SEARCH

SEARCH BY CITATION

Keywords:

  • Fas2;
  • Dlg;
  • basolateral junction;
  • cell migration;
  • tumor invasion;
  • Drosophila

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Epithelial junctions play crucial roles during metazoan evolution and development by facilitating tissue formation, maintenance, and function. Little is known about the role of distinct types of junctions in controlling epithelial transformations leading to invasion of neighboring tissues. Discovering the key junction complexes that control these processes and how they function may also provide mechanistic insight into carcinoma cell invasion. Here, using the Drosophila ovary as a model, we show that four proteins of the basolateral junction (BLJ), Fasciclin-2, Neuroglian, Discs-large, and Lethal-giant-larvae, but not proteins of other epithelial junctions, directly suppress epithelial tumorigenesis and invasion. Remarkably, the expression pattern of Fasciclin-2 predicts which cells will invade. We compared the apicobasal polarity of BLJ tumor cells to border cells (BCs), an epithelium-derived cluster that normally migrates during mid-oogenesis. Both tumor cells and BCs differentiate a lateralized membrane pattern that is necessary but not sufficient for invasion. Independent of lateralization, derepression of motility pathways is also necessary, as indicated by a strong linear correlation between faster BC migration and an increased incidence of tumor invasion. However, without membrane lateralization, derepression of motility pathways is also not sufficient for invasion. Our results demonstrate that spatiotemporal patterns of basolateral junction activity directly suppress epithelial invasion by organizing the cooperative activity of distinct polarity and motility pathways. Developmental Dynamics 236:364–373, 2007. © 2006 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Epithelia are comprised of apicobasally polarized cells that contact each other and surrounding tissues through specialized junctional complexes. The junctional complexes consist of transmembrane receptors bound to cytoplasmic scaffolding proteins that link the complex to the cortical cytoskeleton (Knust,2002). Epithelial cells can transform into migratory cells during embryonic morphogenesis, tissue repair, and regeneration, as well as numerous pathological conditions including cancer (Shook and Keller,2003; Larue and Bellacosa,2005; Thiery and Sleeman,2006). These transitions correlate with extensive remodeling of cellular junctions (Grunert et al.,2003). However, it is not clear to what extent this remodeling contributes to the acquisition of motility. Moreover, it is not clear whether any one junctional complex alone is sufficient to inhibit epithelial invasion.

Drosophila egg chambers provide an in vivo model system for addressing these questions (Goode et al.,2005). They are comprised of a monolayer epithelium of follicle cells that surrounds, yet remains excluded from, the germ cells during early oogenesis (Fig. 1A). Each epithelial follicle cell has five junctions that contain proteins conserved in vertebrate epithelia. Three conserved lateral junctions, the apicolateral, adherens, and basolateral junction (BLJ) (Knust,2002), interconnect follicle cells to each other, while a basal junction connects to the outer basement membrane, and an apical junction connects to germ cells (Fig. 1B). During mid-oogenesis, a small group of epithelial follicle cells, the border cells (BCs), delaminate from the anterior epithelium and migrate between germ nurse cells to the oocyte (Montell,2003).

thumbnail image

Figure 1. Basolateral junctions (BLJs) inhibit tumor invasion. A: Follicle cell development (s, stage). B: Epithelial junctions, and a model of a BLJ supermolecular complex that suppresses invasion. The indicated binding interactions are hypothesized based on this work and their demonstration in other tissues as follows: Fas2-Dlg (Goodman et al.,1997); Dlg-4.1 (Lue et al.,1996); Nrg-Ank (Hortsch et al.,1998); 4.1-Spec/Act and Ank-Spec/Act (Bennett and Baines,2001). Scrib, which is also localized to BLJ, is not shown since we found that its loss does not cause invasion as defined in this work. C,D: Subcellular localization of Fas2 and Baz, and Dlg and Arm; arrowheads point at the adherens junction. E,F: Loss of Fas2 in follicle cells caused invasive tumors (circled). Tumors typically stay attached to the epithelium but sometimes migrate (E; stage 5). G: Three cell Nrg tumor (arrow) invaded between germ cells (arrowhead). H: baz cells change polarity and accumulated in multiple layers (arrows), but did not invade between germ cells (arrowheads).

