The furry gene of Drosophila is important for maintaining the integrity of cellular extensions during morphogenesis

The adult cuticular surface of Drosophila is decorated with large numbers of polarized structures such as sensory bristles and epidermal hairs. The development of these structures has been actively studied as a model for the morphogenesis of highly polarized cellular components (Tilney et al., 2000b; Turner and Adler, 1998). Not surprisingly, these structures contain prominent cytoskeletal elements. Developing hairs and bristles contain actin filaments and microtubules and the function of both is required for normal morphogenesis (Eaton et al., 1996; Turner and Adler, 1998; Tilney et al., 1995).

The large size of the sensory bristle shaft cells has made them an attractive system for microscopic observations (Tilney et al., 1995). Developing bristles contain a series of large bundles of actin filaments located peripherally and juxtaposed to the plasma membrane (Tilney et al., 1996). The bundles are assembled in modules at the distal tip of the shaft. There appear to be two crosslinking systems that are used to assemble the large bundles (Tilney et al., 1995; Tilney et al., 1998). A primary bundling requires the function of the forked (f) gene (Petersen et al., 1994) and a second crosslinking requires the function of the singed (sn) gene (which encodes a fascin; Cant et al., 1994; Bryan et al., 1993). Mutations in either of these genes results in short, twisted and bent bristles. In sn f double mutants, the large bundles of actin filaments are lost, although a small number of actin filaments remain juxtaposed to the plasma membrane (Tilney et al., 1998). Interestingly the phenotype of the double mutant is not substantially stronger than either single mutant. Hence, the large bundles of actin filaments are not essential for bristle outgrowth, but they do appear to be important for the development of a normal length and shape. Microtubules are found centrally located in developing bristles and, based on their location, seem likely to be involved in bristle morphogenesis (Tilney et al., 1995); but genetic data that prove this are not available.

Epidermal hairs are outgrowths of individual cells and are much smaller than the shafts of sensory bristles. Epidermal hairs have primarily been studied on the pupal wing, where each cell forms a single hair (Wong and Adler, 1993). Growing hairs stain strongly for F-actin and tubulin (Eaton et al., 1996; Wong and Adler, 1993; Turner and Adler, 1998). There is no evidence for large bundles of actin filaments, but mutations in genes that encode bundling proteins such as sn and f produce mutant phenotypes (bent hairs), suggesting that actin bundling has a role in hair morphogenesis. Confocal and electron microscopy (EM) studies indicate that the microtubules are centrally located and that F actin is peripherally located in growing hairs (Turner and Adler, 1998), however, it is not clear if the arrangement is as defined as in the bristles. The lateral side branches of the arista are a third type of polarized cuticular structure with many similarities to hairs and bristles. Elongating laterals stain strongly for actin and EM studies shows the presence of bundles of actin filaments. These laterals also contain centrally localized microtubules; thus, developing laterals are quite similar to bristles shafts in morphology (He and Adler, 2001). The development of the arista laterals has not been well studied but many mutations that alter hair and bristle morphology also alter lateral morphology (B. H. and P. N. A., unpublished).

Genetic studies on bristle and hair morphogenesis have been complemented by inhibitor studies using either in vitro organ culture of pupal wings and thoraces or the injection of inhibitors into appropriately aged pupae (Turner and Adler, 1998; Tilney et al., 2000a; Geng et al., 2000). The inhibition of actin polymerization by cytochalasin D (CD) has several effects on wing hair development. The initiation of hair morphogenesis is delayed, hair elongation is slowed, hairs end up shorter than normal, and hair clusters and split hairs are seen (Turner and Adler, 1998). Similarly, bristle outgrowth is delayed, bristle elongation is slowed, and adult bristles are short, bent and frequently split after the application of CD or lantrunculin A (LAT A; Geng et al., 2000; Tilney et al., 2000a). In at least some cases, the splitting appears to be due to ectopic initiation of bristle outgrowths either from ‘budding’ along the shaft of an existing bristle or from the apical surface of the cell. CD treatment also delays and slows arista lateral outgrowth and results in short and frequently split laterals (He and Adler, 2001).

