The Crk gene product was identified approximately 20 years ago in the form of a transforming gene (Gag-Crk) encoded in the genome of a defective avian sarcoma retrovirus called CT10 (

c

t

10

C

r

k

Following the cloning of v-crk or gag-crk, Reichman, et al. reported the isolation of the cellular homolog of Crk by screening a chicken brain library using a portion of the v-crk probe devoid of the retrovirus-derived gag sequences 6. Molecular cloning of chicken cellular crk (c-crk) revealed a molecule with similar structural organization to v-Crk, but containing a ~50 amino acid proline-rich linker and an additional C-terminal SH3 domain (SH3C) (Figure 1B). Matsuda and colleagues characterized two species of Crk in mammalian cells, Crk II and Crk I, derived from alternative splicing from a single gene locus 7. Structurally, Crk I is more analogous to v-Crk and, biologically, significantly more transforming than Crk II when ectopically expressed in NIH 3T3 cells. The oncogenic activation of Crk by C-terminal truncation in the retroviral genome is reminiscent of the mode of oncogenic activation of tyrosine kinases, including the activation of Src in the RSV genome 89. Indeed, it is now clear that the differences in transforming activities of Crk II and Crk I rely, at least in part, on a specific post-translational modification of Crk that involves phosphorylation on Y222 in chicken (Y221 in mouse and human) 10. Both Crk II and CrkL are negatively regulated by autoinhibitory phosphorylation events that ultimately block the ability of Crk to function as an adaptor protein. This is analogous to the C-terminal tyrosine phoshorylation of Src family kinases by C-terminal Src Kinase (CSK) 1112, which allows Src to adopt an intramolecular pTyr-SH2 "closed" structure involving the Src SH2 and a C-terminal phosphorylation motif 13. Phosphorylation of Crk on Tyr 221 (or CrkL on Tyr 207) causes intramolecular binding of the linker region to the SH2 domain, sequestering the SH2 and SH3N and preventing them from binding target proteins 1415. Subsequent studies have shown that the negative regulation of Crk II occurs directly via the activity of the proto-oncogene products Abl and Arg tyrosine kinases, the former originally characterized from the Abelson murine leukemia virus (A-MuLV) 16. Abl and Arg appear to be the predominant tyrosine kinases that negatively regulate Crk II (and CrkL) in cells, as CrkY221 phosphorylation is virtually undetectable in Abl/Arg (-/-) double knockout cells. Current theory holds that v-Crk and Crk I are more transforming than Crk II, mainly because their C-termini end prior to the regulatory Y221 site, although there are mutations in the v-crk gene that also may contribute to oncogenicity 16. Hence, v-Crk has no clear mechanism to downregulate signals, essentially being locked as a constitutively active adaptor in signal transduction.

The Crk and CrkL proteins have central roles in an astonishingly vast number of biological processes, ranging from cell proliferation, cell adhesion and migration, phagocytic and endocytic pathways for apoptotic cells and parasitic organisms respectively, apoptosis, and regulation of gene expression 17. Until the introduction of siRNA technology and knockouts in genetically amenable systems, the majority of studies on Crk relied on identification of binding proteins to specific domains or by overexpression of wildtype or dominant negative Crk proteins in reconstituted systems. Because the SH2 and SH3 domains are modular and retain their binding specificity and affinity when expressed as isolated fragments 418, GST (glutathione S transferase)-pull-down assays, Far-Western blotting, and yeast two-hybrid approaches permitted the initial characterization of Crk binding partners and inference of the participating signal transduction pathways.

