Cyclical bending movements induced locally by successive iontophoretic application of ATP to an elastase-treated flagellar axoneme

Transcription

1 Journal of Cell Science 108, (1995) Printed in Great Britain The Company of Biologists Limited Cyclical bending movements induced locally by successive iontophoretic application of ATP to an elastase-treated flagellar axoneme Chikako Shingyoji* and Keiichi Takahashi Zoological Institute, Graduate School of Science, University of Tokyo and Department of Biology, International Christian University, Tokyo, Japan SUMMARY To elucidate the mechanism of oscillatory bending in cilia and flagella, we studied the effect of protease digestion on the response of axonemes to localized application of ATP. When the axonemes were treated with elastase and then reactivated locally by ATP iontophoresis, a pair of local bends were formed due to localized unidirectional sliding in the vicinity of the ATP pipette. Upon repeated application of ATP, the direction of bending with respect to the sperm head axis changed cyclically from side to side over several cycles. The bends were planar and similar to those observed in axonemes that had not been treated with elastase. In trypsin-treated axonemes, in contrast, repetitive local reactivation did not induce such cyclical bending; instead, it induced a bend that grew only in one direction upon repeated application of ATP. Moreover, the bends were not planar. Electron microscopy of these proteasedigested axonemes showed that both the interdoublet (nexin) links and the radial spokes were disrupted, but the effects of these proteases were different; trypsin disrupted 60-70% of these structures whereas elastase disrupted 20-30% of them. In both cases, spokes no. 3 and no. 8 (and no. 7) were more resistant to digestion than the others, although they tended to be more resistant to elastase than to trypsin. The importance of radial spokes and interdoublet links in the generation of cyclical bending and the determination of the bending plane is discussed. Key words: flagella, elastase, trypsin, cyclical bending, sea urchin sperm INTRODUCTION The bending movement of cilia and flagella is caused by active sliding between the outer doublet microtubules of the axoneme, which is driven by the dynein arms (Satir, 1968; Gibbons, 1981; Brokaw, 1991). The sliding of microtubules can be directly observed under the microscope; when ATP is added to trypsin-treated axonemes, they disintegrate by sliding between the outer doublets (Summers and Gibbons, 1971). Electron microscopic observation has revealed that trypsin treatment disrupts axonemal substructures such as the radial spokes and the interdoublet links (Summers and Gibbons, 1973), suggesting that either the radial spoke/central sheath complex or the interdoublet links are the structures that normally restrict the sliding between the doublet microtubules to a certain limit, thus playing a central role in the conversion of sliding into bending. In a beating sea-urchin sperm flagellum, planar bends are formed and propagated cyclically along its length. In intact flagella, the processes of bend formation, oscillation and propagation are coupled to each other. Our previous experiments using iontophoretic application of ATP, however, demonstrated that these processes could be uncoupled: when ATP was applied iontophoretically to a small region of a demembranated flagellum, a localized, non-propagating oscillation was induced in the vicinity of the micropipette from which the ATP was discharged (Shingyoji et al., 1977). Usually, molecules of ATP had to be discharged to induce the oscillatory response (Shingyoji and Takahashi, 1982). The oscillation lasted longer when the amount of ATP was increased. Smaller amounts of ATP, however, did not bring about oscillation. Instead, they induced a pair of opposite bends indicating local, unidirectional sliding (Shingyoji et al., 1977; Takahashi et al., 1982). In these experiments, the bends induced by ATP iontophoresis were always planar, so that the axoneme behaved as if it consisted of two parallel filaments (or two subsets of doublets) that slid past each other only in the ATP-activated region. To identify the structural components necessary for the generation of planar, oscillatory bends, we have studied the effects of iontophoretic application of ATP on axonemes treated with either trypsin or elastase. These proteases induce disintegration of the axoneme by microtubule sliding in the presence of ATP. Moreover, they differ in their effects on the flagellar waveform of reactivated sperm (Brokaw, 1980; Brokaw and Simonick, 1977), suggesting that they disrupt different structural components. It is therefore of interest to see how the axonemes treated with the two proteases differ in their response to local activation. Here, we show that the elastase-treated axonemes, but not the trypsin-treated axonemes, retain the ability to produce a local, planar bending that can change its direction in a cyclical manner. Electron microscopical study of the axonemes treated

