Mena, like the other Ena/VASP proteins, contains two conserved domains called “EVH1” and “EVH2” and a central unstructured proline-rich region (, and Box 2
). The EVH1 domain mediates protein-protein interactions important for Ena/VASP localization and regulation (Box 3
). The polyproline-rich region and EVH2 interact with the actin monomer binding protein profilin and directly with G- and F-actin, respectively [45
Box 2. Organization of Ena/VASP proteins
The N-terminal EVH1 domain (for Ena/VASP homology) binds to proteins containing a specific proline-rich motif that helps localize Ena/VASP proteins and recruit them into complexes with signaling proteins.
The middle portion of Ena/VASP proteins consists of a proline-rich region that binds a number of SH3- and WW-domain containing proteins including IRSp53, an I-Bar protein and Cdc42 effector that promotes filopodial formation [81
]. The proline-rich region also binds to the actin monomer binding profilin proteins, which play diverse roles in regulating actin dynamics (see recent reviews [83
]), including the ability to transfer bound monomer onto free F-actin barbed ends. Profilin can bind actin monomer and interact simultaneously with Ena/VASP through a high-affinity profilin-binding site (termed “loading site” [85
]) with the consenus PPP[AP]PPLP [68
]. Importantly, profilin:actin complexes have a higher affinity for the loading site than does profilin alone, suggesting that once actin monomer is transferred from profilin to a barbed end, exchange of the profilin bound to the loading site for a new profilin:actin complex would be favored [86
]. Interestingly, while VASP and Evl each have a single loading site, Mena contains four, suggesting it may be capable of more profilin:actin complexes than its paralogs. Importantly, the poly-Pro loading sites in Ena/VASP proteins are located adjacent to actin binding motifs contained in the EVH2 domain.
The C-terminal EVH2 domain of Ena/VASP contains binding sites for G- and F-actin [87
], called “GAB” and “FAB”, respectively. The proximity of a poly-Pro loading site permits Ena/VASP to bind profilin+G-actin complexes through two interfaces simultaneously: profilin-PPP[AP]PPLP and the adjacent G-actin+GAB. The G-actin in this complex is oriented towards the FAB motif of Ena/VASP, presumably positioned to be added on to growing filaments [85
]. The organization of binding sites for profilin, actin monomer and F-actin lead to a model in which profilin:actin binding to the loading site+GAB is followed by direct transfer of the monomer onto the adjacent F-actin barbed end and subsequent exchange of profilin for profilin:actin [83
EVH2-mediated interactions with growing ends of actin filaments are required for stable targeting of Ena/VASP to the leading edge of lamellipodia [42
]. The GAB motif stabilizes Ena/VASP at the tips of filopodia suggesting that it plays a role in recognizing barbed ends analogous to the barbed end capture activity in the highly-related WH2 domain within N-WASP [51
]. Both G- and F-actin interactions are disrupted by phosphorylation at sites within the EVH2 domain [50
], including a protein kinase G site found in both Mena and VASP [40
At the very C-terminus of EVH2, a right-handed coiled-coil mediates both homo-tetramerization and the formation of mixed tetramers containing different family members [92
]. The combination of tetramerization and F-actin binding allows Ena/VASP to bundle actin filaments [87
]; this bundling activity acts to cluster the tips of elongating filaments during filopodial formation and extension [51
], however, a physiological role for Ena/VASP bundling along the length of filaments in cells has not been demonstrated.