Over the past two decades, over 40 cellular proteins have been identified that bind to the SH2 and SH3N domains, and for this reason, it is clearly daunting to assign one specific cellular function to Crk 17. Indeed, a PubMed search combining "Crk" and "signaling" returns over 700 articles, and survey of this literature clearly indicates that the adaptor function of Crk is used repetitively and often as a signal transduction strategy in the context of both physiological and pathophysiological situations (Table 1; and references 2519202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273). Despite such complexity in Crk transduction outcomes, some general rules apply regardless of the specific proteins involved. First, the Crk SH2 domain (the input pathway) binds phosphotyrosine-containing proteins in an inducible on-and-off switch mechanism that involves regulated tyrosine phosphorylation and dephosphorylation. Second, the interactions between the Crk SH3N domain (the output pathway) occur constitutively, and these complexes are generally regulated without post-translational modifications, although there are a few exceptions whereby tyrosine phoshorylation has been shown to impinge on SH3-proline-rich assemblages 74. Therefore, while assigning functions to Crk, it is important to keep in mind that input signals depend on the availability of extracellular hormones and growth factors, the prior history of the cell, including whether the cell has been stressed or re-stimulated, the state of proliferation or differentiation of the cell, and other temporal and spatial epigenetic factors that reflect the physiological and metabolic status of the cell. As the input signals reflect changes in the cellular environment, Crk signals are subject to dynamic and plastic regulation, allowing for rapid and dynamic fluxes through signal transduction pathways that permit pleiotropic changes in the signaling capabilities of cells. Despite a detailed understanding of the types of interactions that exists between Crk and its substrates, there is still a paucity of information as to how these signals are integrated in real time and space.

Another important aspect of Crk signaling that requires definition is to assess which signaling pathways depend exclusively on Crk or CrkL versus those signaling pathways whereby Crk and CrkL can compensate for each other. Despite considerable homology between Crk II and CrkL in their SH2 and SH3 domains, development of knockout mouse models clearly show these gene products have distinct, non-overlapping roles, during embryonic development as both Crk knockout and CrkL knockout mice die perinatally with different developmental defects. Moreover, in Crk II (-/-) fibroblasts, CrkL proteins were not overexpressed in a compensatory manner, and vice versa of CrkII expression in CrkL (-/-) cells, again suggesting little molecular crosstalk at the level of expression in embryonic cells 75. Interestingly, mice homozygous for a null mutation of CrkL show a phenotype characteristic of DiGeorge/velocardiofacial syndrome, in which neural crest derived cells fail to differentiate and migrate leading to defects in cranial and cardiac development 767778. Crk null mice (which lack both Crk II and Crk I) die perinatally due to defects in cardiac and skeletal development 79. Imaizuma and colleagues generated another mouse using an exon trap method that lacks Crk II but still expresses Crk I. Although Crk I expressing mice show no obvious developmental abnormalities, it is not known whether these mice develop age-dependent malignancies, since they die within a few days after birth by unknown mechanisms 80.

However, in contrast to the aforementioned observations showing independent roles for CrkL and Crk II during development, a number of recent studies suggest that Crk II and CrkL are co-expressed in somatic cells and can compensate to each in cellular signaling. An example of such redundancy comes from several independent studies dissecting the Reelin signaling pathway in cortical pyramidal cells of the hippocampus. Reelin, a secreted glycoprotein that binds to receptors on hippocampal neurons such as ApoER2, VLDLR, and α3β1 integrin, controls cell positioning during adult neurogenesis via Src-dependent tyrosine phosphorylation of the scaffold protein Dab1 81. Upon receptor activation, Dab1 becomes tyrosine phosphorylated at multiple sites, and binds a variety of SH2 domain-containing proteins that include both Crk II and CrkL 8283. Studies using shRNA or cre-lox deletion of Crk and CrkL showed that knockdown of both proteins are required for the defects in neuronal positioning, and both proteins appear to converge on the activation of Rap1 through C3G (see below) 848586. How universal Crk and CrkL compensation is in signal transmission still remains an active area of investigation.

Much of the early emphasis on Crk biology focused on the mechanisms of increased cellular tyrosine phosphorylation and models for v-Crk transformation. Although it is still reasonable to consider that v-Crk binds to and "hijacks" a cellular tyrosine kinase, no model in which a specific tyrosine kinase is activated has provided a unifying molecular mechanism for v-Crk and Crk I transformation. However, a number of studies indicate that expression of v-Crk in Src (-/-) or Fyn (-/-) fibroblasts significantly reduces Crk-inducible tyrosine phosphorylation 87, and such studies imply that v-Crk transformation, at least in part, requires initial activation of a Src family kinase member (SFK). Subsequently, Akagi and colleagues demonstrated that focal adhesion kinase (FAK) Y397 phosphorylation (mediated by SFKs), and the recruitment of phosphatidylinositol 3'-OH kinase (PI3K) to FAK was also required for v-Crk-dependent transformation of CEFs 8889. This observation is consistent with recent studies employing Src/Fyn/Yes (SFY)-deficient MEFs, whereby Src was shown to be required for FAK tyrosine phosphorylation and Crk-mediated motility 90.