The application of microtubule antagonists, such as vinblastine (VB) or colchicine (CH) produces an overlapping but distinct set of phenotypes in all three of these cell types. The rate of wing hair elongation is slowed, but in contrast to the results with CD, hair initiation is not delayed (Turner and Adler, 1998). VB and CH treatment leads to the formation of multiple hair cells. The hairs however are not clustered as tightly as those seen after treatment with CD. The injection of VB or CH into pupae results in short and fat bristles (Geng et al., 2000; see Tilney et al., 2000a for a different result). In addition to this dramatic stunting VB also causes bent and occasionally split bristles.

Mutations in the tricornered (trc) gene, which encodes the Drosophila NDR (nuclear DBF2 related) kinase, result in dramatic multiple wing hair, branched bristle and arista lateral phenotypes (Geng et al., 2000). These phenotypes are distinct from those produced by the antagonism of the actin or microtubule cytoskeletons in that trc hairs, bristles and laterals are not shortened and thickened as are those treated with inhibitors. The phenotype of trc bristles also differs from those seen in mutants for actin-bundling proteins such as singed or forked, because in these mutants bristle morphology is abnormal all along the length of the bristle. By contrast, the morphology of trc bristles is usually normal at ‘non-branched’ locations along the bristle shaft. Based on these differences, we previously proposed that trc did not encode a cytoskeletal component required for bristle morphogenesis, but rather that it encoded a protein that had a subtle interaction with the cytoskeleton (Geng et al., 2000). One possibility is that it is involved in coordinating multiple cellular components during morphogenesis.

We now report the recovery and characterizations of mutations in a new cell autonomously acting gene, named furry (fry) that yields a trc-like phenotype in bristles, hairs and laterals. The phenotypic similarities seen in the adult structures extend to the developing pupal epidermis suggesting that trc and fry might function in the same process. Consistent with this possibility, we found that fry trc double mutant cells did not show a stronger phenotype than either single mutant. Compelling evidence that fry results in the splitting of laterals was obtained by in vivo observation of developing laterals. We have cloned and characterized the fry gene using a P insertion allele, RT-PCR, mutant gene sequencing and cDNA clone analysis. We found that fry is a large and complicated gene that encodes two proteins. One protein contains 3479 amino acids and the second is a truncated version of this. Homologs of Fry are found in humans, Caenorhabditis elegans, Arabidopsis and yeast, but nothing is known about the biochemical function of the homologs.

Flies were grown on standard media. Many mutant and Deficiency-containing stocks, and P-insertion mutants were obtained from the stock center at Indiana University. The GMR-sina stock was kindly supplied by Gerry Rubin’s laboratory (Neufeld et al., 1998).

The process of wing hair morphogenesis was studied as described previously (Turner and Adler, 1998). The morphogenesis of other body regions and cell types was studied in a similar manner. Adobe Photoshop was used to compose bitmap figures, and Corel Draw was used to compose line drawings. Confocal images were obtained using a BioRad confocal microscope at the Keck Center for Cellular Imaging at the University of Virginia. Other images were obtained using a Spot digital camera (National Diagnostics) on a Zeiss Axioskop 2 microscope.

The fry cDNA and mRNA were assembled by a combination of cDNA clone isolation and PCR analysis of cDNA preparations. cDNA clones were isolated from the region from exon 3 to 14. The largest of these was more than 4 kb. This clone was sequenced and compared with genomic sequence from the BDGP/Celera project (Adams et al., 2000) and from sequencing in the region that we did independently. A cDNA from the 3′ end of fry (SD10447) was identified as a genome project EST clone. This was obtained from Research Genetics, sequenced and the sequence compared with genomic sequence. To identify the remaining exons and to ensure that all of the exons were linked up as predicted, we scanned the sequence to identify potential coding region and, using primers from these regions, we amplified cDNA from the imaginal disc cDNA library constructed by Brown and Kafatos (Brown and Kafatos, 1988). The PCR products were sequenced and the results used to pick new PCR primers to continue the walk. Eleven overlapping regions were amplified and sequenced in this way. In these experiments, we only obtained evidence for the mRNA that encodes the long form of Fry (fry-l mRNA). Evidence for the alternative fry-s, mRNA came from a cDNA clone. When we used primers specific for fry-s, we were able to detect it by RT-PCR. Northern blot analysis was carried out as described previously (Geng et al., 2000). We also used the Drosophila Rapid Scan cDNA panel (Origene Technologies) to examine the expression of fry.