Download figure to PowerPoint

Discs-large (Dlg), a BLJ scaffolding protein that includes PDZ and other protein-interaction domains, directly suppresses epithelial invasion. Decrease or loss of Dlg results in both faster BC migration (Szafranski and Goode,2004) and development of invasive tumors during early oogenesis (Goode and Perrimon,1997; Goode et al.,2005). We have recently found that two transmembrane cell adhesion molecules of the immunoglobulin superfamily, Fasciclin-2 (Fas2) and Neuroglian (Nrg), are also specifically localized to the BLJ (Szafranski and Goode,2004; Wei et al.,2004). Fas2 directly binds to the first and second Dlg PDZ domains (Goodman et al.,1997). Thus, Fas2, a transmembrane protein, and Dlg, its scaffolding partner, define the minimal components of the BLJ. Other proteins localized to BLJ include Scribble (Scrib) and Lethal-giant-larvae (Lgl) (Bilder et al.,2000; Bilder and Perrimon,2000). Identification of the minimal components of the BLJ (Fig. 1B) allows us now to investigate if suppressing epithelial invasion is a function unique to Dlg, or if other proteins of the BLJ also have this activity. If other proteins in the BLJ have this activity, than we can infer that the BLJ function as a supermolecular complex crucial for suppressing invasion.

In this work, we present systematic analysis of the effect of genetic disruption of each of five epithelial junctions on epithelial invasion. We find that only proteins of the BLJ are sufficient to suppress invasion. Part of the cellular mechanism of this suppression appears to be a specific change in plasma membrane patterning, which we term membrane lateralization, that follows loss of any one BLJ protein. Membrane lateralization is characterized by loss of apical and basal membrane domains, a circumferential distribution of basolateral membrane proteins, and a diffuse distribution of apicolateral and adherens junction proteins. A similar pattern is observed in normal migrating cells during oogenesis, suggesting that lateralization may be universally acquired in some types of epithelial cells in order for them to achieve movement. However, we find that although lateralization is essential, it is not sufficient for acquisition of motility by BLJ mutant cells. In addition, derepression of BLJ motility pathways, which occurs independently of membrane lateralization, is also necessary, although not sufficient for invasion. Thus, we propose that a BLJ supermolecular complex directly suppresses epithelial invasion by organizing the cooperative activity of distinct polarity and motility pathways.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

To determine which junctions suppress tumor invasion, we took advantage of the previous observation that loss of Dlg causes follicle cells to overproliferate, change polarity, delaminate from the native epithelium, and invade between germ cells, all characteristics of invasive tumor cells in humans and other vertebrates (Goode and Perrimon,1997; Goode et al.,2005). The invariant movement of dlg tumors toward the oocyte resembles invasion by BCs (Figs. 1A, 2A,D), but dlg tumor cells start invading early in oogenesis and do not adopt a BC fate (Goode and Perrimon,1997). We used both tumor invasion and BC migration as assays to determine which epithelial junctions are crucial for suppressing invasion in vivo.

thumbnail image

Figure 2. Fas2 inhibits border cell invasion. A,B: Fas2 expression pattern. Fas2 is lost from anterior follicle cells at s8 (brackets). C: Targeted expression of UAS-GFP with c208-Gal4 drove GFP expression in the cells that normally lose Fas2 (compare to B). D: Wild-type s10 egg chamber. BCs (arrow) and the trailing edge of the follicle epithelium (TE, arrowheads) completed migration simultaneously. E: UAS-Fas2/+; c208-Gal4/+ egg chamber. BCs did not invade, even when the TE had completed migration. Similar results were obtained with anterior drivers c606 and 198Y but not posterior driver c324a. F: Fas2 targeted to the late stage 8 anterior epithelium with c208-Gal4 prevented a switch of Dlg localization from basolateral (white brackets) to the cellular circumference. Dlg was absent from the apical surface (blue brackets) and basal membranes (the basal membrane was partially obscured by the muscle sheath that expresses Dlg). G: In contrast, targeting the anterior epithelium with Fas2Δ3, which does not bind to Dlg PDZ domains, and does not block BC migration, did not interfere with the switch of Dlg from a basolateral to a circumferential localization. As in wild-type late-stage 8 BCs, Dlg was present not only in the basolateral membrane (white brackets), but also in apical and basal membranes (blue brackets).

