Structure-function analysis of Drosophila Tensin

Tensin is implicated in linking integrins to the cytoskeleton and signaling pathways. A tensin null was generated and is viable with wing blisters, a phenotype characteristic of loss of integrin adhesion. In tensin mutants, mechanical abrasion is required during wing expansion to cause wing blisters, suggesting that tensin strengthens integrin adhesion. The localization of tensin requires integrins, talin, and integrin-linked kinase. The N-terminal domain and C-terminal PTB domain of tensin provide essential recruitment signals. The intervening SH2 domain is not localized on its own. A model is suggested where tensin is recruited to sites of integrin adhesion via its PTB and N-terminal domains, localizing the SH2 domain so that it can interact with phosphotyrosine-containing proteins, which stabilize the integrin link to the cytoskeleton (Torgler, 2004).

Tensin contains two domains that contribute to its localization to integrin adhesive sites and one that counters this localization. The tensin protein can be divided into three domains: the poorly conserved N terminus (residues 1-456), the SH2 domain (residues 457-556), and the PTB domain (residues 578-720). The ability of individual domains to be recruited to integrin adhesive sites was examined by expressing GFP fusions of each domain separately or in combination. The UAS-GAL4 system was used to express individual domains in specific tissues, even if they caused a lethal dominant-negative phenotype. A GAL4 driver was used that expresses the constructs in the developing wing (Cy2), and another specific for the muscle (Mef2-Gal4), where protein localization could be examined in both the developing embryonic muscles and in the larval muscles where the sarcomeric structure is more established (Torgler, 2004).

As predicted from the distribution of the tensin-GFP rescue construct, expression of full-length tensin localizes to the ends of the embryonic and larval muscles and to the basal adhesion plaques in the developing wing. However, the localization in the muscles is not as tight as when expressed from its own promoter. Two factors contribute to the difference in the appearance of the GAL4-driven tensin-GFP in the muscles. (1) The endogenous gene is expressed on both sides of the muscle attachment site (muscle and epidermal cells), while the GAL4-driven protein is expressed just in the muscles. (2) GAL4-driven protein is expressed at much higher levels than the endogenous protein, judging by the levels of GFP fluorescence. In addition to being localized at muscle ends, overexpressed tensin-GFP was found at Z-lines. Since muscle ends are modified Z-lines, two possible explanations for this difference are (1) that tensin is normally present at the Z-lines, but it is below the limit of detection when tensin-GFP is expressed from its own promoter, or more likely (2) that overexpression of tensin or tensin fragments results in a saturation of the normal sites for recruitment at the muscle ends, so that tensin makes alternative weaker associations with related proteins at the Z-line. The overexpressed full-length protein does not perturb muscle function, judging by the normal mobility and viability of the larvae (Torgler, 2004).

Since SH2 domains are often used to recruit proteins to sites of signaling, the localization of the tensin SH2 domain on its own was examined. The tensin SH2-GFP fusion did not localize to sites of integrin adhesion in the muscle or the wing, and instead appeared distributed throughout the cytoplasm, with some concentration around the nuclei. The distribution of the tensin SH2 domain was compared with proteins phosphorylated on tyrosine, using the monoclonal antibody 4G10. Only a small amount of the SH2 domain colocalized with the major site of phosphotyrosine at the muscle ends. However, the pattern of phosphotyrosine staining varied depending on the antibody used. Thus, it is possible that the SH2 domain distribution in the cytoplasm is due to binding to phosphorylated partners that either are not apparent because of their diffuse distribution or are not recognized by the anti-phosphotyrosine antibodies used. What is clear is that the SH2 domain is not sufficient to recruit tensin to the sites of elevated phosphotyrosine at the muscle ends. Tensin itself does not appear to influence the level of signaling at the muscle ends, since in the absence of tensin, neither the distribution of phosphotyrosine nor the phosphorylation of focal adhesion kinase on tyrosine 397, a hallmark of integrin signaling, were altered. Regulation of the target gene 258 by integrin signaling also occurs normally without tensin, suggesting therefore that tensin is not required for integrin signaling (Torgler, 2004).

