Heterotrimeric Kinesin-II Is Required for the Assembly of Motile 9+2 Ciliary Axonemes on Sea Urchin Embryos

The monoclonal antibodies used in these studies were described previously (the antikinesin, SUK-4 in Ingold et al. [1988]; the anti–kinesin-II mAbs in Cole et al. [1993] and Henson et al. [1995]). The control nonspecific mouse IgG antibody was obtained from Sigma Chemical Co. (St. Louis, MO). Wesson canola oil was obtained locally. All other chemicals were obtained from Sigma Chemical Co., or Fisher Scientific (Pittsburgh, PA). Lytechinus pictus sea urchins were obtained from Marinus, Inc. (Venice, CA) and kept year round in refrigerated aquaria on an extended daylight cycle, fed with dried kelp (Bodega Farms, Bodega Bay, CA), in sea water collected and filtered at the University of California Davis Bodega Marine Laboratory (Bodega Bay, CA). Strongylocentrotus purpuratus sea urchins were collected from tidepools on the northern California coast, north of Bodega Bay, and maintained in 1,000-gallon holding tanks at Bodega Marine Laboratory.

High-speed supernatant of Strongylocentrotus purpuratus was prepared as described (Buster and Scholey, 1991). MTs were prepared from S. purpuratus high-speed supernatant in the presence of AMP-PNP or ATP as previously described (Scholey et al., 1985; Cole et al., 1993; Wedaman et al., 1996). Protein from whole and deciliated L. pictus blastulae was prepared by solubilizing pelleted embryos directly in 5× sample buffer (25% glycerol, 0.156 M Tris-HCl, pH 6.8, 5% sodium dodecyl sulfate, 12.5% β-mercaptoethanol, 0.0025% bromophenol blue). L. pictus blastulae were deciliated, and the cilia were isolated essentially by the method of Stephens (1986), modified to include an extra centrifugation step to reduce the possibility of cell body contamination of the cilia. Isolated cilia were immediately solubilized in 5× sample buffer. Protein concentrations in sample buffer were determined by the method of Minamide and Bamburg (1990). SDS-PAGE was performed in duplicate on 7.5% resolving gels, and one gel was stained with Coomassie blue while the second gel was transferred to nitrocellulose and probed with K2.4 anti–kinesin-II primary antibody (1.0 or 1.5 μg/ml) overnight at 4°C and then with horseradish peroxidase– conjugated goat anti–mouse secondary antibody (1:2,000) for 1 h at room temperature. Blots were developed using the enhanced chemiluminescence detection system (Amersham Corp., Arlington Heights, IL) exposed for 2 min (S. purpuratus MT protein) or 4–20 min (L. pictus blastulae protein). For densitometry, 100 μg of blastulae protein, deciliated blastulae protein, and cilia protein were loaded in adjacent lanes and, assuming equivalent protein transfer occurred in adjacent lanes, K2.4-labeled bands were analyzed on Western blots using a Stratagene (La Jolla, CA) Eagle Eye II gel documentation system.

K2.4 mouse monoclonal antibody for microinjection was prepared from frozen ascites using Affigel protein A and the MAPS II antibody purification system (Bio-Rad Laboratories, Inc., Hercules, CA). SUK-4 was purified essentially as described (Ingold et al., 1988; Wright et al., 1991, 1993). Antibody was then concentrated and exchanged into aspartate injection buffer (AIB: 150 mM potassium aspartate, 100 mM potassium phosphate, pH to 7.2 with potassium hydroxide) using NanoSpin Plus centrifugal devices (10,000 molecular mass cut-off; Gelman Sciences, Ann Arbor, MI). Protein concentration was determined by Bradford microassay with a γ-globulin standard (Bio-Rad Laboratories, Inc.); antibody was diluted to 2.5–10.0 mg/ml in AIB, finally filtered through Ultrafree-MC 0.22-μm filter units (Millipore Corp., Bedford, MA), and stored at 4°C for use within 2 wk. After this period, effects of K2.4 on ciliogenesis began to diminish.

Microinjection of sea urchin embryos was performed by a procedure modified from Kiehart (1982) and Wright et al. (1993). Sea urchin gametes were collected and fertilized according to Wright et al. (1991, 1993). Lateral injection chambers (Kiehart, 1982) allowed high-resolution observation of injections and development and, critical to the success of these experiments, gently restrained the embryos during ciliogenesis at blastula stages. Fertilized eggs were injected with 2.5–5% of cell volume between 20 min after fertilization and the first prometaphase, resulting in an intracellular antibody concentration of ∼500 nM. If the injected cell was in early mitosis, antibody was injected at a point distant from the spindle. Embryos routinely survived the injections and reached blastula stage for analysis. Embryos were observed by differential interference contrast microscopy on an inverted microscope (model IM-35; Carl Zeiss, Inc., Thornwood, NY) using a Plan 40× or Planapo 63× objective and imaged as described (Wright et al., 1993). An Argus 10 Image Processor (Hamamatsu Photonics) was occasionally used for contrast enhancement.