2 1360 C. Shingyoji and K. Takahashi with the two proteases suggests that certain sets of radial spokes and interdoublet links are particularly important for the cyclical bending movement. MATERIALS AND METHODS Preparation of reactivated sperm Sperm of the sea urchins, Anthocidaris crassispina or Hemicentrotus pulcherrimus, were demembranated as described by Gibbons and Gibbons (1972), with some modifications. The sperm were suspended in ten volumes of Ca 2+ -free artificial sea water (465 mm NaCl, 10 mm KCl, 25 mm MgSO 4, and 10 mm Tris-HCl, ph 8.0). Drops of the suspension were added to an extracting solution containing 0.04% (w/v) Triton X-100, 0.15 M KCl, 2 mm MgSO 4, 0.5 mm EDTA, 2 mm CaCl 2, 2 mm dithiothreitol (DTT), and 10 mm Tris-HCl (ph 8.0) and the mixture was gently swirled for seconds at room temperature. After the demembranation, the suspension of axonemes was transferred to 100 volumes of reactivation solution without ATP (0.15 M KCl, 2 mm MgSO 4, 2 mm EGTA, 20 mm Tris-HCl (ph 8.0), 2 mm DTT, and 2% (w/v) polyethyleneglycol). The demembranated sperm could be stored at 0 C for about an hour without appreciable loss of reactivity. The reactivation rate was routinely checked both before and after each experiment. The preparations that showed less than 90% reactivation were discarded. Digestion with elastase or trypsin Demembranated sperm in the ATP-free reactivating solution was treated with µg/ml (about 0.5 unit/ml) elastase (Sigma E- 0127) or 5-7 µg/ml trypsin (Sigma T-8253). We did not use turbidimetry to monitor the digestion process because the head, necessary as a hold in local reactivation experiments, interfered with the assay. Thus, we used an alternative procedure to control the degree of digestion. First, we digested a small portion of the demembranated sperm suspension with elastase or trypsin in the presence of ATP (20 µm), and determined the time (the sensitization time) necessary to induce sliding disintegration of the axonemes. The sensitization time was about 2-4 minutes, depending on the preparation. Then, the rest of the demembranated sperm suspension was treated with elastase (1-10 minutes) or trypsin (0.6-4 minutes) in the absence of ATP in such a way that the ratio of the digestion time to the sensitization time was constant (between 0.5 and 5). Digestion was stopped by dilution with 100 volumes of enzyme- and ATP-free reactivating solution or by the addition of an inhibitor; 50 µg/ml of ovoinhibitor (Sigma T-1886) and 10 µg/ml of soybean trypsin inhibitor (Sigma T-9003) were used for elastase digestion and trypsin digestion, respectively. In the elastase treatment, soybean trypsin inhibitor (6 µg/ml) was added to the digesting solution in order to eliminate the effect of any contaminating trypsin-like enzyme. Specimens were discarded if they did not show almost 100% sliding disintegration in the first assay in the presence of ATP. Local transient reactivation by iontophoresis The methods for local iontophoretic application of ATP were described in previous papers (Shingyoji et al., 1977; Takahashi et al., 1982). A demembranated sperm was held by the head with a polylysine-coated glass microneedle and ATP was iontophoretically applied to its flagellum. A glass-capillary micropipette filled with 10 mm ATP was used for the iontophoresis. The duration of an ATP pulse was milliseconds. The amount of the ATP released from the micropipette is expressed as its electric charge in nanocoulombs (nc). To reactivate the same region of the axoneme repeatedly, the position of the ATP micropipette was readjusted each time to compensate for any movement of the axoneme caused by the previous application of ATP. After each series of local reactivation, the axoneme was immediately tested for its ability to undergo sliding disintegration in a reactivation solution containing 20 µm ATP. The data were discarded if the axoneme failed to show sliding disintegration. All experiments were performed at room temperature (19-21 C for Hemicentrotus, C for Anthocidaris). The axonemes obtained from the two species of sea urchins showed essentially the same response. High-speed cinematography The responses of the axonemes were observed with an inverted microscope (Nikon MD) fitted with phase-contrast optics (objective: Nikon BM 40 ). A xenon short-arc lamp (UXL 500-DO, Ushio Electric Inc.) was used as the light source. The movement of flagellar axonemes was filmed on Kodak Double-X or Fuji Color RT mm film at frames s 1 with a 16 mm high-speed ciné camera (Photo- Sonics 1-PL). The recorded images were analyzed frame by frame by means of a film motion analyzer (nac inc., model MC-OB, with a projecting head, model 160B). Electron microscopy Demembranated sperm were treated with elastase or trypsin as described above for 0 to 2.5 times the sensitization time. A portion (about 50 ml) of the treated sample was centrifuged at 25,000 g for 10 minutes and the resultant pellets were immediately fixed by adding 4 ml of a fixative containing 2% glutaraldehyde, 2 mm MgSO 4, 0.15 M KCl and 10 mm phosphate buffer, ph 7.8. After 45 minutes of fixation, they were washed with fresh buffer, postfixed with 1% OsO 4, dehydrated in ethanol and embedded in Epon 812. Thin sections were stained with uranyl acetate and Reynolds lead citrate. Electron micrographs were taken with a JEOL JEM 100CX electron microscope operated at 80 kv. The rest of the