With respect to the substrates of tyrosine phosphorylation and Crk transformation, a great deal of attention has been placed on the p130Cas/HEF1/Efs proteins, a family of non-enzymatic docking proteins that contain multiple interacting domains and serve as important anchoring points for protein-protein interactions 4691. p130Cas (

C

a

s

Mice with homozygous null mutations of p130Cas exhibit marked growth retardation with poorly developed heart and vasculature and die in utero from congestive heart failure, edema, and hemorrhaging due to breakdown in the regulation of vascular permeability 94. Isolated and immortalized p130Cas (-/-) embryonic fibroblasts display short disorganized actin filaments, show defects in actin bundling, and these cells have small and poorly organized focal adhesions. However, a functional role for p130Cas in non-motile cell types, including neurons, is likely equally important and recent studies in Drosophila showed that targeted loss of p130Cas in CNS neurons induced defects in neurite guidance and target fasciculation 100. In fibroblasts, v-Crk localizes primarily at focal adhesions, and for this reason, the association of Crk with events associated with p130Cas and actin cytoskeleton assemblages has gained considerable acceptance in the Crk transformation field 101102. Specific targeting of Crk to focal adhesions by fusing a FAT sequence to Crk induces p130Cas and FAK phosphorylation and potentiates cell migration 102.

At the molecular level, p130Cas has multiple protein-protein interaction domains including an N-terminal SH3 domain that binds FAK and Pyk2, an interior "substrate domain" characterized by 15 YxxP motifs, a C-terminal Src binding domain, followed by a highly conserved C-terminal region that binds to the Nsp family of proteins (Nsp1, And-34, and Chat) 91 (Figure 2A). Based on the unusual arrangement of repetitive YxxP motifs, there has been continued interest and conjecture as to the reasons why this motif has been duplicated 15 times, since when phosphorylated, they comprise a consensus binding site for the SH2 domain of Crk. Interesting studies by Miller and colleagues showed that the p130Cas substrate region was processively phosphorylated when Src bound to the polyproline-region of p130Cas, initially suggesting that the order of addition of phosphates may organize a pattern to prioritize a signaling mechanism 103. However, their more recent studies suggests that p130Cas phosphorylation is stochastic, meaning that no single "lynchpin" permits phosphorylation at additional sites and that the phosphorylation does not appear to follow an obligatory sequence 104. This would suggest that the juxtaposed repetitiveness of the YxxP motifs in p130Cas would mainly function in signal amplification to maximize the Crk output pathways, but also to diversify signaling, as functional cooperation in the Crk assemblages would also increase the repertoire of output signals, for example during integrin-mediated spreading and migration on extracellular matrix (ECM) molecules 105 (Figure 2C). Assuming that several of the tyrosines in the YxxP motifs are concomitantly phosphorylated at any given time, this might indicate that separate molecular complexes of p130Cas/Crk/DOCK180, p130Cas/Crk/C3G, p130Cas/Crk/Abl, p130Cas/Crk/JNK, and p130Cas/Crk/PI3K could exist simultaneously, but this remains to be seen experimentally (Figure 2C). Clearly, the development of site-specific, phospho-specific antibodies aimed to quantify and map the timing and sequence of p130Cas phosphorylation events could be used to determine whether there are more subtle patterns in the motifs that are phosphorylated, or perhaps more importantly, whether the half lives of individual phosphorylation events vary during signal downregulation. Such multiplicity in p130Cas phosphorylation could achieve extraordinary diversification in signaling by acting as a molecular scaffold to transmit localized subcellular signals.