To ensure that differences observed between conditions were due to differences in injected agent rather than to variation in batches of eggs, or to exact age of development after fertilization, experimental and control conditions were always performed side-by-side on eggs fertilized at the same time and handled identically. Thus, the quantitative results presented were obtained from embryos under experimental and control conditions lying side-by-side in the injection chamber, although qualitatively identical results were consistently obtained with embryos not included in this paper.

Embryos were routinely observed to develop at similar rates and reach ciliogenesis within the same 30-min period after 13–14 h of development whether injected or not. To quantitate this observation, values for “Average duration of cell division cycle” (see Table I) were determined for the fourth through seventh cell division cycles for nine embryos from the same batch of eggs, side-by-side in the same injection chamber. Four K2.4-injected embryos, three SUK-4–injected embryos, and two uninjected embryos were observed sequentially, one every 45 s, continuously over a 5-h period. Onset of first observed nuclear envelope breakdown was used as an unambiguous temporal marker for the beginning of the next cell cycle.

K2.4-injected embryos were scored for “Percent blastulae swimming” by two criteria: First, embryos had to develop into morphologically normal round blastulae, and second, nearby control embryos had to be observed swimming (rotating) in the injection chamber at the same time as the K2.4 embryos remained stationary.

To determine “Percent cilia motile” (see Table II) we visually monitored equatorial regions of embryos that were gently compressed between the coverslips of the injection chamber (Kiehart, 1982), where cilia could move freely if they were motile. Equators of such embryos were videotaped, with care being given to capturing images of ciliary bases where each cilium met the surface of its blastomere. After videotaping, the equatorial surfaces of embryos were traced directly from the video monitor onto plastic transparencies, ciliary bases were marked, and each cilium was scored for motility. Because normal ciliary motility consists of an initiation of bending combined with propagation of the initial bend down the length of the axoneme (Brokaw, 1994), we scored only initiation of bending as a criterion for motility since cilia on K2.4-injected embryos are perhaps too short to exhibit propagation of bending. Specifically, if the first 3 μm of cilium above the blastomere surface showed a 10° change of angle or more from one apogee to the other, then this cilium was defined as “initiating a ciliary bend” and was scored as “motile.” Thus, a cilium was considered motile even if it did not propagate the initiated bend, as is the case for most long apical tuft cilia of the animal pole. By conservatively defining motility as a function of bend initiation, we restrict those cilia described as being “immotile” to those cilia that remain exceptionally still. Control cilia typically initiate a ciliary wave by bending at the base through almost 90°.

To collect data for determination of “Average length of cilia” and “Density of cilia” (see Tables II and III), dozens of cilia were visualized simultaneously by video-enhanced differential interference contrast microscopy of the bottom face of embryos compressed gently against the coverslip in the injection chamber. The cilia were trapped in this region and held still in a single focal plane by the embryo's surface. All cilia whose full length was visible in a single focal plane in these en face regions were traced onto a plastic transparency. Cilia that passed out of the focal plane were clearly distinguished at the periphery of this flattened face by changing focus. Ciliary bases were distinguished from ciliary tips by a broadening of the base visible at the blastomere's surface. For cilia where the entire length was not clearly visible, its base was marked on the transparency with a dot. To generate more length measurements and to ensure that ciliary length data were not biased by sampling only cilia trapped at the embryo surface, cilia lengths were also traced at the embryo equatorial regions described above. Equatorial video footage was viewed frame by frame until a single cilium's entire length, from base to tip, appeared in focus in a single video frame, at which point its full length was traced. The accuracy of this method in reproducibly capturing the full length of a cilium was confirmed by repeatedly tracing cilia that appeared in focus multiple times.

To determine “Average length of cilia,” lengths of traced cilia were measured in millimeters on the transparencies and then converted to micrometers from calibrations of the image dimensions using recorded images of a stage micrometer. To produce an average length for cilia sampled from the entire embryo surface, measurements of cilia from equatorial and en face traces were pooled.