protease-treated sample was stored at 0 C until the end of the fixation of the other portion, when we added the reactivating solution containing 50 µm ATP to ensure that all the axonemes showed sliding disintegration. In evaluating the effect of protease digestion on the radial spokes, those that appeared to be attached to both the outer doublets and the central sheath were scored as intact, and those that did not as disrupted. In estimating the effects on the dynein arms, we counted the inner and the outer dynein arms together. Arms with an apparently normal shape were counted as intact, and those that were shorter or smaller as damaged. To estimate the number of the interdoublet links remaining, the axonemes were dialyzed against a low ionic-strength buffer to remove the dynein arms and part of the radial spokes (Gibbons, 1965; Stephens, 1970). This procedure facilitated observation of the interdoublet links, and at the same time caused disruption of the 9+2 structure at sites where the interdoublet links were missing. The dialysis was done overnight against a Tris-EDTA solution (5 mm Tris-HCl, 0.1 mm EDTA, 0.1 mm DTT, ph 8.0), followed by centrifugation at 35,000 g for 20 minutes. The resultant pellets were fixed for electron microscopy as described above. To count the intact interdoublet links, only those images of the outer doublets that were cut perpendicular to their long axis so as to show their 8-shaped cross-sections clearly were used. As the interdoublet links were not always visible in the sections, a pair of doublets that appeared to be associated with one another by an interdoublet links, as judged by their orientation and distance of separation, was regarded as representing an intact link. To obtain an unprejudiced estimate of the number of disrupted radial spokes and dynein arms, the electron micrographs were given code numbers, shuffled, and then counted. For counting the disrupted radial spokes and dynein arms, 30 to 50 axonemal cross-sections were used for each sample of treated sperm. About 200 doublet microtubules from each sample were examined for counting the interdoublet links disrupted by the protease treatments. RESULTS Local bend formation in protease-treated axonemes Fig. 1A shows typical tracings of a demembranated flagellum

3 Cyclical bending in elastase-treated axonemes 1361 Fig. 1. Responses of intact (A), elastase-treated (B), and trypsintreated (C) flagellar axonemes to local iontophoretic application of ATP. Tracings from a 16 mm ciné film. In each pair of tracings, the upper and lower ones show the same demembranated flagellum before and after an application of ATP, respectively. The position of the ATP micropipette is indicated alongside each of the upper tracings. The upper tracings are superimposed on the lower tracings as broken lines. In the intact axoneme (A), a pair of opposite bends are indicated by p and d. ATP pulses: (A) 0.72 nc, 30 milliseconds; (B) 0.66 nc, 30 milliseconds; (C) 0.80 nc, 40 milliseconds. Hemicentrotus pulcherrimus. Bar, 10 µm. before (upper) and after (lower) an iontophoretic application of ATP. The local bending induced by the application of ATP (lower tracing) has several characteristics. First, during the response, the changes in angle of the proximal and the distal bends (p and d in Fig. 1A) are always equal with opposite signs. Second, the pair of opposite bends are formed in the same plane. Third, this response can be induced in any region along the flagellum except for the extremely distal region where only a single bend is formed (Takahashi et al., 1982). Local application of a small amount of ATP to elastase- or trypsin-treated axonemes also induced a localized bending response. In the elastase-treated axonemes, two opposite bends, a proximal bend and a distal bend, were formed in the same plane (Fig. 1B) as in the untreated axonemes. The angles of the two bends were almost identical. In the trypsin-treated axonemes, in contrast, the two localized bends were not exactly in a single plane and their angles were unequal (Fig. 1C). This behaviour of flagella treated with both enzymes was, however, Fig. 2. Responses of an intact axoneme to repetitive local reactivation. Tracings from a ciné film. ATP pulses of 0.4 nc, 20 milliseconds duration, were applied eight times from the micropipette shown in the tracings. Tracings 1-8 show the stationary form of the axoneme after each application of ATP. The broken line in tracing 1 is the tracing of the original shape of the flagellum just after detergent extraction. Hemicentrotus pulcherrimus. very different from what is observed upon bath-application of ATP: i.e. sliding disintegration into individual doublets (Summers and Gibbons, 1971). Cyclical bending movements induced in intact axonemes We next examined the effect of repeated local application of ATP. Before describing the results obtained with the digested axonemes, we first summarize the response of intact axonemes. In the intact axoneme repeated application of ATP brought about a succession of local, non-propagating bending movements. A notable feature of the response was reversal in the direction of bending as shown in Figs 2 and 3. Thus, in Fig. 2 the proximal bend grew successively first in one direction (upward in tracings nos 2 and 3), relaxed (no. 4) and then increased in the opposite direction (nos 5, 6, and 7). Such cyclical bending was induced in any region along the flagellum. The cyclical bending was asymmetrical with respect to the maximum bend angle, i.e. the bend angle at which the direction of bending reverses. The maximum bend angle in one direction with respect to the axonemal axis in the relaxed position was usually smaller than that in the other direction. Also, in one

