Integrins are the major cell surface receptors mediating cell–extracellular matrix (ECM) adhesion in metazoans. Integrin heterodimers, consisting of an α- and β-subunit, span the plasma membrane to form a link between the ECM and the cytoskeleton. Integrins are highly conserved in animals and are essential for vertebrate, fly and worm development (reviewed in Bökel and Brown, 2002; Meighan and Schwarzbauer, 2008). In general, integrin-mediated adhesion contributes to morphogenesis through two mechanisms: dynamic, short-term adhesion to mediate events such as cell migration and cell rearrangement, and stable, long-term adhesion to mediate tissue maintenance (Bökel and Brown, 2002; Meighan and Schwarzbauer, 2008). During dynamic processes, integrins function in several ways: firstly, they mediate attachment between cells and the ECM, providing traction for cell migration; secondly, integrins act as signaling receptors that regulate cell growth and differentiation; thirdly, integrins contribute to ECM assembly in some tissues (reviewed in Brown et al., 2000; Narasimha and Brown, 2004a). Together these activities help organize cells into distinct tissues and organs. Dynamic processes shape the embryo, but as development progresses and tissues form, the importance of maintaining three-dimensional tissue architecture becomes paramount, requiring stable integrin adhesive contacts.
In Drosophila, both transient and stable integrin-mediated adhesion play essential roles during tissue development and long-term maintenance (Bökel and Brown, 2002). For example, transient integrin-mediated adhesion is required for several dynamic morphogenetic processes, including germband retraction and dorsal closure (Fig. 3; reviewed in Bökel and Brown, 2002; Narasimha and Brown, 2004a). Both processes require transient, integrin-mediated cell–ECM adhesion to drive large-scale epithelial migration, lamellipodia formation and cell shape changes (Brown et al., 2000; Schock and Perrimon, 2002). At the end of embryogenesis, integrins are also required for the maintenance of a diverse array of tissues, including myotendinous junctions (Brown et al., 2000) and terminal tracheal branches (Levi et al., 2006). To provide stable adhesion, integrins at myotendinous junctions (MTJs) recruit a large intracellular adhesion complex (IAC) that includes such proteins as talin (Brown et al., 2002) and PINCH (Clark et al., 2003). This complex enables stable attachment of tensile muscles to the epidermis by linking the muscle cytoskeleton by means of integrins to ECM deposited between the muscle ends and tendon cells; disruption of several IAC components or their association with β-integrin results in loss of the muscle–tendon cell link and subsequent muscle detachment (reviewed in Brown et al., 2000; Narasimha and Brown, 2004a). The formation of MTJs involves many aspects of integrin function, including recruitment and assembly of a large IAC, cytoskeletal attachment, ECM deposition and adhesion to ECM substrate. Drosophila is especially amenable to integrin studies because integrin-mediated adhesion plays diverse roles during fly development that are predominantly fulfilled by just one β subunit (βPS-integrin; encoded by myospheroid or mys), compared with eight in mammalian systems that often function together within the same tissue.
All known β-integrin subunits share a high degree of conservation in both sequence and structure across species (reviewed in Hynes, 2002). β-integrin has three distinct regions, all of which make unique contributions to its function: the extracellular region (∼800 amino acids [aa]), the single-pass transmembrane domain (∼20aa), and the cytoplasmic tail (usually between 13 and 70aa). The I-like domain resides in the extracellular region and contains a Metal Ion-Dependent Adhesion Site (MIDAS) essential for ligand binding (reviewed in Hynes, 2002). The transmembrane (TM) domain is thought to both stabilize inhibitory interactions with the α-subunit and promote β-integrin activation through homomeric β-subunit clustering (reviewed in (Hynes, 2002). The short cytoplasmic tail binds cytoskeletal and signaling adaptors required for integrin function (reviewed in Legate and Fassler, 2009).
Extensive structural analyses and biochemical experiments have shown that the overall strength of integrin-mediated adhesion (“avidity”) is principally regulated in two ways: through changes in receptor affinity and valency (reviewed in Liu et al., 2000; Humphries et al., 2003). The affinity of individual integrin heterodimers for ECM ligands is determined by conformational changes (reviewed in (Askari et al., 2009), while valency, or the number of receptor-ligand bonds, is regulated by β-integrin density or “clustering” (reviewed in Carman and Springer, 2003). Integrins undergo a series of structural changes that result in activation, or a shift from low to high affinity for ECM ligands. This involves extension of the extracellular domain, separation of TM domains and breaking of a cytoplasmic salt-bridge between the α- and β-tails that stabilizes the inactive state (reviewed in Hynes, 2002). Integrin activation can be initiated from inside (“inside-out activation”) or outside the cell (“outside-in activation”) because integrins are allosteric: binding of molecules to an integrin domain on one side of the membrane can induce conformational changes in domains on the other side of the membrane. In outside-in signaling, binding of an extracellular ligand induces long-range conformational changes across intervening domains which lead to separation of the TM and cytoplasmic domains to allow interaction of the cytoplasmic tails with signaling and cytoskeletal molecules; inside-out signaling involves similar changes in the reverse order (reviewed in Askari et al., 2009). Disruption of these regulatory mechanisms results in gross abnormalities in cell architecture and tissue morphology.
During animal development, adhesive processes are regulated to enable progression of dynamic morphogenetic processes and subsequent transition to stable tissue maintenance. Although in vitro studies have provided a wealth of information about the mechanisms that control integrin activation, the roles of these mechanisms during development are poorly understood. Here, we have carried out an in vivo analysis of β-integrin mutations that either activate or deactivate specific functions of integrins in vitro to ask how perturbing these activities affects tissue morphogenesis and maintenance. We replaced wild-type βPS in Drosophila embryos with mutants designed to disrupt regulation of inside-out and outside-in integrin activation. We find that different tissues require distinct modes of integrin function and that conformational changes associated with ligand binding are essential for all developmental processes examined. In addition, we demonstrate that β-integrin must be continuously regulated to support stable adhesion during tissue maintenance. Our approach has allowed us to define roles for various mechanisms that impinge on βPS-integrin activity during Drosophila development.
