling cell adhesiveness during invasion [77].

Importantly, phosphorylation of Ser39, Ser72 and Ser83 is preserved when vimentin is processed and peptides presented associated to major histocompatibility complexes. CD4⁺ T cells distinguish between non-phosphorylated vs. phosphorylated vimentin peptides [78]. Since these phosphorylations are elevated in metastatic cells [78], they could be useful as immunotherapeutic targets.

Is it feasible to target cytoskeletal phosphorylation to treat cancer?

Conclusion

Targeting cytoskeletal phosphorylation has the potential to dramatically alter the mechano-chemical properties of tumor cells, inhibiting their ability to develop the cancer program, at least at a preclinical level. However, these approaches may also have potentially severe side effects, as every cell, normal or cancerous, requires the cytoskeleton. MT inhibitors, discovered over 60 years ago, offered early promise as MT-targeted therapies that improved the outcome of many types of cancer by inhibiting tumor cell division. These therapies are still the first line of chemotherapeutic treatment in many types of cancer. Hence, it is evident that phosphorylation of cytoskeletal components is altered when patients are treated with kinase-targeted therapies. We have only scratched the surface in characterizing these effects. Many of them will be unavoidable consequences of treatments aimed at other fundamental processes, potentially causing side effects that will have to be dealt with. But the potential exists to discover that some cytoskeletal phosphorylation-specific effects underlie unexpected and important effects of current and future targeted therapies.