Differential Activity of Drosophila Hox Genes Induces Myosin Expression and Can Maintain Compartment Boundaries

Compartments are units of cell lineage that subdivide territories with different developmental potential. In Drosophila, the wing and haltere discs are subdivided into anterior and posterior (A/P) compartments, which require the activity of Hedgehog, and into dorsal and ventral (D/V) compartments, needing Notch signaling. There is enrichment in actomyosin proteins at the compartment boundaries, suggesting a role for these proteins in their maintenance. Compartments also develop in the mouse hindbrain rhombomeres, which are characterized by the expression of different Hox genes, a group of genes specifying different structures along their main axis of bilaterians. We show here that the Drosophila Hox gene Ultrabithorax can maintain the A/P and D/V compartment boundaries when Hedgehog or Notch signaling is compromised, and that the interaction of cells with and without Ultrabithorax expression induces high levels of non-muscle myosin II. In the absence of Ultrabithorax there is occasional mixing of cells from different segments. We also show a similar role in cell segregation for the Abdominal-B Hox gene. Our results suggest that the juxtaposition of cells with different Hox gene expression leads to their sorting out, probably through the accumulation of non-muscle myosin II at the boundary of the different cell territories. The increase in myosin expression seems to be a general mechanism used by Hox genes or signaling pathways to maintain the segregation of different groups of cells.


Introduction
During animal development there is a progressive subdivision of the organism into distinct groups of cells that will form different organs and structures. In this process, the cells normally acquire different cellular affinities, which allow both to keep a coherent group of cells with the same fate and to distinguish them from surrounding cells with different identity [1].
The development of the Drosophila wing imaginal disc is a good model to study cell segregation. Wing and haltere imaginal discs are subdivided, early in development, into an anterior (A) and a posterior (P) compartment [2]. The selector gene engrailed (en) is expressed in the P compartment and induces the expression of the Hedgehog (Hh) signaling molecule. Cells from the P compartment, transcribing en and hh, do not mix with cells from the A compartment, lacking the expression of both genes. The boundary separating the two compartments forms a straight border, the line of minimal contact, named the antero-posterior (A/P) compartment boundary [2][3][4].
This strict lineage segregation can be compromised in two ways. First, posterior cells lacking en (and its cognate gene invected), can penetrate into the A compartment [3,5]; reciprocally, if en is ectopically expressed in anterior cells, they can move to the P compartment [3]. Second, anterior cells mutant for smoothened (smo), an obligatory component of the Hh signaling pathway, can cross into the P compartment [6,7]. Although a complete mixing of A and P cells requires changes in the activity of both En and the Hh pathway, eliminating the response to the Hh signal causes a more complete response and predominates over the mechanism depending on changes in en [3].
Wing and haltere imaginal disc are further subdivided into dorsal (D) and ventral (V) compartments. Dorsal cells transcribe apterous (ap), which regulates the expression of Serrate (Ser), a ligand of Notch (N), whereas ventral cells express another N ligand, Delta (Dl). N is active at both sides of the dorso-ventral (D/V) compartment boundary, and it is required to maintain the segregation of D and V cells [8]. Experiments that compared the behavior of cells mutant for N or ap at the D/V boundary [9][10][11][12] suggested that ap has an instructive role, and N a permissive one, in defining the D/V boundary [12,13]. However, an alternative model proposed that N signaling is sufficient to separate D and V cells by creating a ''fence'' [10,14,15].
Segregation between distinct populations of cells also occurs in rhombomeres of the chick vertebrate hindbrain [16]. Rhombomeres have distinguishable cell lineages and express unique combinations of Hox genes [17,18]. These genes specify the main axis in bilaterians [19], and are required to maintain the correct architecture of rhombomeres in the mouse hindbrain [20]. In Drosophila, experiments with imaginal discs in culture have shown that cells with different Hox addresses do not mix [21][22][23]. Moreover, an analysis carried out in the eye-antennal disc suggests that the Hox gene Deformed may be needed to establish a clonal restriction between maxillary and antennal fields [24]. However, the mechanism whereby Hox genes determine different cell affinities in the fly has not been addressed.
We show here that cells with different expression of the Hox gene Ultrabithorax segregate from each other and that this difference is sufficient to maintain A/P and D/V boundaries in the wing, haltere or leg disc. Differences in Ubx activity induce high levels of non-muscle myosin II. Other Hox genes seem to have a similar influence on myosin expression and compartment boundary maintenance. We propose that Hox genes may separate cells with different identity through the control of myosin accumulation.

Genetics
Most of the mutations and constructs are described in Flybase. Other constructs used are UAS-dsUbx [25], UAS-OUbx [26], sqh-GFP [27], baz-GFP and zip-GFP [28]. In the experiments with the Gal4/Gal80 ts system [29] the larvae were changed from 17uC to 29uC at the early third larval instar and kept at 29uC for 24h.