The biological role of p130Cas phosphorylation is also intriguing from the vantage point that a vast variety of extracellular matrix molecules, growth factors, hormones, and metabolic intermediates that induce tyrosine phosphorylation of p130Cas and subsequent coupling to Crk (Table 1). In addition, recent studies have demonstrated that p130Cas phosphorylation is sensitive to physical transduction by mechanical force 51 and this may provide a common element for convergence in signal transduction (Figure 2B). These most interesting studies showed that when cells are cultured on a stretchable substrate consisting of collagen-coated flexible silicone to uniformly and biaxially stretch cells by 10% or more, p130Cas was inducibly tyrosine phosphorylated within 1 minute of the stretch. Although these experiments were designed to mimic integrin-mediated events, it is important to note that tyrosine phosphorylation of p130Cas occurred in the absence of specific integrin ligation and outside to inside integrin signaling. Moreover, using a clever FRET-based assay to monitor stretching of the p130Cas substrate domain (SD) in vitro, these investigators showed extension-dependent phosphorylation of the SD by both Src and Abl tyrosine kinases, but not CSK or ZAP-70. These data suggest that mechanical stretching of the substrate region, presumably by actin-tethering proteins brought together by binding to FAK and/or Src, is sufficient to cause p130Cas phopshorylation and subsequent coupling to downstream pathways (Figure 2C). This data also suggests that physiological events that induce stretching, i.e. adhesion, motility and engulfment of large particles, as well as pathophysiological events such as hypertrophy, are probably sufficient to induce p130Cas tyrosine phosphorylation, independent of the particularities that initiate the signal.

Once extended, the central region of p130Cas is a substrate for activated tyrosine kinases, and p130Cas itself has binding sites for FAK 106 and Pyk2 107 (via the SH3 domain) and SFK members through the C-terminal Src binding region (SBR) 21 (Figure 2A). In vivo, both of these kinases have been proposed to contribute to p130Cas phosphorylation, possibly to permit different mechanisms and different input signals to converge on assembling Crk linkages 108. To gain a better perspective on the specific role of Src versus FAK in mediating p130Cas phosphorylation, and determine whether phosphorylation is sufficient to activate downstream signaling, Sharma and Mayer 109 developed an approach called functional interaction trapping (FIT). Using this strategy, two signaling molecules are brought together by engineering fusion proteins with artificial binding surfaces composed of two leucine zipper coiled-coiled domains of ZipA and ZipB. When FIT was employed to bring together full-length p130Cas or the substrate region of p130Cas and Src, p130Cas was heavily phosphorylated on tyrosine and phosphorylated 130cas was sufficient to activate Rac1 and localize to focal adhesions 109. In apparent contrast, when FAK was fused with ZipB and expressed with p130Cas, p130Cas was weakly phosphorylated. This result appears to resolve a long-standing question as to whether FAK acts principally as a scaffold, since the Tyr397 site in FAK can recruit Src family kinase through its SH2 domain. However, functional studies show that FAK binding is important for p130Cas tyrosine phosphorylation in vivo, as fibroblasts derived from mice expressing a truncated p130Cas that lacks exon 2, and hence the N-terminal SH3 domain that binds FAK, fail to induce p130Cas phosphorylation when fibroblasts were induced to spread on FN 110.

Analogous to the theme described above for p130Cas, other scaffolding molecules that are acted upon by extrinsic signals ultimately interact with Crk, thereby launching alternative versions of a common signaling paradigm. Numerous stimuli, ranging from growth factors and cytokines to ECM, apoptotic cells, microbial products, immune complexes, and endogenous metabolic products activate trans-membrane receptors that inducibly phosphorylate proteins that, in turn, engage the Crk SH2 domain (Table 1). These proteins include Dab1 39, IRS-1 25, Paxillin 111, Gab1 112, and CasL/Efs 91, and analogous to p130Cas, these proteins contain tandem YxxP motifs. As pointed out by Feller, et al., examination of the databases shows that the occurrence of three or more YxxP motifs in 100 kDa intracellular proteins is rare 17, raising the question of how such tandem elements are designed for functional crosstalk between Crk adaptor, or other adaptor protein pathways. As alluded to above, utilization of different scaffolds could allow for localized signaling, by virtue of being targeted by specific kinases, or could control the selectivity of Crk signaling. A good illustration of this comes from a study by Lamorte and colleagues, showing that a switch from a p130Cas/Crk to a Gab1/Crk complex in Met oncogene transformed cells correlates with a change from a motile phenotype to a proliferation phenotype 113. In a few cases, Crk can interact directly with tyrosine kinase receptors. For example, direct recruitment of Crk to the PDGFRα 28 or the VEGFR-3 29 can regulate immediate post-receptor signals from a subset of growth factors at the plasma membrane.