To examine the roles of βPS-integrin regulation during Drosophila development, we selected nine mutations that alter β-integrin activity in vitro and generated corresponding βPS constructs for transgenic in vivo rescue experiments (Fig. 1). These mutations, spanning several key regulatory sites within β-integrin, were previously characterized in structural and/or biochemical studies and their mechanisms of action have been well defined (reviewed in Humphries et al., 2003). We mutated residues within the I-like, TM, and cytoplasmic domains (Fig. 1) and tested their functions during development by removing endogenous βPS-integrin (using the dominant female sterile germline clone technique; Chou and Perrimon, 1996) and replacing it with ubiquitously expressed mutant βPS-integrins. The wild-type (WT) construct rescued embryonic defects associated with loss of integrin function (Table 1). As all mutant transgenes were derived from this WT transgene, we could therefore assess the degree to which the mutant proteins were able to compensate for loss of endogenous βPS-integrin.
Table 1. Phenotypic Penetrance and Rescue Efficiencies of Mutant βPS-Integrin Rescue Transgenesa
Germband retraction defective
Dorsal closure defective
See the Experimental Procedures section for normalized rescue method. The penetrance of defects in germband retraction (GBR), dorsal closure (DC), and myotendinous junctions (MTJs) were scored in null mutant (GLC) embryos rescued with ubiquitously expressed βPS-integrins carrying the mutations indicated in the left column. Given that in null mutant embryos, GBR and DC defects were not completely penetrant and the wild-type (WT) rescue construct typically gave 90-100% rescue, all rescue efficiencies were calculated by normalizing to the rescue obtained with the WT transgene.
Mutational Analysis of Drosophila β-Integrin
Based on published biochemical studies of the closest vertebrate relatives of Drosophila βPS-integrin, β1- and β3-integrin, we designed mutations in βPS-integrin to perturb outside–in signaling by interfering with ECM binding or extracellular domain extension (Fig. 1). Mutations designed to alter inside-out signaling generally disrupted adaptor binding to the cytoplasmic tail or conformational changes in the TM domain.
To address the roles of ligand-induced outside-in signaling during development, we introduced mutations in the I-like domain (Fig. 1). Previous work has identified three I-like MIDAS domain residues critical for ECM binding and/or downstream conformational changes (Bajt and Loftus, 1994). Two residues, D192/S194, were mutated to Alanine, as in vitro studies show that equivalent mutations β3 (D119/S121) and in β1 (D130/S132) abolished binding to ECM ligands (Takada et al., 1992; Bajt and Loftus, 1994; Cierniewska-Cieslak et al., 2002; Chen et al., 2006b). Based on vertebrate studies of β1 and β2, the second I-like domain mutation, S196F (S123 in β3, S134 in β1, S138 in β2), was designed to bind to a subset of ECM molecules but prevent the resulting conformational changes essential to transmit outside-in signals (Bajt and Loftus, 1994; Hogg et al., 1999; Chen et al., 2006b). This allele was identified previously in Drosophila and found to behave effectively as a missense “null” (Jannuzi et al., 2002, 2004), but detailed phenotypic analyses were not performed. In addition, the effect of enhanced outside-in activation was examined using the I-like domain mutation, L211I (L138 in β3; Luo et al., 2009), which promotes vertebrate β3 activation by stabilizing the extended extracellular domain conformation (Luo et al., 2009).
To examine how inside-out signaling and intracellular adhesion complex formation contribute to βPS function during development, we generated mutations in the cytoplasmic tail (Fig. 1). Many signaling and cytoskeletal adaptors bind to few sites in the cytoplasmic tail to regulate integrin activity (reviewed in Legate and Fassler, 2009). We introduced mutations in the three major binding sites: the two NPxY motifs and the membrane-proximal region. We generated the D807R mutant (D723R in β3) because charge reversal of this critical “salt-bridge” residue in the membrane-proximal region (HDRK motif) is reported to cause hyperactivation by disrupting inhibitory interactions with α-integrin (Hughes et al., 1996; Anthis et al., 2009a). This mutant is reported to be among the strongest activating mutations in β-integrin in vitro (Hughes et al., 1996; Anthis et al., 2009a; Bunch, 2009). Another mutant designed to activate inside-out signaling, Y831F/Y843F or YY>FF (Y747/Y759 in β3), disrupts NPxY tyrosine phosphorylation. Although the effects of NPxY tyrosine phosphorylation have been studied extensively, conflicting reports demonstrate differential effects on integrin function (reviewed in Legate and Fassler, 2009). Several lines of evidence indicate that these mutations activate vertebrate integrins by enhancing binding of the adapter protein, talin (Johansson et al., 1994; Datta et al., 2002; Millon-Frémillon et al., 2008; Oxley et al., 2008; Anthis et al., 2009b), which has an established role in activating integrins (reviewed in Calderwood, 2004). We tested the effects of the YY>FF mutation in flies using fluorescence recovery after photobleaching (FRAP) and found that the mobile fraction, a proven method for measuring integrin turnover (Yuan et al., 2010), was lower than WT at MTJs (Supp. Fig. S1, which is available online). A lower mobile fraction reflects increased levels of activation (Cluzel et al., 2005), therefore, these data suggest that blocking phosphorylation at the NPxY motifs (YY>FF) increases βPS-integrin activity in this context. We generated another activating mutation in the TM domain, L796R, which mimics talin-induced conformational changes by shortening the length of transmembrane domain that is buried in the plasma membrane (Wegener et al., 2008). We also introduced a deactivating mutation, N840A, in the distal NPxY motif, which in vertebrates is required to mediate binding of various signaling molecules, notably kindlins, which are required for inside-out integrin activation (Moser et al., 2008; Harburger et al., 2009). In addition, we generated a truncation mutant, 804*stop, which lacks the majority of the cytoplasmic tail (Fig. 1). As this mutant cannot nucleate an intracellular adhesion complex, it should abrogate inside-out signaling and intracellular signal transduction without perturbing extracellular adhesive functions (Ylänne et al., 1993; Kääpä et al., 1999). Finally, we investigated whether integrin clustering would affect development. To this end, we used the TM mutation G792N, shown in vertebrates to enhance clustering by disrupting packing of α and β TM helices (G708 in β3; Li et al., 2003).