Box 3. EVH1-mediated interactions
EVH1 domains bind proteins that contain the consensus: [FL]PX
is any hydrophobic residue [65
]. There are a growing number of proteins with EVH1-binding sites and a full discussion of all such molecules is beyond the scope of this review, therefore only a few examples will be presented. The first characterized EVH1-ligand was ActA, a protein found on the surface of the intracellular bacterial pathogen Listeria monocytogenes that contains four EVH1-binding motifs that recruit host cell Ena/VASP proteins to the bacterial surface [66
]. Listeria employ host cell proteins to trigger actin polymerization on the bacterial surface to produce a propulsive force that drives their movement [67
] and Ena/VASP recruitment by ActA greatly enhances actin polymerization and bacterial movement [68
]. Zyxin, which helps recruit Ena/VASP to focal adhesions and stress fibers, contains four EVH1-binding sites [110
]. Lamellipodin (Lpd), an adaptor protein containing RA and PH domains that bind Ras and PI(3,4)P2, respectively, harbors six EVH1-binding sites and plays an important role in recruiting Ena/VASP to lamellipodia [70
]. Silencing Lpd in B16 cells produces a dramatic reduction in F-actin content, thereby eliminating normal lamellipodial protrusion. Lpd is a target for Abl/Arg tyrosine kinases and is required along with Ena/VASP for PDGF-induced dorsal ruffling in fibroblasts [72
] and the Drosophila Lpd is required for normal epithelial morphogenesis [73
]. Mig-10, the C.elegans Lpd ortholog, is required for cell polarization in response to Netrin and for axon guidance responses to Netrin and Slit [74
]. The Slit receptor, Robo, binds to Ena/VASP through EVH1-binding sites in its cytoplasmic tail [63
]. Palladin, an actin binding protein and EVH1 ligand [77
] has been implicated in metastatic progression; it is upregulated 3.2-fold in the invasion signature [11
], contributes to breast cancer cell invasion [78
] and is a target for the anti-metastatic kinase Akt-1, which blocks Palladin-driven invasion [79
]. Finally, the putative tumor suppressor TES is an unconventional Mena-specific EVH1 ligand that binds via a LIM domain to a region that overlaps with the [FL]PX
P binding pocket [80
Mena has several unique features not found in the other Ena/VASP proteins that endow it with the ability to potentiate carcinoma metastasis dramatically. Importantly, alternate splicing of Mena produces distinct protein isoforms, including an invasion-specific isoform, “MenaINV” (discussed further below), that has no counterpart in VASP or EVL, and which is found exclusively in invasive tumor cells.
Analysis of the invasion signature of mammary carcinoma cells revealed that Mena expression was up-regulated in invasive cells compared to average primary tumor cells [8
]. Increased Mena levels were also observed in invasive human breast cancers compared to normal mammary tissue [32
]. As noted above, perivascular tumor cells expressing high Mena levels are a component of TMEM, a structure whose density in clinical samples correlates with increased risk of metastatic outcome in breast cancer patients [33
]. In addition to breast cancer, Mena up-regulation has been observed in advanced pancreatic, colon and cervical carcinomas [35
Mena has a number of features that its paralogs VASP and EVL do not share. The first is an extended repeat region spanning 70 residues with most of the repeats containing the consensus, [LM]-E-[QR]-[EQ]-[QR] (abbreviated as “LERER” repeat), which is predicted to form a coiled-coil structure [101
]. The repeat is located between the EVH1 domain and the proline-rich region. In addition to this unique feature, the Mena message undergoes extensive alternate splicing to give rise to multiple protein isoforms that are expressed in specific tissues and cell-types [40
] (). In contrast, EVL has 1 alternately included exon and VASP has none. There are 14 constitutively included exons in Mena and 5 alternately included exons that can all encode protein sequence in frame. There has not been a comprehensive analysis of which of the possible combinations of alternately included exons are actually produced as mRNA, nor do we know all of the cell types which produce the various Mena isoforms.
Cloning Mena cDNA from a breast cancer cell line identified the Mena11a isoform [102
]. Analysis of RNA from primary mammary tumor cells collected by FACs, compared to that expressed in invasive mammary tumor cells collected using the in vivo
invasion assay, revealed that the 11a exon is expressed in tumor cells making up the bulk of the primary tumor, but this exon is essentially undetectable in the Mena message from invasive tumor cells [103
]. Consistent with this finding, the 11a exon is specific to Mena isoforms expressed in epithelial cell lines and is not found in mesenchymal cells [100
]. In fact, 11a becomes excluded in human mammary epithelial cells that are driven to undergo epithelial to mesenchymal transition (EMT) by expression of the EMT inducing transcription factor Twist [104
]. The presence of 11a in epithelial cells is driven in part by the activity of the recently identified epithelial-specific splicing factors ESRP1 and ESRP2 [105
]. Mena11a is also expressed in normal ovarian tissue where its inclusion is promoted by the Fox2 splicing factor [106
]. Interestingly, analysis of 21 aggressive ovarian tumors revealed a reduction in Fox2 levels compared to normal tissue and a concomitant loss of 11a inclusion in Mena [106
]. Therefore, Mena11a appears to be included in epithelial cells and primary carcinomas but excluded from mesenchymal cells as well as invasive/aggressive tumor cells.