Download figure to PowerPoint

BLJ But Not Other Junction Proteins Are Sufficient to Suppress Epithelial Invasion

We visualized loss-of-function phenotypes for 13 proteins belonging to each of five epithelial junctions (Fig. 1B). There are three lateral junctions that interconnect follicle cells; from apical to basal, they are the apicolateral junction (vertebrate tight junction) composed of Bazooka/Par-3 (Baz), Par-6, and Crumbs (Crb); the adherens junction containing DE-Cadherin (DE-Cad) and Armadillo/β-Catenin (Arm); and the BLJ containing Fas2, Nrg, Dlg, Lgl, and Scrib (reviewed in Knust and Bossinger,2002). Fas2, Nrg, Dlg, Lgl, and Scrib colocalize in the BLJ, and do not colocalize with adherens junction Arm or apicolateral Baz (Fig. 1C,D) (Bilder et al.,2000; Tanentzapf et al.,2000). In addition to lateral junctions, a basal junction containing Laminin (Lan) (Gutzeit et al.,1991), and Integrin (Int) (Bateman et al.,2001) connects to the basement membrane. An apical junction, containing βH-Spectrin (βH-Spec) (Lee et al.,1997), and enriched in Actin and Notch (Goode et al.,1996), connects to germ cells.

We define invasion as penetration of epithelium-derived cells between germ cells, not merely their delamination from epithelium or epithelial stratification (compare Fig. 1G,H). This strict criterion has not been used in previous investigations to characterize mutations of epithelial junction proteins. By this criterion, mutations in crb, baz, par6, shg (DE-Cad), arm, scrib, N (Notch), and mys (β-Int) cause changes in apicobasal polarity (data not shown) (Peifer et al.,1993; Goode et al.,1996; Tanentzapf et al.,2000; Abdelilah-Seyfried et al.,2003), but do not induce invasion (Fig. 1H, and data not shown). Only mutations in Fas2, Nrg, dlg, and lgl additionally caused tumor invasion (Figs. 1E–G, 3B; see also Fig. 6A). The invasive process appears to be quasi-stochastic since even when all follicle cells were mutant for Fas2, Nrg, dlg, or lgl, only one or two invasions occurred per egg chamber. The quasi-stochastic character of invasion is reminiscent of the movement of tumor cells from mammalian and human in situ carcinomas (Hynes,2003).

thumbnail image

Figure 3. Evidence that Fas2, Dlg, and Lgl function together in a pathway that suppresses epithelial invasion. A: A graph of the frequency of invasion along the anterior-posterior axis of the egg chamber. The pattern of dlg invasion resembles that of Fas2 invasion. B: lglnull egg chamber. As for Fas2 and dlg, lgl invasion occurred predominantly at the poles. C: lgl mutations enhance dlg and Fas2 invasion. Removal of one copy of lgl but not shg (the gene for adherens junction DE-Cad) or Patj (the gene for apicolateral Patj) increased both the frequency and size of Fas2 and dlg invasive tumors (C,D,E). Moreover, we could not recover Fas2, +/+, dlg doubly heterozygous flies, further indicating strong interaction between Fas2 and Dlg. Additional evidence for functional interactions between proteins of this complex can be found in Goodman et al. (1997), Szafranski and Goode (2004), Wei et al. (2004), and Goode et al. (2005).

Download figure to PowerPoint

thumbnail image

Figure 6. Tumor invasion depends on activation of BC motility pathways. A,B: Invasive and noninvasive dlg tumors (circled). C: Graph showing the correlation between BC migration rate and frequency of tumor invasion (dlg and Fas2 mutants are marked in black and red, respectively). The following genotypes were plotted: (1) dlgsw, (2) dlglv55/dlgsw, (3) dlglv55/dlgm35, (4) dlgip20/dlgsw, (5) dlgip20/dlgm35, (6) dlgv59/dlgm35, (7) dlgv59/dlghf for 15 hr at 21°C, (8) Fas2null, (9) dlgip20/dlghf, (10) dlgv59/dlghf, (11) dlglv55/dlghf, (12) Fas2rd1, and (13) Fas2MB2225. Fas2rd1 and Fas2MB2225 (12, 13) had an accelerated rate of BC migration, but unlike Fas2null (8), did not have invasive tumors. D: Rac1 was required for Fas2 and dlg tumor invasion. UAS-RacN17 targeted to all Fas2 or dlg cells using GR1-Gal4 (see Supplementary Figure S2, which can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat, for the GR1-Gal4 expression pattern) inhibits tumor invasiveness (P < 0.005 for inhibition of both Fas2 and dlg invasions; test χ2). Genotypes: Fas2null, UAS-RacN17/+, GR1-Gal4/+ or dlghf/lv55; UAS-RacN17/+, and GR1-Gal4/+.

Download figure to PowerPoint

Several observations indicate that our operational definition of invasion defines active movement, not mere pushing of supernumerary cells in-between germ cells. First, sometimes Fas2 and dlg tumor clusters completely delaminate and migrate to the oocyte (Fig. 1E) (Goode et al.,2005). Second, some dlg alleles only cause overproliferation and multilayering of the epithelium, while others also cause invasion (Goode and Perrimon,1997). Third, although Nrg loss typically results in small, 2- to 3-cell clusters, they nonetheless invade between germ cells (Fig. 1G), unlike even much larger clusters of apicolateral and adherens junction mutant cells (Fig. 1H, and data not shown).