The effects of these mutations on integrin function have been well characterized in vitro by various means, and for most, structural data in addition to activity assays support the proposed mechanistic effects. Using this array of mutants, we systematically analysed how dysregulation of multiple mechanisms of integrin function affect Drosophila development. Comparing rescue with these regulatory mutants to WT transgenic rescue allowed us to discern which aspects of βPS-integrin function were essential for each morphogenetic process examined.
Relative Overall and Cell Surface Expression Levels of Mutant Rescue Constructs
An important criterion in performing rescue experiments is that expression levels among transgenic constructs are similar and that sufficient amounts of protein localize to allow rescue. Quantitative Western blotting was used to determine relative protein expression levels of at least three independent insertion lines per transgene. Insertions expressed at similar levels were chosen for rescue experiments (Fig. 2a). To determine whether the mutant proteins localized, cell surface expression levels were analyzed at MTJs in muscles, where WT integrin normally accumulates (Fig. 2b). All rescue constructs were fused to yellow fluorescent protein (YFP) and detected by YFP immunofluorescence (see Experimental Procedures). We quantified transgenic protein enrichment at MTJs using an established assay that measures the average ratio of cortical to cytoplasmic YFP fluorescence (Tanentzapf, 2006; Tanentzapf and Brown, 2006; Devenport et al., 2007). All constructs localized efficiently to MTJs with the exception of D807R (Fig. 2b,c), and quantitation of surface expression levels showed that mutant and WT proteins at sites of integrin adhesion were generally similar (Fig. 2b,c).
Contrasting our results, D807R in vertebrates (D723R in β3) is reported to exhibit normal surface localization in cultured cells (Hughes et al., 1996; Bunch, 2009). However, in agreement with NMR studies of β3-integrin (Anthis et al., 2009a), this mutation disrupted interaction with talin (see below), which is reported to be required for surface localization of integrins (Martel et al., 2000). Thus, the D807R mutant may localize poorly due to disrupted talin binding.
β-Integrin Dysregulation During Tissue Morphogenesis: Germband Retraction
In the absence of integrin-mediated adhesion, two dynamic morphogenetic movements are disrupted: germband retraction (GBR; Fig. 3e; compare to 3c) and dorsal closure (DC; Fig. 3d; compare to 3b). GBR occurs over ∼2 hours during mid-embryogenesis and serves to unfurl the caudal end of the germband (Fig. 3a; Martinez Arias, 1993). GBR involves integrin-dependent epithelial sheet movements thought to be driven entirely by cell migration and cell shape changes (reviewed in Schock and Perrimon, 2002).
To determine which βPS-integrin regulatory mechanisms are important for GBR, we examined whether mutant integrin transgenes could compensate for loss of βPS during this process. In embryos lacking both maternal and zygotic βPS-integrin, 47% of embryos failed to undergo GBR (n = 94; Fig. 3d; Table 1). We considered this value, 47%, to be the baseline value against which rescue was calculated, equating it to 0% rescue for normalization calculations (see the Experimental Procedures section). In comparison, only 5% of βPS-integrin–deficient embryos ubiquitously expressing transgenic WT βPS exhibited GBR defects (n = 85; Table 1), which we equated to 100% rescue. “Normalized rescue,” or the ability of mutant transgenes to rescue GBR relative to the WT transgene, was determined dividing mutant rescue capacity by WT transgenic rescue capacity (Table 1; see the Experimental Procedures section). For clarity, we generally discuss only normalized rescue efficiencies (see Table 1 for raw data).
Mutations designed to abrogate outside–in integrin signaling disrupted GBR (Fig. 4a; Table 1). βPS-integrin–deficient embryos rescued with the S196F transgene, which cannot transduce outside–in signals, completely failed to undergo GBR (n = 61). Surprisingly, 62.6% normalized rescue was obtained using the D192A/S194A mutant designed to block ligand binding (n = 44). In contrast, the L211I transgene, a putative enhancer of outside–in signaling, rescued GBR (n = 81).
Our data suggest that inside-out integrin signaling is not essential for GBR but may contribute to the process (Fig. 4a; Table 1). Surprisingly, we found that the cytoplasmic tail truncation mutant, 804*stop, exhibited 80.4% normalized rescue of GBR (n = 54). Complete rescue was observed using the activating D807R (n = 55), the phospho-mutant YY>FF (n = 68) and the clustering-inducing G792N (n = 59) transgenes. The activating TM mutant, L796R, conferred moderate but reduced normalized rescue (73.8%; n = 78). Aside S196F, the only mutant that did not substantially rescue GBR was the distal NPxY mutant, N840A, which exhibited 35.5% normalized rescue (n = 50). However, this may reflect dominant-negative effects because it did not rescue as well as the truncation mutant, 804*stop (Table 1). Altogether, these experiments suggest that, while ECM binding and conformational changes associated with outside–in signaling contribute to GBR, regulation of βPS-integrin function through inside-out signaling is not absolutely required for this process.