Mosaic clones were generated using the FLP/FRT system (Golic and Lindquist, 1989; Xu and Rubin, 1993).

Wings were mounted in Euparal (Asco Labs) and examined by bright field microscopy using approaches described previously (Wong and Adler, 1993).

The Sigma Stat program (Jandel) was used for comparing different genotypes or treatments.

In a screen to recover new wing hair polarity mutants, we recovered five EMS-induced alleles of an essential gene we named furry (fry). Clones of fry mutant cells produce a dramatic multiple hair cell phenotype. A more detailed description of this and other fry phenotypes is provided below. The fry mutations were uncovered by Df(3L)AC1 (67A5; 67D7-13) but not Df(3L)29A6 (66F5; 67B1) localizing it to 67B-D. We screened P insertion alleles in this region and identified one l(3)02240 that failed to complement fry point mutations. We confirmed that the P insertion was responsible for the fry mutation, as it was induced to revert at a high frequency by P transposase. In addition to complete revertants, many partial revertants were recovered that displayed a weak fry phenotype. This P insertion mutation has been described as being in a suppressor of the rough eye phenotype that results from driving the expression of sina using the GMR promoter (Neufeld et al., 1998). We confirmed that Df(3L)AC1 suppressed this gain-of-function phenotype, but we did not see any suppression from the P insertion allele or with any of our EMS induced fry alleles.

All of the EMS-induced alleles and the P insertion allele (fry⁶) were recessive lethals. The EMS alleles (as homozygotes, hemizygotes or heteroallelic heterozygotes) typically died as second or third instar larvae with only occasional animals surviving until the prepupal period. Animals that were homozygous for the fry⁶ P insertion allele often died as pharate adults that displayed a strong fry phenotype. Heteroallelic combinations that contained fry⁶ displayed a similar phenotype.

We used fry⁷, a weak viable partial revertant of fry⁶, in an attempts to distinguish between the severity of the EMS-induced alleles. In heteroallelic heterozygotes with fry⁷, three of the four extant EMS alleles resulted in a phenotype that was indistinguishable from the phenotype seen in fry⁷/Df wings. Thus, at least for this phenotype these alleles are phenotypic null alleles. One of the EMS-induced alleles (fry⁴) and the P insertion allele (fry⁶) appeared to be hypomorphic alleles.

We used the P insertion mutation as a tag to identify fry DNA. We analyzed the DNA region surrounding the P insertion allele by DNA sequencing, sequence analysis, RT-PCR, cDNA clone analysis and the sequencing of mutant alleles (details are presented in the Materials and Methods). The P element in fry⁶ was inserted into the first intron of fry (Fig. 1). Genomic Southern blot analyses showed this P was lost in all four of the fry⁶ revertants examined (data not shown). Several partial revertants were examined and in each the P element remained but was altered in structure.

Evidence was obtained for two distinct fry mRNAs from cDNA clone and RT-PCR. One of these (fry-l) was encoded by 28 exons spread out over 46 kb of genomic DNA (Fig. 1). The exons were clustered in three regions along the chromosome. Conceptual translation of this mRNA resulted in a 3479 amino acid open reading frame (AAG41424). We also found evidence for a second form of this mRNA (fry-s) that differed by use of an alternative 5′ splice site for the 14th intron of fry. Conceptual translation of this alternative splice form resulted in an ORF of 1629 amino acids, as there is a stop codon early in the extended exon 14. The first 1623 of these would be identical to those in Fry-L. Based on RT-PCR analysis, this splice form shares downstream exons with fry-l (but not exon 15), but these exons are downstream of a stop codon in the extended exon 14 and thus would not encode protein. Given the size and complexity of the fry transcription unit, it remains possible that there are additional splice forms of the fry mRNA that we did not uncover. We sequenced the open reading frame from two EMS-induced fry alleles. In fry¹, we found a 1 bp deletion in exon 3, resulting in a frameshift mutation that scrambles the amino acid sequence after amino acid 403. In fry² we found a nonsense mutation in exon 22 that truncates the Fry-L protein after amino acid 2298. As the fry² mutation is a strong allele and does not affect the Fry-S protein, we conclude that Fry-L has an important function in the development of hairs, bristles and laterals. The data do not prove that Fry-S also has a function in these cell types. The sequence data confirmed that we had identified the fry gene.