We observed nonnormal distributions of length measurements, so to statistically compare the ciliary length difference between K2.4-injected and control embryos, we performed a Krustal–Wallis test for nonparametric analysis of variance (Zar, 1984) rather than a single factor analysis of variance. By the Krustal–Wallis test, the lengths of cilia on K2.4-injected, SUK-4–injected, nonspecific IgG, and uninjected embryos were found to be highly significantly different (test statistic, H_c, = 171.048; critical value for H = 16.266 at α = 0.001). To determine which mean ciliary lengths were significantly different from the others, we used a nonparametric multiple comparison test similar to the Tukey test (Zar, 1984) but which does not assume population normality or homogeneity of variance. We performed nonparametric multiple comparisons of ciliary lengths by ranking the ciliary lengths and sequentially comparing each condition with the others from the largest difference to the least. We thereby determined that a highly significant difference (α = 0.005) exists between the lengths of cilia on K2.4-injected embryos and every control condition (test statistic, Q, = 12.7552 between K2.4-injected and uninjected embryos; 7.3038 between K2.4-injected and SUK-4–injected embryos; 5.8066 between K2.4-injected and nonspecific IgG-injected; critical value for Q_0.005,4 = 3.342). Such a difference does not exist between the control conditions (Q = 3.1771 between nonspecific IgG–injected and uninjected embryos; 1.5182 between SUK-4–injected and uninjected embryos; 1.3095 between SUK-4–injected and nonspecific IgG-injected embryos).

The “Density of cilia” values were determined from the en face traces as follows. After marking the bases of every cilium visible on the flattened embryo face with a dot on the embryo trace, an irregular polyhedron was drawn connecting the most peripheral dots. In cases where large inclusions (such as the injected oil droplet) obscured a region of the embryo face, a polyhedron was drawn between the most peripheral dots so as to exclude the inclusion from the circumscribed area. The number of ciliary bases circumscribed by the polyhedron, including bases at the polyhedron corners, were counted. To determine the area of the circumscribed polyhedron, a grid of 1-cm² squares was overlaid on it so that the longest side of the polyhedron aligned with a gridline and so the dot at one end of that polyhedron side sat atop an intersection point (vertex) on the grid. With the longest side of the polyhedron away from the scorer, each grid vertex was designated as representing the 1 cm² above it and to its left. The total number of grid vertices within the polyhedron was counted, and any vertex that fell on the polyhedron boundary line was counted if the 1 cm² above it and to its left was within the polyhedron. The total number of grid vertices circumscribed by the polyhedron therefore represented the total number of centimeters squared circumscribed by the polyhedron and was converted to micrometers squared from calibrations of the image dimensions using recorded images of a stage micrometer. Dividing the pooled total number of circumscribed ciliary bases by the pooled total number of circumscribed micrometers squared and multiplying by 100 produced the Density of cilia per 100 μm² values in Table II. We used density per 100 μm² because at ciliogenesis, 100 μm² was the apparent apical surface area of one blastomere.

After microinjection, sea urchin embryos were transferred one by one to separate wells of an observation chamber where their development could be monitored through the time when uninjected control embryos withdrawn from the same injection chamber were ciliated and swimming vigorously. After confirming that all cilia on the K2.4-injected embryos were completely paralyzed, injected embryos were withdrawn individually from the observation chamber and transferred to fixation buffer. Control embryos received identical treatment and processing. Embryos were prepared for EM using a procedure modified from Masuda and Sato (1984) after several other protocols produced either poor ciliary morphology as seen in the light microscope, or inadequate fixation as seen in the electron microscope. Although this protocol uses detergent in the fixation step producing demembranated axonemes attached to the embryo surface, the axonemes remained intact and provided consistently better preserved ultrastructure than other fixations attempted. Ciliated blastulae were fixed for 1 h in 2.5% glutaraldehyde, 0.1 M Pipes, pH 8, 0.8 M glycerol, 0.01 M EGTA, 2 mM MgSO₄, and 1% saponin; washed three times in 0.1 M Pipes, pH 8.0, 0.8 M glycerol, 1 mM EGTA, 0.5 mM MgSO₄; treated for 1 h in the same buffer containing 0.2% tannic acid; washed three times in 0.1 M Pipes, pH 8, 0.8 M glycerol, 0.5 mM MgSO₄; postfixed for 1 h in 1% OsO₄, 0.1 M phosphate buffer, pH 8, 0.3 M sucrose; and finally washed in 0.1 M phosphate buffer, pH 8, 0.3 M sucrose. All steps were performed at 25°C. The embryos were dehydrated through an acetone series that included 1 h in 2% uranyl acetate in 70% acetone and were flat-embedded in Embed-812 (Electron Microscopy Sciences, Fort Washington, PA). 60– 70-nm sections were cut and stained with uranyl acetate and lead citrate and then imaged on an electron microscope (Model 410; Phillips Electronic Instruments, Co., Mahwah, NJ) operated at 80 kV.