4 1362 C. Shingyoji and K. Takahashi Fig. 3. Changes in angle of the proximal bend, upon repetitive application of ATP at two different intensities. ATP pulses of 0.4 nc, 20 milliseconds (A); and 0.8 nc, 20 milliseconds (B). These data were obtained from the same recording as the one shown in Fig. 2. The angles between the tangent to the flagellar segment at the base and the tangent to the segment at a point 5 µm away from the base are plotted against time. Each point was obtained from a single frame of the ciné film. A bend in the direction of the principal bend is given a positive sign and the one in the opposite direction a negative sign. Arrows indicate when ATP was applied. Lines were drawn visually. direction, the maximum bend angle was usually attained by a single step of bending due to an application of any amount of ATP above the threshold, while in the opposite direction multiple steps of bending upon repeated application of ATP were necessary before the maximal bend was attained. The direction in which the bends grew stepwise (the latter case) always corresponded to the direction of the principal bend in a beating flagellum. Fig. 3A shows an example in which the proximal bend angle changed on both sides of the axoneme, that is, in the directions of the principal (positive signs) and the reverse bends (negative signs). Upon each application of ATP, the bend grew rapidly and reached a maximum angle, which was maintained without much change until the next application of ATP. In this example, there are four steps in the principal bend direction, but there are only one or two steps in the opposite direction. The same axoneme responded to application of a larger amount of ATP with a larger step change in bend angle (Fig. 3B). However, we did not systematically analyze the relationships between the amount of ATP and the bend angle or the number of steps. An application of a very large amount of ATP induced a cyclical bending and a further increase of ATP caused propagation of the induced bends. We measured the maximum angles in a number of recordings made with 13 specimens and found that the maximum angle was

5 Cyclical bending in elastase-treated axonemes 1363 Fig. 4. Histogram of the maximum bend angle at which the bending direction reversed upon repetitive local reactivation in the untreated (A) and the elastase-treated (B) axonemes. Data were obtained from 13 (A) and 17 (B) axonemes. variable within a certain limit (Fig. 4A). The maximum angle was not constant even in a single specimen. Cyclical bending movements induced in proteasetreated axonemes A local cyclical bending was also induced in the axonemes treated with elastase (Figs 5A,B and 6A). The induced bends were always planar throughout the response and inducible in any region of the axoneme, except in the very distal region where a local reactivation elicited sliding disintegration. Eleven of the 17 samples examined changed their bending direction on both sides of the axoneme (Fig. 5A,B); in five samples a bend repeatedly grew only in one direction and relaxed (i.e. returning to the original position), and in one sample a bend grew only in one direction without relaxation. The response of the last example was similar to that observed in trypsin-treated axonemes, as described below. As in the intact axonemes, the maximum bend angle was variable within a certain limit, and even in a single specimen (Fig. 4B). In contrast to the elastase-treated axonemes, the trypsintreated axonemes did not show a local cyclical bending response upon repetitive application of ATP; the bending direction did not reverse but the bend grew successively only in one direction (Figs 5C,D, 6B). Moreover, because bends were not planar, it is not clear whether the direction of the successive bending corresponded to the direction of the principal bend or not. The bending responses described above were relatively insensitive to the digestion time as long as it did not exceed 5 times the sensitization time; after longer digestion with Fig. 5. Responses of the protease-treated axonemes to repetitive local reactivation. Superimposed tracings from a ciné film recording. Each number indicates the order of response; no. 1 is before application of ATP, no. 2 the response to the first ATP pulse, and so on. A pair of bends were induced in both directions in the elastase-treated axonemes (A,B). In the trypsin-treated axonemes, the bending direction did not reverse on repetitive reactivation but the bend grew successively in only one direction (C,D). ATP pulses were: (A) 0.26 nc, 20 milliseconds; (B) 0.63 nc,30 milliseconds; (C) 1.2 nc, 30 milliseconds; (D) 0.26 nc, 15 milliseconds. Anthocidaris crassispina (A,B and C); Hemicentrotus pulcherrimus (D).

