![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
Department of Genetics, 2 Howard Hughes Medical Institute, Harvard Medical School, Boston, Massachusetts 02115 USA
 |
Abstract |
|---|
 Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
Drosophila Discs large (Dlg) is a tumor suppressor gene whose loss in epithelial tissues causes disrupted cell polarity and increased cell proliferation. A human Dlg homolog, hDlg, has been implicated in tumorigenic processes via its association with the product of the Adenomatous Polyposis Coli (APC) gene. We show for the first time that Drosophila Dlg is required to block cell invasion. Loss of dlg activity during oogenesis causes follicle cells to change shape and invade in a pattern similar to border cells, a small population of cells that break from the post-mitotic follicular epithelium during wild-type oogenesis, yet dlg mutant cells have not adopted a border cell fate. Both functional and morphological evidence indicates that cooperation between germ cell and follicle cell Dlg, probably mediated by Dlg PDZ domains, is crucial for regulating cell mixing, suggesting a novel developmental mechanism and mode of action for the Dlg family of molecules. These findings suggest that Dlg does not simply inhibit individual cell behaviors during oogenesis, but rather acts in a developmental pathway essential for blocking cell proliferation and migration in a spatio-temporally defined manner. A model for Dlg action in blocking cell invasion is presented.
[Key Words: Cell invasion; cell proliferation; cell shape; Discs large; membrane-associated guanylate kinases (MAGuKs); PDZ domain]
| |
Introduction |
|---|
|
Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
Adjoining tissues must regulate their barriers during development
and adult life. What factors determine whether cells
remain in one place, retaining their associations with their neighbors, or dissociate and move elsewhere? If cells move, what determines where
they go, where and when they stop, and whether or not they associate
with like or unlike cells? These are important questions in
understanding tissue morphogenesis, dynamic physiological processes, and pathologies such as metastasis (Hynes and Lander 1992
; Gumbiner 1996).
Maintenance of homotypic adhesion plays a crucial role in ensuring that
cells do not mix. Blocking E-cadherin function turns cultured cells
from noninvasive to invasive, whereas expressing E-cadherin in cancer
cells reverts their tumorigenic phenotype (Takeichi 1993; Birchmeier
and Behrens 1994). Physiological invasions and pathological metastasis
also require adhesion of invasive cells to foreign matrices and
heterologous cells (Mareel et al. 1991). What cellular mechanisms
underpin the transfer of adhesion from sites of contact between
homotypic cells to sites of contact between heterotypic cells?
An outstanding cell biological and genetic system for analyzing the
problem of cell mixing and migration is the movement of a small cluster
of post-mitotic follicular cells, called border cells (BCs), between
germ cells to the Drosophila oocyte (for review, see Montell
1994; described in Fig. 1). Several molecules have
been identified as crucial players in regulating BC migration. The
slow border cells (slbo) gene is specifically
expressed in BCs and encodes the Drosophila homolog of a
vertebrate basic region-leucine zipper transcription factor, CCAAT
enhancer binding protein (C/EBP; Montell et al. 1992). In
animals harboring strong slbo mutations, BCs fail to initiate
migration, whereas weaker slbo mutations cause delayed
initiation. Thus, slbo appears to play a role in determining
the ability of BCs to disassociate from neighboring epithelial cells,
their ability to attach to germ cells, or both of these processes. The
breathless (btl) gene encodes a FGF receptor tyrosine
kinase homolog that appears to be a direct transcriptional target of
slbo, and mutations in btl dominantly enhance BC
migration defects associated with weak slbo mutations (Murphy
et al. 1995). Whereas slbo and btl are crucial for
triggering BC differentiation and movement, little is known about the
molecules that are involved in positively or negatively regulating the
shape changes, migration, and interaction of BCs with germ cells in a
specific pattern. We show that Discs large (Dlg), known for regulating
cell shape in imaginal epithelia (Woods and Bryant 1991), is likely to
be such a molecule.
|
Dlg belongs to a conserved family of proteins termed membrane
associated guanylate kinases (MAGuKs; Woods and Bryant 1991, 1993).
These proteins share three amino-terminal PDZ domains (named for the
family members PSD-95, Dlg, and
ZO-1), an SH3 domain, and a carboxy-terminal guanylate
kinase (GuK) homology domain (for review, see Anderson 1996; Sheng
1996). Dlg molecules may bind to cytoskeletal proteins through the SH3
domain, a known adaptor motif mediating direct association with
cytoskeletal and signaling molecules (Mayer and Eck 1995). The
GuK-homology domain has not been shown to have kinase activity, and is
missing two amino acids thought to be crucial for catalytic activity
(Woods and Bryant 1993). In Drosophila, the GuK domain has
been shown to play a specific role in prohibiting cell proliferation
(Woods et al. 1996).