The alternately included 11a exon encodes 21 amino acids that are inserted in the EVH2 domain, between the FAB sequence and the coiled-coil tetramerization domain. The Mena paralog EVL also has an alternately included 21 amino acid insertion (“EVL-I”) in an identical relative location as the 11a insertion site, but the sequences share no similarity [107
]. The site of 11a insertion is adjacent to the F- and G- actin binding sites, and the 11a insertion can be phosphorylated [102
], potentially disrupting actin binding. Therefore, it is possible that the 11a inclusion affects the way in which Mena interacts with barbed ends and adds an extra site for phospho-regulation of Mena function.
Three alternately included Mena exons were identified by screening a mouse brain cDNA library [40
]. The largest exon, denoted as “+,” falls adjacent to the proline-rich region and is itself quite rich in proline. Mena+
is a 798 residue protein (the most widely expressed form of Mena, denoted “Menaclassic
,” is 541 amino acids), however, due to their high proline content both Mena+
migrate aberrantly on SDS-PAGE gels at approximately 140kDa and 80kDa, respectively. Western blot analysis of adult tissues has shown that the 140kDa isoform is only readily detected in the brain compared to other organs and tissues [61
]. Two other short exons, denoted “++” and “+++” and encoding 4 and 19 residues, respectively, were identified in brain cDNAs containing the “+” exon. Both ++ and +++ are inserted at the same site just C-terminal to the EVH1 domain and between the LERER repeat. No tissue-specific expression has been identified for Mena++
. Interestingly, the +++ exon is highly conserved in mammals but is not found in other vertebrates.
The majority of Mena mRNA up-regulated in the invasive subpopulation of tumor cells isolated from rat, mouse and human mammary tumors using the In vivo
Invasion Assay contains either the ++ or +++ exon, while strong downregulation of Mena 11a occurs in the same invasive tumor cells. The upregulation of the ++ or +++ exons persists in circulating tumor cells isolated from blood [103
]. These results suggest that Mena+++
are the isoforms that may function in metastatic progression.
This prediction has recently been tested [49
] (Roussos et al.
, unpublished) and findings suggest that expression of Menaclassic
, and Mena+++
(referred to as the “invasion isoform” or MenaINV
) in particular, promotes carcinoma cell invasion in three-dimensional collagen gels and increases carcinoma cell motility in vivo
localize to and stabilize invadopodia, actin-rich protrusions required for degradation and movement through extracellular matrix and possibly invasion across basement membranes, thereby increasing the invasive and metastatic potential of tumor cells.
plays a sensitizing role in the chemotactic and motility responses of tumor cells to EGF as expression of MenaINV
sensitizes mammary tumor cells to EGF signals by at least 25- to 50-fold, causing tumor cells to respond to otherwise undetectable EGF levels [49
] (Roussos et al.
, unpublished). MenaINV
regulates the lifetime of actin filament barbed ends produced by EGF-elicited protrusion; within as little as 20 seconds of stimulation, cells expressing MenaINV
have 80% more free barbed ends than control cells or cells expressing Menaclassic
]. The stimulatory effect of MenaINV
requires cofilin severing but precedes the accumulation of Arp2/3 in lamellipodia, indicating that MenaINV
acts directly on barbed ends generated by cofilin severing. Therefore, we propose that MenaINV
exerts this stimulatory effect by delaying barbed end capping (). This is an important finding because cofilin-generated barbed ends of actin filaments are needed to initiate invasive protrusions during chemotaxis and maintain the motility of crawling tumor cells [10
]. The mechanisms underlying the effect of the additional 19 amino acids in the MenaINV
isoform, and the ability of this isoform to potentiate EGF-dependent motility responses, are under investigation. The present findings, however, indicate that we have identified a master gene that makes breast cancer cells aMENA
ble to metastasis.
Proposed model for Mena anti-capping/elongation activity in carcinoma cell invasion