6 1364 C. Shingyoji and K. Takahashi Fig. 6. Changes in angle of the proximal bends upon a series of ATP pulses to an elastase-treated axoneme (A) and a trypsin-treated axoneme (B). (A) 0.26 nc, 20 milliseconds, Anthocidaris crassispina; (B) 0.26 nc, 15 milliseconds, Hemicentrotus pulcherrimus. In the elastase-treated axoneme (A), plus values were assigned to bends in the principal bend direction. Arrows indicate the time of ATP application. elastase, the axonemes tended to lose their capability to oscillate upon application of a high concentration of ATP; after longer digestion with trypsin, the axonemes tended to undergo sliding disintegration. Even in the axonemes digested with either protease for a short period (0.5-2 times the sensitization time) an application of a very large amount of ATP caused sliding disintegration. Electron microscopy of protease-treated axonemes To understand the structural basis for the difference in response between the trypsin-treated and elastase-treated axonemes, we examined the proteolysed axonemes by electron microscopy. A major change in axonemal structure due to digestion with elastase and trypsin was a detachment of radial spokes from their normal points of attachment to the central sheath and to the doublets (Figs 7, 8). The detachment was observed in about 50% of the spokes in the trypsin-treated axonemes but in about 20% of those in the elastase-treated axonemes (Tables 1, 2). This detachment of the spoke appeared to occur in a nonrandom manner (Fig. 9): spokes no. 3 and no. 8 (and no. 7) were relatively resistant to digestion by both proteases; with elastase treatment, in particular, these spokes remained almost intact (Fig. 9 and Table 2). Spokes no. 1, no. 4, no. 5, no. 6 and no. 9, on the other hand, suffered significant disruption,

7 Cyclical bending in elastase-treated axonemes 1365 Table 1. Axonemal substructures remaining after digestion Dynein arms Interdoublet links Radial spokes Digestion (%) (n) (%) (n) (%) (n) Undigested 97 (60) 88 (244) 100 (30) Elastase 95 (128) 65 (220) 77 (112) Trypsin 89 (84) 30 (199) 53 (84) the degree of which was greater in the trypsin-treated axonemes than in the elastase-treated axonemes. Dynein arms did not appear to be disrupted by either of the proteases (Table 1), although protease nicks in the dyneins cannot be observed by electron microscopy and therefore cannot be ruled out. The protease treatment used in this study was much milder than that used for limited proteolysis for dynein (Ow et al., 1987), so it is unlikely that the function of dynein was significantly affected due to the dynein subunit degradation by the enzyme, in particular by trypsin. The changes in axonemal structures caused by trypsin treatment that we observed were similar to those reported by Summers and Gibbons (1973). Upon dialysis of digested axonemes against a low concentration of EDTA, we found that the trypsin treatment disrupted nearly twice as many interdoublet links as did the elastase Table 2. The percentage of radial spokes remaining after digestion Radial spoke no. Digestion Elastase Trypsin treatment. Fig. 10 shows cross-sectional views of the dialyzed axonemes. A significant fraction of the interdoublet links remained in the elastase-treated samples (B), but not in the trypsin-treated samples (C). An evaluation of the remaining axonemal substructures is summarized in Table 1 (for the evaluation method, see Materials and Methods). A large difference (65% vs 30%) in the number of interdoublet links was found between the products digested with elastase and those digested with trypsin. If we consider the value in the control axonemes, the elastase treatment appeared to have left more than 70% of the interdoublet links intact. Although a 12% loss of the links were observed in the untreated axonemes, the positions of the ruptured links relative to the central pair microtubules suggested that they did not occur constantly between some specific doublets. Fig. 7. Electron micrographs of cross-sections of the undigested (A) and the elastase-treated (B-E) axonemes. (B) Control axonemes kept in the elastase solution containing ovoinhibitor and trypsin inhibitor; (C) axonemes digested with elastase for 1 minutes; (D) 4 minutes (the sensitization time); (E) 10 minutes. In C, the arrow and arrowheads indicate doublet no. 1 and the doublets having persistent spokes (nos 2, 3, 7 and 8), respectively. Bar, 100 nm. Hemicentrotus.