The three PDZ motifs are perhaps the most intriguing aspect of MAGUK
structure. Human Dlg binds via the third PDZ motif to APC, a commonly
mutated tumor suppressor gene in human colon cancer cells
(Matsumine et al. 1996). PDZ motifs have also been shown to bind to
the extreme carboxy-terminal tail of several distinct families of
signaling receptors and channels in a sequence-specific fashion
(Songyang et al. 1997), and genetic evidence for PDZ involvement in
receptor localization at specific subcellular sites has been obtained
in Caenorhabditis elegans (Simske et al. 1996). PDZ domains have thus emerged as novel adaptor modules for specific protein-protein binding, important for clustering membrane proteins as well as linking
signaling molecules in multiprotein complexes at specialized membrane
sites (for review, see Anderson 1996; Sheng 1996).
Drosophila Dlg is essential for prohibiting cell growth and
for maintaining cell adhesion and cell polarity in both embryonic and
adult tissues (Gateff 1978; Perrimon 1988; Woods and Bryant 1989, 1990;
Woods et al. 1996). In animals harboring dlg null mutations,
aberrant cell polarity is revealed in the aberrant organization of both
the actin and microtubule cytoskeletons, the transformation of columnar
epithelial cells to a apolar morphology, and the delocalized
distribution of cell adhesion molecules (Woods et al. 1996). The role
of Dlg in maintaining cell polarity appears to be separable from its
role in proliferation control, because several mutations that eliminate
carboxy-terminal sequences, including the entire GuK domain, cause loss
of proliferation control, without affecting cell polarity (Woods et al.
1996).
In this paper we show for the first time that Dlg is required to block cell invasion and that Dlg activity appears to define a novel developmental pathway. Starting very early in oogenesis, loss of dlg activity causes follicle cells to overproliferate at the poles of the egg chamber and invade germ tissue. Invading follicle cells change shape and move in a pattern similar to BCs, suggesting that the BC migration pathway is established very early in oogenesis. Both functional and morphological evidence indicates that cooperation between germ cell and follicle cell Dlg, probably mediated by Dlg PDZ domains, is crucial for regulating cell mixing. On the basis of these findings, we suggest a model for Dlg action in prohibiting interactions between tissue layers during oogenesis, and for releasing this prohibition in a regulated manner when BCs migrate. We propose that Dlg does not simply inhibit individual cell behaviors during oogenesis, but rather acts in a developmental pathway essential for blocking cell proliferation, shape change, and migration in a spatio-temporally defined manner.
| |
Results |
|---|
|
Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
BC migration
Because our analysis of Dlg function during oogenesis focuses on BC migration, we describe and extend previous observations of this process relevant to our studies. Two postmitotic follicle cell populations migrate at s9 of oogenesis (Fig. 1; see Fig. 5, below, for stages of oogenesis). Most follicle cells surrounding nurse cells move to the oocyte as an epithelial sheet along the outside of the egg chamber, whereas the BCs move through the center of the egg chamber, in concert with the epithelium. Six to seven BCs break from the follicular epithelium and adopt a mesenchymal-like morphology, then migrate to the oocyte, as they contact anterior nurse cells, then posterior nurse cells, before reaching their destination (Fig. 1).
|
The interaction of BCs with posterior nurse cells is particularly dramatic. Nurse cells maintain an invariant quadrihedral-like architecture throughout most stages of oogenesis (Fig. 1e; Fig. 5b, below), but following initiation of BC movement, one nurse cell adjacent to the oocyte extends a cytoplasmic process that contacts the migrating cells (Fig. 1f,g). Nurse cell processes are never observed in s8 egg chambers, preceding BC migration, although BC fate has already been established (data not shown; see also Fig. 4, below). Further, nurse cell processes show a clear directionality. They are never observed to extend from anterior nurse cells to BCs that have moved close to the oocyte at the posterior of the egg chamber. These observations suggest that nurse cells adjacent to the oocyte play an active role in guiding the movement of BCs to the oocyte, a suggestion further supported by analysis described below.
|
Reduction of dlg activity leads to an invasive phenotype starting as early as s1 of oogenesis
To analyze the function of dlg during oogenesis, we took
advantage of temperature-sensitive allele combinations (Perrimon 1988).
dlghf321/dlglv55 animals are
viable and have completely normal egg chambers at 18°C. When these
animals are shifted to 25°C for at least 6 hr, follicle cells exit
the follicular epithelium and intermingle with germ cells at the
anterior and posterior poles of the egg chamber (Fig. 2e; data not
shown). Cells that remain in the epithelium retain
their polarized, epithelial characteristics, whereas cells that have
exited the epithelium have an apolar morphology. Follicle cells that
have exited the epithelium in anterior regions of the egg chamber
invade between nurse cells as they migrate towards the oocyte (Fig. 2,
cf. e and f). The phenotype is evident as early as s1 of oogenesis,
just following the birth of new egg chambers (Fig. 2d).
