Abstract

Eight classes of chemosensory neurons in C. elegans fill with fluorescein when living animals are placed in a dye solution. Fluorescein enters the neurons through their exposed sensory cilia. Mutations in 14 genes prevent dye uptake and disrupt chemosensory behaviors. Each of these genes affects the ultrastructure of the chemosensory cilia or their accessory cells. In each case, the cilia are shorter or less exposed than normal, suggesting that dye contact is the principal factor under selection. Ten genes affect many or all of the sensory cilia in the head. The daf-19 (m86) mutation eliminates all cilia, leaving only occasional centrioles in the dendrites. The cilia in che-13 (e1805), osm-1 (p808), osm-5 (p813), and osm-6 (p811) mutants have normal transition zones and severely shortened axonemes. Doublet-microtubules, attached to the membrane by Y links, assemble ectopically proximal to the cilia in these mutants. The amphid cilia in che-11 (e1810) are irregular in diameter and contain dark ground material in the middle of the axonemes. Certain mechanocilia are also affected. The amphid cilia in che-10 (e1809) apparently degenerate, leaving dendrites with bulb-shaped endings filled with dark ground material. The mechanocilia lack striated rootlets. Cilia defects have also been found in che-2, che-3, and daf-10 mutants. The osm-5 (p802) mutation specifically eliminates the distal segment of the amphid cilia. Mutations in three genes affect sensillar support cells. The che-12 (e1812) mutation eliminates matrix material normally secreted by the amphid sheath cell. The che-14 (e1960) mutation disrupts the joining of the amphid sheath and socket cells to form the receptor channel. A similar defect has been observed in daf-6 mutants. Four additional genes affect specific classes of ciliated sensory neurons. The mec-1 and mec-8 (e398) mutations disrupt the fasciculation of the amphid cilia. The cat-6 (e1861) mutation disrupts the tubular bodies of the CEP mechanocilia. A cryophilic thermotaxis mutant, ttx-1 (p767), lacks fingers on the AFD dendrite, suggesting this neuron is thermosensory. © 1986 Academic Press, Inc.

Introduction

Cilia and flagella are ubiquitous eukaryotic organelles that have been adapted for two seemingly unrelated functions, sensory transduction and cell motility. In the unicellular eukaryotes, Chlamydomonas and Paramecium, for example, they are used for swimming. Similarly, flagella propel the sperm of many animals and lower plants. Arrays of motile cilia line various epithelia, including the respiratory tracts, the oviducts, and the ventricles of the brain, where they propel fluid or particles along the surface.

Sensory cilia are found in the rod and cone cells of the eye, the hair cells of the ear, and the olfactory receptor neurons. In nematodes, cilia are found only in the nervous system where they are sensory receptors specialized for diverse modalities (Ward et al, 1975; Ware et al, 1975). Of the 118 classes of neurons in Centralists elegans hermaphrodites, 24 classes have cilia (White et al, 1986).

The common plan of both motile and sensory cilia is a membrane-bound cylinder of nine doublet microtubules that extend from a centriole. Many cilia have additional structures that adapt them to specific tasks. As they are biochemically complex structures and, in many cases, present in limited numbers, genetic studies have been helpful in understanding the assembly and function of cilia (Afzelius, 1981). In Chlamydomonas and Paramecium, genes coding for ciliary proteins have been identified by selecting for mutants with abnormal swimming (Luck, 1984; Kung et al, 1975). In humans, genetic disorders of ciliary motility produce a syndrome of male infertility and respiratory distress (Afzelius, 1976).

In C. elegans, several collections of mutants have been obtained by selecting for altered sensory behavior (Dusenbery et al, 1975; Hedgecock and Russell, 1975; Lewis and Hodgkin, 1977; Culotti and Russell, 1978; Chalfie and Sulston, 1981; Riddle et al, 1981; Hodgkin, 1983; and Trent et al, 1983). While some of these mutations affect the sensory organs themselves (Lewis and Hodgkin, 1977; Albert et al, 1981; Chalfie and Sulston, 1981; R. Ware, D. Dusenbery, D. Clark, M. Szalay, and R. Russell, personal communication), others presumably disrupt behavior at steps downstream of transduction.

Recently we found that certain sensory neurons in C. elegans accumulate fluorescein when living animals are placed in a solution of this dye (Hedgecock et al, 1985). In this paper, we show that these neurons are chemosensory and that dye uptake occurs through their exposed cilia. We have used this dye-filling technique to identify a subset of behavioral mutants with primary defects in sensory cilia or their support cells. These mutations both prevent dye uptake and disrupt sensory behaviors.

Materials and Methods

Brenner (1974) describes culturing and genetic manipulation of Caenorhabditis elegans. Strains were kindly provided by Martin Chalfie, David Dusenbery, Jonathan Hodgkin, Donald Riddle, Richard Russell, and the Caenorhabditis Genetics Center at the University of Missouri, Columbia.

Chemosensory neurons were stained with fluorescein isothiocyanate and examined by fluorescence microscopy as in Hedgecock et al (1985). New mutants with abnormal staining were induced with ethylmethanesulphonate (Brenner, 1974). The cat-6 (e1861) mutation was separated from the strain CB246.

Animals were fixed for electron microscopy using glutaraldehyde and then osmium as in Sulston et al. (1983). Usually, two or three individuals of each mutant strain were embedded and sectioned together. About 200 contiguous sections, 50 nm thick, were collected from the tips of the heads. One animal, selected for good fixation quality and orientation, was photographed approximately every third section at 5000X magnification to reconstruct the head sensilla. The other animals were examined directly in the microscope.

Results

Description of the Amphid and Phasmid Sensilla
The amphids, a pair of lateral sensilla in the head, are the principal chemosensory organs of nematodes (Fig. 1). In C. elegans, each amphid comprises the ciliated dendrites of 12 sensory neurons plus two support cells called sheath and socket cells (Ward et al, 1975; Ware et al, 1975; White et al, 1986). The sheath and socket cells form a cylindrical channel to the outside (Fig. 2). Of the 12 amphid neurons, 8 (ASE, ADF, ASG, ASH, ASI, ASJ, ASK, and ADL) are evidently chemosensory in that their cilia extend into the channel of the socket cell and are thereby exposed to the external medium. The cilia of three additional neurons (AWA, AWB, and AWC), called wing cells, also share the main lumen of the sheath cell. The wing cilia separate from the others, and invaginate individually into the sheath cell, proximal to where the fascicle of channel cilia enters the socket cell. Finally, the dendrite of a neuron (AFD), called the finger cell, remains separate from the other dendrites in the sheath cell. It has only a rudimentary cilium but, proximal to the cilium, the dendritic membrane expands into about fifty villi, called fingers, that invaginate the sheath cell (Fig. 2). These fingers are about 0.15 micrometer in diameter and 2 micrometer long. No internal microfilaments or microtubules have been seen in them but they tend to be oriented anteriorly or posteriorly in the sheath cell.

The phasmids, a pair of lateral sensilla in the tail, are similar but smaller chemosensory organs (Sulston et al 1980; Hall and Russell 1986; White et al, 1986). In newly hatched larvae, each phasmid comprises two ciliated dendrites (PHA and PHB), a sheath cell, and a socket cell. The neurons resemble the amphid channel neurons in that their cilia extend into a socket channel that is open to the external medium.

Below the cilia, the sensory dendrites are joined to the sheath cell by belt junctions (Fig. 2). These have been described both as tight junctions (Ward et al, 1975) and as desmosomes (Ware et al, 1975) and may have properties of both. Similar belt junctions encircle the channels, joining the sheath and the socket cells together. Finally, belt junctions join the socket cells to the surrounding hypodermis.

The channels of the amphid sheath and socket cells appear to originate by different mechanisms (Wright, 1980). The sensory dendrites deeply invaginate and, excepting the AFD neuron, completely penetrate the sheath cell so that it is topologically a solid torus with 11 holes. In contrast, the socket cell wraps around the receptor channel and forms a typical intercellular belt junction where it meets with itself. Thus topologically it has no hole.

The channel of the socket cell is lined with cuticle that is continuous with the external cuticle (Fig. 3a). The sheath cell channel is not lined with cuticle. Instead, the anterior sheath channel, in the region where the cilia draw together into a tight fascicle, has a characteristic dark lining (Fig. 3b). More posterior, nearer the bases of the cilia, the dark lining is interrupted by matrix-filled vesicles fusing with the lumen. The cytoplasm adjacent to the anterior sheath channel contains longitudinally aligned microtubules and intermediate filaments. These filaments may form a scaffold for the receptor channel (Wright, 1980). A much thinner scaffold, joined at its ends to the self-junction, wraps around the socket channel (Fig. 3a).

In glutaraldehyde fixed animals, a dark matrix surrounds the cilia in the posterior sheath channel (Figure 4a). The matrix material appears to be synthesized at lamellae posterior to the cilia and transported forward in membrane-bound vesicles which later fuse with the channel lumen (Wright, 1980). The matrix material of the amphid sheath cells, and a similar material in the other sensilla, is not well preserved in animals fixed with Os0₄ alone. In consequence, several published reports erroneously describe an empty space around the cilia or empty vesicles in the sheath cytoplasm. The matrix material, though separating the cilia in the posterior channel, gradually thins until the membranes of the channel cilia are in direct apposition in the anterior sheath and the socket channel (Fig. 3). The pattern of fasciculation of the channel cilia is invariant in wild-type animals (Ward et al, 1975; Ware et al, 1975).

