Apical cell shape changes during Drosophila imaginal leg disc elongation: a novel morphogenetic mechanism

The shape of most multicellular animals is determined largely by changes in the dimensions and topography of epithelial sheets during development (Trinkaus, 1984; Fristrom, 1988). Epithelial tissues can change shape in two general ways: symmetrically, such that the proportions of the tissue are not changed and asymmetrically, such that one dimension (e.g. length, width or circumference) changes size relative to the other dimensions of the tissue. Symmetric changes in epithelial surface area often result from changes in apical-basal cell height (cell flattening or columnarization). Altering the relative proportions of an epithelium, however, cannot result from symmetric changes in cell area. One means of changing the proportions of an epithelial sheet is by changing the arrangement or packing of cells within the epithelium. Some well-studied examples of epithelial cell rearrangement during morphogenesis include elongation of the neural plate (Jacobson, 1981), convergent extension in amphibian gastrulation (Keller, 1978) as well as archenteron elongation in sea urchins (Hardin, 1989). Alternatively, the relative proportions of an epithelium can be altered by changing the shape of the component cells. A clear example of cell shape changes during epithelial morphogenesis is seen in the nematode, where the epidermal cells contract circumferentially and extend longitudinally to generate the final body shape of the worm (Priess and Hirsh, 1986).

During metamorphosis of Drosophila, the external structures of an adult fly are generated from imaginal discs and histoblasts. Imaginal epithelia undergo elaborate changes in shape in response to the steroid hormone 20-hydroxyecdysone, and employ a variety of morphogenetic mechanisms in the process (Fristrom, 1988). During the early stages of imaginal leg disc morphogenesis, the proportions of the disc change dramatically; the leg elongates in the proximal-distal axis, and constricts circumferentially. Although genes likely to be involved in metamorphosis have been identified (Spencer et al. 1982; Natzle et al. 1986, 1988; Osterbur et al. 1988; Brower and Jaffe, 1989), a detailed description of the cellular basis of disc elongation has been hampered by the small size of imaginal cells, extensive apical and basal extracellular matrices, and the convoluted geometry of the tissue (Fig. 1A). Because previous research did not reveal significant changes in cell shape during leg disc elongation, this process has been thought to be mediated by cell rearrangement (Fristrom, 1976). In the work reported here, the significant increase in resolution afforded by scanning confocal laser microscopy has provided a new perspective on the cellular mechanism of leg disc elongation.

We used an Oregon R strain maintained in our laboratory since 1964 and reselected for fecundity in 1986. Imaginal discs were dissected from larvae and prepupae reared at 25 °C on standard cornmeal-molasses medium supplemented with yeast paste. Prepupae were staged from puparium formation (white prepupal stage). Third instar larvae were staged from hatching (±2h). Leg discs were dissected at room temperature in modified Ringer’s (130mM NaCl, 5mM KC1, 1.5mM CaCl₂, 5mM TES, 0.2% BSA) and fixed 2h in 4% formaldehyde in PBS at 4°C. For trypsin treatment, leg discs were dissected from white prepupae and exposed to 0.01 % trypsin (TPCK treated; Worthington) in Robb’s culture media for 15 min. Prior to fixation, the peripodial epithelium was dissected from both control and enzyme-treated discs. For phalloidin staining, fixed discs were permeabilized with 0.5 % Triton X-100 in PBS (l h), stained 2–12h in 4% formaldehyde containing 0.3 μM fluorescein-phalloidin (Molecular Probes), rinsed 2h in PBS-Triton, then mounted in 1:9 glycerokPBS containing 0.15% Hanker-Yates reagent (Sigma), and viewed as whole-mounts with a BioRad MRC500 scanning confocal laser microscope (SCLM).