To quantitate the frequency of occurrence of central pair apparati in cross and oblique sections of ciliary axonemes by EM, we only scored axonemes in which outer doublets were distinct. For cross sections, axonemes were scored as having a central pair when nine outer doublet MTs could be localized at the same time that 20-nm-diam rings or circles could be localized in the axoneme interior (for example, Fig. 6,E). Axonemes were scored as having centralized amorphous material, but no recognizable central pair, when no central 20-nm-diam rings or circles were observed in the axoneme interior but electron-dense material still appeared focused at the center of the cavity (for example, Fig. 6, G and H). Axonemes were scored as completely lacking central pairs when neither 20-nm rings or circles nor centralized amorphous material was observed (for example, Fig. 6 F). In addition, oblique sections (which cut completely across the axoneme) were scored, while longitudinal sections were not, to avoid incorrectly scoring any grazing sections that did not cut into the central pair. For oblique sections, axonemes were scored as having a central pair when linear walls of any central pair MT were visible in the central density. Data were derived from 21 cross or oblique sections from 4 control embryos, and 39 cross or oblique sections from 5 K2.4-injected embryos.

The experimental monoclonal anti–kinesin-II antibody used in the microinjection experiments described here, K2.4, was generated against purified Strongylocentrotus purpuratus kinesin-II and shown to react specifically with the AMP-PNP–enhanced, ATP-sensitive MT-binding SpKRP₈₅ subunit of S. purpuratus kinesin-II (Fig. 1,A; Cole et al., 1993; Henson et al., 1995). To ensure that K2.4 reacts specifically with kinesin-II in the Lytechinus pictus embryos that were used in these microinjection studies at the stage where its inhibitory effect was observed (below), we probed whole L. pictus blastula-stage embryos (before and after deciliation) and isolated cilia (Stephens, 1986) with K2.4. We observed a strong, specific reaction with SpKRP₈₅ in blastulae before and after deciliation and a fainter reaction with cilia (Fig. 1 B), indicating first that K2.4 is specific for kinesin-II in L. pictus blastulae, and second that the majority of the kinesin-II is stockpiled in the blastula cell bodies with a small fraction being present in cilia. Blot densitometry indicated that cilia contain between 5–10% as much kinesin-II per milligram total protein as is contained in cell bodies.

The three controls for the microinjection of anti–kinesin-II monoclonal antibodies were uninjected embryos, embryos injected with a nonspecific mouse IgG, and embryos injected with the antikinesin mAb SUK-4, which is the same isotype as K2.4, binds L. pictus kinesin, blocks kinesin motor function in vitro (Ingold et al., 1988; Wright et al., 1991, 1993), and blocks the delivery of exocytic vesicles to the plasma membrane (Steinhardt et al., 1994; Bi et al., 1997).

Kinesin-II immunolocalizes to punctate, vesicle-like structures in sea urchin embryonic mitotic spindles, suggesting that it might participate in mitosis or cell division (Henson et al., 1995). To investigate this possibility, we examined experimental and control embryos for effects on progression through mitosis and cytokinesis during cleavage-stage development (Fig. 2). We observed that embryonic cells containing anti–kinesin-II mAbs displayed identical average cell cycle duration to the uninjected controls and no significant difference from nonspecific antibody-injected controls (Fig. 2, Table I). This result did not support the hypothesis that kinesin-II has an essential function in mitosis or cell division.

Kinesin-II immunolocalizes to punctate, detergent-insensitive particles in flagellar axonemes, and with basal body– like structures in the sperm midpiece (Henson et al., 1997), suggesting that it might participate in axoneme assembly or function. In support of this hypothesis, we observed that the microinjection of anti–kinesin-II mAbs produced a dramatic inhibition of the assembly of motile cilia on blastula stage embryos. An obvious consequence of this effect was that K2.4-injected embryos remained stationary in the injection chamber indefinitely, while uninjected or control-injected embryos formed normal motile cilia with which they swam around in the sea water of the injection chamber (Fig. 2, final panel; Table I).

Visual examination of the surfaces of K2.4-injected embryos revealed that the embryos' inability to swim resulted from defects in ciliogenesis leading to the presence of short, completely paralyzed cilia on the blastula surface (Fig. 3). Thus, high magnification observations of the surfaces of swimming control blastulae revealed the presence of long beating cilia, whereas the anti–kinesin-II–injected embryos possessed equal numbers of very short, immotile cilia over their outer surfaces (Fig. 3). The effects of anti– kinesin-II antibody microinjection on ciliogenesis were quantitated in detail (Table II). This analysis revealed that virtually all cilia on the K2.4-injected embryos were completely immotile, and they averaged typically less than half the length of cilia on control embryos (7.3 μm long versus 12–17 μm long for controls; Table II). When analyzed statistically as described under Materials and Methods, ciliary lengths on K2.4-injected embryos were highly significantly different than all three control conditions, while the control conditions were not highly significantly different from each other. In addition, the short cilia on the anti–kinesin-II–injected blastulae all had a conspicuous, knoblike bulb at their distal tips, apparently irrespective of the length or age of the cilium (Fig. 3 C). These knobs were found at the tips of virtually all the K2.4-injected cilia, where they remained stationary and stable, but they were exceedingly rare on control cilia.