Description of Other Head Sensilla
In addition to the amphids, four classes of cuticular sensilla (cephalic, inner labial, outer labial quadrant, and outer labial lateral) are found in the tip of the head (Ward et al, 1975; Ware et al, 1975). They resemble the larger amphid sensilla in having two support cells, a sheath and a socket, that form channels around the ciliated portion of the dendrites. They differ from the amphids in that the socket channels are not lined with cuticle. Most of the structural components of the amphid sensilla described above are also found, reduced in size, in these sensilla.

The tip of the head has six symmetrically arranged lips (2 dorsal, 2 ventral, and 2 lateral). An inner labial sensillum is found on the apex of each lip. These sensilla each contain two ciliated dendrites (IL1 and IL2) (Fig. 1). The dorsal and ventral lips also contain a cephalic and an outer labial quadrant sensillum. The cephalic sensilla have a single dendrite (CEP) in hermaphrodites and an additional dendrite (CEM) in males. The outer labial quadrant sensilla have a single dendrite (OLQ). The lateral lips contain, in addition to an inner labial and an amphid sensillum, an outer labial lateral sensillum. The outer labial lateral sensilla have a single dendrite (OLL).

After passing through the socket channels, the IL1, CEP, OLQ, and OLL cilia end embedded in the subcuticle and are believed to be mechanosensory. In contrast, the tips of the IL2 and CEM cilia completely penetrate the cuticle and are believed to be chemosensory.

Finally, two classes of ciliated dendrites (BAG and FLP) found in the lateral lips are not surrounded by support cells (Fig. 1). Their cilia end somewhat behind the cuticle in bag and flap-shaped sheets, respectively, that envelop short projections from the inner labial socket cells.

Ultrastructure of Amphid Cilia
The dendrites of amphid channel neurons ASE, ASG, ASH, ASI, ASJ, and ASK each end with a single cilium about 7.5 micrometer long in adults (Ward et al, 1975; Ware et al, 1975). The dendrites of channel neurons ADF and ADL are similar but each ends in a pair of cilia (Figs. 2, 3). Three segments can be distinguished in these cilia. The proximal segment, which corresponds to the transition zone of the motile flagella in Chlamydomonas (Ringo, 1967), is a constriction at the base of the cilium about 0.27 micrometer in diameter and up to 1.0 micrometer in length. It comprises nine doublet-microtubules joined to the membrane by Y-shaped links (Gilula and Satir, 1972) and drawn inward by attachments to a central cylinder (Fig. 4b). A variable number of singlet microtubules are attached to the inner surface of the cylinder. The central cylinder may correspond to the apical rings found in the transition zones of cilia in some organisms. The inner singlets in C. elegans differ from the central pair of microtubules found in motile cilia in that they originate at the base of the transition zone rather than above it. The inner singlets, like axonal microtubules in C. elegans, have only 11 protofilaments whereas the A and B subfibers of the peripheral doublets have 13 and 11 protofilaments, respectively (Chalfie and Thomson, 1982).

The middle segment differs from the transition zone in lacking the central cylinder. The doublets, still linked to the membrane, spread apart somewhat and the cilium flares in diameter (Fig. 4a). The Y-shaped bases of the membrane links are no longer apparent, perhaps relaxing against the membrane in the absence of inward tension on the doublets. The inner singlet microtubules continue, unattached, in the center of the cilium. The middle segment of the channel cilia corresponds to the flagellar shaft in Chlamydomonas and continues for about 4 micrometer.

The B subfibers of the doublet microtubules are gradually lost near the end of the middle segment (Fig. 3b). The distal segment, about 2.5 micrometer long, contains only A subfibers and inner singlet microtubules (Fig. 3a). The membrane links are probably also lost. The distal segment, roughly the portion in the socket channel, may be the transducing region of the cilium.

The amphid cilia, like sensory cilia in nematodes generally have no apparent basal bodies (Wright, 1980). The cilia terminate proximally in connections from the peripheral doublets to the cell membrane (Fig. 4c; see also Figs. 5g, 7c). These terminal connections may be equivalent to the transitional fibers seen in other organisms (Reese, 1965; Ringo, 1967). As they have complex substructure, they conceivably also contain some residue of the nematode centriole.

Unlike the channel cilia which are all cylindrical, each of the amphid wing cilia (AWA, AWB, and AWC) has a unique shape (Ward et al, 1975; Ware et al, 1975). The AWC cilium spreads vertically into two enormous sheets, resembling wings. These wings and the surrounding sheath cell, fill much of the left and right hemisectors at the tip of the animal (Fig. 1). The AWA and AWB cilia are smaller than AWC and comparable in size to the channel cilia. The distal segments of the AWA cilia split into several small projections each containing one or more of the original nine doublet microtubules (Fig. 2). The AWB dendrite, like ADF and ADL, ends in a pair of cilia. The distal segments of the AWB cilia do not split like the AWA cilia but are somewhat flattened and irregular.

None of the amphid dendrites contain striated ciliary rootlets. Instead, an amorphous gray material extends posteriorly from the centers of the channel and wing cilia for about a micrometer (Fig. 4d). This material gradually thins, revealing a fascicle of 3 to 12 neurofilaments that continue at least several micrometers further (Fig. 4e). It is likely that these neurofilaments are actually embedded in the amorphous root and extend to the base of the cilia. The amorphous root is reduced or absent in the AFD dendrites. In those cells, a fascicle of neurofilaments extends to the base of the cilia. Finally, numerous coated pits and vesicles are found in all the amphid dendrites just proximal to the cilia (Fig. 4d).

Ultrastructure and Specialization of Mechanocilia
The transition zones of the various mechanocilia resemble those of the amphid cilia. In particular, central structures, probably short cylinders, join the inner faces of the doublets. In many of the mechanocilia, some peripheral doublets terminate just distal to the transition zone. In the CEP and OLL cilia, for example, usually only five membrane-linked doublets continue in the middle segments (Figs. 5e,6). The distal segments of the CEP and OLL cilia contain an amorphous dark material and associated microtubules common to proved mechanocilia in insects (Ward et al, 1975; Ware et al, 1975; Thurm et al, 1983). In the CEP cilia, the microtubules are interspersed with the dark material and mold it into irregular rods (Fig. 5a). In the OLL cilia, the dark material is not interspersed with microtubules but forms a large aggregate surrounded by a single layer of microtubules. In both the CEP and OLL cilia, the outermost microtubules appear to have fine attachments to the membrane. The microtubules in the distal segments are all singlets and, at least a majority, are supernumerary in that they do not derive from the nine-doublet microtubules of the axoneme nor are they central singlet microtubules arising at the apical ring as in the amphid cilia. The supernumerary microtubules and the dark tubule-associated material are confined to the region embedded in the cuticle (Fig. 6).

The OLQ cilia are unique in two respects. First, the A and B subfibers have filled cores giving the doublets an exceptionally dark appearance. Second, exactly four of the nine doublets extend through the cilium (Figs. 5a-e). These four doublets are not membrane linked but are joined along their lengths by thick cross-bridges to form a square. Fine radial arms join these doublets to a small hub in the center of the square. The corners of the square always point radially or circumferentially in the wild type. In the distal segment, embedded in the subcuticle, one or two small aggregates of amorphous dark material, resembling the tubule-associated material of the CEP and OLL cilia, flank the doublet microtubules at the circumferential corners, but not the radial corners (Figs. 5a,b). This material may also be connected to the membrane.

The tips of the IL1 cilia contain a disc of dark material attached on both faces to the ciliary membrane (Fig. 7a). This dark material is positioned in the cuticle in such a way as to be compressed by outward radial deflections of the papillary protrusions caused by head-on contact with external objects.

The distal segments of the CEP, OLL, and OLQ cilia are anchored in cuticle by a small dark nubbin (Ward et al, 1975; Ware et al, 1975). In the CEP and OLQ neurons the nubbin occurs at the base of the transducing region (Figs. 5b, 6). The OLL cilia differ in that the nubbin is at the distal tip of the cilium and the supernumerary microtubules and tubule-associated material are proximal to the nubbin.

Finally, three classes of sensory cilia (BAG, IL1, and OLQ) in the hermaphrodite have large striated rootlets (Ward et al, 1975; Ware et al, 1975). The rootlets extend into the center of the transition zone (Figs. 7b,c).

Fewer than nine peripheral doublets have been reported for some classes of cilia in C. elegans (Ward et al, 1975; Ware et al, 1975). Using glutaraldehyde-fixed adults, we consistently found nine doublets in the transition zone of the BAG, CEP, IL1, and OLQ cilia. Since not all nine doublets extend into the shaft in some of these classes, they could be overlooked in a coarse series. All the IL2 cilia examined in wild-type adults have fewer than nine doublets in the shaft and no well-formed transition zone.

Mechanism of Dye Filling
When living C. elegans are placed in solutions of 5-fluorescein isothiocyanate (FITC), six pairs of neurons in the head and two pairs in the tail fill with dye (Fig. 8a). Their cell bodies and processes become visible within 5 min and reach a maximum brightness within about 2 hr when stained in 0.1 mg/ml FITC. Dye filling proceeds equally well at 0° as at 20°. Once filled with 5-fluorescein isothiocyanate, the neurons remain brightly stained for many hours in the absence of dye. Staining with fluorescein, in contrast, reverses completely in the course of an hour. Presumably, 5-fluorescein isothiocyanate, but not free fluorescein, can combine with amino groups within the cell and become either immobile or impermeant to cell membranes. In support of this, 5-fluorescein isothiocyanate, when coupled to bovine serum albumin, cannot enter the neurons from the outside.