Once the Crk proteins are engaged by tyrosine phopshorylation via the SH2 domain in a particular time and place, they can transmit signals downstream through constitutive binding to the Crk SH3 domain. Although many Crk SH3N binding proteins have been identified and have been the subject of an excellent review 17, some of the better characterized p130Cas signaling pathways are described below.

Since the discovery of v-Crk and c-Crk II, it is clear that the C-terminal region exerts negative regulation on Crk signaling and transformation. Whereas models for protein interactions mediated by the SH2 and SH3N have been well documented to explain how Crk functions as an adaptor protein 14168174, how the C-terminus of Crk imposes negative regulation has only recently begun to be unraveled 15175176177178. Earlier biochemical studies suggested that the Crk SH3C could impose regulation on the adaptor protein function of Crk. For example, when expressed in Rat 3Y1 cells, the SH3C domain suppressed p130Cas phosphorylation and the transforming potential of Crk II, and negatively regulated EGF signaling to Ras 179. Mutations in the SH3C of Crk also lead to increased association of Abl with the SH3N, suggesting that the SH3N and SH3C functionally interact 180. This latter study also revealed that disruption of the boundary between the SH3 linker and the SH3C domain (aa 240–296 in chicken Crk) regulated the activation of FAK and increased the numbers of focal adhesions in cells. This paradigm for autoinhibition of the SH3 linker and SH3C domain suggested that regions independent of the Y221 might contribute to the negative regulation of Crk.

NMR studies on various Crk proteins that contain the SH3C domain and SH3 linker have begun to offer explanations as to the nature of its regulatory aspects. Like the Y221YAQP motif, the SH3C contributes to the negative regulation of Crk II and CrkL by intramolecular domain crosstalk. Although the Crk SH3C is highly conserved from C. elegans to mammals, there are notable differences in the Crk SH3C compared to canonical SH3 domains that bind PxxP motifs. First, although the amino acids that comprise the structural elements of the hydrophobic core are highly conserved, the amino acids that comprise the surface of the SH3C, in particular, where the PxxPxK is expected to bind, are highly divergent 165176. The charged residues on the surface of the SH3 domain that interact with the P1, P3, and K6 residues are replaced by unusually polar and bulky amino acids, precluding a socket structure for the prolines and providing a clear explanation for why CrkSH3C does not bind PxxPx(K, R) ligands. Indeed, inspection of the solution NMR structure of the isolated murine SH3C does not show an obvious binding grove, suggesting that this domain may not bind to, or bind with low affinity, to target proteins 176. Indeed, the only known protein that has been shown to interact with the Crk SH3C domain is the nuclear export protein Crm1, which is posited to bind to a LALEVGELVKV sequence in the Crk SH3C, and has broad similarity to a nuclear export sequence (NES) 181. However, in the monomer Crk proteins, the LVK motif is buried in the hydrophobic core of the SH3C and not likely to form a contact surface. Interestingly, in the case of the CrkL SH3C, partial SH3 unfolding that may occur during monomer to dimer transition in the nucleus, specifically may expose this LVK sequence for Crm1 binding 175. This suggests that, in the context of nuclear Crk, regulated unfolding of the SH3C domain may provide a novel and unexpected regulatory mechanism for Crk efflux. Future studies aimed at identification of the molecular mechanisms of Crk unfolding, and whether this reflects an anti-apoptotic mechanism warrant further investigation 182.