The simplest interpretation of these results is that the microinjection of the anti–kinesin-II antibody interferes with the assembly of motile cilia. However, it could be argued that the short ciliary lengths observed in K2.4-injected embryos is an artifact resulting from fortuitously measuring ciliary lengths at the start of the 60-min-long cell cycle (Table I) when ciliogenesis has just begun; cilia measured at this time would obviously be transiently shorter than normal because of the timing of the measurements of ciliary lengths. To rule this out, we analyzed cilia on the same embryos at two different time points, 40 min apart (Table III), and observed that the average ciliary length in K2.4-injected blastulae plateaus at a short value, suggesting that the short ciliary length on K2.4-injected embryos represents an “end point” that does not change significantly over subsequent time. The fact that the density of cilia per unit area increases significantly between these observation times (Table III) confirms that the blastomeres were continuing to divide while the average ciliary length had reached a plateau.

To control for potential differences in the response to microinjection between individual embryos, we injected one cell of a two-cell embryo with the anti–kinesin-II antibody (Fig. 4). Remarkably, the resulting blastula was covered on the side derived from the K2.4-injected blastomere with shortened, paralyzed cilia, and on the other, uninjected side with normal, beating cilia. These “half-paralyzed” blastulae did indeed swim, with the motile half pulling the paralyzed half around the injection chamber, and on these embryos, normal, motile cilia beat side-by-side with paralyzed cilia on adjacent blastomeres. Thus, individual K2.4-injected blastomeres fail to assemble normal cilia, while adjacent blastomeres ciliate normally.

Cilia on anti–kinesin-II–microinjected embryos were homogeneously short with an average length that was two-fold shorter than that of control embryos at the same stage (Table II). Control embryos not only possessed longer cilia than K2.4-injected embryos, but in addition, the ciliary lengths on control embryos were heterogeneous, with cilia at the animal pole normally growing to three times the length of cilia on the rest of the embryo (Burns, 1973, 1979). Significantly, in analyzing the length distributions on control embryos, we observed a discrete subpopulation of cilia whose average length corresponds to that of the short cilia on K2.4-injected embryos.

To investigate these patterns of recurring ciliary lengths in detail, we analyzed the entire pool of length data as length distribution histograms (Fig. 5). The lengths of cilia from the anti–kinesin-II–injected embryos are all clustered into a positively skewed distribution with a single mode at ∼7 μm of length (Fig. 5, upper panel) and with only 20% of the cilia 11 μm or longer. However, ciliary lengths from the control blastulae are spread over a much wider distribution of longer lengths than cilia from K2.4-injected blastulae, with between 62 and 78% of cilia 11 μm or longer, and they appear to fall into three subpopulations that differ in their modal length. One subpopulation of cilia has a predominant mode at ∼18 μm. A second subpopulation of control cilia displays a prominent mode at 7 μm, which is coincident with the mode of the short cilia on anti–kinesin-II–injected embryos. In addition, another subpopulation of cilia on control blastulae appears to constitute a minor third mode at ∼30 μm of length, but this group of longer cilia appears less discrete than those that are 18 and 7 μm long.

The light microscope provided clear evidence that the cilia on K2.4-injected embryos were morphologically and behaviorally different from those on control embryos. To identify a potential ultrastructural basis for the ciliary paralysis, we developed a system for processing single embryos from microinjection through embedding for electron microscopy. Several methods of fixation of intact cilia were tried, including those used for Chlamydomonas flagella (Kozminski et al., 1993), but they resulted either in artifactual production of conspicuous knobs on distal tips of control cilia or in obscured axonemal structure. However, we found that a modified protocol of Masuda and Sato (1984) developed for sea urchin gave consistent preservation of axonemal architecture on the surface of intact blastulae. EM analysis of sections cut through the axonemes of paralyzed K2.4-injected embryos revealed that, although they contain nine outer doublet MTs, they completely lack central pair MTs (Fig. 6). This 9+0 structure of the paralyzed axonemes is in stark contrast to the clear 9+2 structure of the control embryo axonemes where the central pair and its associated structures are distinct (Fig. 6, A, E, and J). Paralyzed cilia also possessed amorphous material within and around the nine outer doublet MTs that sometimes appear disorganized (Fig. 6, F–I, and K). The lack of central pair MTs in paralyzed cilia is apparent in both cross (Fig. 6, F–I) and oblique (Fig. 6, B, C, and K) sections. Scoring 39 such sections revealed that none contained central pair microtubules: 54% completely lacked any ultrastructural feature in the center of the axoneme (e.g., Fig. 6, F and K), while the remaining 46% contained indistinct but still centralized amorphous material (e.g., Fig. 6, G and H). Sometimes the outer doublets collapsed inward as seen in an extreme case in Fig. 6,I, where the nine outer doublets appear to spiral into the axoneme's central cavity. In control axonemes, 86% of 21 scored sections possessed two distinct central pair MTs (Fig. 6,E), while the remaining 14% contained centralized material with a less distinct ultrastructure. These results suggest that kinesin-II is required for the assembly or maintenance of central pair singlet MTs. In contrast, and as suggested by the normal cilia density on the surface of paralyzed embryos (Table II), kinesin-II is apparently not required for the positioning of basal bodies or formation of singlet cytoplasmic MTs and striated rootlets which appeared indistinguishable from control in the anti–kinesin-II–microinjected cells (Fig. 6, B–D).