β-Integrin Dysregulation During Tissue Morphogenesis: Dorsal Closure
DC, another integrin-dependent morphogenetic event, lasts ∼2 hr and serves to cover the discontinuity that remains in the embryonic dorsal epidermis after GBR. During DC, lateral epidermal cells elongate along the dorsal–ventral axis and spread over the dorsal surface as the underlying amnioserosal cells constrict. The amnioserosa then invaginates while filopodia from the overlying epithelia “zipper” together to facilitate closure along the midline (reviewed in Jacinto et al., 2002).
To determine which βPS-integrin regulatory mechanisms are important for progression of DC, we analyzed whether mutant integrin transgenes could compensate for the absence of βPS-integrin during this process. While 67% of embryos lacking βPS failed to undergo DC (n = 94; Fig. 3e; Table 1; 0% normalized rescue), only 7% of mutant embryos ubiquitously expressing transgenic WT βPS exhibited DC defects (n = 100; Table 1; 100% normalized rescue).
DC was strongly disrupted in embryos rescued with mutant βPS constructs designed to perturb outside–in integrin signaling (Fig. 4b; Table 1). Embryos rescued with the D192A/S194A and S196F transgenes exhibited severe DC defects similar to or stronger than integrin-deficient embryos (n = 35 and 73, respectively). The activating mutant, L211I, rescued DC (n = 76).
In addition, most mutations designed to alter inside-out signaling failed to rescue DC (Fig. 4b; Table 1). Neither of the activating mutants, D807R or YY>FF, ameliorated DC defects (n = 63 and 41, respectively). Moreover, the 804*stop transgene enhanced the penetrance of DC failure to 95% (n = 39), approximately double that of the null mutant. In contrast, the deactivating mutant, N840A, gave 68.7% normalized rescue (n = 62). The activating L796R TM domain mutant and clustering-promoting G792N mutant nearly completely rescued DC (n = 81 and 49, respectively). Overall, these results demonstrate that proper modulation of both inside-out and outside–in integrin signaling is essential for DC.
Loss of integrin function at MTJs disrupts stable adhesion between muscles and tendon cells. MTJs fail at the onset of muscle contraction during stage 16, giving rise to a characteristic muscle detachment phenotype by stage 17 (Fig. 3g; Wright, 1960; Brown, 1994). To determine which βPS regulatory mechanisms are important for MTJ maintenance, we examined whether WT and mutant integrin rescue constructs could compensate for loss of endogenous integrin in muscles in mid- to late-stage 17 embryos. While 100% of embryos lacking βPS-integrin exhibited MTJ defects (n = 51; Fig. 3g), only 5.7% of mutant embryos ubiquitously expressing the WT transgene exhibited MTJ defects (n = 53; Table 1; 100% normalized rescue).
Disrupting outside–in integrin signaling resulted in severe but variable muscle detachment defects in mid- to late-stage 17 embryos (Fig. 4c; Table 1). The S196F rescue transgene conferred just 9.7% normalized rescue (n = 55). Surprisingly, the D192A/S194A mutant conferred 22.5% normalized MTJ rescue (n = 78) at mid- to late-stage 17, however, MTJ defects were fully penetrant by early larval stages. Therefore, muscle detachment was delayed in embryos rescued with the D192A/S194A transgene. Conversely, the activating mutant L211I rescued MTJ defects (n = 65).
Mutants designed to alter inside-out signaling gave little or no rescue of MTJ defects (Fig. 4c; Table 1): the D807R, YY>FF and 804*stop constructs completely failed to maintain muscle attachments (n = 53, 75, and 56, respectively). In contrast, the deactivating mutation, N840A, gave 36% normalized rescue (n = 47). Nearly complete rescue was obtained using the activating L796R and clustering-inducing G792N mutants (n = 69 and 48, respectively). Altogether, these results demonstrate that ongoing regulation of inside-out and outside–in βPS-integrin signaling are required for long-term maintenance of stable adhesion at MTJs.
βPS-Integrin Regulation Maintains MTJ Morphology
To further examine MTJ defects in βPS-integrin mutant rescue embryos, we analyzed MTJ morphology. Since deposition of the ECM component Tiggrin (Fogerty et al., 1994) reflects MTJ morphology, morphometric analyses of the lengths and areas of ventral longitudinal muscle attachments were done using confocal projections of mid- to late-stage 17 embryos immunolabeled for Tiggrin (see Experimental Procedures; Supp. Table S1). In βPS-integrin-deficient embryos expressing the WT transgene, average MTJ length and area were 30.8 μm and 69.1 μm, respectively (Fig. 5a,i,j). In embryos lacking endogenous βPS-integrin, MTJs were shorter and smaller overall, at 10.4 μm and 18.6 μm (Fig. 5b,i,j), just 33.6% of the length and 26.9% of the area of WT MTJs on average.
Reductions in both MTJ length and area were measured in embryos rescued with βPS-integrin transgenes designed to disrupt outside-in integrin signaling: MTJs in D192A/S194A rescues were 58.3% and 56.8% of the length and area of those of the WT transgenic rescue, and MTJs in the S196F rescues were 77.0% and 70.3% of the length and area of the WT rescue (Fig. 5i,j; Supp. Info. Table S1). By contrast, the activating mutant L211I rescued MTJ length and partially rescued MTJ area.