|
Several morphological characteristics suggest that dlg invasive cells behave like BCs during wild-type oogenesis. First, BCs always migrate as a cluster of interconnected cells (Fig. 1), and likewise, dlg invasive cells always migrate as streams of interconnected cells (Fig. 2d-f). Furthermore, like BCs, invasive cells always migrate through the center of the egg chamber. They never diverge on their path to move between lateral nurse cell membranes (Figs. 2, 3, 4, 6, 7, and 9). That invasive cells are attracted to the oocyte is further supported by simple temperature-shift experiments. Temperature shifts of increasing duration result in increasingly larger streams of follicle cells that move increasingly closer to the oocyte (data not shown). Furthermore, if dlghf321/dlglv55 animals are shifted to the restrictive temperature for 6 hr, then placed at the permissive temperature for several hours, a small clump of follicle cells separates from the anterior epithelium; 24 hr later, the small clump of cells is typically found adjacent to the oocyte, and never outside the center of the egg chamber (data not shown).
|
|
|
|
The same invasive phenotype, with varying degrees of severity, is found
for many dlg mutant combinations (Table 2, below). The degree
of severity directly correlates with that described for imaginal disc
tissue (Table 1; Woods et al. 1996). For example, the
stage at which the dlg invasive phenotype is first manifested in dlghf321/dlg1P20 (or v59 or
X1-2) animals directly correlates with the degree of GuK
truncation associated with the dlg1P20,
dlgv59, and dlgX1-2 mutations (Table
1), with the weakest allele combinations only disrupting the latest
stages of oogenesis, and increasingly stronger allele combinations
disrupting increasing earlier stages of oogenesis. A similar temporally
graded pattern of phenotypic expression has been observed for mutations
disrupting components of other signaling and adhesion pathways during
oogenesis (Goode et al. 1992, 1996a,b). Because a similar invasive
phenotype is observed in many dlg mutant combinations,
including clones of a null dlg mutation (see below), we
conclude that the invasive phenotype is caused by a reduction or loss
of dlg activity.
|
|
Invasive follicle cells have not adopted a BC fate
Because
dlghf321/dlglv55 follicle
cells behave like BCs, we determined whether they have adopted a BC
fate. The slbo gene (Montell et al. 1992) is specifically
expressed in BCs at s8, before they break from the epithelium and
initiate migration to the oocyte, and slbo continues to be
expressed in BCs during their migration to the oocyte at s9 (Fig.
3a-c). We analyzed the expression of slbo
in dlg mutant egg chambers. slbo is expressed in the
wild-type pattern in s8 and s9
dlghf321/dlglv55 mutant egg
chambers and is not expressed in invasive follicle cells (Fig. 3d,e).
slbo expression never initiates before s8 in dlg
mutant animals (not shown). We also analyzed expression of FasIII
(Patel et al. 1987), which is specifically expressed in two polar cells
at the anterior of the egg chamber starting during mid oogenesis. The
polar cells are included among the cells that will go on to migrate as
border cells (not shown). We did not find an increase in the number of
FasIII-positive cells in
dlghf321/dlglv55 mutant egg
chambers, or in egg chambers harboring follicle cell clones of the null
dlgm52 mutation, further suggesting that
overaccumulating follicle cells have not acquired a border cell fate
(data not shown). Finally, we note that whereas dlg-invasive
cells share an apolar morphology with border cells, other aspects of
their morphology, such as absence of lamellipodia-like structures
(Figs. 1, 2), are not shared with border cells. On the basis of these
observations, we conclude that dlg invasive follicle cells
have not adopted a BC fate, yet, as described below, these cells
undergo and participate in several morphogenetic transitions
characteristic of BC invasion.
The oocyte and nurse cells attached to the oocyte extend processes that contact invasive follicle cells
Although dlghf321/dlglv55 follicle cells have not adopted a BC fate, they resemble BCs in their apolar morphology and migration pattern. We sought to determine if dlg follicle cells also behave like BCs in their pattern of interaction with nurse cells attached to the oocyte (Fig. 1e-g). To perform these experiments, dlghf321/dlglv55 animals were shifted to the restrictive temperature for 6-12 hr, so that invasive follicle cells would initiate migration just past the first nurse cells (Fig. 4a), the point at which BCs come in contact with a process extended by an oocyte-associated nurse cell (Fig. 1e-g). In dlghf321/dlglv55 mutant egg chambers, nurse cells extend processes to meet invasive follicle cells. This is clear by s6, when the egg chamber has acquired an elongated shape (Fig. 5a), but as for the invasive phenotype, these interactions probably start as early as s1 (see also Fig. 2e). We confirmed these interactions for other allele combinations. For example, the interactions are also observed in dlghf321/dlg1P20 animals, in which cells do not invade before s7 (Fig. 4b; Table 2).