To investigate the persistence of this effect on ciliogenesis, we monitored the development of anti–kinesin-II antibody–injected embryos for several days after control-injected and uninjected embryos had ciliated and begun to swim around in the injection chamber and surrounding sea water. After formation of their short, immotile cilia, anti– kinesin-II–injected embryos underwent mesenchymal cell ingression, gastrulation, and spiculogenesis at comparable times to the same developmental events occurring in uninjected embryos inhabiting the same injection chamber (Fig. 7). The only noticeable difference from controls at these later stages was that K2.4-injected embryos, unlike controls, underwent a gradual swelling and increase in blastocoel volume. But despite the continuation of many normal developmental events, the cilia on anti–kinesin-II– injected embryos never increased their length or motility. Instead, the K2.4-injected embryos perpetually retained short immotile cilia on their surfaces for several days of observation, indicating that the microinjection of anti– kinesin-II antibodies precluded normal ciliogenesis indefinitely.

We have demonstrated that the microinjection of the anti– kinesin-II–specific mAb, K2.4, causes a dramatic inhibition of the formation of motile, 9+2 cilia of normal length on sea urchin blastula-stage embryos, suggesting that kinesin-II plays an important role in ciliogenesis. We demonstrated, by Western blotting, the presence of kinesin-II in whole blastulae, deciliated blastulae, and in isolated cilia, in which it is found at ∼5–10% of the level in intact or deciliated blastulae (Fig. 1). To elucidate the function of kinesin-II in living sea urchin embryos, we microinjected the anti–kinesin-II antibody K2.4 into fertilized eggs and monitored their development by light microscopy. Although K2.4 injection had no apparent effect on mitosis or cytokinesis (Fig. 2), we observed a dramatic effect on ciliogenesis at the swimming blastula stage; cilia formed but were completely immotile and were consistently short compared to cilia on control embryos (Figs. 3–5). Histograms of ciliary length distributions revealed that the paralyzed cilia on K2.4-injected embryos fall into a single mode at 7 μm of length, coincident with the shortest of three modes seen for control embryo cilia (Fig. 5). This suggests that cilia on control embryos may progress through quantized assembly steps of which the first corresponds to the 7-μm length of the experimental cilia. Moreover, quantitation of the density of paralyzed cilia on the surface of K2.4-injected blastulae revealed similar densities to controls (Table II), suggesting that proper basal body localization was occurring in the absence of kinesin-II function. Therefore, we hypothesize that the dramatically shortened cilia resulted from defective assembly of the distal segment of cilia, so that in the absence of kinesin-II function, cilia achieved only the first of three discrete ciliary assembly states.

To investigate the ultrastructural basis for the paralysis of the cilia on K2.4-injected embryos, we analyzed individual microinjected embryos by electron microscopy. EM of axonemes on control and K2.4-injected embryos confirmed that the localization of basal bodies and the formation of associated cytoplasmic MTs and striated rootlets was occurring normally in K2.4-injected embryos (Fig. 6). Yet unexpectedly, and in striking contrast to the clear 9+2 structure of the control axonemes, axonemes on the surface of K2.4-injected embryos completely lacked central pair microtubules (Fig. 6). The paralyzed phenotype of the 9+0 cilia in the anti–kinesin-II–injected embryos may result from the absence of central pair MTs, like axonemes seen in many paralyzed Chlamydomonas mutants that have one or no central pair MTs (for review see Dutcher, 1995).