Spatio-Temporal Expression of Fas2 Predicts Epithelial Invasions

To further investigate the specificity of BLJs in suppressing invasion, we determined which junction proteins change expression in the anterior epithelium just before BC invasion (Fig. 1A). Only Fas2 changed dramatically. Fas2 was expressed at high levels in the early follicular epithelium (Fig. 2A). Just before BC delamination, Fas2 is specifically lost from BCs and immediately adjacent cells (Fig. 2A,B). Fas2 is a cell adhesion molecule of the immunoglobulin superfamily (IgSF) and an ortholog of human NCAM, which is also localized to BLJs (Marmorstein et al.,1998).

To determine whether Fas2 loss is essential for invasion, we used an anterior follicle cell driver to target Fas2 to cells from which it is normally lost preceding BC invasion (Fig. 2B,C). Ectopic Fas2 blocked invasion of approximately 90% of BC clusters (n > 1,000 stage-9 egg chambers) (Fig. 2E). Two other anterior drivers gave similar results, whereas a posterior driver failed to block migration. Further, targeting the anterior epithelium with other BLJ IgSF members, Nrg or Kekkon-1, with BLJ scaffolding proteins Dlg or Lgl, with adherens junction DE-Cad, Arm, or with apicolateral scaffolding protein Baz, had little effect on BC invasion, underscoring Fas2 specificity (data not shown). The functional characteristic that only Fas2, but not Nrg, Dlg, or Lgl, block BC migration stems from its unique contribution to controlling polarity as described below. The Fas2 tumor invasion and blocked BC migration phenotypes indicate that Fas2 is both necessary and sufficient to inhibit invasion, and that its pattern of expression predicts epithelial invasion.

Fas2, Dlg, Lgl, and Nrg Function Together to Suppress Invasion

Molecular and functional data indicate that Fas2 functions in a BLJ supermolecular complex with Nrg, Dlg, and Lgl (Fig. 1B) to suppress invasion. First, all four molecules precisely colocalize to the BLJ (Szafranski and Goode,2004; Wei et al.,2004). Second, Fas2 directly binds Dlg (Goodman et al.,1997). Third, molecular interactions between Nrg, Ankyrin, and Spectrin (Dubreuil et al.,1996; Hortsch et al.,1998) suggest a manner for Nrg to be linked to Fas2-Dlg through membrane protein 4.1 (Fig. 1B). Furthermore, although direct binding between Lgl and Dlg has not been reported, Lgl clearly appears to be part of the scaffold, because it precisely colocalizes with Dlg and Fas2 in wild-type follicle cells, and when Fas2 is ectopically expressed (data not shown) (Szafranski and Goode,2004). Forth, Fas2, Nrg, dlg, and lgl have similar BC and tumor phenotypes (Figs. 1E–G, 3A,B,D,E; see also Fig. 6A,B) (Szafranski and Goode,2004; Wei et al.,2004; Goode et al.,2005). Fifth, we observe strong enhancing genetic interactions between Fas2 and dlg, Fas2 and lgl, dlg and lgl, and dlg and Nrg (Fig. 3C–E) (Szafranski and Goode,2004; Wei et al.,2004). These interactions indicate a specific functional relationship because we do not observe similar genetic interactions between basolateral junction mutants and null alleles of other junction genes, such as Patj, which encodes a PDZ-scaffolding protein in the apicolateral junction, or adherens junction shg, the gene for DE-Cad (Fig. 3C–E) (Szafranski and Goode,2004; Wei et al.,2004).