Clonal Analysis
Clones of the following genotypes were induced at 24-48 h and 48-72 h (smo, smo Ubx and Abd-B clones) or 48-72 h and 72-96 h (Ubx clones) after egg laying with a one-hour heat-shock given at 37uC.
To determine the crossing of the A/P boundary by smo mutant clones, each investigator scored the clones ''blind''. Only in those cases in which the three researchers agreed we considered the clones as crossing or not crossing the compartment boundary.

Immunochemistry
Antibody staining was done according to standard protocols. The antibodies used are: mouse anti-Ubx at 1/10 [30], rabbit anti-

Adult Cuticle Analysis
It was done following standard procedures.

Results and Discussion
The Drosophila Ultrabithorax (Ubx) Hox gene determines the development of the third thoracic segment (T3), and Ubx mutants transform this segment into the second thoracic one (T2) [31]. Ubx is expressed in the haltere discs, which will form the dorsal adult T3, but not in the wing disc (but for the peripodial membrane), which develops into the dorsal T2 [32]. As observed in the adult    [33], cells lacking Ubx expression in the haltere disc do not mix with Ubx-expressing cells: Ubx mutant clones induced in this disc are round, with smooth borders and segregate from the surrounding epithelium (Fig. 1A). This segregation is also evident in clones expressing Ubx ectopically (Fig. 1B). These observations confirm that Ubx activity provides specific cell affinities.
Ultrabithorax can Maintain the Antero-posterior Compartment Boundary in the Absence of Hh Signaling Anterior clones mutant for smo, when abutting the A/P compartment boundary of the wing disc, frequently cross into the P compartment [6,7] (Fig. 1C, C9). In the haltere disc, the A/P boundary is not so straight as in the wing disc (Fig. 1D), but we have also observed a similar behavior of many anterior smo clones (Fig. 1E-E99 and Table 1; see also Fig.1d in Ref. 7).
The observation that either Ubx activity or Hh signaling can provide the cells with particular cell affinities, and that Hh is needed to maintain the separation of cells from A and P compartments, suggested the possibility that Ubx activity may be sufficient to maintain compartment boundaries. Anterior clones double mutant for smo and Ubx, if induced close to the A/P boundary of wing or haltere discs, can penetrate into the P compartment ( Fig. 2A-B99). This seems to suggest that differences in Ubx activity cannot compensate for the absence of hh signaling. However, in this experiment both A and P compartments express Ubx, and Ubx + cells from both compartments may equally reject the mutant cells. We wondered if a different activity of Ubx in A and P compartments could be sufficient to separate A and P cells when Hh signaling is compromised. To answer this question we induced anterior smo 2 clones in bithorax (bx) haltere discs, in which only the P cells express Ubx [31,32]. These clones, when abutting the P compartment, respect the A/P boundary (Fig. 2C-C99 and Table 1). The comparison of the distribution of smo clones induced in wildtype and bx 2 discs suggests that the different Ubx activity in A and P cells significantly contributes to maintain the A/P compartment boundary in the absence of Hh signaling.
We decided to study if the same result applies to the second and third leg discs. The rationale for these experiments is that Ubx is expressed in both compartments of the third leg disc but only in the P compartment of the second leg disc [34,35]. The results presented in Fig. 2D-E99 and in Table 1 show that most smo 2 clones induced in the A compartment of the third leg disc cross into the P compartment, whereas similar clones induced in the second leg disc respect the boundary. As also observed in the haltere disc, a few smo 2 clones in the third leg disc do not readily cross into the P compartment (Table 1). This may be due to Ubx being present at higher levels in the P compartment of this disc than in the A compartment [32,36], or represent a coincidental event. Collectively, the data strongly suggest that Ubx can maintain the A/P boundary in the absence of Hh signaling.

Ultrabithorax Maintains a Smooth Dorso-ventral Boundary in the Absence of Notch Signaling
Dorsal and ventral cells of the wing and haltere discs are separated by a straight D/V boundary (Fig. 3A, B). When ap expression is substantially reduced (ap-Gal4 UAS-GFP/ap UGO35 discs), and N signaling, therefore, compromised, the boundary in the wing [12] (Fig. 3C) and haltere (Fig. 3D) disc is uneven. However, if we express Ubx in the dorsal compartment of an ap 2 wing disc (ap-Gal4 UAS-GFP/ap UGO35 ; UAS-Ubx/tub-Gal80 ts larvae; n = 14), so that there is Ubx activity only in the dorsal side, or if we reduce Ubx activity in the dorsal compartment of the haltere disc (ap-Gal4 UAS-GFP/ap UGO35 ; UAS-dsUbx Df109/+ larvae; n = 11), and therefore keep Ubx activity only in the ventral compartment, the smooth D/V border is largely restored in both discs (Fig. 3E, F). These results suggest that strong differences in Ubx expression between dorsal and ventral compartments (that is, an on/off Ubx situation) can maintain a smooth D/V boundary in discs lacking N signaling.