Measurements for Table 1 were made from SCLM images of Fl l C-phalloidin stained whole-mounts of second leg discs at 0 and 6h AP using BioRad MRC500 and Image 1.2 image analysis software. Measurements were confined to the presumptive basitarsus and distal tibia. Because cell boundaries are only seen in those parts of the tissue that he more or less parallel to the plane of the optical section, cells cannot be clearly viewed around the entire circumference of a segment in a single specimen. By making measurements on a number of differently oriented specimens, the entire circumference of these segments was sampled thus giving a much more comprehensive measure of cell shapes than was obtained using SEM (Fristrom, 1976). Estimates of cell dimensions were made by counting the cells intersecting sets of measured Unes drawn parallel to proximal-distal and circumferential axes. The proximal-distal length of the segment was measured along the dorsal midline from a longitudinal section. The circumference of the segment was determined by approximation to an ellipse using the dimensions of the segment measured at the proximal-distal midpoint. The mean number of cells in the length and circumference of the segment was estimated by dividing the segment dimensions by the average cell dimensions in each axis.

Filamentous actin in whole-mounted imaginal discs was stained with fluorescein-conjugated phalloidin (FITC-phalloidin) and viewed by scanning confocal laser microscopy (SCLM). The cortical actin filaments associated with the zonula adherens (Fristrom, 1988) stained particularly intensely and serve to define the apical cell shapes (Fig. 1B). Intense signal was also seen at the basal cell surface, another area of well-defined actin filaments (Fristrom and Fristrom, 1975). Lateral cell boundaries were marked by weak staining suggesting that actin is also present in these areas. No other intra- or extra-cellular components were stained by phalloidin. The ability of FITC-phalloidin to outline each epithelial cell combined with the optical sectioning power of SCLM enabled us to examine the threedimensional shapes and arrangements of cells throughout imaginal discs at all stages of development.

Our most striking observation was that the apical profiles of cells in the dorsal-lateral regions of late third instar discs are highly asymmetric (Fig. 1B). These circumferentially elongated cells are oriented such that the long axis of the cell is perpendicular to the eventual axis of disc elongation. Our previous observations using scanning electron microscopy (SEM) (Fristrom, 1976) had not revealed anisometric apical profiles because they occur in regions of the disc that are covered by apical extracellular matrix (Fig. 1A). In the ventral quadrant of the disc where apical cell profiles can be seen by both SEM and SCLM (e.g. between arrows in Fig. 1A and B), cells are isometric in shape. Finding highly anisometric cells was surprising because apical cell profiles in epithelial sheets are generally isometric due to the equal tension exerted on all cells by their neighbors (Honda et al. 1984; Keller and Trinkaus, 1987).

Comparing optical sections from different focal planes revealed that cell shape varies between apical and sub-apical profiles of cells in the same region of the disc (Fig. 2). Cells were most anisometric in their apical-most 1– 2 microns, in the region of the zonula adherens (Fig. 2A-C). Below this level cells were elongated, but the elongation was less pronounced (Fig. 2D-F). This raised the question of how cells are packed within the epithelium. Optical sections were taken at 0.5 or 1 gm steps from the apical surface in regions of apically elongated cells (dorsal basitarsus, tibia) at three different developmental stages. Examination of individual cells and their neighbors through their apical-most 5–6 gm revealed that some cells have different sets of neighbors basally and apically. This indicates that the apical and basal ends of cells are packed differently, presumably to accommodate their different dimensions. In contrast, in regions where cells are isometric apically, their shape and packing are the same apically and basally (see also Fig. 9).

To determine when in disc development anisometric cells first arise, we examined FTTC-phalloidin-stained discs dissected at 6h intervals throughout the third instar (48-96h after hatching (AH); Fig. 3). There are at least six rounds of cell division during this period, and the discs increase in size substantially (Nôthiger, 1972; Bryant, 1987). Circumferentially elongated cells were first observed in the early third instar (54 h AH), prior to the formation of distinct folds in the disc epithelium (Fig. 3A) and persisted as the discs increased in size by cell division (Fig. 3). Large round profiles of mitotic cells were frequently observed in fields of elongated cells (Fig. 3A,B). By 60h of development, the columnar epithelium of the disc proper has begun to fold, and elongated cells were conspicuous in the presumptive proximal and dorsal regions of the disc (Fig. 3B). By mid third instar (72 h AH), the characteristic folds of the mature disc have been established, and large fields of elongated cells were observed in the periphery (presumptive proximal regions) of the basitarsus, tibia and femur (Fig. 3C).