Tumor Cells and BCs Undergo Similar Membrane Lateralization

To obtain mechanistic insight into how BLJ mutant cells invade, we compared the polarity of BLJ tumor cells, and motile, wild-type BCs as they began to invade (Fig. 4). We found that BLJ tumors and BCs showed similar changes in membrane patterning that we term membrane lateralization. From basal to apical, the specific defects we observe are: loss of basal LanA, β-Int, and α-Tubulin (α-Tub) (Fig. 4C); circumferential localization of basolateral Dlg, Lgl, and α-Spec (Fig. 4D); diffuse, circumferential distribution of adherens junction DE-Cad and Arm, and apicolateral Baz, Par-6, and Crb (Fig. 4E,F); and loss of apical βH-Spec (Fig. 4G). However, BCs differed from BLJ tumors in that the zonula adherens and apicolateral proteins were still localized in junctions that interconnect BCs, which likely accounts for the organized BC rosette and its predictable delamination compared with disorganized and quasi-stochastic patterns of BLJ mutant invasion (Fig. 3A). The lateralized pattern is consistent with polarity changes previously described for mutations in BLJ scaffolding proteins (Bilder et al.,2000; Bilder and Perrimon,2000), but our study also used mutants in transmembrane receptor proteins and addressed perturbance of apical and basal domains. Our data thus suggest that the BLJ does not act merely as a fence (Bilder and Perrimon,2000), preventing basal movement of apicolateral proteins, or as a targeting site for vesicles alone (Bilder et al.,2000; Bilder and Perrimon,2000), but rather suppresses transformation to a lateralized pattern.

thumbnail image

Figure 4. Epithelial junction protein patterns. A,B: Schematic diagrams of a BC cluster and invasive tumors. CG: Localization of epithelial junction proteins in the stationary follicular epithelium, BCs, and invasive Fas2 tumor cells. Note that apicolateral proteins were localized in discrete puncta between each follicle cell (arrowheads), marking the location of the intercellular junctions. In contrast, as indicated by the absence of these puncta, apical βH-Spec was only distributed along the apical membrane. This was further seen in apical cross-sections, where the apicolateral proteins localized in a circumferential pattern, but βH-Spec was seen throughout the membrane (data not shown). See the text for a further explanation of the localization patterns.

Download figure to PowerPoint

Membrane Lateralization Is Necessary But Not Sufficient for Invasion

The similarity of the lateralized membrane pattern in BCs and BLJ tumor cells (Fig. 4) led us to ask if it is sufficient for invasion. We compared localization of epithelial markers for invasive versus noninvasive tumors. Noninvasive dlg cells have a lateralized pattern (Fig. 5) that is indistinguishable from that of invasive cells (Fig. 4). Thus, a lateralized membrane pattern is not sufficient for invasion. However, several observations suggest that lateralized membrane is essential for cell movement. First, BCs are determined before they acquire a lateralized pattern, but do not move until they have completed this transformation (Szafranski and Goode,2004). To test the importance of this change for BC movement, we asked if it occurs in BCs when ectopic Fas2 blocks their motility (Fig. 2E). In these BCs, we found that Dlg fails to switch to circumferential distribution (Fig. 2F), suggesting that loss of Fas2 permits lateralization and movement. Another possibility is that ectopic Fas2 is blocking movement through Ig-mediated adhesion. To test this, we completed the same experiment with Fas2Δ3, which is missing three carboxyl-terminal amino acids that bind Dlg (Goodman et al.,1997). Fas2Δ3 still localizes to the membrane, where it can mediate adhesion, but fails to block BC migration, and also fails to block lateralization (Fig. 2G). Likewise, ectopic expression of Nrg, Dlg, or Lgl does not block membrane lateralization, and does not block BC migration (data not shown), further indicating the necessity of lateralization for movement.

thumbnail image

Figure 5. Cell membrane patterning of noninvasive dlgip20/dlgm35 tumor cells. Localization of junctional proteins in noninvasive dlg tumor cells resembles their localization in invasive cells (Fig. 4; and data not shown).

Download figure to PowerPoint

The importance of membrane lateralization is also indicated by detailed analysis of tumor invasion in Fas2, dlg, lgl, or Nrg egg chambers. Most mutant epithelial cells do not invade. We typically observe that only one or two tumor invasions occur per egg chamber. As for BC migration, we found that cells only invade once they have adopted a lateralized pattern. To test the importance of this observation, we analyzed the polarity of follicle cells for two Fas2 mutants, Fas2rd1 and Fas2MB2225, in which the Fas2 enhancer region, but not Fas2 protein, is mutated (Cheng et al.,2001). Fas2 is expressed at <10% of wild-type levels in Fas2rd1 and Fas2MB2225 follicle cells (Cheng et al.,2001; Szafranski and Goode,2004). Unlike Fas2null cells, Fas2rd1 and Fas2MB2225 cells never become lateralized, and never invade (n>2000 egg chambers), suggesting that lateralization is crucial for invasion. Another possibility was that there are no cells to invade because Fas2rd1 and Fas2MB2225 follicle cells do not overproliferate. However, this possibility was ruled out because we found that Fas2rd1 and Fas2MB2225 cells overproliferate as much as Fas2null follicle cells (see Supplemental Figure S1, which can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat). Lastly, we considered that Fas2rd1 and Fas2MB2225 follicle cells might not invade because unlike Fas2null, they do not derepress motility. However, both mutations cause faster BC migration similar to Fas2null (Szafranski and Goode,2004). Thus, the simplest explanation for why Fas2rd1 and Fas2MB2225 follicle cells do not invade is that they do not become lateralized. We conclude that membrane lateralization is necessary, but not sufficient for invasion.