We tested a variety of other fluorescent dyes and none, except certain fluorescein derivatives, accumulate in the amphid and phasmid neurons. The fluorescein derivatives that stain the neurons are weak acids and exist as both neutral and anionic forms within the physiological range of pH values. In their uncharged forms, favored by lower pH, they can probably diffuse across cell membranes.

The FITC-filled neurons in the head and tail were identified as amphid channel neurons (ADF, ASH, ASI, ASJ, ASK, and ADL) and phasmid channel neurons (PHA and PHB), respectively (Hedgecock et al, 1985).These cells stain in larvae of all stages and in adults. To learn whether fluorescein enters these neurons through their exposed sensory cilia, we killed the phasmid support cells in newly hatched larvae using a laser microbeam (Sulston and White, 1980). These animals were tested as adults for dye uptake into the phasmid neurons. Killing the socket cell (2 animals), which presumably disconnects the sheath and cilia from the cuticle, or the sheath cell (1 animal) abolished filling of the ipsilateral neurons without affecting the neurons of the contralateral phasmid sensillum. Control ablations of neighboring cells did not affect dye uptake.

The amphid channel neurons ASE and ASG, the IL2 neurons, and the various male-specific chemosensory neurons do not appear to fill with fluorescein dyes. Thus access of the sensory dendrites to the dye is apparently necessary but not sufficient to ensure filling. Apparently a physiological property, shared by some but not all sensory neurons, is also required for filling. A simple suggestion is that for dye to fill the entire neuron, the rate of dye entry through the sensory receptor must be greater than the rate of dye leakage into the body cavity from the sensory process. The rate of entry is controlled by the geometry, and possibly, membrane properties of the exposed dendrites. The rate of leakage from the processes might depend on membrane potential or intra-cellular pH.

Identification of Behavioral Mutants with Impaired FITC Uptake
Mutants with sensory defects have been isolated by selections involving chemotaxis toward Na+ or Cl- ions (tax and che genes: Dusenbery et al, 1975; Lewis and Hodgkin, 1977), thermotaxis (ttx genes: Hedgecock and Russell, 1975), male mating (Lewis and Hodgkin, 1977, Hodgkin, 1983), avoidance of solutions of high osmotic strength (osm genes: Culotti and Russell, 1978), dauer larva formation (daf genes: Riddle et al, 1981), coarse mechanical stimulation (mec genes: Chalfie and Sulston, 1981), egg-laying (egl genes: Trent et al, 1983), and formaldehyde-induced fluorescence (FIF) to visualize catecholamine (dopamine) containing mechanosensory neurons (CEP, ADE, and PDE) (cat genes: Sulston et al, 1975).

We examined alleles of all the published cat, che, daf, mec, osm, tax, and ttx genes for defects in FITC uptake into chemosensory neurons. All of the cat, ttx, and mec mutants, with the exceptions of mec-1 and mec-8, were essentially normal in dye filling. In contrast, all of the osm mutants and some of the che, daf, and tax mutants are defective in dye uptake, affecting both the amphid and phasmid neurons (Fig. 9, Table 1).

We tested whether any of these mutations, isolated in different laboratories, fail to complement. Indeed, the mutations che-3 (el 124), che-8 (e1253), and osm-2 (p801) on linkage group I all fail to complement for FITC uptake. Similarly, mutations daf-10 (el387) and osm-4 (p821) on linkage group IV represent a single gene. Finally, the unmapped tax mutation, a83 (formerly RS3, Dusenbery et al, 1975) is an allele of osm-1.

We also isolated nine new mutants with reduced dye uptake. These fall in two of the known osm genes and five new genes designated che-10 through che-14. Excluding the mec-1 and mec-8 alleles, there are now 25 mutations, defining 14 complementation groups, which reduce or eliminate FITC uptake by amphid and phasmid neurons (Table 1, Fig. 9). A spectrum of behaviors was tested for each mutant (Table 1).

Dye Filling of Mutant Mechanosensory Neurons
Mechanosensory neurons do not normally fill with FITC. In some chemosensory mutants, however, certain mechanosensory neurons, including CEP, ADE, and PDE neurons, occasionally stain brightly (Table 1). In many of the mutants showing occasional staining of mechanosensory neurons in hermaphrodites, occasional ray neurons also stain in males (Table 1). We examined ray staining in detail in osm-1 (p808) males. It appears that neurons from each of the 18 ray sensilla are capable of staining. Apparently only one neuron per sensillum can fill with dye. We speculate that the stained cells are RnA neurons, rather than RnB neurons, as the RnB dendrites are externally exposed, yet nonstaining, in wild-type males (Sulston et al, 1980).

a The following mutants were found to have normal FITC uptake: che-5 (e1073), che-6 (e1126), che-7 (ell28), daf-1 (e1287), daf-2 (e1370), daf-3 (e1376), daf-4 (e1364), daf-5 (e1386), daf-7 (el372), daf-8 (e1393}, daf-9 (e1406), daf-11 (m47), daf-12 (m20), daf-13 (m66), daf-14 (m77), daf-15 (m81), daf-16 (m26), daf-17 (m27), daf-18 (e1375), and daf-20 (m25). Heat-sensitive alleles were tested at nonpermissive temperature (25°). In che-1 (e1034) mutants, an additional class of amphid neurons often stains.

b The frequency and intensity of staining of neurons is indicated qualitatively: 3, usually or always stains; 2, frequently stains; 1, occasionally stains; 0, rarely or never stains. A suffix w indicates that the staining intensity is much weaker than in wild-type.

c Avoidance of concentrated NaCl (osmotic, OSM) was tested with a population assay (Culotti and Russell, 1978). Attraction (chemotaxis, CTX) was tested individually using dilute gradients of NaCl (Ward, 1973). Dauer larva formation (DAF) was tested on crowded, starved plates using sodium dodecyl sulfate to kill nondauer larva (Cassada and Russell, 1975). The cuticles of survivors were examined using Nomarski optics to confirm the presence of dauer-specific alae. Ability to follow isotherms (thermotaxis, TTX) was tested individually in radial temperature gradients (Hedgecock and Russell, 1975). Touch sensitivity (mechanosensory, MEC) was tested with an eyebrow hair (Chalfie and Sulston, 1981). Males were obtained from him-5 (e1490) double mutants and their mating ability (MAT) was tested by the procedure of Hodgkin (1983). All behaviors, except mating, were scored either (-) no response, (±) intermediate response, or (+) essentially wild-type response. Male mating ability was scored according to Hodgkin (1983): 4, very efficient mating (30-100% of wild-type efficiency); 3, efficient mating (10-30% of wild-type); 2, poor mating (1-10% of wild-type); 1, very poor mating (less than 1% of wild-type); and 0, no detected matings.

d For each amphid sensillum in che-14 (e1960), either all six neurons stain or none stain. In addition to the CEP neurons, unidentified sensory neurons with cell bodies anterior to the nerve ring frequently stain in che-14 (e1960). In the OSM assay, about 10% of the che-14 (e1960) animals failed to avoid concentrated NaCl.

e The daf-19 (m86hs) mutants form dauer larvae constitutively, particularly at high temperature (D. Riddle, personal communication). There is no FITC staining at either permissive (15°, adults and dauers) or nonpermissive temperature. (25°, dauers only).

f The phasmid neurons were examined in forty mec-1 (e1066) mutants. Both neurons stained brightly in 58 sensilla, only one neuron stained in 15 sensilla, and no neurons stained in 7 sensilla. In comparison, both phasmid neurons stained in 78 sensilla and no neurons stained in 2 sensilla in 40 wild-types.

Mutants of two genes, cat-6 and che-14, show a much higher frequency of dye filling by mechanosensory neurons. In cat-6 mutants, the amphid and phasmid neurons stain normally, but the CEP, ADE, and PDE neurons also stain brightly in many animals. The proportion of these mechanosensory neurons staining is greatest just after molts (Fig. 10). In che-14 mutants, the phasmid neurons never stain and the amphid neurons frequently fail to stain (Table 1). The CEP, ADE, and PDE neurons stain brightly in many animals as do additional, unidentified sensory neurons in the head. As shown below, the CEP dendrites, and presumably the other classes that stain, have abnormal access to the external medium in cat-6 and che-14 mutants.

Mutants of two genes, cat-6 and che-14, show a much higher frequency of dye filling by mechanosensory neurons. In cat-6 mutants, the amphid and phasmid neurons stain normally, but the CEP, ADE, and PDE neurons also stain brightly in many animals. The proportion of these mechanosensory neurons staining is greatest just after molts (Fig. 10). In che-14 mutants, the phasmid neurons never stain and the amphid neurons frequently fail to stain (Table 1). The CEP, ADE, and PDE neurons stain brightly in many animals as do additional, unidentified sensory neurons in the head. As shown below, the CEP dendrites, and presumably the other classes that stain, have abnormal access to the external medium in cat-6 and che-14 mutants.

Mutants with Short Axonemes in all Classes of Cilia
Mutations in three genes, che-13 (e1805), osm-1 (p808), and osm-5 (p813), shorten the axonemes of all classes of sensory cilia in the head. Singlet or doublet microtubules, joined to the membrane by Y links, assemble below the transition zones. The various distal specializations of the mechanocilia also assemble ectopically in these mutants.

The peripheral doublets of the amphid channel cilia end within about 2 micrometer of the transition zone (Fig. 11). The inner singlets do not extend beyond the apical ring. The wing cilia are similarly affected. Interestingly, the AWC cilium fails to spread into sheets and the surrounding sheath cell is correspondingly reduced. The AFD cilia, although fairly short in wild-type, are reduced further and often tilted. The AFD fingers themselves are unaffected in number or appearance.