We also found that invasive follicle cells show unusual interactions with the oocyte. Once invasive cells have migrated to the most posterior nurse cells, the oocyte elongates to contact the cells (Fig. 4c,d; see also Fig. 8d). Although the oocyte does not extend during BC migration in wild-type animals, this result documents the selective adhesive forces acting between the oocyte and BCs, and suggests that dlg plays a role in maintaining oocyte architecture during wild-type oogenesis.
|
Dlg expression
To help establish the basis for the invasive behavior of follicle
cells in dlg mutant egg chambers, we analyzed the
spatio-temporal and subcellular distribution of Dlg proteins throughout
oogenesis. Western analysis reveals three polypeptide species of about
91, 100, and 102 kD (Fig. 5a), approximately the same mass described for Dlg proteins in imaginal discs and larval muscles (Lahey et al.
1994; Woods et al. 1996). Dlg is expressed in both germ and follicle
cell tissues from the time that the germ cell cyst becomes surrounded
by follicle cells in the germarium (Fig. 5b). Dlg appears to be
expressed at equivalent levels in both tissues throughout the growth
phases, as the germ cell cyst expands in size and the number of
follicular epithelial cells increases >10-fold. Following cessation
of follicle cell proliferation, levels of Dlg protein appear to
dramatically decrease in germ cells, corresponding to the time just
preceding and including BC migration to the oocyte (Fig. 5b).
At the cellular level, Dlg is localized to sites of contact between
follicle cells and to sites of contact between germ cells, but appears
to be excluded at sites of contact between germ cells and follicle
cells (Fig. 5c,d). Likewise, before and during BC migration, Dlg is
expressed at sites of contact between border cells, but not at points
of contact between border cells and germ cells (data not shown).
Considerable focus has been placed on the localization and function of
Dlg in epithelial septate junctions (Woods and Bryant 1991, 1993), but
our analysis does not address the function of Dlg in these structures,
because septate junctions are not found in the germ cell cyst or
follicular epithelium throughout the stages preceding and including BC
migration (Mahowald 1972).
Dlg is required in both germ cells and follicle cells
Because Dlg is expressed in both germ cells and follicle cells, we sought to determine the respective function of germ cell Dlg and follicle cell Dlg in the genesis of the invasive phenotype. To determine where Dlg is required, germ cell and follicle cell clones were generated with the genetic null mutation dlgm52 (Fig. 6b). Eliminating dlg function in germ cells alone did not cause any defects, except for occasional misshapen germ cells (Fig. 6c, not shown). Eliminating dlg function in follicle cells alone led to overaccumulation of follicle cells at the anterior and posterior poles of the egg chamber (as in dlghf321/dlglv55 egg chambers, see Fig. 2e), but these cells invade germ tissue only rarely (Fig. 6d; invasion occurs in ~1/50 egg chambers, n = 425, Fig. 6f). When dlg function is eliminated in both germ cells and follicle cells, follicle cells always invade along the BC pathway (Fig. 6e). We conclude that Dlg acts in both germ cells and follicle cells to block cellular invasion.
Dlg appears to be required for prohibiting cell proliferation at the poles of egg chambers
We considered two possibilities for the overaccumulation of follicle cells at the poles of dlg mutant egg chambers (Fig. 2e; see below). One possibility is that follicle cells have redistributed from lateral regions of the egg chamber to the egg chamber poles. This does not seem likely because even after 36 hr at the restrictive temperature, and the massive invasion of follicle cells, lateral regions of the follicular epithelium do not appear to differ in cell density from wild-type epithelia (Fig. 2, c and f). A second possibility is that follicle cell overaccumulation results from follicle cell overproliferation.
To obtain evidence for overproliferation of follicle cells, we compared the degree to which follicle cells at the poles of wild-type and mutant egg chambers incorporate BrdU, a marker for newly dividing cells, as a function of distance along the anterior-posterior axis. Wild-type follicle cells incorporate BrdU at the same frequency independent of anterior-posterior position (Fig. 7a,c). In contrast, the relative frequency with which dlghf321/dlglv55 follicle cells incorporate BrdU is higher at the anterior and posterior poles of the egg chamber (Fig. 7b,c). The difference between mutant and wild type was even more pronounced if the degree of incorporation of BrdU in follicle cells thought to be most susceptible to loss of dlg activity, those residing strictly within the plane of invasive cells, was analyzed (Fig. 7c).
To further substantiate that follicle cells overproliferate in dlg egg chambers, the total number of follicle cells in dlgsw versus wild-type s5-s6 egg chambers was compared. In dlgsw egg chambers, in which follicle cells accumulate at the poles of the egg chamber without invading (Fig. 7d), there are on average ~150 more follicle cells per egg chamber compared with wild type (Fig. 7e). Further, these follicle cells appear to overaccumulate at the anterior and posterior poles of the egg chambers, where we observed increased BrdU incorporation in dlghf321/dlglv55 egg chambers (Fig. 7f). Combined, these findings strongly indicate that BrdU incorporation at the poles of dlghf321/dlglv55 egg chambers results from loss of proliferation control, and not from a redistribution of follicle cells.
dlg mutations that map to the SH3 and GuK domains do not confer premature cell mixing
dlg mutations that behave as genetic nulls, such as
dlgm52 (described above), cause loss of
proliferation control, cell polarity, and adhesion in imaginal
epithelia (Woods et al. 1996; Table 1), and an invasive phenotype when
removed from germ cells and follicle cells during oogenesis (Fig. 6).
