Intermicrotubular actin filaments in the transalar cytoskeletal arrays of Drosophila

Filaments with diameters of 6 nm are concentrated within bundles of microtubules in the epidermal cells of Drosophila wing blades during the final stages of pupal development. Filament diameters correspond to those of actin microfilaments but the square cross-sectional profiles that they exhibit in high-resolution transmission electron micrographs are similar to those found for certain intermediate filaments (Tucker et al. 1986). This study was undertaken to ascertain whether the filaments bind the SI subfragment of myosin (SI) in the arrowhead decoration pattern that is generally characteristic of actin filaments, and if this proved to be the case, to assess filament polarities. These two issues are of importance in terms of filament function and assembly. The filaments are aligned with microtubules that are somewhat unusual (most of them are composed of 15 rather than 13 protofilaments). The filaments and microtubules are the main components of cell surface-associated transcellular bundles, which span the apicobasal axes of cells and occupy the basal extensions of the cells. Pairs of bundles form transalar cytoskeletal arrays that link dorsal and ventral wing surfaces v/a sets of basal attachment desmosomes. Bundles seem to have a supportive rather than a contractile function (Johnson & Milner, 1987). Are the filaments really composed of actin? Assembly of most of the microtubules is apparently nucleated at apical plasma membrane-associated sites (Mogensen & Tucker, 1987). If the filaments are composed of actin, is the orientation of SI arrowheads and filament polarity consistent with filament elongation from cell apices?

Developing wings of Drosophila melanogaster (Oregon S) were dissected from pupae (that had been maintained at 25 °C) after immersion of pupae in a Drosophila tissue culture medium and at a point 87 h after the start of pupariation as described (Tucker et al. 1986). At this stage two or three small cuts (100-150 μm long) were made in each wing with tungsten needles to facilitate penetration by solutions. Wings that had been freshly isolated in this way were prepared for electron microscopy as in an earlier study (Mogensen & Tucker, 1987).

Decoration with SI was carried out using the procedure described by Schliwa & van Blerkom (1981) with slight modifications. The freshly isolated wings were rinsed for 30 s in PHEM buffer (60mM-Pipes, 25 mM-Hepes, 10mM-EGTA, 2mM-MgC1₂, pH 6·9) and then immersed in a solution of 1% Triton X-100 in PHEM for 20 min at room temperature. After rinsing in PHEM (30 s) wings were incubated for 2 h at 24°C in a solution (pH 6’2) containing rabbit muscle myosin subfragment SI (3·3mgml⁻¹) (Sigma Chemical Co.), 50% glycerol, 0·5M-KC1 and 0·025 M-K2HPO4. After two 30-s rinses in PHEM (pH 6·9) specimens were fixed for 30 min in a solution containing 2% glutaraldehyde and 2% tannic acid in PHEM (pH 7·4).

Material destined for embedding in Araldite was then rinsed for 30 s in PHEM (pH 7·4) and postfixed for 30 min in a solution of 1% OsO₄ in PHEM (pH 7·4) before dehydration with ethanol. Wings that were to be embedded in the water-soluble melamine resin Nanoplast FB 101 (Kit A204S, Polaron Equipment Ltd, Watford, UK) were rinsed for 30s in distilled water after fixation with the glutaraldehyde-containing solution. Block staining was effected at this point by incubating wings at 40°C for 17 h in an aqueous solution of 1% uranyl acetate. Then the undehydrated wings were flat embedded in a mixture of resin (50 parts) and catalyst (1 part). Curing was conducted for 48 h at 40 °C in a sihca-gel-containing desiccator before final hardening for 48 h at 60°C.

Extraction of epidermal cell components was progressively less advanced at increasing distances from the cuts that had been made in wing blades (Fig. 1). Cuts evidently provide a conduit through the impermeable (adult) cuticle for the permeabilizing detergent and other reagents in solution, which were applied before fixation. The cuticle is fairly substantial at the relatively late stage of wing morphogenesis that has been examined (87 h after the start of pupariation). Cuticle on the dorsal and ventral surfaces of a wing blade effectively sandwiches the epidermal bilayer, which forms a cellular filling. In unextracted wings, each cell contains a large transcellular bundle of microtubules (900 on average) and 6 nm filaments that are mainly concentrated between the microtubules (Fig. 2) (Mogensen & Tucker, 1987). Bundles are relatively resistant to extraction compared with other epidermal cell components. Most of them remain oriented at right angles to the plane of the adjacent wing surface (Figs 1, 3). In most cases the basal end of a bundle is still associated with the partly extracted basal attachment desmosome complex that connects it to the base of a bundle in the epidermal layer on the opposite side of a wing blade (Fig. 4). The apical ends of bundles remain positioned near the apical (cuticle-secreting) surfaces of cells and apparently retain connection to hernidesmosomes, which anchor them to cell apices. Hence, pairs of transcellular bundles still form transalar cytoskeletal arrays that span the thickness of a wing blade even in the immediate vicinities of cuts (Fig. 1).