In contrast to the SH2 domain alone, a construct combining the N terminus with the SH2 domain results in strong localization at muscle ends and the Z-lines, suggesting that the N terminus contributes to localization. Since the N terminus alone confers a similar pattern of localization, it is sufficient for localization. Dividing the N terminus into two fragments shows that the localization signal resides within the first 229 amino acids; the N-terminal half retained strong localization, while the second half (230-449) was distributed throughout the cytoplasm, with a hint of association with the Z-lines visible in the larval muscles. Thus, the N terminus of tensin contains a targeting signal that results in the enrichment of tensin at the muscle attachment sites and Z-lines. Since the N terminus has actin binding activity, this enrichment could be due to direct binding to actin ends (Torgler, 2004).

The N-terminal domain does not confer identical localization to the full-length protein. This suggests that the full-length protein contains an additional localization signal, most likely the PTB domain. Consistent with this hypothesis, a construct expressing the SH2 domain with the PTB domain was recruited to sites of integrin adhesion, although still not as tightly as full-length tensin. Since the SH2 domain itself did not localize to the muscle ends, this suggested that the PTB domain provided the localization signal. Indeed, the PTB domain alone is recruited to larval muscle ends, with a small amount at Z-lines. However, the PTB domain alone was not detectable in embryonic muscles or developing wing. These findings suggest that the PTB domain contains a localization signal that requires the SH2 domain to function effectively (Torgler, 2004).

Tests were performed to see whether the influence of the SH2 domain on the PTB domain was due to the ability to bind to phosphotyrosine-containing peptides. Mutation of a highly conserved arginine residue within the FLVR motif of SH2 domains to glutamic acid abolishes the binding of this domain to phosphotyrosine peptides and in consequence abolishes the function of the SH2 domain. A comparable mutation (R883E in full-length tensin) was made in the SH2PTB construct and its distribution was examined. It was found expressed at levels similar to the SH2PTB domain, demonstrating that it is the additional sequence of the SH2 domain rather than phosphotyosine binding that improves the function of the PTB domain. More importantly, the mutant SH2*PTB product was localized much more tightly at muscle ends compared to the wild-type SH2PTB. Thus, contrary to signaling adaptors such as Grb2 where the SH2 domain plays a key role in the recruitment of the protein, the binding of the SH2 domain of tensin to proteins throughout the cell appears to be overcome by the recruitment activity of the PTB and/or N terminus (Torgler, 2004).

Analysis of localization signals within the tensin molecule has identified an activity for the N terminus and PTB domain, but not the SH2 domain. This then raises the question of how the SH2 domain contributes to tensin function. The function of the different domains was examined in two ways by testing whether (1) overexpressed domains had dominant-negative activity and (2) tensin molecules lacking individual domains were able to substitute for tensin (Torgler, 2004).

High-level expression of a central fragment of chicken tensin inhibits the formation of fibronectin fribrils by fibroblasts (Pankov, 2000), showing that expression of portions of tensin can perturb integrin function, possibly through dominant-negative effects on the function of endogenous tensin. No perturbation of muscle structure or function was detected by overexpression of any of the constructs described above in the muscles. Expression of the full-length protein or the N terminus did not cause wing blisters or wrinkles. However, expression of the PTB domain with the Cy2 Gal4 driver, alone or in combination with the SH2 domain, caused small bubbles and wrinkles in the wing; similar results were obtained with another driver, 69B. These are not identical to the large distal wing blisters that result from the absence of tensin, suggesting either that it is a weaker phenotype or that the PTB domain inhibits the function of proteins other than tensin itself. To distinguish between these, the effect was examined of reducing or increasing the amount of wild-type tensin on the dominant-negative effects caused by either the PTB or SH2PTB constructs; this did not substantially alter the penetrance of the dominant phenotypes. These constructs do not compete with tensin for interactions with target proteins, but it has been suggested that they perturb the function of another protein (neomorphic rather than antimorphic). Thus, endogenous tensin is not easily perturbed by overexpression of single tensin domains, but overexpression of the PTB domain inhibits a process required for wing development (Torgler, 2004).

The role of the different domains in tensin function was examined, using either Cy2-driven UAS constructs or genomic rescue constructs to test their ability to rescue the null blistery allele. Neither the construct deleting both SH2 and PTB domains nor the construct deleting the PTB domain alone was able to rescue by, demonstrating that the PTB domain is essential for function. The N terminus localizes to muscle ends and Z-lines in the absence of endogenous tensin, ruling out the possibility that it becomes localize by binding to wild-type tensin. Similarly, the combined SH2 and PTB domains still localized in a tensin mutant background. To test if expression of the SH2PTB construct was capable of rescuing the wing blister phenotype, despite its dominant-negative effects, the construct was expressed in the developing wing of by flies. Presumably due to uneven expression, only partial rescue was achieved with the control wild-type UAS-tensin construct: 22.5% were fully rescued; 59.5% were partially rescued, with either small blisters on both wings or a blister on just one wing; 18% were not rescued, with blisters on both wings. Expression of the SH2PTB construct did not rescue the blister phenotype at all, demonstrating that the N-terminal domain is essential for the function of tensin (Torgler, 2004).