In contrast, dlg mutations that specifically disrupt the SH3
and GuK domains cause loss of proliferation control, but have no effect
on cell polarity (Table 1; Woods et al. 1996). To assay the requirement
of the SH3 and GuK domains in blocking cell invasion, we analyzed the
phenotype of SH3 and GuK mutations by employing the germ cell-follicle
cell clone technique described (Fig. 6). In egg chambers in which both
germ cells and follicle cells are mutant for dlglv55
or dlgX1-2 mutations, which partially and almost
completely eliminate the GuK domain, respectively, or for the
dlgm30 mutation, which disrupts the SH3 domain,
follicle cells accumulate at the poles of the egg chamber, but do not
invade between germ cells (Fig. 7g,h). Further, in animals heterozygous
for dlg1P20 or lv55 (that disrupt GuK sequences),
and dlgm52 (that disrupts the SH3 domain), follicle
cells accumulate at the poles of egg chambers without invading (Table
2; Fig. 7i). These results indicate that Dlg molecules lacking the
activity of the GuK domain, and probably the activity of the SH3
domain, retain the Dlg activity that prohibits follicle cell invasion.
Dlg prohibits premature BC migration
Our analysis indicated that Dlg is required for inhibiting several interactions and activities in growing egg chambers that appear to be fundamental to BC migration. Does Dlg play a similar role during the process of BC migration? To analyze the role of dlg during post-mitotic stages, we used extremely weak mutant combinations, such as dlghf321/dlgm35 and dlghf321/dlgsw, in which no detectable defect on egg chamber development is found until s8-s9, when BCs are determined and migrate (Table 1). At s8, the only observable defect is the occasional premature breaking and movement of BCs from the anterior follicular epithelium (Fig. 8a,b). We do not think it is likely that this phenotype results from delayed migration of the follicular epithelium towards the oocyte, because we do not observe defective epithelium migration in stronger dlg mutant combinations. At s9, in addition to the premature arrival of BCs to the oocyte, we sometimes observe cytoplasmic extensions between the oocyte and BCs, suggesting excessive interaction between these cells (Fig. 8c,d). These extensions are not found during BC migration in wild-type egg chambers. These results indicate at least two similarities for the role of Dlg during BC migration and the earlier stages of oogenesis. First, the premature breaking of BCs suggests a functional connection to the early invasive phenotype, and second, the pulling of the oocyte by the BCs is reminiscent of the cytoplasmic extensions observed between posterior germ cells and invasive follicle cells in egg chambers mutant for stronger dlg mutations (Fig. 4c,d).
kek-1 is expressed in BCs and in overproliferating and invasive follicle cell populations
Our analysis indicates that follicle cells at the poles of egg chambers throughout the growth phases, and BCs during the migratory phases, respond differentially to loss of dlg during oogenesis. This finding was unexpected because Dlg protein is expressed at uniform levels in all follicle cells throughout oogenesis (Fig. 5). To determine if follicle cells sensitive to loss of Dlg function might be genetically distinct, we screened enhancer trap lines. We found that the Drosophila kekkon-1 (kek-1) gene (Mussachio and Perrimon 1996) is specifically expressed in fields of follicle cells at the poles of the egg chamber, and in BCs before and during their migration to the oocyte (Fig. 9). This finding suggests that BCs and follicle cells at the poles of the egg chamber are functionally distinct, on the basis of sensitivity of these cells to loss of Dlg function, and genetically distinct, on the basis of their expression of the kek-1 gene.
| |
Discussion |
|---|
|
Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
Dlg acts in a developmental pathway that prohibits the cell shape changes and morphogenetic transitions characteristic of BC migration
We show for the first time that Dlg is required to prohibit cell
invasion, in a pattern resembling border cell migration. BCs are
distinguished from other follicle cells by expression of a
C/EBP-like factor, encoded by the slbo gene
(Montell et al. 1992). Following C/EBP expression, BCs
become further distinguished by their adoption of an apolar morphology,
their migration, and their interaction with germ cells in the posterior
of the egg chamber (Fig. 1). We found that the Dlg plays no role in BC
specification, but rather is likely to function following
C/EBP expression to regulate morphogenetic activities of
both BCs and germ cells. In support of this hypothesis, we find that
BCs migrate prematurely in animals mutant for weak Dlg mutations (Fig.
8). Premature migration may result from both decreased Dlg activity
between BCs and adjacent follicle cells, as indicated by the expression
of Dlg at sites of contact between these cells (Fig. 5), and from the
excessive interaction of the BCs with the germs cells, as indicated by
the hypertrophic cell extensions between BCs and the oocyte (Fig. 8).