The existence of a large fry mRNA (>10 kb) was confirmed by northern blot analysis (Fig. 2A). This band was reduced in amount (and perhaps slightly larger in size) in the fry⁶ homozygous pupae (P allele). Only one band was seen on a northern blot with either a 5′- or 3′-specific probe, consistent with both splice forms being of similar size (or one being much rarer than the other). We used RT-PCR to examine the developmental distribution of the fry mRNA (Fig. 2B). fry mRNA was seen in cDNA preparations from all developmental stages examined. It was most abundant in adult heads.

The BDGP/Celera analysis of the Drosophila genome sequence predicted three independent genes that we have found to comprise the fry transcription unit (Adams et al., 2000). These are the genes CG6774, CG14171 and CG 6780 (Fig. 1). The genome project predicted 4 genes that are located in introns of fry. All of these predicted genes are transcribed on the opposite strand of DNA from fry.

Fry is a member of a conserved family of proteins found in mammals, C. elegans, Arabidopsis and yeast (Fig. 3). The similarity is concentrated in five regions separated by short regions of little or no similarity. The similarity is highest in the N-terminal most region that we call the fry domain. In this 630 amino acid region, the Fry and the human CAB4244 protein are 63% identical and 79% similar. The fly, human and worm proteins contain all five of the similarity regions, suggesting these regions may represent functional domains. The Arabidopsis and yeast proteins only contain the first three similarity regions. Several additional database sequences share similarity to part of the C-terminal region of Fry. These sequences are likely to represent partial cDNAs. The Fry-S protein contains the Fry domain and about half of the second homology segment. It is unclear whether other organisms produce a protein that is equivalent to the Fry-S protein.

The Fry proteins do not contain any protein motifs that are insightful with regard to function. There is no information about the biological function of the homologs, as all were found in genome or EST sequencing projects. The homolog from yeast (Tao3) has been found to be nonessential. Two phenotypes have been identified with the knockout of this gene. One is altered transcription of the OCH1gene and the second is the clumping of mutant cells (Saccharomyces Genome Database).

We examined the phenotype of strong fry alleles in mitotic clones in adult wings (Fig. 4, see also Fig. 9). Even small clones induced in all regions of the wing showed a strong phenotype, indicating that fry needs to function in all wing cells. Most fry cells formed a cluster of hairs and we could see examples of hairs that were split distally. On average, the fry clone cells produced 5.7 hairs per cell (a branched hair with two distal ends was counted as two hairs in these experiments), although the cellular phenotype was quite variable and some cells formed more than a dozen hairs. The fry hairs were often almost orthogonal to the wing surface, appearing as dots in the light microscope. A similar phenotype was seen in pharate adults that were homozygous or hemizygous for the fry⁶ allele (Fig. 4C,D). Weaker versions of this phenotype were seen on the wings of fry⁷/fry^x (where x is any of the EMS alleles) flies. Based on the phenotype seen in clones, it appeared quite likely that fry acted cell autonomously, but to test this in a rigorous way we induced clones marked by the loss of the N-Myc epitope tag and examined these in pupal wings that were also stained to show the actin cytoskeleton. In all cases, only clone cells showed a mutant phenotype confirming that fry acted cell autonomously in all regions of the wing (Fig. 4B). We also examined the wings of fry⁶ homozygous flies and found that late in hair development (>36 hours after white pupae) the phenotype resembled that seen in adult wings (Fig. 4D). At earlier times (e.g. 33 hours after white pupae) the phenotype seemed somewhat weaker, suggesting that the hair splitting occurred progressively throughout hair morphogenesis. This is similar to what we had seen previously in trc (Geng et al., 2000). Compelling evidence for the phenotype getting more severe as development proceeded was obtained for arista laterals, as described in some depth below. It is worth noting that fry hairs stained as strongly for F-actin, as did wild-type hairs.