Derepression of Motility Pathways Is Necessary But Not Sufficient for Invasion

If lateralized membrane is necessary, but not sufficient for invasion, what additional functions do BLJs regulate to suppress movement? The observation that Fas2, dlg, and lgl BCs migrate faster than wild-type cells (Szafranski and Goode,2004), led us to question if BLJs also suppress motility pathways in tumor cells. If tumor invasion depends on derepression of motility pathways that control BC migration, we would expect a correlation between faster BC migration and an increased tumor invasion frequency in BLJ mutants. We compared 13 Fas2 and dlg genotypes, which vary in invasion frequency. We found a significant linear correlation between faster BC migration and a higher tumor invasion incidence (Fig. 6C; Spearman correlation = 0.89; P = 0.0002; see Experimental Procedures section), suggesting that BLJs actively suppress motility pathways to prevent tumor invasion. Indeed, loss of function of proteins from the other four epithelial junctions either had no impact or delayed or blocked BC migration (data not shown) (Peifer et al.,1993; Niewiadomska et al.,1999; Zarnescu and Thomas,1999; Abdelilah-Seyfried et al.,2003; Pinheiro and Montell,2004), further indicating that repression of motility is a unique function of the BLJ. To obtain further insight into how BLJs control invasion, we expressed a dominant-negative allele of Rac1, a G-protein essential for BC migration (Geisbrecht and Montell,2004), in Fas2 and dlg epithelium. Dominant-negative RacN17 inhibited invasion of tumor cells (Fig. 6D) but not their overproliferation or changes in polarity, suggesting that Rac acts specifically downstream of a BLJ in a pathway that suppresses movement.

Is derepression of motility pathways sufficient for invasion? This seems unlikely because, as noted above, two Fas2 alleles, Fas2rd1 and Fas2MB2225, cause faster BC migration similar to Fas2null, but do not cause tumor invasion (Fig. 6C). In contrast to Fas2null, epithelial polarity is maintained in these mutants. Thus, BLJs control at least two pathways that cooperate to suppress invasion, one that suppresses transformation to lateralized membrane, and another that directly suppresses motility.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

BLJs Are Sufficient to Suppress Epithelial Invasion

We found that four BLJ proteins, Fas2, Nrg, Dlg, and Lgl, but not proteins of other epithelial junctions, are sufficient to suppress epithelial invasion in vivo. We propose that these four proteins form a supermolecular complex at the BLJ that functions to suppress epithelial invasion (Fig. 1B). The BLJ inhibits invasion by both preventing adoption of a lateralized membrane pattern and repressing motility pathways that depend on Rac signaling.

We suggest that the similar lateralized membrane pattern of BLJ mutant cells and differentiated BCs indicates that this transformation is not adequately described as loss of polarity, but rather reflects adoption of a new pattern rich in information, essential for motility. Each membrane domain in epithelial cells acts as a targeting site for specific exocytotic and endocytotic pathways (Mellman,2000; Nelson,2003). Thus, the exclusion of apical and basal domains, and the predominance of the lateral membrane, may create a unique targeting niche essential for motility. BLJ-specific docking motifs in vesicular proteins such as SNARE (Pinheiro and Montell,2004) may be important for rapid membrane cycling in moving cells (Mellman,2000). Acquisition of the lateralized niche appears to be precisely timed in BCs, but is quasi-stochastic in BLJ mutant tumors (Fig. 3A).

Interestingly, the lateral membrane of mammalian keratinocytes also plays a role in migration events that lead to de novo epithelial sheet formation (Vasioukhin et al.,2000). The presumptive lateral surface of these cells extends filopodia, which contact adjacent cells, and facilitate adherens junction formation, an essential step preceding membrane zippering. We speculate that transformation of stationary follicle cells to invasive cells may involve a similar, modified epithelial zippering process that includes the following steps. First, BLJ mutant follicle cells decrease adhesion to each other, because as we have shown, they fail to form adherens junctions. Likewise in mammals, Dlg-1 has been shown to be required for adherens junction formation (Firestein and Rongo,2001). Second, BLJ transformed cells develop novel adhesion interactions with germ cells, because they develop basolateral membrane proteins around their circumference, and because not only proteins of the adherens junction, but also those of BLJ (with the exception of Fas2) are present on germ cell membranes (data not shown). Third, the BLJ mutant cells undergo a repetitive, recursive epithelial zippering-like process, similar to that described in keratinocytes, as they invade the germ cells, but their inability to form adherens junctions drives invasion rather then epithelial zippering. A similar mechanism may also drive BC migration.