We speculate that simultaneous reduction and loss of Dlg activity in
BCs and germ cells, respectively, allows BC migration in wild-type animals. BC differentiation, induced by slbo, may lead to
reduction of Dlg activity specifically in BCs (see below), whereas
unknown factors apparently lead to a loss of Dlg protein in germ cells (Fig. 5).
That reduction of Dlg in germ and follicle cells is essential for BC migration is indicated by the cellular invasion phenotype observed when Dlg activity is lost or reduced in both germ cells and follicle cells, starting as early as s1 of oogenesis. The cellular invasion phenotype is associated with several cellular events indistinguishable from those guiding BC movement. Both BCs and dlg invasive cells undergo an epithelial-mesenchyme-like transition, in which they move from adjacent epithelial cells as they shift their morphology from a polarized epithelial phenotype to an apolar morphology. dlg invasive follicle cells then move to the oocyte, along a path that is indistinguishable from the BC pathway. Invasive cells never turn from the central path, between lateral nurse cell membranes. The fidelity of this migration pattern is likely to result from specific interaction of invasive cells with posterior nurse cells, as indicated by the hyperplastic extension of a cytoplasmic process by a nurse cell adjacent to the oocyte (Fig. 4), resembling the contacts established between BCs and an oocyte-adjacent nurse cell during normal oogenesis (Fig. 1). Loss or reduction of Dlg appears to be required in both cell types to permit these interactions, because the invasive phenotype is always manifested when dlg activity is eliminated or reduced in both germ cells and follicle cells, but is typically not manifested when one tissue is mutant. This functional requirement is consistent with morphological evidence indicating that both germ and follicle cells play a role in guiding BC movements.
Significantly, our findings imply that Dlg does not simply inhibit individual cell behaviors during oogenesis, but rather may be a component in a developmental pathway essential for blocking cell proliferation and migration in a spatio-temporally defined pattern. dlg mutant egg chambers indicate that this pathway acts in germ and BCs and in fields of follicle cells at the anterior and posterior poles of egg chambers. Functionally, this is indicated by anterior and posterior follicle cell shape change, overaccumulation, and movement towards the oocyte (Figs. 2, 6, 7, 8; the movement of apolar posterior cells towards the oocyte is best documented in Fig. 2e), and genetically, this is indicated by kek-1 expression in BCs and follicle cells at the poles of the egg chamber. Strikingly, slbo is also expressed at both the anterior and posterior poles of the egg chamber (Fig. 3), further suggesting a relationship between polar follicle cells.
BCs and dlg invasive cells are distinguished in that the latter are supplied from a field of overproliferating follicle cells that migrate as a continuous stream, whereas BCs, derived from a postmitotic epithelium, completely dissociate from the follicular epithelium. These distinctions result from the requirement of Dlg in proliferation control (see below), and from the inability of dlg mutant cells to completely break from the follicular epithelium. These two processes appear to be intimately connected. In animals harboring dlg temperature-sensitive mutations, follicle cell overproliferation and cell invasion can be turned off simply by placing the animals at the permissive temperature (not shown). Under these circumstances, the epithelium heals, while invading cells continue to move towards the oocyte. Thus, the breaking of BCs from the follicular epithelium during wild-type oogenesis may involve a cycle of reduced Dlg activity for the BCs to exit the epithelium, followed by a return to higher activity to reseal the follicular epithelium.
Whereas we have presented several lines of positive evidence
implicating Dlg in a pathway that inhibits the invasion characteristic of BCs, specificity implies that Dlg should not be required in many
other cellular interactions during oogenesis. Both anterior-posterior and dorsal-ventral specification of follicle cells, which establish the
oocyte axes, are known to depend on intimate interactions between germ
and follicle cells (Gonzalez-Reyes et al. 1995; Roth et al. 1995), yet
we do not observe any defects in these processes in dlg mutant
egg chambers. Furthermore, we have not observed a role for Dlg in the
movement of the follicular epithelium along the outside of the egg
chamber during s9. Many aberrant interactions between germ cells and
follicle cells could be imagined, such as simultaneous invasion of
follicle cells between the interface of each nurse cell, but such
patterns are not observed. Finally, we suggest that the requirement for
Dlg in prohibiting follicle cell proliferation may not be distinct from
dlg functions maintaining epithelial integrity, because the
loss of epithelial characteristics of follicle cells during the early
stages of oogenesis occurs in the environment of a mitotically active
epithelium. Thus, the requirement of Dlg for inhibiting follicle cell
division may reflect an indirect consequence of the protein's
requirement in maintaining appropriate contact between the epithelial
cells, which may be indirectly tied to the regulation of follicle cell
proliferation.