βPS-integrin mutants designed to disrupt or enhance inside-out activation conferred varying effects on MTJ morphology (Fig. 5i,j). Moderate rescue was attained using the activating cytoplasmic mutant, D807R, and the deactivating 804*stop mutant: MTJs in D807R rescues were 67.6% and 54.8% of the length and area of MTJs in WT rescues, and MTJs in 804*stop rescues were 69.0% and 70.9% of the length and area of WT rescues (Fig. 5i,j; Supp. Info. Table 1). The activating mutation YY>FF conferred better rescue of MTJ length and area, as MTJs were 88.0% of the length and 75.8% of the area of the WT rescue. Interestingly, although MTJ length was fully rescued by the deactivating mutation, N840A, a mild but significant reduction in MTJ area was observed (74.4% of WT rescue; Fig. 5i,j; Supp. Info. Table 1). The activating mutant, L796R, and the clustering-inducing mutant, G792N, conferred strong rescue of MTJ length and area (Fig. 5i,j; Supp. Info. Table 1). Overall, these results suggest that persistent regulation of both inside-out and outside-in integrin signaling are essential for maintenance of MTJ morphology.
Intra- and Extracellular βPS-Integrin Regulatory Mechanisms Control IAC Assembly
To investigate how MTJ defects arise, we examined whether βPS-integrin dysregulation impinged on IAC assembly. We assayed whether each mutant rescue construct recruited IAC components to MTJs using immunofluorescent labeling for talin (Fig. 6) and PINCH (Fig. 7). In βPS-integrin–deficient embryos rescued with the WT transgene, talin colocalized with integrin at MTJs (Fig. 6a,b), while in βPS-integrin mutants, localization of talin was lost (Fig. 6c).
Intriguingly, transgenes designed to either disrupt or enhance outside-in signaling (D192A/S194A, S196F, L211I) rescued talin recruitment to MTJs (Fig. 6a,d,e). Conversely, some mutants designed to interfere with inside-out integrin signaling did not affect talin localization (N840A, YY>FF; Fig. 6a,h,i). While the D807R mutation would be expected to significantly reduce talin recruitment to MTJs (Fig. 6a,g) according to vertebrate studies (i.e., Anthis et al., 2009b), this mutant localized very poorly; therefore, we could not properly assess its affects on talin recruitment (Fig. 2). As expected, the cytoplasmic truncation mutant, 804*stop failed to maintain talin at MTJs. Together, these data suggest that talin recruitment to MTJs is only partially dependent on outside-in integrin signaling.
We next examined localization of PINCH, an essential component of the IAC (Clark et al., 2003), by determining the ratio of PINCH to αPS2-integrin fluorescence at MTJs (Fig. 7). This method was previously used to quantify PINCH enrichment and thereby IAC assembly at muscle attachments (Devenport et al., 2007). Counterstaining for αPS2 was used to label integrin heterodimer localization. In WT transgenic rescues, PINCH localized to MTJs (Fig. 7a,b), while in the absence of βPS-integrin, localization of both PINCH and αPS2 was lost (Fig. 7c). In comparison, severe reductions in PINCH localization to MTJs were observed in embryos rescued with the D192A/S194A transgene (Fig. 7a,d). However, the S196F transgene conferred significant rescue of PINCH recruitment, although at a reduced level compared with the WT transgene (Fig. 7a,e). Interestingly, the outside-in activating mutation L211I only partially rescued PINCH recruitment (Fig. 7a,f). With the exception of 804*stop (expected not to bind any intracellular components), mutants that disrupted inside-out regulation generally did not impinge upon PINCH localization (Fig. 7a,g–i; 804*stop data not shown). Notably, PINCH recruitment was significantly enhanced in embryos rescued with the phospho-mutant, YY>FF (Fig. 7a,h). Altogether, these data suggest that maintenance of the IAC at MTJs requires ongoing regulation of inside-out and outside-in integrin signaling.
In this report, we investigate the importance of integrin-mediated inside-out and outside-in signaling during distinct morphogenetic processes in Drosophila embryos. Our results suggest essential and differing roles for outside–in and inside-out integrin signaling during tissue formation and maintenance in Drosophila. We demonstrate that βPS-integrin must be regulated by different mechanisms among different tissues; although integrin-dependent morphogenetic movements such as GBR and DC may appear mechanistically similar at the tissue level, they in fact require different modes of βPS regulation. We find that long-term adhesion and tissue maintenance also require ongoing regulation of integrin adhesion. Overall, our data highlight the importance of transmitting information by means of integrins between the outside and inside of the cell to ensure proper progression of morphogenesis.
Differential Regulation of Integrin Activation Is Essential for Dynamic Morphogenetic Processes
Integrin-mediated adhesion is known to play essential roles during dynamic morphogenetic processes that involve cell migration through collective epithelial sheet movement (reviewed in Schmidt and Friedl, 2010). We found that two similar, dynamic morphogenetic processes in the embryo, GBR and DC (Schock and Perrimon, 2002; Narasimha and Brown, 2004a), required distinct aspects of both inside-out and outside-in integrin activation. We found GBR to be significantly more robust than DC as a process, as it was only partially affected by mutations that impinge on integrin regulation while DC was highly sensitive to modulation of all integrin functions assayed. Strikingly, GBR could be rescued by integrins designed to completely abolish all intracellular functions (804*stop) and partially rescued by a mutant designed to disrupt ECM binding (D192A/S194A). Two potential explanations for this surprising observation are that the roles of βPS-integrin in GBR may center on domains of the protein that we did not disrupt, or that ECM binding and cytoplasmic domain functions may be redundant during this process. Moreover, these results suggest that the molecular functions of βPS in GBR may be different from other contexts, in which ECM binding and/or cytoplasmic domain functions are absolutely required, such as in tracheal development (i.e., Levi et al., 2006), wing epithelial adhesion (i.e.., Domínguez-Giménez et al., 2007), and midgut migration (i.e., Li et al., 1998; Narasimha and Brown, 2004b). Understanding how integrins function during GBR requires an understanding of the cause(s) of the null phenotype; however, it is currently unclear why loss of βPS-integrin causes GBR failure and this limits interpretation of our GBR data. However, we propose that at the molecular level, outside-in activation of βPS is important GBR while inside-out activation is dispensable, and that βPS functions differently in this process than in its “canonical” roles requiring ECM binding and cytoskeletal linkage.