8 1366 C. Shingyoji and K. Takahashi Fig. 8. Electron micrographs of cross-sections of the trypsin-treated axonemes. (A) Control axonemes kept in the trypsin solution containing trypsin inhibitor; (B) axonemes digested with trypsin for 45 seconds; (C) 2 minutes (the sensitization time); (D) 4 minutes. In C, the arrow and arrowheads indicate doublet no. 1 and the doublets having persistent spokes (nos 4, 6, 7 and 8), respectively. Bar, 100 nm. Hemicentrotus. DISCUSSION Localized sliding and bend formation in proteolysed axonemes We have shown that localized transient reactivation of intact axonemes and elastase- and trypsin-treated axonemes produces a localized unidirectional sliding to form a pair of opposite bends, without causing sliding disintegration of even the protease-treated axonemes. Moreover, in intact and elastasetreated axonemes, repeated application of ATP induced a localized cyclical bending without accompanying bend propagation. If the amount of ATP is increased, this localized bending movements can develop into a localized oscillatory bending motion in the absence of wave propagation, and a further increase in the amount of ATP induced propagating waves in untreated axonemes. The present study thus demonstrates that localized cyclical sliding movements can be induced in the absence of wave propagation. We may regard a short segment of the flagellar axoneme as an independent oscillator. In the previous study we observed that the local reactivation of the intact axoneme induced a pair of bends in the same plane (Shingyoji et al., 1977; Takahashi et al., 1982). On the basis of this observation, we suggested that an unknown mechanism controls the microtubule sliding such that the axoneme behaves as if it consisted of two subsets of outer doublets. A simplified view would be that the axonemal outer doublets are functionally divided into two groups, with borders between doublets no.3 and no.4, and between doublets no.7 and no.8 (Sale, 1986; Satir, 1985; Tamm and Tamm, 1984; Wais-Steider and Satir, 1979). Iontophoretic application of ATP should cause a localized sliding between the two groups and produce the bends. The locally applied ATP activates only dynein arms in Fig. 9. The percentage of radial spokes that remained apparently intact in the axonemes after digestion with elastase (A) and trypsin (B). In the elastase treatment (A), axonemes were digested for 0 times (open bar), 0.25 times (hatched bar), 1 time (filled bar) and 2.5 times the sensitization time (dotted bar); in the trypsin treatment (B), digestion was 0 times (open bar), 0.4 times (hatched bar), 1 time (filled bar) or 2 times (dotted bar).

9 Cyclical bending in elastase-treated axonemes 1367 the vicinity of the ATP pipette, while leaving the other dynein arms inactive; those inactive dyneins most likely form rigor cross bridges that resist the shear (Gibbons and Gibbons, 1974; Johnson, 1985). In short, a pair of bends is formed due to the unequal shear along the length of the axoneme. According to this view, the most important functional features of the axoneme necessary for bend formation are force generation by dynein-microtubule interaction and the mechanism that differentiates the outer doublets to form two groups. Our present finding that even protease-treated axonemes can produce normal planar bends indicates that these two important features can survive proteolysis to some extent, although, in trypsin-treated axonemes, which showed three-dimensional bends, the mechanism that defines the two outer-doublet groups seem to be impaired. It is not expected that elastase-treated axonemes would retain the ability to generate localized cyclical sliding movements that cause cyclical planar bends, despite their loss of the ability to propagate bending waves. For cyclical bending, those dynein arms that are positioned between the two counteracting doublet groups (e.g. doublets no. 3 and 4 versus doublets no. 7 and 8) must reciprocally undergo activation and inactivation. Such cyclical behaviour must be crucial to the mechanism that produces normal axonemal beating, but little is known about how it occurs; what turns on inactive dynein arms and turns off active dynein arms? The present finding that elastase-treated axonemes can reverse their bending direction, while trypsin-treated axonemes cannot, may well provide a basis for an experimental approach to this problem. For example, a simple comparison between Fig. 6A and B indicates that it is not the absolute value of bend angle (that is, the absolute amount of shear between the two doublet groups) that directly triggers the on/off switching of the dynein activity. It may be that trypsin destroys the components that convey the information about the bend angle to the mechanism that regulates the activity of dynein. The maximum angle did not exceed rad in the direction of the principal bend and 1.0 rad in the other direction, both in the axoneme not treated by proteases and in the elastase-treated axonemes (Fig. 4A and B). It thus appears that there are certain angles beyond which a bend cannot grow. These angles are almost half the maximum bend angles observed in beating flagella. Fig. 10. Electron micrographs of the cross-sections of the axonemes proteolyzed and dialyzed against Tris-EDTA solution. Most dynein arms were