SI decorates filaments in bundles that are situated within about 40 μm of cuts. Material with a tweedy herringbone-like pattern is apparent in between microtubules and aligned alongside them in longitudinal sections of such bundles (Fig. 4). Confirmation that the filaments are composed of actin was obtained from examination of control preparations, which showed that the intermicrotubular decorated material described above was not present if pyrophosphate (10mM) was included in the SI-containing solution. Cross-sections of bundles in regions of wings where longitudinal sections reveal that decoration of filaments has occurred show a densely stained reticulum of material that is situated in the spaces between microtubules. Most of this reticulum probably represents the cross-sectional profiles of decorated fila-ments, since a substantial reticulum is lacking in the control preparations. The reticulum is mainly confined to the spaces between microtubules and often it is not concentrated right around the microtubules flanking the sides of bundles (Fig. 5, arrows); in this respect its distribution is similar to that of the 6 nm filaments in the unextracted cells of freshly isolated wings (compare Figs 2, 5).

Distinct head-to-tail arrays of arrowheads were rarely apparent in the decorated preparations. Only eight examples of portions of decorated filaments in which polarity could be distinguished were found during a search that included over 100 longitudinal sections of bundles with decorated filaments. In all cases arrowheads were oriented with their sharp ends pointing basally (Fig. 6) (i.e. away from the cuticle-secreting ends of cells); these examples were found at all levels within bundles (some closer than 1 μm to the basal attachment desmosomes).

The detergent-extracted preparations used in this investigation also showed that the basal ends of longitudi-nal profiles of microtubules sometimes appear to termin-ate abruptly (rather than to be passing gradually out of sections) at points where a discrete concentration of dense material fills their cores (Fig. 4, short arrows). These closed or plugged microtubule ends may have escaped detection in earlier studies because they were masked by the dense filamentous and desmosomal ma-terial that is concentrated at the bases of transcellular bundles (Tucker et al. 1986). Extraction of desmosomes,release of microtubule ends from their desmosomal anchorages, wider spacing of microtubules and actin filaments, and some relaxation of the highly folded configuration of the basal attachment desmosome com-plexes promoted by detergent treatment may all have helped to reveal the microtubule ends in the present examinations.

Wings were embedded in Nanoplast resin in an attempt to improve visualization of arrowheads and filament polarity. This resin has been used to obtain a high degree of structural resolution for cross-sectional profiles of insect muscle filaments (Westphal & Frosch, 1984). Image contrast in thin sections was very low (unlike the situation in the study of muscle cited above), unless sections were double stained with lead citrate and uranyl acetate in the same manner as for specimens embedded in an epoxy resin. Transcellular bundles have a substantially different appearance in wings embedded in Nanoplast from that described above for wings embedded in epoxy resin. In the vicinity of cuts, image contrast is lower and the pattern of SI decoration is less distinct than is the case after embedding in Araldite. However, image contrast is greater in portions of wings sited more distantly from cuts (more than about 100 f.tm away) where detergent extraction of epidermal cells is less marked and SI decoration of filaments does not occur (presumably because of failure of the SI subfragments to gain access to filaments). In these wing portions the protofilamentous structure of microtubules is particularly distinct (Fig. 7). Thus, it seems that if extraction of cells and their components has been substantial then discrimination of structural detail in cytoskeletal fibres fixed in the presence of tannic acid is reduced compared with that found in less extensively extracted regions. Since cell extraction and permeabilization is essential for access of the SI probe we have reached an impasse so far as this particular approach to assessment of actin filament polarity is concerned.

The improved discrimination of microtubule protofilaments after embedding in Nanoplast is a valuable advance that has also been noted by Westphal & Frosch (1986). Protofilaments were not distinct in the vast majority of cross-sectional profiles of microtubules in the wings used in this study (Fig. 5), or an earlier one (Tucker et al. 1986), that had been fixed in the presence of tannic acid and embedded in Araldite. The yield of profiles with unequivocally countable protofilaments around entire microtubule walls is increased in Nano-plast-embedded material to the extent that protofilament number can often be monitored for several microtubules within a section of part of a bundle (Fig. 7), whereas in our earlier study searches of many sections of many bundles were required to find unequivocally countable profiles.

Decoration with SI reveals that the transcellular fila-ment/microtubule bundles in late pupal Drosophila wing epidermal cells contain large quantities of F actin. Are actin filaments concentrated between the microtubules in vivo, or do the decorated filaments described here represent re-located actin derived from elsewhere in the cells that has accumulated around the microtubules during preparation? The decorated filaments are concentrated throughout the interiors of the bundles and are not more markedly aggregated around microtubules at the sides of bundles. This coincides with the arrangement of 6 nm filaments in the bundles of unextracted cells and hence these filaments probably represent the original source of the decorated actin.