To better understand the mechanism by which the SH3C and linker regulate the SH3N, Kalodimos and coworkers used NMR analysis on a SH3N-linker-SH3C fragment of chicken Crk 178 (Figure 5). In solution at pH 5.5, this fragment showed characteristics of two species in slow exchange originating from cis-trans isomerization at the proline from the PFY motif at the boundary of the linker-SH3C, a region previously identified as an important negative regulatory region in chicken Crk 180. Moreover, the cis and trans forms appear to have entirely different regulatory functions: the cis isomer is locked in a configuration that tethers the SH3N, preventing ligand binding, while the trans configuration appears extended and flexible (Figure 5). Support of this model comes from the fact that mutation at either Pro238 or F239 abrogates cis-trans isomerization, favoring the trans form and leading to increased association of Abl with the SH3N 178. By sequence analysis, this proline was expected to be a substrate for Cyclophilin A (CypA), a protein that has peptidyl prolyl isomerase activity and catalyzes the prolyl isomerization of peptide bonds. Interestingly, this suggests that enzymatic-catalyzed cis-trans proyl isomerization may influence Abl and Arg dependent phosphorylation and hence the rate at which Crk proteins are negatively regulated. Future studies should investigate the physiological significance of CypA on Crk function, as well as whether cyclosporin A (inhibitor of CypA) may be a specific inhibitor of Crk pathways in vivo. Another interesting aspect of this regulatory mechanism involving the PFY motif in chicken Crk is that F239, adjacent to P238, was absolutely required to achieve cis-trans isomerization, and mutation of F239 to Ile or Val abrogated this regulatory mechanism. This is most intriguing, given the fact that chicken is the only species of Crk with the PFY motif. This motif diverges to PIY in human and PVY in mouse (Figure 6).

Recently, Inagaki and colleagues performed NMR of the entire structure of Crk I and Crk II and the phosphorylated form of Crk II without the SH3C (amino acids 1–228) 15 (Figure 7A). Interestingly, while the SH2 and SH3 domains of Crk I are flexible 15, conversely, Crk II forms a compact structure. Solution structure of Crk II showed that the three SH domains were assembled to a central short sequence, residing in the inter SH3 region (amino acids 224–237) called the inter SH3 core region (ISC), and that this compact structure was stabilized mainly by the SH3C domain. These studies also found that the SH3N is semi-closed and covered by the bottom of the SH2 domain, mimicking the interaction of SH3 and PxxPxK. These structures in many respects provide confirmation of previously observed differences in the biological activity of Crk I versus Crk II, including the fact that the affinity of Crk II for the SH3N target polyproline motifs is lower than the affinity of Crk I. Upon phosphorylation, the entire Crk structure is dramatically shifted, and the SH2 binds to phospho-Y221, as reported previously. In phosphorylated Crk II, the SH3N surface was shown to be blocked by an internal sequence between the SH2 and SH3 domains (Arg122- Glu133), suggesting that phosphorylated Crk II cannot bind to either of the SH2 or SH3N targets (Figure 7A). Again, these observations are consistent with previous biochemical and cell biological studies, which predict that: (i) Crk I is constitutively active, (ii) Crk II is regulated by the intramolecular SH3N interaction, and (iii) phosphorylated Crk appears to be completely shut off for binding effector proteins (Figure 7B and additional File 1). Moreover, although details remain to be determined, it appears that the mode of negative regulation for mammalian and chicken Crk II proteins has somewhat diverged.