extracted upon dialysis. (A) Undigested axonemes. The axonemes retained almost complete 9+2 structures. (B) Axonemes digested with elastase for the sensitization time (4 minutes). Outer doublets appeared in groups. (C) Axonemes digested with trypsin for the sensitization time (2 minutes). Most doublets were separated from each other. Bar, 100 nm. Hemicentrotus. Structural basis for axonemal oscillation and the direction of bending In an attempt to understand the structural basis for the cyclical bending movement, we compared the structures of axonemes treated with trypsin or elastase. Electron microscopy of thinsectioned samples showed that protease digestion disrupted mainly the interdoublet links and the radial spokes, to different extents depending on the time and the species of protease used. Under our experimental conditions, trypsin caused more extensive degradation than did elastase. The degree of the structural disruption is thus correlated with that of the functional loss caused by the two proteases. It is not easy to determine which structural disruption caused a particular functional loss, since interdoublet links and radial spokes (except for no. 3, no. 7, and no. 8) underwent degradation simultaneously with either protease. Thus the loss of the abilities to propagate bending waves, to oscillate locally, and to form planar bends might well be caused by different degrees

10 1368 C. Shingyoji and K. Takahashi of disruption of interdoublet links or radial spokes (except for no. 3, no. 7, no. 8), or both. It is tempting to speculate, however, that the disruption of the radial spokes is not responsible for the loss of the ability of cyclical bending movements, since there are ample examples of axonemes that can bend and propagate bending waves in the absence of radial spokes (Luck et al., 1977; Brokaw et al., 1982; Goldstein, 1982; Goldstein and Schrével, 1982; Huang et al., 1982; Ishijima et al., 1988). A recent report (Smith and Sale, 1992), however, indicates that a possible role of radial spokes is to activate dynein-induced microtubule sliding activity. Correlation between the spokecentral sheath attachment cycle and the switching of arm activity has been suggested (Satir and Matsuoka, 1989). Also, localization of kinesin-related proteins to the central pair apparatus has been reported (Fox et al., 1994; Johnson et al., 1994). These findings suggest a need for further investigation of the role of the central pair-radial spoke complex. Thus, although the role of radial spokes in cyclical bending movement cannot be ruled out, the failure of the trypsin-treated axoneme to reverse its bending direction might be caused by the disruption of interdoublet links. The radial spokes, on the other hand, may be involved in the production of planar waves, because the central pair/radial spoke complex has been implicated in the mechanism that determines the bend plane (Gibbons, 1961; Gibbons and Gibbons, 1973; Ishijima et al., 1988; Lindemann et al., 1992; Tamm and Horridge, 1970). In this respect, it is interesting to note that the radial spokes that remained almost intact after the elastase treatment are on the same plane as the central microtubules. It is possible that their resistance to digestion is related to the structural stability of the two doublet groups (Holwill and Satir, 1994). Although our present study cannot rule out other possibilities, these ideas can explain the observed behaviour of the elastase-treated and the trypsin-treated axonemes. Finally, local bends induced by repetitive local reactivation are asymmetric, i.e. they differ in the bend angle, indicating different sliding distances for opposite sliding directions. Thus, the bend asymmetry (principal and reverse bends) observed in normal flagellar movement does not seems to result from the asymmetric initiation of bending at the basal region of flagella, but seems to reflect some functional asymmetry in the nine doublets. Our recent experiments imposing a planar vibration on the head of sea urchin sperm while rotating the vibration plane (without rotating the axoneme itself) have provided evidence that the flagellar beat plane and the direction of the bend asymmetry are not inherent in a given orientation of the nine outer doublets (Gibbons et al., 1987; Shingyoji et al., 1991). The mechanism that produces this asymmetry, however, is not known. The present study indicates that the mechanism responsible for this exists at least in elastase-treated axonemes. Radial spokes no. 3, no. 7 and no. 8, which remained almost intact, may well be involved in the mechanism underlying the asymmetry. We thank Dr Ritsu Kamiya (University of Tokyo) and Dr Charlotte K. Omoto (Washington State University) for critical reading of the manuscript. REFERENCES Brokaw, C. J. and Simonick, T. F. (1977). Motility of Triton-demembranated sea urchin sperm flagella during digestion by trypsin. J. Cell Biol. 75, Brokaw, C. J. (1980). Elastase digestion of demembranated sperm flagella. Science 207, Brokaw, C. J., Luck, D. J. L. and Huang, B. (1982). Analysis of movement of Chlamydomonas flagella: The function of the radial-spoke system is revealed by comparison of wild-type and mutant flagella. J. Cell Biol. 92, Brokaw, C. J. (1991). Microtubule sliding in swimming sperm flagella: Direct and indirect measurements