Dlg PDZ domains appear to be required for prohibiting follicle cell invasion
We found that the Dlg SH3 and GuK domains do not appear to play a
role in blocking cell invasion (Fig. 7), indicating that the PDZ
domains inhibit germ cell-follicle cell mixing. We have not tested
this hypothesis directly, because dlg mutations that specifically disrupt the PDZ domains do not exist. What role might the
PDZ domains play in blocking cell mixing? PDZ domains bind to the
cytoplasmic tail of transmembrane molecules and localize them to
specific regions on the cell surface (for review, see Anderson 1996;
Sheng 1996). During oogenesis, Dlg is localized to sites of contact
between germ cells and to sites of contact between follicle cells, but
Dlg is absent from sites of contact between germ cells and follicle
cells (Fig. 5). This localization almost certainly ensures that
presumptive Dlg-bound cell surface molecules retain tissue-specific
segregation, and further suggests that Dlg PDZ domains serve as
components of a germ cell-follicle cell tissue barrier (Fig.
10). Because the germ cell-follicle cell interactions that we observe in dlg mutant egg chambers
specifically resemble BC interactions (see previous section), we
suggest that Dlg binds to a class of cell surface molecules that have
the potential to mediate the cellular interactions characteristic of BC
development, starting at s1 (Fig. 10). We propose that at the time of
BC migration, via a regulated step that allows the redistribution of at
least some of the cell surface molecules bound by Dlg PDZ domains to the interface between germ cells and follicle cells, germ cells and BCs
intermingle. Obvious candidates for this regulated step are loss or
reduction of Dlg in germ cells, and expression of slbo in BCs.
Loss or reduction of Dlg in germ cells might lead to the diffusion of
previously bound molecules to cellular regions other than the interface
with other germ cells. We can only speculate how slbo might
elicit a freeing of Dlg bound molecules, but precedent for regulated
interactions involving MAGuKs is the interaction between the MAGuK
PSD-95 and the inward rectifier K+ channel Kir 2.3, which appears to
be blocked by phosphorylation of a serine within the consensus
PDZ-binding motif of Kir 2.3 by protein kinase A (Cohen et al. 1996).
We suggest that in dlg mutant egg chambers, molecules bound by
PDZ domains are freed prematurely, in the absence of specific signals,
resulting in patterns of cell transformation and intercellular
interaction resembling BC development. This model suggests that the BC
migration pathway has been established very early in oogenesis and
provides an explanation for the variegated adhesion phenomena mediated by Dlg, apparently promoting adhesion between follicle cells, while
inhibiting contact and excessive interaction between germ cells and
follicle cells.
|
Dlg may help to pattern follicle cell divisions
Loss or reduction of Dlg activity in the follicular epithelium
causes follicle cells to overaccumulate at the anterior and posterior
poles of the egg chamber (Figs. 2, 6, 7, and 9). We do not believe that
this overaccumulation results from redistribution of follicle cells,
because we did not find discontinuities in lateral regions of the
epithelium (Figs. 2, 6, 7, and 9). Further, egg chambers appear to have
a greater total number of follicle cells, and follicle cells
incorporate BrdU higher levels in the region of the epithelium that
contains overaccumulating cells (Fig. 7). The spatial specificity of
inhibited follicle cell proliferation does not reflect an underlying
pattern in follicle cell divisions, because follicle cells in the
wild-type follicular epithelium appear to enter both S phase and M
phase at a uniform frequency relative to the anterior-posterior axis of
the egg chamber (this study; Goode et al. 1996b). Rather, the
differential requirement of Dlg would appear to reflect an underlying
distinction between follicle cells at the poles of the egg camber. The
fields of cells at the poles are distinguished by their expression of
kek-1, and perhaps by greater intercellular tension caused by
their position in a curved region of the epithelium. Dlg may be a
component of a signaling/adhesion pathway that helps to
establish a spatially uniform rate of cell proliferation by
differentially inhibiting cell proliferation at the egg chamber poles.
A Drosophila model for cellular invasion
How do dlg invasive phenotypes compare with normal and
pathological cell invasions in other organisms? The cell invasions that
we have analyzed have no relevance to those in which cells traverse
extracellular matrix, because basement membrane does not exist between
either nurse cells or between the follicular epithelium and nurse
cells. Rather, our data may be useful for understanding how cells move
between heterologous cells during developmental processes such as
migration of gonocytes, adult processes such as movement of leukocytes
to specific destinations throughout the organism, and during
pathological states such as cancer. Whereas most malignant cancer cells
traverse an extracellular matrix before reaching their destination,
some cancerous cells, particularly in the central nervous system,
apparently metastasize strictly through cellular environments, such as
axon fascicles (Mareel et al. 1991).