Doublet microtubules, joined to the membrane by Y links, assemble below the cilia in these mutants. These doublets are not continuous with the nine peripheral doublets of the cilium (Fig. 12a). The ectopic doublets do not generally cross the neuron/sheath junctions but instead create a posterior projection within the sheath cell (Fig. 12b). Like normal cilia, these projections are topologically distal to the junctions. They strikingly mimic the middle segment of a normal cilium (Fig. 12c). They end blindly within the sheath cell and are usually filled with vesicles where they terminate (Fig. 12d). The occasional doublets that cross the neuron/sheath junction, lose their membrane links below the junction.

As judged by the hooks on microtubules with partial B subfibers, the ectopic doublets have the opposite clocksense to the nine ciliary doublets in adjacent sections. As the ectopic tubules project posteriorly and the cilia project anteriorly, both classes of doublets have the same relative clocksense. In particular, the B subfibers are counterclockwise of their respective A fibers for a viewer looking from proximal to distal.

The amphid sheath channel in these mutants contains more matrix than wild-type and much of the space normally occupied by cilia is filled with matrix instead. Abnormal large matrix-filled vesicles accumulate in the anterior cytoplasm of the sheath cell. Often these vesicles are partially fused with the channel.

The CEP, IL1, IL2, OLL, and OLQ axonemes are greatly reduced in length in che-13 (e1805), osm-1 (p808), and osm-5 (p813) mutants (Fig. 13). The dendrites themselves, however, continue and may reach the cuticle. In particular, the CEP, OLL and OLQ dendrites form cuticle-attached nubbins. Empty tunnels in the subcuticle are found anterior to CEP and, less often, the OLQ dendrites suggesting that these dendrites once extended somewhat further but have retracted, usually to the nubbin.

The transition zones of the CEP, IL1, IL2, OLL, and OLQ cilia, although normal in structure, are frequently mispositioned along the dendrite either anteriorly, to the level of the socket channel or beyond (Fig. 14a), or posteriorly, to the level of the neuron/sheath junction (Fig. 16a) or even into the ectopic posterior projections. As in the amphid cilia, membrane-linked microtubules assemble ectopically behind the cilia. These microtubules are generally fewer and shorter than in the amphid cilia and are more often singlets than doublets. Again, these ectopic membrane-linked microtubules do not cross the neuron/sheath junction but instead create a posterior projection within the sheath cell.

The supernumerary microtubules and associated dark material normally found in the distal segments of the CEP and OLL cilia were present but positioned irregularly along the dendrites, both distal and proximal to the residual cilia. Large, ball-shaped aggregates of the tubule-associated material were often found in the ectopic posterior projections of the CEP cilia (Fig. 14b).

The joined square of doublets is formed in the OLQ cilia but generally fails to extend past the sheath channel. In many cilia, the corners of the square do not point radially and circumferentially. In a few cases, five rather than four doublets were joined by cross-bridges to make an irregular pentagon with two central hubs (Fig. 15).

The dark material that normally flanks the circumferential corners was fragmented and mispositioned.

The dark membrane-attached discs normally found at the tips of the IL1 cilia were present but displaced posteriorly in these mutants, often to the level of the transition zone (Fig. 16a).

The striated ciliary rootlets of the IL1, OLQ, and BAG neurons are normal in these mutants and attach properly to the transition zone. Interesting, the ectopic membrane-attached microtubules found in these mutants also recruit small rootlets (Figs. 16b, c).

In an unexpected contrast to wild-type, well-formed transition zones comprising a tight ring of nine Y-linked doublet microtubules were found in all classes of cilia, including IL2, in che-13, osm-1, and osm-5 mutants.

The osm-6 (p811) mutant has a similar, though perhaps less severe, ultrastructural phenotype than the che-13, osm-1, and osm-5 mutants. The microtubules of the various classes of cilia extend further than in the other mutants but ectopic membrane-attached microtubules still assemble proximal to the cilia. The large wings of the AWC cilia are reduced but not eliminated. The transition zones of the mechanocilia in osm-6 (p811), in contrast to the other three mutants, are positioned normally along the dendrites. The dark discs in the IL1 dendrites are also positioned normally at the tips but another mechanosensory specialization, the supernumerary microtubules and dark tubule-associated material of the CEP dendrites, assembles ectopically. Possibly significant, the amphid sheath cytoplasm contains an excess of small, unfused matrix-filled vesicles rather than the large vesicles found in the other mutants. The osm-6 (p811) vesicles resemble the unfused matrix-filled vesicles found in wild type except for their greater numbers.

daf-19 Mutants Lack All Classes of Cilia
The sensory dendrites in daf-19 (m86) mutants entirely lack cilia including the transition zones. Vestigial centrioles, without membrane attachments, are found in a few of the amphid dendrites (Fig. 17). No ectopic membrane-linked microtubules are found in the amphid dendrites. A few membrane-associated singlet microtubules are found in the CEP, IL1, and OLQ cilia. The amphid dendrites, and most of the mechanosensory dendrites, terminate in club-shaped endings after invaginating, and forming belt-shaped junctions with their respective sheath cells. The CEP dendrites, though not the OLL, and OLQ dendrites, extend through their socket channels to end in cuticle-attached nubbins. Supernumerary microtubules and associated dark material are present, though mispositioned, in CEP and OLL dendrites. Similarly, the disc-shaped accessories normally found at the tips of the IL1 cilia are present in the mutant dendrites immediately distal to the neuron/sheath junctions. Striated rootlets are present in IL1 and OLQ dendrites, some in their normal position and others in ectopic posterior projections of the dendrite distal to the neuron/ sheath junctions. The fingers of the AFD neuron are normal. Abnormal large matrix-filled vesicles accumulate in the amphid sheath cell.

che-11 Cilia Contain Abnormal Ground Material
In contrast to the mutants mentioned above, the amphid wing and channel cilia in che-11 (e1810) are nearly normal in length and arrangement of microtubules. However, these cilia contain abnormal dark ground material interspersed among the microtubules of the axoneme (Fig. 18). Some of the cilia are slightly enlarged in diameter and irregular in contour. The dendrites below the cilia also contain dark ground material and few, if any, membrane-attached microtubules. The AWC cilia fail to spread into wing-shaped sheets. Abnormal large matrix-filled vesicles accumulate in the amphid sheath cell. In one sensillum examined, many of the intermediate filaments in the sheath scaffold are oriented circumferentially rather than longitudinally.

The CEP cilia in che-11 (e1810) mutants are reduced in length and largely resemble the cilia in che-13, osm-1, osm-5, and osm-6. Dark material and associated microtubules assemble in both rod- and ball-shaped aggregates along the dendrites and in ectopic posterior projections. The transition zones are often displaced. Empty tunnels are present in the subcuticle distal to the cuticle-attached nubbin. In contrast to the other four mutants, the posterior projections are filled with dark ground material and numerous vesicles.

The IL1, IL2, OLL, and OLQ axonemes are nearly normal in length and the transition zones are positioned correctly in che-11 (e1810). The joined squares in the OLQ cilia are oriented normally but the flanking dark material is fragmented and mispositioned. The distal segments of some OLQ cilia have unattached singlet microtubules in addition to the joined square. The dark discs of the IL1 cilia and the amorphous dark material in the OLL cilia were positioned normally. A few membrane-attached singlet microtubules were found below the IL1 cilia.

che-10 Mutants Lack Amphid Cilia and Striated Rootlets
Most of the amphid wing and channel dendrites in che-10 (e1809) mutants have no recognizable transition zones or axonemes. These dendrites generally have enlarged bulb-shaped endings filled with dark ground material (Fig. 19b). However, usually one or two dendrites per sensillum have well-formed cilia with normal transition zones and nearly full-length axonemes (Fig. 19a). The wing-shaped sheets of the AWC cilia are present. The AFD cilia are absent or tilted but the fingers are normal. Abnormal large matrix-filled vesicles accumulate in the sheath cell.

The striated rootlets normally found at the base of the cilia in the IL1 (Fig. 20), OLQ, and BAG neurons are entirely missing in the mutant che-10 (e1809). The distal specializations of these cilia, and the other mechanosensory cilia of the head, are normal.

osm-3 Specifically Required for Amphid and Phasmid Cilia
The distal segments of the amphid channel neurons are absent in osm-3 (p802) mutants (Fig. 21). Both the transition zones and middle segments are normal in length and contain a full complement of membrane-linked doublet and central singlet microtubules. The cilia end abruptly, however, in the region where the B subfibers normally terminate. Thus the distal segments, containing only A subfibers and central singlets, are entirely truncated and the socket channel is empty of cilia.

Because the channel cilia in osm-3 (p802) have normal middle segments, they are substantially longer than the che-13, osm-1, osm-5, and osm-6 cilia. Moreover, the cilia are not displaced forward in the sheath cell as in the mutants without middle segments. Finally, no ectopically assembled membrane-linked microtubules are found in osm-3 (p802) dendrites.