In contrast to GBR, DC was highly sensitive to perturbations to both outside-in and inside-out βPS-integrin signaling. Thus, DC depends on multiple aspects of integrin function and requires tight regulation of integrin-mediated adhesion. This assertion fits well our knowledge of the many roles integrins play in DC, which include filopodia-mediated epidermal zippering (Jacinto et al., 2000), amnioserosal cell contraction (Narasimha and Brown, 2004b), adhesion to laminin-rich ECM (Narasimha and Brown, 2004b), amnioserosa-yolk sac adhesion (Reed et al., 2004), and force sensing and response (Kiehart et al., 2000; Hutson et al., 2003). The differential abilities of βPS-integrin mutants to rescue GBR and DC may reflect the increased number of adhesive roles integrins play in DC compared with GBR: during GBR, integrins are only known to mediate cell-ECM adhesion (Schöck and Perrimon, 2003), while in DC, integrins also mediate both cell–cell adhesion and ECM assembly (reviewed in Schock and Perrimon, 2002; Narasimha and Brown, 2004a).
A controversial question in integrin biology concerns the role of conformational changes in integrin activation, as in vitro studies have shown that integrins can bind certain ECM ligands in both the extended and bent conformations (Askari et al., 2009). Comparing the rescue capacities of the S196F and D192A/S194A mutants allowed us to uncouple ECM binding from ligand-induced outside-in conformational changes to determine their relative contributions to morphogenesis: while the D192A/S194A mutant should not bind ECM, the S196F mutant should retain minimal ECM binding capacity but fail to relay outside-in signals due to conformational constraints (Bajt and Loftus, 1994; Hogg et al., 1999; Chen et al., 2006b). While GBR, DC, and MTJs were differentially sensitive to perturbation of some aspects of integrin function, we found that all failed in S196F rescue embryos. Overall, our data suggest that conformational changes in βPS-integrin due to outside-in signaling can be as important as the ability to bind ECM in the context of Drosophila development, and that the various activities of integrins must be differentially regulated in a tissue-specific manner to ensure proper morphogenesis.
Regulation of Integrin Activation Is Essential for Long-term, Stable Adhesion During Tissue Maintenance
The MTJs form late in embryonic development (Volk and VijayRaghavan, 1994) and must be maintained over approximately 5 days through stable integrin-mediated adhesion (Wright, 1960). We find that such stable adhesion requires ongoing outside-in and inside-out integrin signaling, as mutants with disrupted I-like domain function (D192A/S194A, S196F) or altered cytoplasmic tail function (804*stop, YY>FF, N840A) failed to rescue MTJ defects.
To examine why muscles detach in βPS-integrin mutant rescue embryos, we looked at recruitment of two essential IAC components, talin and PINCH, to MTJs. Talin is required for MTJ integrity and is among the first in a hierarchy of IAC proteins recruited by βPS-integrin during adhesive junction formation (reviewed in Brown et al., 2000). Our data show that talin recruitment is not generally affected by mutations that impinge on integrin-mediated ECM binding, outside-in and inside-out signaling, but that disruption of these functions interfered with recruitment of IAC components downstream of talin. While embryos lacking βPS-integrin failed to localize talin, all mutant integrins tested recruited talin to MTJs. This result suggests that interactions between talin and integrin may be largely independent of other aspects of integrin regulation (i.e., ECM binding) and is consistent with the suggested role of talin as a core, nucleating component of the IAC (Brown et al., 2002; Giannone et al., 2003; Tanentzapf and Brown, 2006; Tanentzapf et al., 2006).
In contrast, we found that localization of PINCH, an essential IAC component recruited downstream of talin (Clark et al., 2003), was dependent on integrin ECM binding activity. With the exception of the cytoplasmic truncation mutant 804*stop (not expected to recruit cytoplasmic components), all mutants recruited PINCH to the IAC except the putative ECM binding mutant D192A/S194A (Fig. 7). Intriguingly, the phospho-mutant, YY>FF, recruited significantly more PINCH to the IAC, suggesting that phosphorylation of these residues may regulate PINCH binding.
It is curious that the S196F mutation, designed to block outside-in integrin signaling, recruited PINCH. How might we reconcile the requirement for ECM binding but not outside-in signaling in PINCH localization? Mutations that interfere with ECM binding may reduce transmission of tensile force, which could be important for IAC assembly. We therefore propose that PINCH may be recruited in response to tension imposed on the nascent IAC by ECM binding. This idea is supported by previous work showing that a mutant βPS chimera that cannot transduce tension (comprising the cytoplasmic tail fused to an extracellular domain that cannot bind ECM ligands) can recruit talin but not PINCH or tensin (Tanentzapf et al., 2006). In addition, a recent Drosophila study illustrates that force may be required for the recruitment of IAC components, as the extracellular domain of αPS2-integrin is required for intracellular tensin recruitment in ovarian epithelial cells (Delon and Brown, 2009). Thus, force may play crucial roles in recruiting certain IAC components during tissue morphogenesis. Such a role for tension in vivo fits well with recently published in vitro studies highlighting the importance of force in the maturation of integrin adhesive contacts (reviewed in Puklin-Faucher and Sheetz, 2009).