Finally, an examination was performed to determine whether the ability of tensin to bind to phosphorylated proteins through its SH2 domain is essential for tensin function. A point mutation, which should inactivate the tensin ability to bind phosphotyrosine, was tested; this was engineered into the full rescue construct. Transgenes encoding this mutant tensin were localized properly in the tensin null mutant but were largely inactive in rescuing the wing blister phenotype: depending on the line, 67%-100% of the flies had the by mutant phenotype of blisters in both wings; 0%-14% had a partial rescue with a blister in one wing; and 0%-19% were rescued, with no blisters (5 lines tested, n = 528). This demonstrates that SH2 function is important for tensin function. The observation that there is some rescue suggests that the other domains contribute a function independent of the SH2 domain, in addition to recruiting the SH2 domain to the right place. Thus, each domain of tensin contributes to its normal function in the developing animal (Torgler, 2004).

A function for tensin in strengthening the integrin adhesive complex is consistent with the finding that tensin localization requires the prior recruitment of talin and ILK. The protein PINCH, which binds to ILK, is not required, consistent with the finding that ILK is still recruited to muscle ends in the absence of PINCH (Clark, 2003). Functionally, tensin is added after talin and ILK, but this does not correspond to a distinguishable temporal sequence, since they were simultaneously recruited to the developing muscle attachments. In mammalian cells, the spatial separation of the early and late adhesions allows such a distinction: the initially formed focal complexes have little tensin, and the fibrillar adhesions, which arise from the focal complexes, have high levels of tensin (Zamir, 1999). However, these results appear to contradict the finding that tensin is associated with integrins prior to ligand binding, and prior to the association of talin with integrins (Miyamoto, 1995). This can be resolved by postulating that tensin initially associates with unbound integrins, is displaced when integrins bind extracellular ligand in the focal complex, and then reassociates with the integrin contacts as they mature into fibrillar adhesions. A putative dominant-negative construct of tensin inhibits the formation of fibrillar adhesions (Pankov, 2000). This is a stronger defect than would be predict from the Drosophila tensin mutant, suggesting either that fibrillar adhesions are particularly sensitive to the loss of tensin or that the dominant-negative construct may inhibit other proteins, as was found when overexpressing the Drosophila PTB domain (Torgler, 2004).

A stabilizing function for tensin also fits well with the phenotype of the tensin1 knockout mouse (Lo, 1997). These mice are viable, but show defects in kidney structure that are enhanced with time. As in Drosophila, the integrin junctions appear to be formed normally, but were not stable. Kidneys could be particularly sensitive to the loss of tensin, as are Drosophila wings, or this could be the one tissue that lacks expression of the other tensin genes. Judging from the by phenotype, it is predicted that a triple knockout of tensin genes in mouse may extend the loss of integrin junction stability to additional tissues, rather than producing a more severe phenotype at integrin junctions (Torgler, 2004).

By replacing the endogenous tensin with mutated forms, this study shows that three different regions of the protein are essential for function. Deletion of the N terminus or PTB completely inactivates the protein. For the SH2 domain, a point mutation was used that eliminates binding to phosphotyrosine, which almost completely inactivates tensin, showing that binding to phosphorylated proteins is essential for tensin function. Partial rescue by this mutant indicates an additional function independent of the SH2 domain. Examination of the distribution of the different domains has provided further insight into how tensin functions (Torgler, 2004).

From the analysis of tensin domain distribution, it is concluded that the N terminus and PTB domains each bind to a protein(s) localized at sites of integrin adhesion. The difference in distribution of these two domains suggests that they bind to different proteins. The preferential localization of the PTB domain at muscle ends suggests that it interacts with a protein associated with the integrin cytoplasmic domain. In contrast, the N terminus is more equally distributed between the muscle ends and the Z-lines, where barbed ends of actin are situated, suggesting that tensin binds a protein concentrated at this end of actin, or actin itself. The N terminus binds filamentous actin in vitro, supporting the idea of direct binding. Tensin may therefore function as an adaptor between the integrin complex and actin, independent of the SH2 domain. The combination of these two localization signals led to a greater fraction of tensin colocalizing with integrins. This suggests that it is the combination of two low-affinity binding sites that achieves the normal tight localization (Torgler, 2004).