Differences in Ultrabithorax Expression Induce Accumulation of Myosin
It has been proposed that N signaling creates a ''fence'' that prevents cells from the D and V compartments from mixing [10,14,15]. Accordingly, elevated levels of F-actin and of the regulatory chain of non-muscle myosin II, encoded by the gene spaghetti-squash (sqh), are observed at this boundary in the wing disc [14,15]. Significantly, a role for actomyosin in maintaining the A/ P boundary in the wing disc [37] and in the embryo [38] has been described. Moreover, absence of zipper (zip), encoding the nonmuscle myosin II heavy chain, prevents the maintenance of a normal D/V boundary in the wing disc [15].
To check if these proteins might also play a role in the separation of cells with different Ubx expression, we induced Ubx clones in the haltere disc and observed the expression of sqh-GFP and of zip-GFP. Most of these clones are surrounded by a ring of sqh-GFP, zip-GFP and phalloidine staining (Fig. 4A-B999). A lower level of bazooka (baz), a gene required to establish apico-basal cell polarity in Drosophila [39], was also reported at the wildtype D/V boundary of the wing disc [14], but in cells surrounding Ubx mutant clones baz expression is also increased (Fig. 4C-D999). Similar results are seen at the A/P compartment boundary of bx 3 hh-lacZ/TM2, Ubx 130 haltere discs, in which only the A compartment lacks Ubx protein (Fig. 4E-E999). By contrast, in bx 3 hh-lacZ/TM6B control discs the higher expression of sqh-GFP is not observed (Fig. 4F-F999). These results suggest that when the A and P compartments have different Ubx expression, the higher levels of myosin II at the Ubx 2/ Ubx + border may contribute to the maintenance of the A/P boundary in the absence of Hh signaling.
Although haltere and wing imaginal discs are physically separated throughout development, cells of different discs (as wing and haltere discs) form a continuous layer of adult cuticle during pupation. The different affinities in imaginal discs provided by Hox genes may prevent mixing of cells when this fusion takes place. In agreement with this idea, adult flies defective for Ubx occasionally present abnormal contralateral fusion of the T2 and T3 segments (Fig. 4G; G. Morata, personal communication).

Different Hox Genes can Induce High sqh Levels
To see if other Hox genes may also induce cell segregation, we induced Abd-B mutant clones and look for myosin expression in the genital disc, where this gene is required [40]. As shown in Fig. 5A, A9, increased expression of sqh is observed surrounding these clones, but not in control Abd-B clones induced in the wing disc, where Abd-B is not required (Fig. 5B, B9). Consistently with these results, the expression of Abd-B in the dorsal compartment of the wing pouch is sufficient to maintain a straight D/V boundary when N signaling is compromised ( Fig. 5C; n = 11). Moreover, the expression of an onycophoran Ubx (UAS-OUbx) [26] is also sufficient to maintain this boundary ( Fig. 5D; n = 12). Taken together, the results suggest that myosin may have a role in the segregation of cells with different Hox activity.

Hox Genes and Cell Segregation
It was previously proposed that Hox genes confer different affinities to cells [41]. Studies in cultured Drosophila imaginal discs showed that cells from wildtype haltere and wing disc segregate, but that Ubx mutant haltere discs mix with wildtype wing discs [23]. We have shown here that Ubx is sufficient to maintain to a large extent the A/P and D/V boundaries in the absence of Hh and N signaling (see Fig. 6 for a summary of results). The mechanism whereby Ubx sorts out cells may be similar to that used by Hh and N signaling, and probably involves the accumulation of myosin where cells expressing and lacking Ubx are juxtaposed. This prevents territories with different properties to mix freely, and helps to get coherent patches of cells with distinct fates.
The sorting out of cells with distinct Hox activity in Drosophila has been reported before [24,33,40,[42][43][44], and in the case of the Hox gene Deformed a possible function in cell segregation has been assigned to such activity [24]. We have observed some cases that show that Ubx is needed to maintain segregation of cells from different segments during pupation. It is possible that Drosophila Hox genes may have a function in cell segregation during this pupal stage, where cells from different discs and histoblast nests fuse to develop the adult cuticle. The mechanism of segregation seems to rely on the confrontation of cells with different Hox function and not on the absolute levels of Hox expression. This implies that Hox activity in neighboring cells may be checked through proteins at the cell membrane whose expression or levels must be controlled by Hox genes. In the embryo, the Hox gene Abd-B has been shown to regulate molecules like cadherins [45], and such proteins may mediate segregation between adjacent cells with distinct Hox input.
In vertebrates, cells from different rhombomeres are also almost completely prevented from freely mixing [17]. As we have shown here for Drosophila, it has been proposed that the tension provided by the activity of actomyosin molecules, controlled by Hox genes, could prevent mixing of cells in the vertebrate's rhombomeres [46]. Hox-directed cell segregation, therefore, prevents cells with different Hox code to intermingle, and therefore the appearance of homeotic transformations. This function of Hox genes may be an old one in evolution, required in animals in which development of different body regions is not coupled to the mechanisms of segmentation [47]. In Drosophila, this role of Hox genes may not be needed in cells that are physically separated during most of development (as in imaginal discs and histoblasts from different segments) or superseded by the activity of proteins like Engrailed and Hedgehog, but the maintenance of different affinities by Hox genes and signaling pathways through myosin accumulation may be a general mechanism to segregate cell populations in different species.