Overall, our data show that talin localization is not dependent on ECM binding or outside-in activation, while PINCH recruitment requires that integrins be bound to ECM, and is also sensitive to phosphorylation of the cytoplasmic NPxY motifs. Therefore, different features of βPS-integrin function are required for localization of talin and downstream IAC components to maintain long-term integrin-mediated adhesion at MTJs.
Inside-out and Outside-in Integrin Signaling Can Coordinately Regulate Tissue Shape
We observed that mutations that disrupt regulation of integrin outside-in or inside-out signaling disrupted the characteristic three-dimensional shape of MTJs. Both strongly activating and deactivating mutations decreased overall length and area, while mild activating mutations rescued MTJs to WT morphology. Intriguingly, we found that, despite complete loss of the cytoplasmic βPS-integrin tail (804*stop), abrogation of ECM binding (D192A/S194A), or outside-in signaling activity (S196F), all mutants conferred partial rescue of MTJ length and area. As both the βPS-integrin I-like and cytoplasmic domains are required for complete rescue, these results suggest that functions of both domains contribute in distinct, nonredundant ways to the maintenance of MTJ morphology.
How Does Tyrosine Phosphorylation at the Conserved Cytoplasmic NPxY Motifs Influence βPS-Integrin Function?
The conserved membrane-distal and membrane-proximal NPxY (or NxxY) motifs in the β-integrin cytoplasmic tail are recognition sites for phosphotyrosine binding (PTB) domains in many signaling and cytoskeletal adaptor molecules (reviewed in Legate and Fassler, 2009). Phosphorylation of NPxY tyrosine residues can regulate which adaptors bind the β-integrin tail, thereby regulating integrin activity. Although the effects of NPxY tyrosine phosphorylation have been studied extensively, conflicting reports illustrate differential effects on integrin function (reviewed in Legate and Fassler, 2009). In this study, we examine the effects of rendering both NPxY motifs non-phosphorylatable using the YY>FF βPS-integrin rescue transgene. We found that this mutant exhibited lower turnover (by FRAP) and recruited more PINCH to MTJs, suggesting that these were activating mutations in this context. A previous study in flies characterized the ability of the same YY>FF mutant to rescue DC and muscle detachment defects and found less striking differences between mutant and WT transgenic rescues than we found (Li et al., 1998). A key difference between our study and the previous study is that we utilized a ubiquitous promoter to ensure robust expression while Li et al. used a genomic promoter that allowed expression of 6–20% of endogenous protein levels (Grinblat et al., 1994; Li et al., 1998). Such drastically reduced expression levels permitted only partial rescue even with the WT transgene, while we obtained nearly complete rescue of all phenotypes analyzed. This enabled us to more rigorously test the differential rescue capacities of mutant constructs compared with WT. Moreover, our analyses were more extensive, as we examined additional developmental and molecular processes. Given that the YY>FF transgene conferred poor rescue of DC and MTJ defects and appeared to affect PINCH recruitment to MTJs, we propose that phosphorylation of NPxY tyrosine residues is an important integrin regulatory mechanism in the fly. Similar results have been described in studies using the equivalent YY>FF mutation in mouse β3-integrin, which results in loss of integrin activation in platelets (Law et al., 1996; Blystone et al., 1997); however, the same mutation does not affect β1 function (Chen et al., 2006a; Czuchra et al., 2006). The NPxY motifs were also shown to be important during Ceanorhabditis elegans development (Xu et al., 2010). Drosophila βPS-integrin is close in sequence to both the vertebrate β1A and β3 integrin but may be closer in function to β3 integrin in its dependence on phosphorylation of NPxY tyrosine residues.
How Does Ectopic Activation of β-Integrin Affect Morphogenesis?
In Drosophila, a strong hyperactivating mutation in βPS-integrin was also shown to induce deleterious phenotypes in clones of cells in vivo (Jannuzi et al., 2004). We wanted to test the possibility that different modes of βPS-integrin hyperactivation might hinder progression of developmental processes. However, we found that the activating mutations we tested, which promote extracellular domain extension (L211I), enhance clustering (G792N) or induce activating conformational changes in the TM domain (L796R) behaved as WT during tissue morphogenesis (Fig. 3; Table 1). These mutations have been shown to induce relatively mild activation compared with the D807R salt-bridge mutant; however, because D807R localized poorly to the plasma membrane, we could not assess the relative effects of this strongly activating mutation in vivo. Overall, our data suggest that mild hyperactivation of integrins is not deleterious to morphogenesis.
Hyperactivation of integrins is a hallmark of many cancers, and it has been intuitively argued that cancer cells are no longer dependent on integrin signaling (Hanahan and Weinberg, 2000). Our study may provide etiological insight into human pathologies that arise as a result of defective adhesion to the ECM. With relevance to clinical treatment of such diseases, our results suggest that milder βPS-integrin activating mutations centering on the βPS-integrin transmembrane and extracellular domains (i.e., L211I, G792N, L796R) as opposed to stronger activating mutations in the cytoplasmic tail (i.e., YY>FF) may prove therapeutically valuable in treating diseases attributable to integrin dysfunction.
We have found that integrin adhesion and activation through inside-out and outside-in signaling are required to both form and maintain tissues and that these mechanisms play distinct roles during different morphogenetic processes in Drosophila. Although we tested a subset of the regulatory mechanisms that potentially control integrin activity, we have been able to assign specific functions to these mechanisms in the context of development and show that different tissues require different regulatory inputs. Using this approach, we provide in vivo developmental context for structural and biochemical insights in integrin biology that span more than 20 years.