Thus, we propose a model (Fig. 8) in which ciliary growth in the sea urchin (Masuda, 1979; Masuda and Sato, 1984; Stephens, 1995) begins with formation of a short 7-μm-long assembly intermediate, or “procilium,” which forms by a kinesin-II–independent pathway. We hypothesize that the plus end–directed kinesin-II motor (Cole et al., 1993) functions in the elongation of the 7-μm-long procilium and in establishment or stabilization of the central pair MTs, perhaps by transporting ciliary components outwards towards the assembling plus ends of axonemal MTs located at the distal ciliary tip (Euteneuer and McIntosh, 1981; Johnson and Rosenbaum, 1992).

If, as we propose in our model, kinesin-II transports ciliary components along the axoneme to the ciliary tip for assembly, what could be the cargo of kinesin-II in sea urchin cilia whose absence would result in the formation of short and paralyzed 9+0 cilia? The short length of the cilia could arise if kinesin-II delivers structural or regulatory components required for the assembly or maintenance of the distal segment of the axoneme. Because the length of cilia is probably controlled by regulating the length of axonemal MTs (Burns, 1973), perhaps the shortened phenotype that we observed resulted from insufficient delivery of tubulin (Johnson and Rosenbaum, 1992) or tektin-A (Linck and Langevin, 1982; Stephens, 1989; Norrander et al., 1995) to a site of assembly at the ciliary tips (Johnson and Rosenbaum, 1992). The ciliary paralysis we observed in kinesin-II–blocked embryos might result if one of its cargoes is axonemal dynein. This hypothesis is supported by studies showing that FLA10, a Chlamydomonas homologue of a kinesin-II subunit, delivers inner dynein arms to the growing ends of axonemes (Brokaw and Kamiya, 1987; Piperno and Ramanis, 1991; Piperno et al., 1996). Further work will be required to biochemically identify molecules missing in cilia formed on K2.4-injected embryos.

Because K2.4-injected sea urchin embryos form short but stable cilia, unlike the Chlamydomonas fla10 mutant, which undergoes flagellar retraction at nonpermissive temperature (Kozminski et al., 1995), K2.4 injection of sea urchin embryos offers the potential opportunity to identify axonemal components targeted to cilia by kinesin-II at the ultrastructural level. This type of analysis has now revealed that axonemes on K2.4-injected embryos lack central pair MTs, which strongly suggests that ciliary components required for assembly or maintenance of the central pair MTs are potential cargoes of kinesin-II. Candidates include structural components of the central pair apparatus itself, such as the sea urchin homologue of the PF16 protein found on one central pair MT in Chlamydomonas (Smith and Lefebvre, 1996), or factors that stabilize the central pair, such as MT-capping proteins (Dentler, 1990; Miller et al., 1990). Sea urchin kinesin-II contains a large accessory subunit, SpKAP₁₁₅, that is localized at the presumptive cargo-binding carboxyl termini of the heterodimerized motor subunits (Wedaman et al., 1996), where it may function as an adaptor for cargo attachment (Cole et al., 1993; Wedaman et al., 1996). Interestingly, both SpKAP₁₁₅ and PF16 (Smith and Lefebvre, 1996) contain armadillo repeats, a protein motif proposed to mediate protein–protein interactions (Gindhart and Goldstein, 1996). It is therefore tempting to speculate that PF16 might be a component of kinesin-II's cargo by direct interaction.

Also of significance to this study is the observation that the pf16 mutant exhibits a destabilized central pair MT that can be lost during preparation of axonemes (Smith and Lefebvre, 1996). This raises the possibility that the 9+0 axonemal structure we observe in the kinesin-II– blocked embryos may simply result from loss of an unstable central pair during fixation rather than from an inhibition of central pair formation. Although we can not formally rule out this possibility at this time, we favor the hypothesis that the cilia on K2.4-injected embryos never form a central pair because (a) not a single central MT was observed in K2.4-injected embryo axonemes and (b) several behavioral and morphological similarities are shared between the paralyzed cilia on living K2.4-injected blastulae before fixation and primary cilia in other cell types that are known to be 9+0 (Menco, 1980; Poole et al., 1985; Roth et al., 1987, 1988; Wheatley, 1995). Further work will be required to discriminate between inhibition of de novo assembly versus destabilization of central pair MTs in K2.4-injected embryos.

We believe that the three modes in the control cilia length distribution histogram (Fig. 5) represent three discrete ciliary assembly states; cilia form initially as (a) the 7-μm procilia, which quickly mature and grow to become (b) the 18-μm-long fully assembled and beating cilia, of which only the ones at the very animal pole will continue growing to become (c) the 30-μm-long cilia of the “apical tuft.” The apparent entrapment of nascent cilia into the first assembly state by the inhibition of kinesin-II function raises interesting experimental predictions. First, the positively skewed distribution of cilia lengths on the K2.4-injected embryos (Fig. 5), with far more cilia of 7–8 μm than 1–2 μm, suggests that, in the absence of kinesin-II function, assembly of these 9+0 cilia to a length of 7 μm is somehow favored over shorter lengths. Further investigation will be necessary to test this hypothesis and to test its implication that short, 9+0 procilia may be found on normal, control blastulae. Initial formation of an immotile procilium could explain the observation that, in some urchin species, a minimum ciliary length is required before ciliary beating commences (Masuda and Sato, 1984).