A potential binding partner for the tensin PTB domain is the cytoplasmic tail of the integrin β subunit. All β subunits contain at least one PTB binding motif, NPXY, and many contain two. The colocalization of tensin with integrins that were clustered but not binding ligands has suggested that tensin may bind to the integrin tail prior to ligand binding (Miyamoto, 1995). Using an in vitro assay for protein binding, the PTB domain of human tensin2 was shown to bind strongly to the cytoplasmic tails of the human integrin subunits β3, β5, and β7 and weakly to β1A (Calderwood, 2003). This shows that the tensin PTB domain can bind to the NPXY motifs present in the β cytoplasmic tails and that it can bind nonphosphorylated sites. However, it seems unlikely that tensin is directly bound to the integrin tail in the mature integrin adhesion junctions in Drosophila muscle and wing. This is because tensin is not colocalized with integrins in the absence of talin or ILK. Neither talin nor ILK has an NPXY motif, suggesting that they do not recruit tensin directly. This suggests that the PTB interactor has yet to be identified (Torgler, 2004).

The distribution of tensin SH2 domain constructs led to the idea that one role of the two flanking domains is to concentrate the SH2 domain at the right location, where it can associate with the appropriate target proteins. Surprisingly, the distribution of the SH2 domain did not correspond to sites of strong phosphotyrosine staining, even though this is coincident with the site of integrin adhesion; indeed, the SH2 domain was widely distributed in the cell when expressed independently. This indicates either that it binds to proteins that are also widely distributed or that its affinity for the correct target is weak. This is different from the classical way of thinking about SH2 domains as the primary domain that recruits a protein complex to sites of active signaling at the membrane. However, it is consistent with the finding that SH2 domains have high specificity for phosphorylated versus nonphosphorylated tyrosine, but only weak discrimination for the flanking residues of a particular phosphotyrosine containing sites, suggesting that SH2 domains must rely on other domains in the protein to achieve sufficient specificity. Mutation of the arginine residue that binds phosphotyrosine has in all cases been shown to perturb interaction with phosphorylated targets. The finding that this mutation in the SH2PTB construct improves its localization to sites of integrin adhesion shows that the nonlocalized cytoplasmic distribution of the SH2 domain is due, at least in part, to binding phosphoproteins. This 'mislocalization' activity appears to be overcome normally by the combination of the N-terminal domain and the PTB domain (Torgler, 2004).

These findings have been combined into a model of tensin function. The two ends of tensin bind to proteins that are associated with integrins and the ends of actin. The binding partners of tensin require the prior association of talin and ILK to the integrin adhesive complex before they can recruit tensin. The recruitment of tensin restricts its SH2 binding activity to this site, where it binds to proteins phosphorylated on tyrosine. This is the only function of tensin that this study identified that qualifies as a signaling activity, since loss of tensin did not perturb the integrin signaling pathways examined. The combination of the crosslinker function provided by the N- and C-terminal binding sites of tensin and the activity of the associated phosphorylated protein results in the stabilization of the integrin adhesive complex (Torgler, 2004).

This structure-function analysis of Drosophila tensin is fully in accord with the recent findings of Chen (2003), who analyzed the ability of tensin domains to localize to sites of integrin function in mammalian cells in culture. Similar to what has been described in this study, Chen found two focal adhesion localization sites within tensin, one at the N terminus, and one at the PTB domain. Thus, even though the primary sequences of the N termini of vertebrate and Drosophila tensin are poorly conserved, they both have similar activities in associating with the ends of actin filaments. As was found for Drosophila tensin, the N terminus associated more broadly with actin, while the PTB domain localized to focal adhesions and caused a dominant-negative effect. Chen also found that the SH2 domain did not play a role in recruitment, but in that system the mislocalization activity of the SH2 domain that is described here for Drosophila could not be distinguished (Torgler, 2004).

blistery: Biological Overview | Evolutionary Homologs | Developmental Biology | Effects of Mutation | References

Home page: The Interactive Fly © 1997 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.