We were particularly interested in the possible contribution of changes in cell shape to leg morphogenesis. Rapid elongation of leg discs occurs at the onset of metamorphosis, between 0 and 6h after puparium formation (AP). During this period, the regions of the disc that give rise to the basitarsus, tibia and femur, elongate proximally-distally and constrict circumferentially. In contrast, the tarsal segments distal to the basitarsus do not change dimensions substantially (Fig. 4). At Oh AP (Fig. 5A), elongated cells were observed in the basitarsus, tibia and femur (basaitarsus shown in Fig. 5D). Asymmetric cells were oriented circumferentially; i.e. with their long axis perpendicular to the axis of disc elongation. Six hours later (Fig. 5B), elongated cells had become isometric (Fig. 5E). During morphogenesis, as cells became isometric both apically and subapically, the difference in cell packing along this axis also decreased. Sequential optical sections of the basitarsus at 6h AP (comparable to those made at Oh AP; Fig. 2) showed that isometric cells contact the same set of neighbors throughout their apical-basal extent (not shown).

Mild trypsin treatment of Oh AP discs (Fig. 5C) accelerates appendage elongation (Poodry and Schneiderman, 1971; Fekete et al. 1975). Trypsin-accelerated elongation was accompanied by the rapid (15 min) conversion of anisometric cells to isometric ones (Fig. 5F), suggesting that changes in cell shape drive leg elongation. The dependence of appendage elongation on changes in cell shape is also supported by observations on mutants (broad and Stubble) in which leg discs fail to elongate properly and disc cells do not change shape. In 6h AP Stubble discs, the basitarsus and tibia have not extended normally, and cells in these segments are highly anisometric (Fig. 6). In the distal tarsal segments, which do not elongate between 0 and 6h AP in either mutant or wild-type discs (Fig. 4, Fig. 5), cells are isometric throughout this period (Fig. 6).

We estimated the contribution of cell-shape changes to disc morphogenesis by measuring cell and tissue dimensions in the basitarsus and the distal tibia; the two leg segments that show the greatest elongation between Oh and 6h AP. The length of the segment along the dorsal midline and the circumference of the segment at the proximal-distal midpoint were measured before and after elongation (Fig. 4). The average cell dimensions in both axes were also measured (see Materials and methods). In both the basitarsus and the tibia, the changes in tissue length and circumference paralleled the changes in cell dimensions (Table 1). Estimates of the number of cells in the circumference and long axis of the tissue indicated that there is little or no cell rearrangement between Oh and 6h AP. Moreover, the agreement between the ratios of cell shape and tissue shape changes confirms that changes in cell shape are sufficient to account for most of the tissue elongation and constriction observed in both segments during this period.

Elongated cells are not a prominent feature of all imaginal discs. Anisometric cells were frequently observed in leg and antennal discs, but occurred much less frequently (or not at all) in wing and haltere discs (Fig. 7). The morphologies of the various disc types reflect, at least in part, the different morphologies of their constituent cells. The initial and final shapes of leg and antennal discs are similar; both are short, fat cylinders that become elongate, narrow cylinders during metamorphosis (Poodry and Schneiderman, 1970; Postlethwait and Schneiderman, 1971). In contrast, wing and haltere discs are initially shallow domes that expand in area during metamorphosis and, in the case of wing discs, flatten to form an epithelial bilayer. Neither wing nor haltere discs elongate as much as leg discs. One expects distinct morphogenetic mechanisms to be used by tissues with radically different initial and final shapes. For example, cell flattening during morphogenesis (also a form of apical cell shape change) is more pronounced in wing discs than in leg discs (Waddington, 1940; Fristrom and Fristrom, 1975), and adhesion of basal cell processes is crucial for the correct flattening and apposition of wing blades, yet plays no role in the morphogenesis of leg discs (Waddington, 1940; Fristrom and Fristrom, 1975; Wilcox et al. 1989). The conversion of anisometric cells to isometric shapes may be specifically involved in the elongation and circumferential constriction of an epithelial tube.