The amphid wing cilia are essentially normal in osm-3 (p802). Similarly, the AFD dendrites, and the various mechanosensilla, are also normal. The only defect in osm-3 (p802) besides the distal truncation of the amphid channel cilia, is an accumulation of abnormal, large matrix-filled vesicles in the anterior cytoplasm of the sheath cell (Fig. 22).

che-12 Affects the Amphid Sheath Matrix
The matrix vesicles of the amphid sheath cell appear pale or empty in che-12 (e1812). The lumen of the sheath channel and the extracellular space surrounding the AFD fingers are devoid of matrix. The amphid wing and channel cilia, particularly near the membrane, are abnormally dark (Fig. 23). The channel cilia are shorter than normal and only extend partway through the socket channel. Unlike other mutants with shortened cilia, no large matrix vesicles accumulate in the sheath cytoplasm.
Irregular vesicles are present between the two layers of the adult cuticle in che-12 (e1812) (Fig. 23).

che-14 Affects the Joining of the Amphid Channels
The amphid channel is abnormally large in diameter and poorly aligned at the join between the sheath and socket cells in che-14 (e1960) mutants. The socket scaffold is disorganized and some of the intermediate filaments are oriented circumferentially rather than longitudinally. The socket cytoplasm contains abnormal vesicles and the cuticle lining of the channel is abnormally thin. The sheath scaffold is apparently stretched thin near the join and the dark lining of the channel is absent. More posteriorly in the sheath cell, the scaffold and dark lining appear normal. The belt junction between the sheath and socket cells is normal.

In some cases, the socket channel fails to connect with the sheath channel and ends as a blind, cuticle-lined pocket (Fig. 24). When the cilia, which form a normal fascicle in the sheath, reach an obstructed socket channel, they are either deflected sideways in the sheath cell or invaginate the socket cell without obtaining access to the externally open channel (Fig. 25). Matrix accumulates in the sheath around the distal ends of the deflected fascicles.

The cuticle at the tip of the head in che-14 (e1960) is thin and irregular. The hypodermis, which is pale and somewhat distended, reveals numerous aggregates of longitudinal intermediate filaments (Fig. 26). Presumably similar filaments are present in the wild-type hypodermis.

The cuticle-embedded specializations of certain mechanocilia are abnormal in che-14 (e1960). The discs at the tips of the IL1 cilia are tilted. The nubbins of the CEP and OLQ dendrites are recessed in cuticular tunnels (Fig. 27). The joined squares of the OLQ cilia are sometimes misoriented and, even when the squares are oriented normally, the dark material that normally flanks the circumferential corners occurs in abnormally small pieces and is positioned randomly.

Like the CEP neurons, the ADE and PDE neurons fill with fluorescein in che-14 (e1960) mutants (Table 1). This suggests that defects in hypodermis and cuticle may extend along the entire length of the animal.

The various mechanosensilla of the head are normal in mec-8 (e398).

cat-6 Affects the CEP Specializations
The transition zones and middle segments of the CEP cilia in cat-6 (e1861) are positioned slightly anterior of normal but are normal in length. The distal specializations, supernumerary microtubules and associated dark material, form normal rod-shaped aggregates. These rods, however, are not confined to the distal segments but assemble along the entire cilia as well as ectopically, proximal to the cilia (Figs. 30, 31d, e). The cuticle-attached nubbins may also contain rods separated from the ciliary shaft. Such nubbins are enlarged and often extend completely through the cuticle (Figs. 30, 31a-c).

The OLQ cilia in cat-6 (e1861) may have a reduced amount of dark material flanking the circumferential corners of the square of doublet microtubules. The other classes of mechanocilia, including IL1 and OLL, and the amphid sensilla appear normal.

ttx-1 Thermosensory Mutants Lack the AFD Fingers
The fingers of the AFD neurons in the cryophilic mutant ttx-1 (p767; formerly EH67, Hedgecock and Russell, 1975) are entirely missing. Instead, a fingerless sack of membrane protrudes from the dendrites just below the cilia (Fig. 32). The cilia are about 4 micrometer long, three times their normal length, and are tilted ventrally at their bases, away from the anteriorly projected sack (Fig. 33). The amphid wing and channel cilia and the various mechanosensilla of the head are normal.

Discussion

Development of the Sensilla
The assembly of an individual sensillum requires interactions between at least four cell types. One or more neurons invaginate and form junctions with the sheath cell, the sheath cell forms junctions with the socket cell, and the socket cell forms junctions with adjacent epidermal cells. Presumably a specific cell-cell adhesion is required before any permanent junction can form. In the embryo, many sensilla are assembled concurrently in a small region of the head. Thus each cell must adhere to its correct partners despite possible competition from nearby cells of similar type. This specificity of attachment is not absolute as hybrid sensilla can form when normal partners are removed (Sulston et al., 1983).

Invagination may be an early step in neuron/sheath interaction. In daf-19 mutants, where cilia are apparently not formed, the sensory dendrites still invaginate their sheath cells and form normal belt junctions.

The amphid dendrites show specific neuron/neuron interactions in addition to the neuron/sheath interactions observed in all sensilla. The 12 dendrites normally invaginate the sheath cell roughly in register. The channel cilia form a tight fascicle in the sheath cell which extends into the socket channel. Curiously, the arrangement of cilia within this fascicle is invariant in wild-type animals (Ward et al, 1975; Ware et al, 1975). It is unknown whether the ciliary pattern is inherited from the more complex pattern of the papillary nerves.

In mec-8 mutants, the amphid dendrites invaginate the sheath cell at irregular levels and their cilia do not fasciculate fully. A similar, if milder, defect in amphid fasciculation has been observed in mec-1 mutants (Lewis and Hodgkin, 1977; Chalfie and Sulston, 1981). Conceivably the mec-1 and mec-8 genes specify adhesive molecules that determine pairwise affinities of the amphid dendrites or their cilia. In addition, the mec-1 and mec-8 mutations disrupt the function of certain nonciliated mechanosensory neurons (Chalfie and Sulston, 1981). The mec-1 mutations were shown to prevent the normal attachment of these neurons to the hypodermis.

The lining and scaffold of the amphid sheath channel assemble around the fascicle of cilia. The sheath channel forms correctly in mutants with truncated or missing cilia suggesting that the dendrites, and not exclusively their cilia, can induce these structures. Small fascicles or isolated cilia in mec-8 mutants form separate channels that can accrete a scaffold and dark lining resembling the normal sheath channel.

The sheath matrix material appears to be synthesized at the lamellae, transported forward in membrane bound vesicles, and secreted from these vesicles into the sheath channel near the base of the cilia (Wright, 1980). The cilia themselves appear to induce the deposition of the matrix material. In mec-8 mutants with displaced cilia, the matrix material still deposits along them. It is also deposited around the ectopic cilia-like projections found in che-13, osm-1, and osm-5 dendrites.

In mutants with short or absent cilia, matrix material accumulates in large vesicles in the anterior sheath cytoplasm. Abnormal accumulations of large matrix vesicles have also been reported in the amphid sheath cells of che-2, che-3, and daf-6 mutants (Lewis and Hodgkin, 1977; Albert et al, 1981). It may be that matrix material is normally discharged from the cilia through the amphid openings. This would explain why it accumulates in the che-14, daf-6, and mec-8 mutants that have apparently normal cilia but obstructed channels.

The che-12 mutation appears to disrupt the synthesis or secretion of matrix by the sheath cells. Interestingly, empty vesicles still form at the lamellae, transport forward, and fuse with the channel lumen. Presumably the abnormal darkening of the channel cilia in the che-12 mutants is a degenerative change resulting from the loss of matrix normally surrounding the cilia. These mutants also have a defect in cuticle secretion by the epidermis.

The socket channel has a rather different origin than the sheath channel (Wright, 1980). The socket cells can wrap around and form junctions with themselves, thus creating a channel, even when there are no cilia to envelop. The scaffold that assembles around the channel cilia in the sheath cell may be important in joining the sheath and socket channels. In the absence of a well defined sheath channel, the socket channel sometimes ends in a blind pocket in mec-8 mutants. The sheath channel appears nearly normal in che-14 mutants but the join with the socket channel is defective. Conceivably the primary defect is in the sheath or socket scaffolds. The published description of daf-6 (e1377) mutants suggests they, too, may be defective in the joining of the sheath and socket channels (Albert et al, 1981). Consistent with this idea, Herman (1984) has shown that the genetic focus of the daf-6 phenotype is probably the sheath or the socket cell, or possibly both, but not the neurons (Table 2).

Assembly of Sensory Cilia
All classes of sensory cilia are absent in daf-19 mutants. Vestigial centrioles, without membrane attachments, were found in a few dendrites. No membrane-linked microtubules assemble ectopically in these mutants, suggesting (see below) that the wild type daf-19 product directly affects the peripheral doublets or their Y links. A mutation disrupting doublet-microtubules has been described in Chlamydomonas (Goodenough and St. Clair, 1975).

The che-13, osm-1, osm-5, and osm-6 mutations shorten the axonemes of all classes of cilia. Microtubules, attached to the membrane by Y links, assemble ectopically in these mutants. The number and lengths of these ectopic microtubules vary by neuron type and roughly parallel the normal lengths of cilia in these cells. We suggest these ectopic microtubules are misassembled components of the axoneme. Thus the peripheral doublets and Y links can apparently self-assemble and the wild-type products of these four genes are needed to ensure they assemble only on the ciliary template. Interestingly, the transition zones are fairly normal in these mutants. Also, the OLQ axonemes are probably less affected than other classes. It may be that additional structures, such as the apical ring in the transition zone and the filled microtubules or crossbridges in the OLQ axonemes, increase the stability of these segments even in the absence of normal che-13, osm-1, osm-5, and osm-6 products. Consistent with the idea that these genes encode ciliary components, Herman (1984) has shown that the wild-type osm-1 gene must be expressed in the neurons themselves for normal cilia as judged by FITC uptake.