Our data do not address whether transformation to a lateralized pattern precedes activation of motility pathways or vice versa, or if these changes occur simultaneously on the path to invasion. Independent of the sequence of these cellular transformations, the data indicate that BC invasion and BLJ tumor invasion share significant molecular and functional homology, suggesting that in addition to repressing early epithelial invasion, the BLJ also plays a dominant organizing role in controlling BC and most likely other epithelial movements (Fig. 7). We propose that in both BCs and in the early epithelium, the BLJ directly suppresses invasion by organizing the cooperative activity of distinct polarity and motility pathways (Fig. 7).

thumbnail image

Figure 7. Homology between BC and tumor invasion. Both BCs and BLJ invasive tumor cells switch to a lateralized membrane pattern. BC migration is not only under control of BC transcription factors Slbo and Jing but also depends on Fas2 loss. Early loss of BLJ proteins is sufficient to trigger invasion in the absence of Slbo or Jing, in a quasi-stochastic pattern (Fig. 3A). Motility via either pathway depends on Rac signals. Polarity and motility pathways cooperate to suppress invasion (see text).

Download figure to PowerPoint

BLJ proteins also suppress epithelial invasion in other contexts in flies. In eye disks that express activated RasV12, loss of dlg, lgl, or scrib causes tumor metastasis (Pagliarini and Xu,2003). Surprisingly, we found that loss of scrib in the ovarian epithelium does not cause epithelial invasion. There are several possible explanations for why Scrib does not suppress invasion in the ovary. Scrib may only suppress invasion in the context of activated Ras, or its function in suppressing invasion may depend on tissue context, the eye versus the ovary. Another possibility is that some molecules of the BLJ such as Scrib may function primarily to control polarity and suppress proliferation (Bilder,2004), while others like Dlg, Lgl, Fas2, and Nrg, additionally suppress invasion.

In contrast to BLJ proteins, loss of adherens junction DE-Cad in the RasV12 eye disk (Pagliarini and Xu,2003), or in the wild-type follicular epithelium (Niewiadomska et al.,1999), does not trigger cell invasion. This appears to be at odds with studies in mammals, where E-Cad expression is down regulated at epithelial-mesenchyme transitions, and in many human metastatic cancers. However, E-Cad loss has only been shown to trigger invasion in tumor cells, suggesting that motility may depend on the context of oncogenic signals. Thus, our work suggests that at least in the Drosophila follicular epithelium, only proteins of the BLJ, and not other junction proteins, are sufficient to suppress invasion in vivo, in the absence of large-scale genomic imbalances, or additional confounding oncogene or tumor suppressor mutations.

Relevance to Human Carcinoma Invasion

Our findings in Drosophila suggest that the BLJ may act as a conserved supermolecular complex that is crucial for suppressing invasion across phylogeny. An important finding in our model is that the spatiotemporal pattern of Fas2 predicts which follicle cells will invade (Fig. 2A,B), suggesting that tightly controlled spatiotemporal patterns of BLJ activity are crucial for suppressing motility. Thus, it is probably significant that mammalian Fas2 and Nrg orthologs, NCAM and L1, are also localized to the BLJ (Marmorstein et al.,1998; Knust,2002), along with several other IgSF members implicated in invasive cancer (Fogar et al.,1997; Roesler et al.,1997; Bazzoni,2003). Direct evidence supports NCAM's importance for suppressing invasion. NCAM is down regulated in normal migrating cells, and decreasing NCAM levels are correlated with increased tumor malignancy in human pancreatic and colorectal cancer (Fogar et al.,1997; Roesler et al.,1997; Perl et al.,1999). NCAM is also progressively lost in invasive pancreatic tumors in mice, and, like Fas2, its expression is sufficient to block invasion (Perl et al.,1999). In general, BLJ proteins are conserved in vertebrate epithelia (Knust,2002), and several BLJ proteins are lost in invasive carcinomas (Fogar et al.,1997; Roesler et al.,1997; Perl et al.,1999; Bazzoni,2003; Huang et al.,2003; Schimanski et al.,2005). hDlg-1, which directly interacts with the most commonly mutated protein in colorectal cancer, APC, is lost in murine invasive ovarian cancer (Huang et al.,2003), and loss of human ortholog of lgl contributes to colorectal cancer invasion (Schimanski et al.,2005). Greater than 90% of cancers are of epithelial origin, but usually do not pose a serious threat to life until tumor cells start to invade. The BLJ may thus be an important target for anti-cancer therapeutics because it can be manipulated from the cell surface, and is sufficient to suppress epithelial invasion. As a precedent for this idea, antibodies to L1, a human ortholog of Nrg, suppress ovarian epithelial cancer dissemination (Arlt et al.,2006). Future work will determine the degree to which the BLJ is sufficient to suppress epithelial invasion in other tissues and organisms.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Drosophila Strains and Genetics