We primarily examined the fry bristle phenotype in pharate adults of strong hypomorphic genotypes that contained fry⁶. A wide range of bristle phenotypes was seen (Fig. 5). In all body regions, some bristles were indistinguishable from wild type (Fig. 5A). In the abdomen, legs, notum and triple row on the wing, the most common phenotype was the presence of one or a few locations where the bristle was branched (Fig. 5A,C). Such bristles were relatively normal in shape and length except for the branching. In some cases, several branches were seen at the same proximal/distal location along the shaft. Bristles with extensive splitting were occasionally seen and these were often shorter than normal. There were also some bristles that were bent, distorted and shorter than normal without being extensively split (Fig. 5B). This was common only on the head. We examined developing bristles in fry pupae, where we stained the actin cytoskeleton. The phenotypes seen overlapped those seen in fry adult bristles. The large bundles of actin filaments seen in wild-type bristles were present in fry bristles, but some abnormalities were seen. In some bristles, the bundles appeared to become disorganized distally, so that there was no longer a series of parallel bundles (Fig. 5E). This is similar to the phenotype that is induced by treating cultured thoraces with the broad-spectrum kinase inhibitor stauroporine (Tilney et al., 2000a). In some cases we found pupal bristles where the shaft was clearly split (Fig. 5D). In these bristles, most bundles of actin filaments were continuous both proximally and distally from the branch-point.

In weak fry genotypes that were adult viable, we found frequent branched arista laterals. Except for the branching, the morphology of the laterals was close to normal. In some cases, mutant laterals appeared slightly thinner than normal in distal regions and/or slightly curled. The overall phenotype was similar to that of weak trc mutants (Geng et al., 2000). In pharate adults homozygous or hemizygous for fry⁶ a stronger lateral phenotype was seen (Fig. 6B). There was extensive branching of essentially all laterals and the weak distal curling and thinning phenotype noted above was more noticeable. Many of the branches were very thin and only visible at higher magnification (see Fig. 6B insert). Equivalent observations were made on developing laterals in pupae (Fig. 6D). As was noted for hairs and bristles above the actin staining of developing laterals was strong and equivalent to that seen in wild type.

The antenna is located at the anterior end of the pupae and we have found it possible to image developing laterals in vivo by cutting a window in the pupal case. We used this approach to determine the timing and properties of splitting laterals in weak fry mutants (fry⁷) (Fig. 7). In describing these data we have paid particular attention to the locations of the branch-points and the lateral segments distal to the branch-points (we refer to these as arms) and their distance from the central core of the arista. Substantial variation was seen in the natural history of lateral splitting. In this set of experiments, branched laterals could be put into one of two phenotypic groups. It is notable that branched laterals typically did not grow dramatically more slowly than normal laterals in the same pupae (either ipsilateral or contralateral).

The most common observations were examples (15/18 laterals (from 17 aristae and 14 pupae)) where an elongating lateral was normal when first seen but where it subsequently split at or near the distal tip (Fig. 7, right). In these cases, it is possible that the branching was a result of an abnormality at the lateral tip. This could happen at any time during lateral elongation. As morphogenesis proceeded, both of the resulting arms continued to elongate, and the location of the branch-point also moved distally. The relative extent of branch-point movement (i.e. distance from the central core to the branch-point) compared with arm elongation was quite variable. Typically, the extent of branch-point movement was much greater than that of arm elongation, as was the case for the lateral shown in Fig. 7 (right side). The relative extent of arm elongation was also variable. In some cases both arms elongated similar extents, while in other cases one grew more than the other.

A second pattern for lateral branching was the de novo formation of a new arm well proximal to the distal tip (Fig. 7, left side; 3/18 cases – all three from separate animals). In these cases, it is likely that the branching is due to a fraying of the lateral. We sometimes saw a bend or bulge in the lateral at the location of a future branch. As was seen when the branching took place near or at the elongating tip, both of the resulting arms typically continued to elongate and the location of the branch-point moved distally. Once again there was substantial variation in the relative extent of elongation.