Drosophila Stocks and Transgenic Lines
Germline clone females lacking βPS-integrin were generated using the FRT/FLP OvoD method (Chou and Perrimon, 1996) and the following stocks: y,w,OvoD1, FRT101/y,w,C(1)DX;; hsFLP38 (Bloomington), mysXG43, FRT101/FM7,KR::GFP (Nick Brown, Gurdon Institute, Cambridge). Female flies of the genotype mysXG43, FRT101/OvoD1, FRT101;;hsFLP/+ were heat-shocked, and their female progeny crossed to males carrying transgenic rescue constructs.
pUbi::βPS-integrin-YFP was created from the pHSβPS-Venus plasmid (gift of Thomas Bunch, University of Arizona, Tucson, AZ), which contains a cDNA clone of βPS with an internal Venus YFP fusion protein inside a nonconserved, serine-rich N-terminal region of the Hybrid domain, replacing residues 113–134 with a nonnative serine and threonine residue, followed by the Venus YFP-coding sequence. A 4783 base pair section incorporating the hsp70 promoter, the βPS-YFP fusion protein and the tubulin polyA site was excised from the plasmid using the restriction enzyme XbaI and inserted into the SpeI site downstream of the ubiquitin-63E promoter in the pWRpAUbiP (pUBI) plasmid (Tanentzapf and Brown, 2006). The hsp70 promoter was then excised using the SacII and SgrAI restriction enzymes. All other β-PS-integrin mutant rescue constructs were derived from this construct. Transgenic lines were generated by BestGene (Chino Hills, CA).
Western Blot Analysis
Protein was isolated from 10 adult flies (5 male + 5 female) of each genotype, aged 3–5 days, and blotted according to standard procedures using primary antibodies to GFP (A6455, 1:4,000, Invitrogen) and actin (8224, 1:5,000, Abcam) and secondary antibodies conjugated to InfraRed (IR) dyes, IR680 and IR800 (LiCor Odyssey, 1:15,000). Blots were scanned on a LiCor Odyssey Imaging System.
Rescue Efficiency Calculation
Given that in null mutant embryos, GBR and DC defects were not completely penetrant and the WT rescue construct typically gave 90–100% rescue, all rescue efficiencies were calculated by normalizing to the rescue obtained with the WT transgene. Therefore, the calculation used in these cases was:
For example, 47% of null integrin mutants exhibited GBR defects and 53% had a normal phenotype. In embryos rescued with the WT transgene, 4.7% of embryos exhibited GBR defects, while 95.3% had a normal phenotype. In comparison, GBR penetrance in embryos expressing the D192A/S194A transgene was 20.5%, while 79.5% of embryos had a normal phenotype. We thus calculated normalized rescue as follows:
Immunofluorescence and Microscopy
Embryos were collected over 24 hr at 25°C on apple juice agar plates and fixed with heat treatment (Tanentzapf and Brown, 2006) or paraformaldehyde (4%) for 20 min at room temperature. Immunostaining was carried out using standard procedures using the following antibodies: anti-βPS (1:20; Nick Brown, Gurdon Institute, Cambridge, UK), anti-αPS2 (7A10, 1:10), anti-PINCH (1:1,000; Mary Beckerle, Huntsman Cancer Institute, Utah), anti-Cher (C-fil, 1:1,000; Lynn Cooley, Yale University, New Haven, CT), anti-Tiggrin (1:500; Liselotte Fassler, University of California, Los Angeles), and anti-GFP (A6455, 1:1,000, Invitrogen). Fluorescently conjugated Alexa 488, Cy3 and Cy5 secondary antibodies were used at 1:400 dilution (Molecular Probes, Eugene, OR). Rhodomine-conjugated phalloidin was used to stain filamentous actin (1:500; Invitrogen). Images were collected using an Olympus FV1000 confocal microscope with a UplanSApo 60/1.35 oil objective and a UplanFL N 40×/1.30 oil objective and processed using Adobe Photoshop.
Fluorescence Recovery After Photobleaching (FRAP)
Embryos were collected from apple juice plates, dechorionated in 50% bleach for 3 min, and washed and mounted in phosphate buffered saline (PBS) on glass slides. FRAP analysis was carried out at room temperature on whole-mount stage 17 embryos 2 hr after mounting. Embryos were heterozygous for each transgene. Photobleaching was performed using the 405 nm laser at 30% power with the Tornado scanning tool (Olympus) for 2 s at 100 msec/pixel. Fluorescence recovery was recorded over 5 min at 1 frame every 4 s. To control for movement in and out of focus, multiple regions of interest (ROI) were selected in nonphotobleached regions; only samples for which intensities within control ROIs remained steady throughout the FRAP experiment were used. Recovery data were further analyzed using Prism software (GraphPad, La Jolla, CA): mobile fraction was calculated as previously described (Reits and Neefjes, 2001). Statistical tests (t-test, analysis of variance [ANOVA] test) were carried out using Prism.
Z-stack projections (∼12–16 μm thick) through the ventral longitudinal muscles of at least four representative embryos of each genotype immunolabeled for Tiggrin were selected for morphometric analyses. Length and area measurements, as well as localization analyses (Figs. 2, 5–7) were performed using ImageJ (National Institute of Health, Bethesda, MD). Statistical tests (t-test, ANOVA) were done using Prism software.
We thank Dan Kiehart, Mary Beckerle, Lynn Cooley, and Lisa Fessler for antibodies. We thank members of the laboratory for helpful discussions and comments on the manuscript. M.J.F. holds an NSERC Alexander Graham Bell Canada Graduate Scholarship. G.T. is a CIHR New Investigator and Michael Smith Foundation for Health Research Scholar. Research in the laboratory of G.T. is supported by an HFSP Career Development Award.