In recent years, a growing and unexpected body of evidence suggests that Crk may contribute to bacterial pathogenesis by influencing entry into cells or by serving as targets of bacterial toxins that disrupt essential cellular function through Crk-dependent mechanisms. A role for Crk in bacterial uptake pathways was first noted by Weidow and colleagues, who demonstrated that Yersinia pseudotuberculosis infection into human epithelial cells activated the p130Cas-Crk-DOCK180-Rac1 pathway via binding of the bacterial invasin to the β1 integrin receptor 63186. This pathway appears be somewhat analogous to the phagocytic pathway for apoptotic cells, although the phagocytic pathway involves β3 and β5 integrins. More recent studies by Burton and colleagues found that Crk is also implicated in Shigella flexneri infection into NIH 3T3 cells using a curious pathway involving Abl-Arg-mediated Crk Y221 phosphorylation, whereby Crk Y221F mutation blocked S. Flexneri infection 64. In very provocative studies involving Pseudomonas aeruginosa infection 187 and Helicobacter pylori infection 66, Crk appears to acquire a gain-of-function activity by co-opting activities with bacterial virulence factors. In the former study, a novel role for Abl and Crk phosphorylation was identified and found to be essential for Pseudomonas aeruginosa internalization 188. In the case of H. pylori, Crk associates with tyrosine phosphorylated CagA, and the CagA/Crk complex regulates Erk signaling and Rac1-mediated cytoskeletal assemblages 66. Although in H. pylori-associated gastric cancer, CagA-SHP2 signaling is known to play a central role, the functional significance of the CagA/Crk complex for cellular infection should be evaluated. Finally, exoenzyme T (ExoT) of P. aeruginosa ADP-ribosylates Crk I and Crk II at specific residues in the SH2 domain (Arg20) that impairs binding of Crk to p130Cas 187. Although to date, there has been very little to report on the role that Crk proteins play in viral entry and pathogenesis, this may be an area of active ferment as a recent study showed that CrkL can bind to NS1 proteins in Influenza A infected cells 189.

A number of studies have suggested that Crk may play an important role in human cancers. Immunohistochemical analysis of human cancers showed the over-expression of Crk in adenocarcinomas of lung, breast, and stomach, and also sarcomas, followed by the demonstration of increased levels of Crk mRNA in more aggressive phenotypes of lung adenocarcinoma 62190191192193194. In an analysis of gene-expression profiles of 86 primary lung adenocarcinomas, increased Crk expression was shown to be a predictive factor, contributing to poor prognosis and shorter survival 190. Overexpression of Crk in tumor cells or in experimental model systems leads to increases in tyrosine phosphorylation of p130Cas and the activation of an intracellular loop that further enhances the activity of Crk and induces increased motility and the aggressive potential of cancer cells. Therefore, Crk proteins are not simply conduits for intracellular signal transduction but also can control the amplitude of signaling.

Based on these observations, several groups have begun to investigate the effects of Crk knockdown on the phenotypic behavior of human tumor cells. Simultaneous downregulation of both Crk I and Crk II by siRNA demonstrated the essential role of Crk in the malignant features of human ovarian cancer cells 192, synovial sarcoma cells 112, and brain tumors, such as glioblastoma cells 58. Recent studies also suggest that Crk II is a likely target for MiR-126, and that overexpression of MiR-126 in lung cancer cells decreased adhesion, spreading, and invasion 195. In Crk knockdown cells, formation of focal adhesion was decreased, and cells had cytoskeletal disorganization, whereby formation of laemellipodial structures at the leading edges of cells was impaired (Figure 8, additional file 2 and additional file 3). For example, knockdown of Crk in human synovial sarcoma cells showed a suppression of HGF-dependent activation of Rac1 and decreased motility 192. In glioblastoma cells, Crk knockdown affected the early attachment to laminin. In addition, Crk also regulated Rho-dependent phosphorylation of ezrin-radixin-moesin proteins that direct migration towards hyaluronic acid in the brain 58. Increased levels of Crk I mRNA are frequently observed in WHO grade III and IV malignant gliomas 191193. Thus, Crk I expression may correlate with poor prognosis, and this is a novel example of splicing-dependent control of human cancer malignancy. As Crk knockdown in cell lines was not lethal, Crk may be an effective therapeutic target for decreasing the metastatic potential of human tumor cells (see additional file 2 and additional file 3).