In addition to their 9+0 axonemal core, the short and paralyzed cilia of the anti–kinesin-II–injected embryos share some interesting morphological similarities with primary cilia found on a variety of cell types (Menco, 1980; Poole et al., 1985; Roth et al., 1988; Wheatley, 1995; Wheatley et al., 1996). Primary cilia are also immotile (Sorokin, 1968; Roth et al., 1988), and in some epithelial cell lines up to a quarter of the cilia display varicosities very similar to the swellings that we observe on the tips of the K2.4-injected cilia (Fig. 3). The ultrastructure of the swellings on cilia of anti–kinesin-II–injected embryos was not revealed by our EM studies but will remain a goal of future work. Others have attributed the swellings on primary cilia of epithelial cells to the abnormal assembly of the primary ciliary axoneme (Jensen et al., 1987; Roth et al., 1988), as appears to be the case with the short cilia that grow on anti–kinesin-II–injected sea urchin embryos described here. In the light of these results, it will be interesting to determine if cells containing primary cilia lack functional kinesin-II.

How does the proposed function of kinesin-II in sea urchin embryonic ciliogenesis relate to studies of its close relatives in other organisms? Homologues of the motor subunits of kinesin-II have been found in Drosophila melanogaster, Caenorhabditis elegans, fish, mouse, frog, and Chlamydomonas rheinhardtii, where they are implicated in axonal transport, pigment granule dispersion, and flagellar assembly (Aizawa et al., 1992; Kondo et al., 1994; Pesavento et al., 1994; Walther et al., 1994; Kozminski et al., 1995; Tabish et al., 1995; Yamazaki et al., 1995; Beech et al., 1996; Cole and Rosenbaum, 1996; Rogers et al., 1997; for review see Scholey, 1996). These proteins are members of a family of motors named the hetero-kinesins (Vale and Fletterick, 1997). The microinjection of K2.4 produces a phenocopy most similar to mutations in the Chlamydomonas FLA10 gene, which encodes a homologue of a subunit of kinesin-II (Rashid et al., 1995). The fla10 mutation in C. rheinhardtii causes defects in flagellar assembly and stability (Lux and Dutcher, 1991). The FLA10-kinesin is concentrated in cell bodies relative to flagella and localized with centrioles, mitotic spindles, basal bodies, and axonemes (Vashishtha et al., 1996). These results, combined with our previous immunolocalization results (Henson et al., 1995, 1997) and our identification of kinesin-II in sea urchin cilia at 5–10% of the level found in cell bodies, are consistent with the hypothesis that kinesin-II is stockpiled in cell bodies with a fraction moving out along the axoneme to deliver components that are required for ciliogenesis. Furthermore, recent biochemical studies show that the FLA10-kinesin is also a subunit of a heterotrimeric motor complex (Cole and Rosenbaum, 1996) that is thought to drive the intraflagellar transport of 16S cargo complexes (Cole and Rosenbaum, 1996; Piperno and Mead, 1997), which form rafts beneath the flagellar membrane (Kozminski et al., 1993, 1995). Thus, work on the FLA10-kinesin together with our studies on kinesin-II provides a nicely consistent picture of a family of heterotrimeric kinesins playing important roles in axonemal assembly and stability in different systems.

The formation of normal motile cilia on sea urchin blastomeres is essential for embryonic motility and feeding (Masuda, 1979; Masuda and Sato, 1984; Stephens, 1995). Taken together, the data presented here provide strong evidence that the heterotrimeric kinesin-II motor protein plays a critical role in the assembly of motile 9+2 cilia on blastula-stage sea urchins and is thus essential for embryonic viability.

We would like to thank David Meyer for assistance in immunoblotting; Grete Fry, Robert Munn, and Paul Lee for expert technical advice and support on EM; William Dentler (University of Kansas, Lawrence, KS), Elizabeth Smith (University of Minnesota, Minneapolis, MN), and Samuel Bowser (Wadsworth Center, Albany, NY) for helpful discussions on cilia; and Shari Morris and the members of the Scholey lab for discussions and encouragement.

MT

microtubule

SpKAP₁₁₅

Strongylocentrotus purpuratus kinesin-related protein of molecular mass 85 kD

SpKRP₃₅

Strongylocentrotus purpuratus kinesin accessory protein of molecular mass 115 kD