Mutations in a number of genes such as multiple wing hairs (mwh), fuzzy (fy) and ultrahairA (ultA), which are thought to function in different processes result in a multiple hair cell phenotype (Wong and Adler, 1993; Turner and Adler, 1998; Eaton et al., 1996; Geng et al., 2000; Adler et al., 2000). We addressed the possibility that fry functioned in the same pathway as several of these genes using a genetic test. If mutations were in genes that encoded components of a linear pathway, then a double null mutant should not have a more severe phenotype than the most severe single mutant. However, if two genes function in parallel pathways, then a double null mutant might display an additive or synergistic phenotype.

One group of genes where mutations have multiple hair cell phenotypes are downstream components of the fz pathway in the wing such as multiple wing hairs (mwh; Wong and Adler, 1993; Shulman et al., 1998). We used FLP/FRT to generate mwh fry¹ wing clones and found that these had a much stronger phenotype than either single mutant (mwh clones had 3.54 hairs/cell (Fig. 8E); fry¹ had 5.7 hairs/cell (Fig. 8A)) consistent with these genes being in parallel pathways. The large number of very small hairs formed by cells in these clones were difficult to count, but we estimate that on average the clone cells contained more than 15 hairs/cell (Fig. 8F; our actual count was 16.35 hairs/cell). We also generated fry¹ mutant clones in fuzzy and fritz mutant wings (Collier et al., 1997; Collier and Gubb, 1997). Once again we found the doubly mutant (e.g. fy; fry) cells had a stronger phenotype than either single mutant (data not shown).

Mutations in the trc gene result in a multiple hair cell phenotype that is quite similar to that of fry (compare Fig. 8A with C; Geng et al., 2000). Clones of fry¹trc⁷ mutant cells did not show a stronger phenotype (Fig. 8D) than the single mutants (fry¹trc⁷ clones had 5.37 hairs/cell and trc⁷ 5.74 hairs/cell (P=0.46, t-test)). This is in sharp contrast to what we saw for mwh fry¹ mutant cells. This suggests that fry and trc may function in the same pathway. As a further test of this hypothesis, we also examined mwh trc⁷ clones. As expected, we found the mwh trc⁷ cells had a much stronger phenotype than either single mutant alone (Fig. 8B).

The detailed morphology of fry mutant cells is indistinguishable from trc mutant cells, suggesting they might function in the same cellular process. Consistent with this hypothesis, we found that cells doubly mutant for fry and trc showed a phenotype that was indistinguishable from either single mutant. Previous data has shown that trc encodes the Drosophila NDR kinase (Geng et al., 2000). This raises the possibility that fry might encode a substrate for this kinase or a protein involved in modulating trc activity. A target peptide sequence for the human NDR kinase has been reported (Millward et al., 1995), but this sequence is not found in Fry. However, as Millward and colleagues noted, this synthetic peptide does not appear to be an optimal substrate for NDR.

Mutations in genes such as mwh, a downstream component of the fz pathway in the wing, also gives rise to multiple hair cells (Wong and Adler, 1993). The morphology of these differs from the fry mutant cells in a number of ways. Notably, in fry the multiple hairs are clustered more tightly, there is much more evidence of splitting. Furthermore, the prominent polarity abnormalities of genes such as mwh are missing in fry. We found that mwh fry¹ cells had a much stronger phenotype than either single mutant, a result that stands in sharp contrast to the lack of additivity for the fry and trc. We interpret this result as mwh causing the formation of multiple independent prehair initiation sites and each of these giving rise to a cluster of hairs due to the fry-dependent splitting. Indeed the phenotype of the doubly mutant cells approximated the multiplicative phenotype this model predicts. Similar results were also found for mwh trc⁶ mutant cells. We also carried out similar experiment where we induced fry clones in a fy (Collier et al., 1997) or frtz (Collier and Gubb, 1997) mutant background. These two genes, which also appear to be components of the fz pathway in the wing have a much weaker multiple hair cell phenotype than mwh. In both of these cases we found the doubly mutant cells also had a stronger phenotype than either single mutant.