The above studies suggest that Crk may have a central role in cell motility and metastasis in highly aggressive motile cancer cells. There are a plethora of studies showing that p130Cas-HEF and the FAK-Src circuitry is involved in tumorigenesis, and Crk may indeed be an attractive target due to its central integrative downstream role in signaling by these molecules. Recent studies now suggest that DOCK180 and/or ELMO may be oncogenes, or have oncogenic potential, and Crk may be an attractive drug target in tumor cells that use Rac1 activation to acquire motile or metastatic potential. DOCK180, for example, is overexpressed at the borders, but not at the centers, of human malignant glioma, and together with ELMO1, it regulates Rac activity to enhance tumor invasion 196. Although the relation of C3G and human cancer is controversial, an increase in C3G expression is observed in human small cell carcinoma 197, and a Crk-C3G-Rap1 pathway leading to B-RAF and Erk activation is involved in transformation downstream of Ret in papillary thyroid carcinoma 124. In addition, C3G activation of Rap1 regulates a number of cell-matrix or cell-cell interactions through integrins and cadherins 130, pointing to a role for C3G overexpression in the modulation of adherens junctions, leading to tumor cell dissemination 130198. It will be important to experimentally observe whether siRNA toward Crk can influence the motile behavior of cells that express upregulated C3G and/or Rap-1. Finally, a most curious function for CrkL has been recently proposed, whereby a secreted form of CrkL was shown to bind to the plexin-semaphorin-integrin (PSI) domain of β1 integrin at the extracellular domain in order to promote cell growth and survival [213]. These authors made the provocative suggestion that targeting extracellular Crk proteins may have therapeutic value.

Another area of clinical investigation is the role of tyrosine phosphorylation of CrkL and Crk II in CML 199200201. This cancer results from the chromosomal translocation termed Philadelphia chromosome (Ph1) as t(9;22), generating the aberrant chimeric protein Bcr-Abl. Current understanding is that CrkL is not essential for the development of CML, because CrkL knockout mice developed CML by the introduction of Bcr-Abl. However, the full role of CrkL and Crk II in CML has yet to be explained.

In recent years, both Crk II and CrkL have been implicated in nuclear signal transduction events. For CrkL, elegant studies have shown that CrkL binds tyrosine phosphorylated Stat5 in Bcr-Abl expressing cells, and in a variety of cytokine-stimulated hematopoietic cells 202203. In many, if not all, of these cases, the complex translocates to the nucleus to bind STAT5-responsive elements 204205. Although Crk II can also bind to STAT5, this complex does not appear to enter the nucleus, nor do anti-Crk II antisera supershift STAT5 DNA binding complexes 34. While these studies may indicate that CrkL has a role in transcription of interferon regulated genes, interesting studies by Kornbluth and colleagues indicate that Crk may actively and directly participate in apoptosis in Xenopus by activating caspases 181. Depletion of Crk II from egg extracts prevented apoptosis, and these studies further demonstrated that Crk binds to the cell cycle regulatory protein Wee1, implying that this pro-apoptotic function of Crk II may be due to nuclear localization 206. Evidence in support of this model includes the fact that Crk has a putative NES in the SH3C domain, that when mutated increases cell death, and that ionizing radiation increases complex formation between Wee1 and Crk II. Furthermore, evidence for a direct role of Crk in apoptosis is supported by the findings that specific targeting of Crk to the nucleus by fusing NLS sequences, spontaneously activates apoptosis and potentiates stimuli that induce apoptosis 182. Recent structural studies with CrkL indicate that the Crk NES may only function in the dimeric state of the protein, mediated by a homotypic interaction in the SH3C domains 175. These studies predict that the Crk NES motif may become exposed conditionally under certain circumstances, an event that would actively promote efflux of a nuclear Crk pool into the cytoplasm. However, the presence of an NES sequence in Crk II does not explain how Crk II initially trafficks to the nucleus, since Crk does not appear to possess any NLS sequences. It will be equally important to characterize the physiological signals that result in Crk translocation into the nucleus, and whether there exists a regulatory signal that controls the interaction between Crk and Crm1.

The Crk gene was identified about 20 years ago, and the characterization of Crk and its mode of cell signaling has arguably revolutionized the way we think about intracellular signal transduction. Despite significant advances and an enormous wealth of new information, the field is now faced with the challenging problem of how these complex signaling networks are controlled in time and space to produce biological responses. Efforts will need to be redirected from the relatively simple task of identifying specific effector proteins to an understanding of when and where these complexes are formed in cells. Development of FRET technology that reports specific protein-protein interactions, as well as the generation of a battery of phosphospecific antiseras against specific post-translational epitopes, should help provide a better understanding of Crk biology. This exciting field can expect substantial progress as the new observations continue. One thing is certain: the field of Crk biology will continue to grow over the next 20 years.