All classes of sensory cilia in che-2 and che-3 mutants have been previously shown to have normal transition zones and truncated axonemes (Lewis and Hodgkin, 1977; Albert et al, 1981). It was reported that microtubules assemble ectopically below the transition zones in both of these mutants. Whereas the truncated amphid cilia in che-2, che-13, osm-1, osm-5, and osm-6 mutants are normal in diameter, the amphid cilia in che-3 mutants have enlarged, bulb-shaped endings filled with dark ground material (Lewis and Hodgkin, 1977).

The amphid channel cilia of che-11 and daf-10 mutants are irregular in contour, variably enlarged in diameter, and may contain dark ground material in the center of the axoneme (Albert et al, 1981). The che-11 channel cilia are nearly normal in length whereas the daf-10 cilia are described as foreshortened (Albert et al, 1981). Both mutants also affect the CEP cilia and, at least for che-11, probably other cilia. Perhaps related, dark ground material has been observed in the center of the axonemes in the bronchial epithelium of a human subject with immotile cilia (Afzelius, 1976).

The amphid cilia are usually absent in che-10 mutants. Instead, the dendrites have large, bulb-shaped endings filled with dark ground material. Occasional dendrites have well-formed cilia, suggesting the amphid defect is degenerative rather than developmental. The mechano-cilia appear normal but lack striated rootlets. It may be interesting to examine the amphid dendrites in embryos or L1 larvae of this strain.

The osm-3 mutation specifically eliminates the distal segments of the amphid channel cilia, leaving the middle segment and the transition zone unaffected. The distal segment differs from the middle segment in that the B subfibers of the peripheral doublets, and the membrane-links, are absent. The osm-3 product may be a protein specific to the distal segments of these cilia. Alternatively, it may affect the entire cilium, perhaps being associated with the A subfibers, but the distal segment is most vulnerable to its absence.

Dissociation of the IL2 Cilia
In wild-type adults, the IL2 neurons, and possibly some mechanosensory neurons, have incomplete cilia comprising fewer than nine doublets (Ward et al, 1975; Ware et al., 1975). Interestingly, in the che-13, osm-1, osm-5, and osm-6 mutants with truncated cilia, the transition zones of the IL2 cilia and the various mechanocilia are actually longer and better formed, in the sense of showing nine Y-linked doublet microtubules drawn together in a ring, than in wild type. We speculate that when they form all classes of cilia have complete transition zones, but certain classes, particularly the IL2 cilia, later dissociate or rearrange, leaving fewer microtubules and no recognizable nine-fold organization. Rearrangement might be expected if, for example, the neurons undersynthesize ciliary proteins during dendrite growth. Mutants which destabilize the axoneme might actually leave more material available for maintaining the transition zone than in the wild type. It may be interesting to examine the IL2 cilia in embryos or L1 larvae.

Striated Rootlets
Striated rootlets are frequently associated with the basal bodies of both sensory and motile cilia but their function is unknown. Salisbury and Floyd (1978) have shown that the rootlets of certain flagellate alga are contractile and that contraction is induced by calcium. A calcium-binding phosphoprotein of 20,000 MW is the principal component of the contractile rootlets from Tetraselmis striata (Salisbury et al, 1984). Striated rootlets have also been purified from several other sources (see Salisbury et al, 1984). In each case, only one or two proteins account for most of the protein in the purified rootlets. However, the molecular weights of these proteins vary widely and it remains to be seen how they are related.

The daf-19 mutation eliminates cilium formation, but striated rootlets still assemble in the appropriate dendrites. Interestingly, certain sensory neurons in C. elegans males contain striated rootlets but no associated cilia (Sulston et al., 1980). The rootlets are attached to dark plates, resembling hemidesmosomes, at the dendritic tips. These observations imply that rootlet and cilium formation can occur independently and an unknown mechanism ensures that the distal end of the rootlet attaches to the base of the cilium. A similar conclusion has been reached using basal body defective and rootlet defective mutants of Chlamydomonas (Goodenough and St. Clair, 1975; Wright et al., 1983). Interestingly, the ectopic membrane-linked microtubules in the che-13, osm-1, osm-5, and osm-6 mutants can recruit small rootlets. Presumably, the same affinity exists between the rootlets and microtubules of normal cilia.

The che-10 mutation eliminates striated rootlets from the mechanosensory cilia of the head. The wild type che-10 product may be a rootlet component. The amphid cilia, which lack striated rootlets in the wild-type, are badly degenerated in this strain. A possible explanation, that the strain harbors two mutations, one responsible for the rootlet defect and one for the amphid defect, can be resolved by isolating a second, independent che-10 mutant and examining its rootlets.

Mechanosensory Specializations and Modalities
The distal tips of the IL1 cilia each contain a dark membrane-attached disc. These discs are positioned in the cuticle at the base of the papillae in such a way as to be compressed by head-on contacts of the animal. The mutants with truncated or missing IL1 cilia show that the discs can assemble normally in the absence of cilia but they require the cilia to position them at the tip of the dendrite.
The supernumerary microtubules and associated dark material of the CEP cilia closely resemble the tubular bodies found in proved mechanocilia of insects (Thurm et al, 1983). The dark material, itself amorphous, appears to be molded into rods by the associated microtubules. A similar dark material is found at the tips of the OLL cilia where it forms a ball. The difference in shape may reflect the comparative paucity of microtubules in the OLL cilia. In mutants with truncated or missing cilia, the supernumerary microtubules and dark tubule-associated material in the CEP neurons assemble ectopically along the dendrite in both rod- and ball-shaped aggregates. This shows that these specializations can self-assemble but the cilia are needed to position them at the distal tip of the dendrite. In the cat-6 mutants, both the CEP cilia and their specializations appear normal but the association between them is specifically disrupted.

Amorphous dark material is also present in the OLQ cilia as small lumps that flank the circumferential corners of the joined square of microtubules. These lumps may also be connected to the membrane. In the wild-type, the corners of the square always point radially and circumferentially. There is no obvious structure joining the OLQ axoneme to the support cells or the cuticle that might provide orientation. A simple suggestion is that the square and associated dark material are aligned passively by repeated deformation of the cuticle as might occur during stimulation. Interestingly, the OLQ squares are sometimes misoriented and the dark lumps are fragmented and mispositioned in che-14 mutants which have abnormally thick subcuticle adjacent to the cilia. Perhaps relevant, the cilia of the respiratory epithelia, which normally are oriented to beat in a common direction, are randomly oriented, as judged by their basal feet, in subjects with immotile cilia (Afzelius, 1981). Here, it is thought that cilia form in random orientations and become oriented through a mechanism involving active beating.

As the CEP, OLL, and OLQ cilia are situated somewhat posterior to the IL1 cilia, they probably detect radial, rather than axial, displacements. The geometry of the OLQ cilia suggests they have substantial directional discrimination. The adjacent CEP cilia may be lower threshold, isotropic detectors.

The dark, cuticle-embedded nubbins of the CEP, OLL, and OLQ dendrites presumably provide mechanical anchorage of the dendrite to the cuticle. They are not a ciliary specialization as such since they persist in mutants with truncated cilia and, at least for the CEP dendrites, in the daf-19 mutants without cilia. In the cat-6 mutants with enlarged CEP nubbins or the che-14 mutants with abnormally thin cuticle, the nubbins can completely penetrate the cuticle and expose the dendrite to the medium.

A similar cuticle-embedded nubbin occurs in males at the distal tips of the CEM cilia. Here, it penetrates the cuticle and is believed to provide access of the CEM dendrite to the chemical environment. As there is no cuticular opening in the cephalic sensilla of hermaphrodites which lack the CEM neurons, the openings in males must be created by the CEM dendrites and not the cephalic socket cells. In contrast, the raised cuticle and pore of the inner labial sensilla appear to be formed by the innerlabial socket cell. They persist in mutants with truncated or missing IL1 and IL2 cilia.

Chemosensory Behaviors
C. elegans has at least five distinct chemosensory behaviors. First, it is attracted or repelled by a variety of small molecules at low concentrations (10 ^-3 M or below) (Ward, 1973; Dusenbery, 1974, 1975, 1976a, 1980a,c). Second, it is repelled by very high concentrations of various chemically unrelated solutes, including NaCl and fructose (Ward, 1973; Culotti and Russell, 1978). Third, when starved under crowded conditions, young larvae may differentiate into dauer larvae, a non-feeding, arrested stage, adapted for long-term survival and dispersal (Cassada and Russell, 1975). Crowding is sensed by the accumulation of a fatty-acid-like pheromone made constitutively by all animals (Golden and Riddle, 1982, 1984a,b). Fourth, chemical cues influence egg laying (Horvitz et al, 1982; Golden and Riddle, 1982; Trent et al, 1983). Fifth, males are attracted to hermaphrodites by an unknown attractant (H. Horvitz and J. Sulston, personal communication). The cephalic companions (CEM), a class of chemosensory neurons found only in males, may be detectors for an hermaphrodite pheromone.

The amphid and phasmid sensilla have long been suspected of mediating many of these chemosensory behaviors (Wright, 1980). Each class of neurons in these sensilla has distinct synaptic outputs, suggesting their cilia may detect different chemicals (Hall and Russell, 1986; White et al, 1986). The clearest evidence that the ten classes of channel cilia are required for chemotaxis, osmotic avoidance, and dauer larvae formation comes from the mutants osm-3, (p802), and daf-6 (el377). The ultrastructural defects in the heads of these mutants are confined to the amphid channel cilia and amphid sheath cell, respectively (Albert et al, 1981). These mutants fail to form dauer larva in response to pheromone, to avoid concentrated NaCl or fructose, or to chemotax toward dilute NaCl (Culotti and Russell, 1978; Albert et al, 1981).