In addition to identifying anisometric cell shapes, SCLM of phalloidin-stained discs has revealed many highly reproducible cell patterns in specific regions of the leg disc. Some of these are mentioned here because they may provide useful markers for studying other aspects of disc development. The most striking of these patterns was seen at the tip of the tarsus (Fig. 8A). A circular region enclosed by a ring of approximately 40 cells is bisected by a line of cells that extends down the side of the tarsus and corresponds in position to the anterior-posterior compartment boundary. On either side of this boundary, the cells are organized into whorls of very small cells. A very similar pattern has also been observed from the basal surface in silver-stained preparations of leg discs (Larsen and Zom, 1989). Other striking patterns were seen in the presumptive tibia (Fig. 8B). Triangular patches of cells (arrow) at the junction of the tibia and basitarsus occur at the approximate point where the two tibial apodemes will form after pupation (Poodry, 1980). Whorls of cells such as the one shown in Fig. 8A were seen in consistent locations in both leg and wing discs. The locations of some of these whorls correlate with positions where sensillae will differentiate (Cole and Palka, 1982; Jan et al. 1985; Held, 1990).

During leg disc elongation, the dimensions of the tissue change significantly; the circumference of the disc decreases and the length greatly increases. Our previous view of prepupal (0–6 h AP) leg disc morphogenesis supposed that the appendage elongation initiated by 20-HE occurred by a process of active cell rearrangement whereby cells intercalated longitudinally to increase the number of cells in the long axis and decrease the number of cells encircling the disc. This conclusion was based on SEM observations of evagi-nated discs and limited regions of unevaginated discs which revealed only isometric cells. The possibility that cell elongation plays an important role in appendage elongation has previously been suggested (e.g. Fristrom el al. 1969; Mandaron et al. 1977), but was not well supported by the published data. Our current estimates suggest that changes in cell shape account for most of the disc elongation that occurs in prepupae. Imaginal leg discs undergo considerable morphogenesis both before and after the period we have examined, and the contribution of cell shape changes and other morphogenetic processes (e.g. cell rearrangement and cell flattening) to early and late disc morphogenesis has not been determined.

The changes in cell shape and basal cell packing that occur during leg disc morphogenesis are summarized in Fig. 9. Prior to disc elongation, cells are highly anisometric, such that their long axis is perpendicular to the eventual axis of disc elongation. Cells are elongated throughout their entire apical-basal extent, but cell elongation is most pronounced apically. This difference in cell shape could be due to any number of factors: the forces responsible for generating anisometric cell shapes may be more pronounced at the level of the apical junctions or basal cell shape may be less easily distorted due to the rigidity of the cell nucleus. To accommodate the difference in apical and subapical cell shape without folding the epithelial sheet, the basal ends of cells are repacked relative to their apices (Fig. 9A). During disc elongation, apical and basal cell shapes become isometric, resulting in a decrease in the circumference of the tissue, and a concomitant increase in its length. Once the cells have assumed isometric proportions, we no longer observe differences in cell packing along the apical-basal axis (Fig. 9B). Since we do not observe significant apical cell rearrangement during this period (Table 1), it is likely that the basal portions of the cells repack to come into register with their apical cell contacts.

Surprisingly, elongated disc cells arise well before the onset of morphogenesis and persist in tissue that is still undergoing mitosis (Fig. 3). Although anisometric cell shapes are observed transiently during the morphogenesis of some epithelial tissues (Keller and Trinkaus, 1987) and in a few terminally differentiated epithelia (Wright and Lawrence, 1981; Priess and Hirsh, 1986), a mitotically active population of elongated epithelial cells has not been described. The early appearance of anisometric cells indicates that the final shape of the leg is ordained well before the end of the third instar. Moreover, discs exhibit a variety of characteristic local patterns of cells (rows, whorls and clusters) prior to the onset of hormone-induced morphogenesis (Fig. 8). Fate maps of imaginal discs indicate that disc cells are genetically determined during the third instar (Schu-biger, 1968). Here, in the distinctive and reproducible patterns of cell shape, size and arrangement, we see the first cellular expression of these programmed differences in cell fate.