All experiments were completed at 25°C unless otherwise indicated. Mosaics were generated as described (Szafranski and Goode,2004) using FRT-Flp system. The following alleles were used (see Goode and Perrimon,1997; Szafranski and Goode,2004; Wei et al.,2004) for Drosophila stock sources other than indicated): arm1 (null) (Perrimon and Mahowald,1987), bazEH171 (null) (Abdelilah-Seyfried et al.,2003), dlgsw, dlgv59, dlglv55, dlgip20, dlgm35, dlgm52 (null), Fas2EB112 (null), Fas2rd1, Fas2MB2225 (Cheng et al.,2001), kst2H-Spec) (Thomas et al.,1998), LanA9-32 (null) (Deng and Ruohola-Baker,2000), lgl4 (null), and mysXG43 (β-Int, null) (Bunch et al.,1992). We used the following UAS lines: Baz (Wodarz et al.,2000), Dlg, Fas2, Fas2Δ3, Kek, Lgl, Nrg, DE-Cad (Sanson et al.,1996), Arm, and RacN17 (Luo et al.,1994), and Gal4 lines (Manseau et al.,1997): 198y, c208, c324a, c606, and GR1.

Histochemical Analysis and Imaging

Antibody and phalloidin staining were performed as previously described (Goode and Perrimon,1997). The following primary antibodies were used at the indicated concentrations: mouse anti-Arm (N2 7A1, 1:500; DSHB), rabbit anti-Baz (1:500; A. Wodarz), rat anti-DE-Cad (DCAD2, 1:500; DSHB), rat anti-Crb (1:1,000; U. Tepass), rabbit anti-Dlg (1:500; K.-O. Cho), mouse anti-Fas2 (1D4, 1:1,000; DSHB), rabbit anti-β-Gal (1:2,000; Cappel), rabbit anti-β-Int (CF.6G11, 1:500; D. Brower), rabbit anti-LanA (1:1,000; L. Fessler), sheep anti-Lgl (1:2,000), rabbit anti-phospho-Histone H3 (1:1,000; Upstate), rabbit anti-Par6 (1:500; J.A. Knoblich), mouse anti-α-Spec (3A9, 1:500; DSHB), rabbit anti-βH-Spec (1:500, G. Thomas), and mouse anti-α-Tub (1:500; Sigma). Cy5-, FITC-, or rhodamine red-X-conjugated donkey secondary antibodies (Jackson ImmunoResearch Labs) were used (1:2,000). Alexa488 & 568-phalloidin (1:10; Molecular Probes) were used to visualize Actin. Images were acquired with a Zeiss LSM 510 confocal microscope and processed using Photoshop software (Adobe).

Cell Migration Rates

To determine BC migration rates, images of stage-9 egg chambers were captured with a conventional epifluorescence Zeiss Axioplan-2 microscope equipped with a Hamamatsu ORCA digital camera, and the data were analyzed as described (Szafranski and Goode,2004). The statistical significance of the correlation between frequency of Fas2 or dlg tumor invasion and rates of BC migration was analyzed by applying linear regression analysis.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

We thank D. Brower, N.H. Brown, K.-O. Cho, L. Fessler, J.A. Knoblich, L. Luo, N. Perrimon, H. Ruohola-Baker, T. Schupbach, U. Tepass, G. Thomas, D. Van Vactor, J.-P. Vincent, A. Wodarz, and Developmental Studies Hybridoma Bank for reagents; M. Zhao for data on dlghf/ip20; lglnull/+ invasion; and K.-W. Choi, P.R. Hiesinger, A. Rajkovic, M. Anderson, C. Hall, and M. Zhao for critically reading the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

The Supplementary Material referred to in this article can be viewed at www.interscience.wiley.com/jpages/1058-8388/suppmat

FilenameFormatSizeDescription
jws-dvdy.21020.fig1.tif839KSupporting Information file jws-dvdy.21020.fig1.tif
jws-dvdy.21020.fig2.tif1280KSupporting Information file jws-dvdy.21020.fig2.tif

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.