Interestingly, osm-3 (p802) mutants are still responsive to some chemicals including pyridine, C0₂ and H+ (Dusenbery, 1980b). This may reflect some residual responsiveness of the shortened channel cilia. Alternatively, the amphid wing neurons or the IL2 neurons may be the principal detectors for these chemical species. Significantly, the osm-1 (p808) mutation, which affects the assembly of all classes of cilia, abolishes even these responses (Dusenbery, 1980b). Hence, all known chemical attractants, repellants, and pheromones are apparently sensed through ciliated receptors in C. elegans.

Competing levels of food and crowding pheromome are believed to regulate both entry into and exit from the dauer larval stage. Mutants with abnormal amphid cilia are generally incapable of forming dauer larva in response to crowding and starvation (Tables 1 and 2) or direct application of pheromome (Golden and Riddle, 1984a). Two of these mutants, daf-6 and daf-10, have been induced to form dauer larva by introducing second mutations which favor dauer formation (Albert et al, 1981; Riddle et al, 1981). In these genetic backgrounds, the daf-6 and daf-10 mutations inhibit exit from the dauer stage perhaps by preventing detection of a food signal. Paradoxically, the daf-19 mutants with no sensory cilia form dauer larva constitutively in the absence of crowding or starvation (D. Riddle, personal communication). This suggests that mutations affecting the sensory cilia can shift decisions to enter or exit the dauer stage in either direction

Horvitz et al. (1982) have reported that che-3, daf-10, and osm-3 mutants lack biochemically detectable levels of octopamine, a presumptive neurotransmitter found in the wild-type. The common defect of these three mutants is a disruption of the amphid and phasmid cilia. This suggests that these neurons either make octopamine or regulate the neurons which do.

Mating Behavior
Mating by C. elegans males is a complex behavior involving ten classes of male-specific sensory neurons (Ward et al, 1975; Sulston et al, 1980; Hodgkin, 1983). Ten genes that affect mechanosensory receptors in the head, che-2, che-3, che-10, che-11, che-13, daf-10, daf-19, osm-1, osm-5, and osm-6, are also required for mating (Table 1). These mutations likely prevent mating by disrupting male-specific sensilla in the tail. Most of these mutants show occasional fluorescein uptake into ray neurons, indicating that the ray sensilla are abnormal. The mating defect in che-10 (e1809) may be the consequence of missing striated rootlets normally found in dendrites of the ray, hook, and postcloacal sensilla (Sulston et al, 1980).

As expected, the various neurons of the amphid and phasmid sensilla are probably not important for male mating behavior as che-12, daf-6, osm-3, and ttx-1 mutants all mate efficiently (Table 1). Similarly, the efficient mating of cat-6 males implies that the ADE, CEP, and PDE neurons are not involved.

Possible Thermosensory Role of Amphid Finger Neuron (AFD)
The AFD dendrites are unique among sensory receptors in C. elegans in having numerous fingers that invaginate the surrounding sheath cell. These fingers, which are topologically proximal to the AFD cilia, do not depend on the cilia for formation since mutants with reduced axonemes (che-13, osm-1, osm-5, and osm-6) or no cilia (daf-19) have normal fingers.

R. Ware has suggested, based on his unpublished observations on ttx-1 (p767) mutants, that the AFD neurons may be thermosensory. As confirmed here, the AFD fingers are entirely missing and the AFD cilia are longer than normal in ttx-1 mutants. The other sensory receptors in the head appear ultrastructurally normal.

When placed in a thermal gradient, wild-type animals move toward the temperature at which they were previously raised (Hedgecock and Russell, 1975). The ttx-1 thermotaxis mutants seek the cold regardless of their thermal history. These mutants are also hyper-responsive to dauer-inducing pheromone (Golden and Riddle, 1984a,b). Elevated temperatures are known to lower the pheromone threshold for dauer-larva formation in wild-type animals. This suggests that both the cryophilia and heightened pheromone sensitivity of the ttx-1 mutants may reflect a common sensory defect in which the animal perceives a higher temperature than actual. Other sensory behaviors, including chemotaxis and mating, are normal in this strain (Hedgecock and Russell, 1975; Dusenbery and Barr, 1980).

A reduction in the number of fingers on the AFD dendrites has also been reported for the che-1 mutants, e1034 and a74 (formerly DD74) (Lewis and Hodgkin, 1977; R. Ware, D. Clark, M. Salzay, and R. Russell, personal communication). No thermotaxis defects were detected in population assays of che-1 mutants (Hedgecock and Russell, 1975). However, the finger abnormality in che-1 mutants is variable and comparatively mild. About half of the che-1 (e1034) animals examined by Lewis and Hodgkin (1977), for example, had normal or nearly normal AFD dendrites.

It may be possible to confirm a role for the AFD neurons in thermal behavior by killing these cells with a laser microbeam (Sulston and White, 1980) and testing the animals in individual thermotaxis assays (Hedgecock and Russell, 1975).

Photosensory Behavior
Burr (1985) has reported that C. elegans responds to light by reversing, and consequently changing the direction of movement, more frequently than in the dark. This is a nonoriented response but, in principle, could be used to keep animals away from lighted areas (Fraenkel and Gunn, 1961). The light appears to act directly and not by radiant heating.

In nematodes with true phototaxis, the ocelli comprise a pair of amphid dendrites plus nearby pigment spots in the pharynx which provide shadowing (see Burr, 1985). Although C. elegans lacks obvious photopigments or shadowing pigments, the AWC neurons are plausible candidates for photoreceptors as their cilia have extremely large membrane areas. Tests of mutants such as daf-19 may help ascertain whether the photoresponse in C. elegans is mediated by ciliated sensory neurons.

Evolution of Sensory Cilia
Motile cilia, found in unicellular eukaryotes, lower plants, and animals, are believed to be ancient organelles. The sensory cilia of animals probably arose by later modification of motile cilia. In nematodes, the motile functions of cilia have apparently been lost. Their spermatozoa are nonflagellated and move by extending contractile pseudopodia (Ward et al, 1982), and there are no ciliated epithelia. In contrast, the sensory functions of cilia are highly elaborated. Wright (1983) has suggested, that since there is no selective pressure to maintain ciliary structures used strictly for motility, nematode cilia may be simpler than in other animals. For example, the dynein and nexin arms, radial spokes, and central pair of singlet microtubules that generate the sliding force in motile axonemes and control the flexion are all apparently absent in nematode cilia.

The absence of basal bodies seems a paradox as they are believed to have two functions, one of which is essential. First, they are the templates for the ninefold structure of the axonemes. The nine doublet microtubules of the axoneme are a direct extension of the A and B subfibers of the nine triplet-microtubules in the basal body. Second, basal bodies are attachment points for cytoplasmic microtubules which anchor the cilium to the cytoskeleton. This coupling is essential for transmitting force from a beating cilium into cell motion. It may also be useful for holding cilia erect from the cell surface.

In nematodes, the mechanical role of the basal body is probably not needed. The template role would be filled if the centriole is present only transiently to initiate the cilium and then disappears. Alternatively, the transitional fibers themselves could be the residue of the centriole. Importantly, nematode centrioles are composed of singlet microtubules, plus some attachments that may be vestiges of B and C subfibers, rather than triplets (D. Albertson, A. Crowther, and J. N. Thomson, personal communication). Finally, in view of these departures from what are usually regarded as universal characteristics of centrioles and cilia, it is worth mentioning that microtubules themselves may be unusual in nematodes. Cytoplasmic microtubules in nerve processes contain only 11 protofilaments rather than the more usual 13 protofilaments (Chalfie and Thomson, 1982).

Acknowledgments

Our many colleagues who generously provided strains are mentioned under Materials and Methods. We thank R. Ware for sharing his unpublished observations on the sensilla of chemosensory and thermo-sensory mutants; J. Weiss for illustrations; and E. Aamodt, P. Albert, D. Albertson, A. Burr, M. Chalfie, D. Dusenbery, L. Gremke, D. Hall, R. Herman, J. Hodgkin, C. Kenyon, B. Menco, D. Riddle, R. Russell, S. Siddiqui, J. Sulston, S. Ward, J. White, and K. Wright for ideas and discussions. In sadness, we acknowledge the assistance and kindness of Kay Buck who died unexpectedly during the course of this work. Part of this research was supported by a Basil O'Connor starter grant from the March of Dimes Foundation and by NIH Grants NS16510 and NS20258 to J.C. E.H. was recipient of postdoctoral fellowships from the Muscular Dystrophy Association of America and the NIH.

References

APZELIUS, B. A. (1976). A human syndrome caused by immotile cilia. Science 193, 317-319.

AFZELIUS, B. A. (1981). Genetic disorders of cilia. In "International Cell Biology 1980-1981" (H. G. Schweiger, ed.), p. 440. Springer-Verlag, Berlin.

ALBERT, P. S., BROWN, S., and RIDDLE, D. L. (1981). Sensory control of dauer larva formation in Caenorhabditis elegans. J. Comp. Neurol 198, 435-451.

BRENNER, S. (1974). The genetics of Caenorhabditis elegans. Genetics 77, 71-94.

BURR, A. H. (1985). The photomovement of Caenorhabditis elegans. A  nematode which lacks ocelli. Proof that the response is to light not radiant heating. Photochem.  Photobiol 41, 577-582.