Regardless of interactions with matrices, at a cellular level, normal
and pathological cell invasions involve common transitions, including
loss of homotypic cell adhesion, epithelial-mesenchymal conversion,
gain of heterotypic cell adhesion, and cell migration. In our studies,
we have shown that Dlg plays a role in mediating all of these cell
behaviors. These cell activities provide a common link between normal
and pathological cell invasions, and have led to the suggestion that
developmental genes that block invasion are switched off in malignant
cells, whereas genes that promote invasion become activated on
developmental maturation (Mareel et al. 1991). Although our
developmental studies do not make a direct connection to adult
malignancy, human Dlg has been implicated in the development of the
tumorous state via its association with APC, the most commonly mutated
gene in human colon cancer (Matsumine et al. 1996). Further, we have
shown that dlg invasive cells resemble malignant cells in
several regards, including loss of homotypic adhesion, loss of
proliferation control, and cell polarity, and enhanced heterotypic
interactions and migratory ability. The Dlg invasive phenotype may
serve as an excellent genetic and cell biological model for analyzing
intercellular interactions underlying cell shape and cell mixing
phenomena in a variety of normal and pathological cell invasions.
| |
Materials and methods |
|---|
|
Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
Dlg mutations
dlg mutations used in this study have been described
previously (Table 1; Perrimon 1988; Woods and Bryant 1989, 1991). All flies were reared on standard media at 25°C or were reared at 18°C, and then shifted to 25°C, as described in the text.
Histochemistry
Fluorescein-phalloidin staining was performed as described
(Goode et. 1996a). Immunohistochemistry, by use of FITC-conjugated secondary antibodies, was performed as described previously (Goode et
al. 1996b). The following sera were used in this study: Dlg (Woods and
Bryant 1991), Arm (Peifer et al. 1993), and FasIII (Patel et al. 1987).
All fluorescein-labeled tissues were analyzed with a Bio-Rad laser
scanning confocal microscope, attached to a Zeiss Axiophot microscope.
BrdU labeling and analysis of follicle cell divison patterns
BrdU was incorporated into ovarian tissue in vivo. Females that
had been reared at 18°C were placed with an equal number of males
for a few days and starved for 2 hr at 18°C (to induce hunger). The
flies were then shifted to 25°C onto a yeast paste mixture consisting of 0.5 grams of dry yeast:1.5 ml of BrdU-labeling
solution (Amersham). After feeding for 12 hr, females were sacrificed
and their ovarian tissue was fixed in Carnoy's fixative, washed in PBT, hydrolyzed in 2 N HCl for 1 hr, and rinsed twice in PBT. Ovarian tissue was then treated with 50 µg/ml of
RNase in PBS for 2 hr. Following RNase, the tissue was stained with
antibodies to BrdU (Amersham), by use of a FITC-conjugated secondary
antibody, as described above. The tissue was then stained for 20 min
with propidium iodide and rinsed with PBS. The relative position of follicle cells that had incorporated BrdU was established with a laser
scanning confocal microscopy, as described previously (Goode et al.
1996b).
Clonal analysis
To analyze the phenotypic consequence of introducing dlg
mutations in germ cells, follicle cells, or both tissues
simultaneously, FRT-mediated recombination was induced with a heat
shock FLPase (hsFLPase) chromosome (Golic and Lindquist 1989). Clones
were marked by loss of a constitutively expressed lacZ gene
driven by a tubulin promoter, TUBlacZ [(Fig. 6; originally X37
lacZ (D. Harrison, pers. comm.)]. A nuclear localization
signal is fused to the 
-gal protein produced by the X37
lacZ gene, thus targeting expression to both germ and follicle
cell nuclei. FRT101 was used in all experiments. The progeny
of dlg FRT101/FM7, B animals crossed
to TUBlacZ FRT101/Y;
hsFLPase/+ animals were heat shocked for 2 hr on
consecutive days during early pupal development. Adult B+ females
were allowed to lay eggs for several days before dissecting ovaries.
Ovaries were stained for lacZ expression by standard methods.
Western analysis
Western analysis was performed as described (Woods et al. 1996).
| |
Acknowledgments |
|---|
Dan Woods provided Dlg sera and stimulating and helpful discussion throughout this study. We thank Kevin Fitzgerald, Denise Montell, and Dan Woods for helpful comments on the manuscript. We thank Beth Noll and Betsy Wilder for technical advice concerning the BrdU labeling protocol and Allan Spradling for helpful discussion concerning follicle cell counts. Doug Harrison kindly provided the X37 tubulin lacZ marker. Arm sera were a gift of Mark Peiffer, and FasIII sera were a gift of Nipam Patel. S.G. was supported by a National Institutes of Health postdoctoral fellowship (GM17511-02) and is currently a Leukemia Society of America Special Fellow (3232-98). N.P. is an investigator of the Howard Hughes Medical Institute.
The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC section 1734 solely to indicate this fact.
| |
Footnotes |
|---|
Received June 12, 1997; revised version accepted August 5, 1997.
1 Corresponding author.
E-MAIL goode{at}rascal.med.harvard.edu; FAX (617) 432-7688.
| |
References |
|---|
|
Top
Abstract
Introduction
Results
Discussion
Materials & Methods
References
|
|---|
to produce the follicular epithelium.
Dev. Biol.
178:
35-50[CrossRef][Medline].
noncoding
region of kek-1 (Musacchio and Perrimon 1996