The fry phenotype in wing hairs, bristles and arista laterals shows striking similarities, suggesting a common mechanism is involved. Observations on developing pupal aristae showed that in fry mutants, laterals can split at a variety of stages. The splitting can be early or late in lateral morphogenesis, near the distal tip or far from it. Observations on fixed pupal bristles and adult cuticular bristles suggest this is the case in this cell type as well. This argues that fry functions to maintain the integrity of these structures during their morphogenesis. The function of fry does not appear to be absolutely essential for their morphogenesis, as at least some bristles in fry mutants are indistinguishable from wild type. The situation is less clear for wing hairs, as many fry cells produce clusters of hairs with only a minority of hairs being obviously split. It is possible that a lack of fry function could cause the formation of independent hair initiation centers in wing cells or to hair initiation centers that are too large to ensure a single hair is formed (Adler et al., 2000). It would not be surprising if the assembly and crosslinking properties of actin and tubulin might function to reduce the size of an initiation center to insure that a single hair was formed. fry mutations could interfere with this process, leading to a larger center and hair clusters. Alternatively the clusters could be due to splitting that occurs early in hair morphogenesis or the splitting of the initiation center prior to actual hair outgrowth. This latter hypothesis provides a common explanation for the phenotypes seen in all three cell types, hence is in some ways more appealing.

The morphology of branched hairs, laterals or bristles is typically normal except for the region of the branch point. This suggests that fry does not encode an integral component of these cellular extensions. In this way, fry and trc differ in a fundamental way from mutations in actin cytoskeleton components such as crinkled (Ashburner et al., 1999; Kiehart et al., 1999), sn (Cant et al., 1994) or f (Petersen et al., 1994), which result in abnormal morphology in all regions of the structure. We previously suggested that trc might encode a component of a system that either coordinates or organizes the growth of different subcellular components during morphogenesis (e.g. membrane, actin cytoskeleton, etc.), or monitors the ‘quality’ of the developing structure to insure its integrity (i.e. it functions in a pathway that is analogous to a morphogenetic checkpoint) (Geng et al., 2000). It would also be appealing to explain the function of fry in the same way.

The outgrowth of a cellular extension is a feature of many cells. In several systems, studies on the mechanisms of elongation has led to the conclusion that the polymerization of actin at the membrane drives extension (Alberts et al., 1994). Indeed, the ultrastructure of actin bundles in developing bristles argues that the large bundles of actin filaments are assembled in smaller units at the distal tip of the elongating bristle (Tilney et al., 1996). It is not clear, however, how other cellular components (e.g. plasma membrane) are added to such outgrowths. A model of growth only from the tip does not easily accommodate our observations on split elongating laterals (Fig. 9). If growth were only from the tip, then we would not expect the branch-point to move distally as growth proceeded; however this was a routine and prominent result. We note however, that a similar distal movement of ‘blebs’ was reported in a classic paper on the growth of the Thyone acrosomal process (Tilney and Inoue, 1982). This is an example where actin polymerization at the tip is believed to drive extension. In addition, our observations do not support a simple model of growth being restricted to the proximal base of the lateral. In this case we would not expect the length of an arm distal to a branch-point to grow but this was also routinely seen. These data could be explained by growth taking place at both the distal and proximal ends of the lateral; however, this model cannot explain an increase in the distance between branch-points on multiply branched laterals. Such multiply branched laterals were rare in the fry⁷/fry¹ mutants examined, but they were seen frequently in equivalent experiments on trc mutants where multiple branch-points are much more common (P. N. A., unpublished). The data can most simply be explained by growth taking place at all locations along the proximal distal axis of the laterals. An alternative hypothesis is that growth is normally from the tip and that growth throughout the lateral is a consequence of the fry mutation. The examination of additional ‘split lateral’ mutants should determine if this latter hypothesis is tenable. It remains to be established whether all cellular components are added to a growing lateral (or bristle) in the same way. It should be possible to follow the addition of different cellular components by the induction of transgenes that encode tagged proteins during lateral elongation (e.g. GFP-actin). For example, if actin is added only at the tip we predict that the tip will be the principal location of GFP fluorescence shortly after transgene induction. Such experiments are in progress.

This work was supported by a grant from the NIH (GM53498) to P. N. A.