CASSADA, R. C., and RUSSELL, R. L. (1975). The dauerlarva, a post-embryonic developmental variant of the nematode Caenorhabditis  elegans. Den. Biol 46, 326-342.

CHALFIE, M., and SULSTON, J. (1981). Developmental genetics of the mechanosensory neurons of Caenorhabditis elegans. Dev. Biol. 82, 358-370.

CHALFIE, M., and THOMSON, J. N. (1982). Structural and functional diversity in the neuronal microtubules of Caenorhabditis elegans. J. Cell Biol 93,15-23.

CULOTTI, J. G., and RUSSELL, R. L. (1978). Osmotic avoidance defective  mutants of the nematode Caenorhabditis elegans. Genetics 90, 243-256.

DUSENBERY, D. B. (1974). Analysis of chemotaxis in the nematode Caenorhabditis elegans by countercurrent separation. J. Exp. Zool. 188, 41-47.

DUSENBERY, D. B. (1975). Avoidance of D-tryptophan by the nematode Caenorhabditis elegans. J. Exp. Zool. 193, 413-418.

DUSENBERY, D. B. (1976a). Attraction of the nematode Caenorhabditis elegans to pyridine. Comp. Biochem. Physiol C 53, 1-2.

DUSENBERY, D. B. (1976b). Chemotactic behavior of mutants of the nematode Caenorhabditis elegans that are defective in their attraction to NaCl. J. Exp. Zool 198, 343-352.

DUSENBERY, D. B. (1980a). Appetitive response of the nematode Caenorhabditis elegans to oxygen. J. Comp. Physiol 136, 333-336.

DUSENBERY, D. B. (1980b). Chemotactic behavior of mutants of the nematode Caenorhabditis elegans that are defective in osmotic avoidance. J. Comp. Physiol 137, 93-96.

DUSENBERY, D. B. (1980c). Behavior of free-living nematodes. In "Nematodes as Biological Models," (B. M. Zuckerman, ed.), Vol. 1, pp. 127-158. Academic Press, New York.

DUSENBERY, D. B., and BARR, J. (1980). Thermal limits and chemotaxis in mutants of the nematode Caenorhabditis elegans defective in thermotaxis. J. Comp. Physiol 137, 353-356.

DUSENBERY, D. B., SHERIDAN, R. E., and RUSSELL, R. L. (1975). Chemotaxis-defective mutants of the nematode Caenorhabditis elegans. Genetics 80, 297-309.

FRAENKEL, G. S., and GUNN, D. L. (1961). The Orientation of Animals. Dover, New York.

GILULA, N. B., and SATIR, P. (1972). The ciliary necklace: A ciliary membrane specialization. J. Cell Biol 53, 494-509.

GOLDEN, J. W., and RIDDLE, D. L. (1982). A pheromone influences larval development in the nematode Caenorhabditis elegans. Science 218, 578-580.

GOLDEN, J. W., and RIDDLE, D. L. (1984a). A pheromone-induced developmental switch in Caenorhabditis elegans: Temperature-sensitive mutants reveal a wild-type temperature-dependent process. Proc. Natl Acad. Sci. USA 81, 819-823.

GOLDEN, J. W., and RIDDLE, D. L. (1984b). The Caenorhabditis elegans dauer larvae: Developmental effects of pheromone, food, and temperature. Dev. Biol. 102, 368-378.

GOODENOUGH, U. W., and ST. CLAIR, H. S. (1975). Bald-2: A mutation affecting the formation of doublet and triplet sets of microtubules in Chlamydomonas reinhardtii. J. Cell Biol 66, 480-491.

HALL, D. H., and RUSSELL, R. L. (1986). Electron microscopic reconstruction of the posterior nervous system of the nematode Caenorhabditis elegans. Philos. Trans. R. Soc., in press.

HEDGECOCK, E. M., and RUSSELL, R. L. (1975). Normal and mutant thermotaxis in the nematode Caenorhabditis elegans. Proc. Natl Acad. Sci USA 72, 4061-4065.

HEDGECOCK, E. M., CULOTTI, J. G., THOMSON, J. N., and PERKINS, L. A. (1985). Axonal guidance mutants of Caenorhabditis elegans identified by filling sensory neurons with fluorescein dyes. Dev. Biol 111, 158-170.

HERMAN, R. K. (1984). Analysis of genetic mosaics in the nematode Caenorhabditis elegans. Genetics 108,165-180.

HODGKIN, J. (1983). Male phenotypes and mating efficiency in Caenorhabditis elegans. Genetics 103, 43-64.

HODGKIN, J., HORVITZ, H. R., and BRENNER, S. (1979). Nondisjunction mutants of the nematode Caenorhabditis elegans. Genetics 91, 67-94.

HORVITZ, H. R., CHALFIE, M., TRENT, C., SULSTON, J. E. and EVANS, P. D. (1982). Serotonin and octopamine in the nematode Caenorhabditis elegans. Science 216, 1012-1014.

KUNG, C., CHANG, S.-Y., SATOW, Y., VAN HOUTEN, J., and HANSMA, H. (1975). Genetic dissection of behavior in Paramecium. Science 188, 898-904.

LEWIS, J. A. and HODGKIN, J. A. (1977). Specific neuroanatomical changes in chemosensory mutants of the nematode Caenorhabditis elegans. J. Comp. Neurol 172, 489-509.

LUCK, D. J. L. (1984). Genetic and biochemical dissection of the eukaryotic flagellum. J. Cell Biol 98, 789-794.

RAND, J. B., and RUSSELL, R. L. (1984). Choline acetyltransferase-deficient mutants of the nematode Caenorhabditis elegans. Genetics
106, 227-248.

REESE, T. S. (1965). Olfactory cilia in the frog. J. Cell Biol 25,209-230.

RIDDLE, D. L., SWANSON, M. M., and ALBERT, P. S. (1981). Interacting genes in nematode dauer larva formation. Nature (London) 290,668-671.

RINGO, D. D. (1967). Flagellar motion and fine structure of the flagellar apparatus in Chlamydomonas. J. Cell Biol 33, 543-571.

SALISBURY, J. L., and FLOYD, G. L. (1978). Calcium-induced contraction of the rhizoplast of a quadrilflagellate green alga. Science 202, 975-977.

SALISBURY, J. L., BARON, A., SUREK, B., and MELKONIAN, M. (1984). Striated flagellar roots: Isolation and partial characterization of a calcium-modulated contractile organelle. J. Cell Biol 99, 962-970. SULSTON, J., DEW, M., and BRENNER, S. (1975). Dopaminergic neurons in the nematode Caenorhabditis elegans. J. Comp. Neurol. 163, 215-226.

SULSTON, J. E., and HORVITZ, H. R. (1977). Postembryonic cell lineages of the nematode Caenorhabditis elegans. Dev. Biol 56,110-156.

SULSTON, J. E., and WHITE, J. G. (1980). Regulation and cell autonomy during postembryonic development of Caenorhabditis elegans. Dev.
Biol 78, 577-597.

SULSTON, J. E., ALBERTSON, D. G., and THOMSON, J. N. (1980). The Caenorhabditis elegans male: Postembryonic development of nongonadal structures. Dev. Biol 78, 542-576.

SULSTON, J. E., SCHIERENBERG, E., WHITE, J. G., and THOMSON, J. N.(1983). The embryonic cell lineage of the nematode Caenorhabditis elegans. Dev. Biol. 100, 64-119.

SWANSON, M. M., EDGLEY, M. L., and RIDDLE, D. L. (1984). Genetic Maps 3, 286-299.

THURM, U., ERLER, G., GODDE, J., KASTRUP, H., KEIL, TH., VVOLKER,W., and VOHWINKEL, B. (1983). Cilia specialized for mechanoreception. J. Submicrosc. Cytol 15, 151-155.

TRENT, C., TSUNG, N., and HORVITZ, H. R. (1983). Egg-laying defective mutants of the nematode Caenorhabditis elegans. Genetics 104,619-647.

WARD, S. (1973). Chemotaxis by the nematode Caenorhabditis elegans:Identification of attractants and analysis of the response by use of mutants. Proc. Natl Acad. Sci. USA 70, 817-821.

WARD, S., THOMSON, N., WHITE, J. G. and BRENNER, S. (1975). Electron microscopic reconstruction of the anterior sensory anatomy of the nematode Caenorhabditis elegans. J. Comp. Neurol 160, 313-338.

WARD, S., ROBERTS, T. M., NELSON, G. A., and ARGON, Y. (1982). The development and motility of Caenorhabditis elegans spermatozoa. J.
Nematol. 14, 259-166.

WARE, R. W., CLARK, D., CROSSLAND, K., and RUSSELL, R. L. (1975).The nerve ring of the nematode Caenorhabditis elegans: Sensory input and motor output. J. Comp. Neurol 162, 71-110.

WHITE, J. G., SOUTHGATE, E., THOMSON, J. N., and BRENNER, S. (1986). The nervous system of Caenorhabditis elegans. Philos. Trans. R. Soc., in press.

WRIGHT, K. A. (1980). Nematode sense organs. In "Nematodes as Biological Models." (B. M. Zuckerman, ed.), Vol. 2, pp. 237-295. Academic Press, NewYork.

WRIGHT, K. A. (1983). Nematode chemosensilla; form and function. J.Nematol 15,151-158.

WRIGHT, R. L., CHOJNACKI, B., and JARVIK, J. W. (1983). Abnormal basal-body number, location, and orientation in a striated fiber-defective mutant of Chlamydomonas reinhardtii. J. Cell Biol 96,1697-1707.