Decoration indicates that most actin filaments may be oriented with their pointed ends directed apicobasally with respect to the longitudinal axes of transcellular bundles (Fig. 8). Actin filaments with opposing polarities of SI arrowhead decoration have been demonstrated in the chinchilla organ of Corti where they are concentrated between microtubules to form transcellular bundles in certain supporting cells (Slepecky & Chamberlain, 1983). The very small sample of filaments found exhibiting detectable arrowhead polarity in the present study does not rule out the possibility that intermicrotubular actin filaments with opposite polarities are also reasonably common in the transcellular bundles of Drosophila. A predominant apicobasal orientation of arrowheads may be an indication that many of the filaments start to assemble at cell apices as a consequence of nucleation and monomer addition to their barbed plus ends at or near the cuticle-secreting surfaces of the cells in the manner proposed for certain other cell types (Tilney et al. 1981; Mooseker et al. 1982). This parallels evidence that assembly of the microtubules is also initiated at cell apices. The microtubules apparently have a polarity that is opposite to that of most of the actin filaments and elongate from plasma membrane-associated nucleating sites in an apicobasal direction, with their plus ends pointing basally during bundle assembly (Fig. 8) (Mogensen & Tucker, 1987).

The actin filaments may not be employed as part of an actomyosin system to promote active contraction of the transcellular microtubule bundles in which they are situated. The bundles do not seem to be actively involved in the final flattening and thinning of wing blades that occurs shortly after eclosion; apparently, they are simply dismantled to permit these events to proceed. Epidermal lysis occurs (except in the vicinities of wing veins) as the dorsal and ventral cuticular layers flatten out and become closely juxtaposed so that the space left by demolition of the lysed epidermal bilayer is virtually eliminated (Waddington, 1941). Fine-structural examination of the procedure has shown that cell lysis and fragmentation of the microtubule/actin bundles are initiated before flattening of the previously highly folded wing blade has advanced significantly, so that it has been suggested that the pairs of dorsoventrally associated bundles act as transalar ‘pitprops’ (Johnson & Milner, 1987). The role of the bundles seems to be that of acting as rigid pillars to maintain the basal cell extensions and hence keep open the intercellular haemocoelic space between them and the two layers of epidermal cell bodies during the final stages of pupal wing morphogenesis while wing blades are highly folded and are increasing in surface area. If these suggestions are correct then the actin filaments are serving to provide a stiff intermicrotubular matrix and may perhaps be crosslinked to each other and the microtubules to reduce shearing within bundles. The actin filaments could also be employed, perhaps in association with the material that plugs microtubule ends, as part of a system that anchors microtubule tips to the basal desmosomes. The basal desmosome complexes may be more closely related to adherens junctions in the category that provides membrane anchorage sites for actin-containing microfilaments such as fasciae adhérentes (Geiger et al. 1987), rather than to the intermediate filament-associated true desmosomes (maculae adhérentes) of vertebrates (Franke et al. 1987).

There are some other situations where intermicrotubular actin filaments may function in a very similar fashion to that proposed above. Epidermal muscle attachment cells in a range of arthropods exhibit an ultrastructural phenotype that is very similar to that of the wing epidermal cells studied here. They possess transcellular bundles of microtubules that are attached to apical hemidesmosomes and basal attachment desmosomes (Poodry, 1980). It has been shown that the microtubules are mainly composed of 15 protofilaments in one instance (Nagano & Suzuki, 1975) and concentrations of 7 nm filaments between microtubules have been noted in several cases (e.g. see Lai-Fook, 1967). Presumably, these filaments are composed of actin and in this case also there is no reason to suppose that bundles contract; the probable role of the ‘tendinous’ bundles is to transmit tension (rather than to resist compression as seems to be the case for wing cells). Furthermore, bundles of actin filaments (which do not include microtubules) seem to serve as non-contractile intracellular tendons in the ctenophore Beroë (Tamm & Tamm, 1987). The design of wing epidermal cells also exhibits some marked similarities to that of certain supporting cells (Dieters’ cells and pillar cells) in the mammalian organ of Corti. These cells possess transcellular bundles of 15 protofilament microtubules (Saito & Hama, 1982) and intermicrotubular actin filaments, which may, perhaps, modulate cell length (Slepecky & Chamberlain, 1983). However, as in wing cells, the prime role of the bundles is apparently to act as stiff supportive pillars (and anchor the sensory hair cells to which they are joined by desmosome-like junctions). Several forms of motility are associated with the framework of colinear microtubules and actin filaments in the amoeba Reticulomyxa. There are indications that actin-based contractility may mediate one of these (Koonce & Schliwa, 1986).

Support from the Science and Engineering Research Council (UK) (grant no. GR/D/00733) is gratefully acknowledged.