on sea urchin and tunicate sperm flagella. J. Cell Biol. 114, Fox, L. A., Sawin, K. E. and Sale, W. S. (1994). Kinesin-related proteins in eukaryotic flagella. J. Cell Sci. 107, Gibbons, B. H. and Gibbons, I. R. (1972). Flagellar movement and adenosine triphosphatase activity in sea urchin sperm extracted with Triton X-100. J. Cell Biol. 54, Gibbons, B. H. and Gibbons, I. R. (1973). The effect of partial extraction of dynein arms on the movement of reactivated sea-urchin sperm. J. Cell Sci. 13, Gibbons, B. H. and Gibbons, I. R. (1974). Properties of flagellar rigor waves formed by abrupt removal of adenosine triphosphate from actively swimming sea urchin sperm. J. Cell Biol. 63, Gibbons, I. R. (1961). The relationship between the fine structure and direction of beat in gill cilia of a lamellibranch mollusc. J. Biophys. Biochem. Cytol. 11, Gibbons, I. R. (1965). Chemical dissection of cilia. Arch. Biol. 76, Gibbons, I. R. (1981). Cilia and flagella of eukaryotes. J. Cell Biol. 91, 107s- 124s. Gibbons, I. R., Shingyoji, C., Murakami, A. and Takahashi, K. (1987). Spontaneous recovery after experimental manipulation of the plane of beat in sperm flagella. Nature 325, Goldstein, S. F. (1982). Motility of 9+0 mutants of Chlamydomonas reinhardtii. Cell Motil. Suppl. 1, Goldstein, S. F. and Schrével, J. (1982). Motility of the 6+0 flagellum of Lecudina tuzetae. Cell Motil. 4, Huang, B., Ramanis, Z. and Luck, J. L. (1982). Suppressor mutations in Chlamydomonas reveal a regulatory mechanism for flagellar function. Cell 28, Holwill, E. J. and Satir, P. (1994). Physical model of axonemal splitting. Cell Motil. Cytoskel. 27, Ishijima, S., Sekiguchi, K. and Hiramoto, Y. (1988). Comparative study of the beat pattern of American and Asian horseshoe crab sperm: Evidence for a role of the central pair complex in forming planar waveform in flagella. Cell Motil. Cytoskel. 9, Johnson, K. A. (1985). Pathway of the microtubule-dynein ATPase and the structure of dynein: A comparison with actomyosin. Annu. Rev. Biopys. Biophys. Chem. 14, Johnson, K. A., Haas, M. A. and Rosenbaum, J. L. (1994). Localization of a kinesin-related protein to the central pair apparatus of the Chlamydomonas reinhardtii flagellum. J. Cell Sci. 107, Lindemann, C. B., Orlando, A. and Kanous, K. S. (1992). The flagellar beat of rat sperm is organized by the interaction of two functionally distinct population of dynein bridges with a stable central axonemal partition. J. Cell Sci. 102, Luck, D., Piperno, G., Ramanis, Z. and Huang, B. (1977). Flagellar mutants of Chlamydomonas: Studies of radial spoke-defective strains by dikaryon and revertant analysis. Proc. Nat. Acad. Sci. USA 74, Ow, R. A., Tang, W.-J., Mocz, G. and Gibbons, I. R. (1987). Tryptic digestion of dynein 1 in low salt medium. J. Biol. Chem. 262, Sale, W. S. (1986). The axonemal axis and Ca 2+ -induced asymmetry of active microtubule sliding in sea urchin sperm tails. J. Cell Biol. 102, Satir, P. (1968). Studies on cilia. III. Further studies on the cilium tip and a Sliding filament model of ciliary motility. J. Cell Biol. 39, Satir, P. (1985). Switching mechanisms in the control of ciliary motility. Modern Cell Biol. 4, Satir, P. and Matsuoka, T. (1989). Splitting the ciliary axoneme: Implications for a switch-point model of dynein arm activity in ciliary motion. Cell Motil. Cytoskel. 14, Shingyoji, C., Murakami, A. and Takahashi, K. (1977). Local reactivation of Triton-extracted flagella by iontophoretic application of ATP. Nature 265, Shingyoji, C. and Takahashi, K. (1982). Iontophoresis of ATP: Determination of transport number by bioassay with demembranated flagella. J. Fac. Sci. Univ. Tokyo, Sec. IV 15, Shingyoji, C., Katada, J., Gibbons, I. R. and Takahashi, K. (1991). Rotating

11 Cyclical bending in elastase-treated axonemes 1369 the plane of imposed vibration can rotate the plane of flagellar beating in seaurchin sperm without twisting the axoneme. J. Cell Sci. 98, Smith, E. F. and Sale, W. S. (1992). Regulation of dynein-driven microtubule sliding by the radial spokes in flagella. Science 257, Stephens, R. E. (1970). Isolation of nexin. - The linkage protein responsible for the maintenance of the nine-fold configuration of flagellar axonemes. Biol. Bull. 139, 438a. Summers, K. E. and Gibbons, I. R. (1971). Adenosine triphosphate-induced sliding of tubules in trypsin-treated flagella of sea-urchin sperm. Proc. Nat. Acad. Sci. USA 68, Summers, K. E. and Gibbons, I. R. (1973). Effects of trypsin digestion on flagellar structures and their relationship to motility. J. Cell Biol. 58, Takahashi, K., Shingyoji, C. and Kamimura, S. (1982). Microtubule sliding in reactivated flagella. In Symp. Soc. Exp. Biol. vol. 35, Prokaryotic and Eukaryotic Flagella (ed. W. B. Amos and J. G. Duckett), pp Cambridge: Cambridge University Press. Tamm, S. L. and Horridge, G. A. (1970). The relation between the orientation of the central fibrils and the direction of beat in cilia of Opalina. Proc. Roy. Soc. B 175, Tamm, S. L. and Tamm, S. (1984). Alternate patterns of doublet microtubule sliding in ATP-disintegrated macrocilia of the ctenophore Beröe. J. Cell Biol. 99, Wais-Steider, J. and Satir, P. (1979). Effect of vanadate on gill cilia: Switching mechanism in ciliary beat. J. Supramol. Struct. 11, (Received 22 August Accepted 17 January 1995)