The mechanisms by which anisometric cell shapes are established and maintained during imaginal disc development is unknown. Oriented cell divisions that add cells predominantly to the radial (future proximal-distal) axis of the disc (Bryant, 1969) could contribute to circumferential cell elongation by generating circumferential stress as the disc increases in size. However, the observation that cutting discs in half, thereby relieving any tissue-wide compression or tension exerted on cells, does not result in immediate disc elongation (Schu-biger, 1968) suggests that some cellular mechanism is required to maintain elongated cell shapes. Moreover, discs that have been isolated from the larvae (and thereby from any forces generated outside of the discs themselves) and isolated disc fragments (Fristrom and Chihara, 1978) will elongate normally in vitro, suggesting that local forces in discs and/or cell autonomous mechanisms act to control cell shape both before and during tissue elongation. In C. elegans, elongated hypodermal cell shapes are generated by a circumferential contraction of microfilaments, and subsequently stabilized by extracellular matrix (Priess and Hirsh, 1986). Any mechanism employed by imaginal leg discs to maintain anisometric cell shapes must be dynamic with respect to cell division; allowing cells to round up, divide, and resume an elongated form.

We have frequently observed rounded apical profiles, characteristic of epithelial cell mitoses, in regions of the disc that otherwise contain only elongated cells (Figs IB, 3B, 3C). This indicates that after each mitosis the isometric daughter cells reestablish an elongated shape. In doing so, cells may establish new apical contacts with neighbors located lateral to the position of their nucleus at the time of division. In discs as in other epithelia, the zonulae adherens and septate junctions that maintain the contiguity of the epithelial sheet are located in the apical-most 2–3μm of the cell. Once junctional contact has been established between adjacent cells, the junctions can rapidly accommodate changes in the extent of that contact (Fristrom, 1982). As disc cells return to isometric dimensions during metamorphosis, elongation and circumferential constriction of the leg occur without requiring any alteration in apical neighbor relations.

At metamorphosis, elongated cells become iso-metric, decreasing the circumference and increasing the length of the tissue. Cell shape change may simply involve the release of whatever constraints have been imposed on cell shape during disc development. Leg elongation, however, is a cytochalasin-B-sensitive, energy-dependent process that presumably requires intact actin filaments. A role for the active contraction of the circumferential microfilaments associated with the zonula adherens seems likely (Owariba et al. 1981). Whether these filaments exist throughout the third instar in a state of tension or whether they are induced to contract as part of the response to ecdysone remains to be determined. As apical cell dimensions change, basal cell adhesions must loosen to permit the basal ends of cells to come into register with the apical ends. Repacking of the basal ends of cells could be a relatively

‘passive’ response to changes in cell shape, or could in part ‘drive’ cell shape changes. This latter possibility would require both directed migration of the basal ends of cells in the long axis of the tissue, and some additional mechanism for translating basal cell position into an alteration of overall cell shape. Alternatively, and we feel more likely, a circumferential contraction of apical actin filaments could convert anisometric cells to isometric dimensions without requiring positional information to be available to the cells during disc elongation.

Trypsin could play several roles in accelerating disc elongation. Trypsin removes the basal lamina from discs and loosens basal intercellular adhesions (Poodry and Schneiderman, 1971), which may facilitate the repacking of the basal ends of the cells during leg elongation. Trypsin also removes apical extracellular matrix (Fristrom, unpublished) which may release extracellular constraints involved in maintaining anisometric cell shapes. There is an intriguing correlation between the distribution of apical extracellular matrix and the distribution of anisometric cell shapes in late third instar imaginal discs (Fig. 1). Moreover, trypsin must have access to the apical surface to be effective in accelerating leg elongation (Birr et al. 1990). Elimination of cellular contacts to the apical matrix, perhaps by endogenous proteolysis, may be required for normal elongation. Consequently, it is interesting that imaginal discs release proteases when exposed to 20-HE in vitro (Pino-Heiss and Schubiger, 1989) and that at least one disc apical cell-surface protein is proteolytically cleaved during elongation (Birr et al. 1990).

The cellular mechanism of disc morphogenesis presented here invokes striking changes in cell shape that primarily affect the apical surface of the epithelium. This suggests that the apical surface, including the zonula adherens and apical actin filaments, is particularly important in disc morphogenesis. Our results showing that many ecdysone-induced molecular changes also affect the apical surface (Birr et al. 1990; Paine-Saunders et al. 1990) further emphasize the importance of the apical surface in disc morphogenesis.

We thank Drs H. J. Yost, R. Keller and others for their critical input on this manuscript. Thanks also to Janet Duerr for technical advice and to George Oster for many useful discussions and help with Fig. 9. M.L.C. is the recipient of a postdoctoral fellowship from the American Cancer Society. This work was supported in part by a USPHS grant (GM-19937).