The development of ciliary band pattern in the doliolaria larva of Florometra serratissima is described based on scanning and transmission electron microscopy. The uniformly ciliated epithelium of the post-hatching larva develops four regularly spaced bands over a period of approx. 20 h generating an epithelial pattern that is, essentially, a series of stripes. The first visible events of pattern formation progress over the larval surface in a posterior-to-anterior and dorsal-to-ventral sequence, but the initial pattern is not, in fact, striped. It instead consists of a close-packed array of oval interband domains separated and surrounded by belts of band cells. Secondarily the interband domains expand laterally and coalesce to form continuous, broad stripes, while the bands remain as narrow stripes between them. Two possible explanations for this unusual sequence of events are discussed: (1) that it can be understood in evolutionary terms with reference to band pattern in other echinoderm larvae, and (2) that it is a morphogenetic necessity because limitations inherent in the patterning mechanism prevent the direct formation of regular stripes.

The locomotory cilia of the barrel-shaped doliolaria larva of crinoids are arranged in a series of regularly spaced circumferential bands, four or five in number depending upon the species. The pattern arises in what is at first, in the newly hatched larva, a uniformly ciliated epithelium. Since epithelial cells in echinoderm larvae are at most uniciliate, the banding in fact represents the formation of a series of stripes containing, alternately, ciliated and non-ciliated cells. While general descriptions of larval development are available for about a dozen species (e.g. Mladenov & Chia, 1983; Holland & Kubota, 1975; Mortensen, 1920; Bury, 1888), the development of the banded pattern has not been examined in any detail. A number of points remain unclear; whether, for example, the bands form simultaneously or in sequence; whether they coalesce initially as complete stripes or arise through some more complicated sequence of pattern-forming events. Sea cucumbers also have a doliolaria stage, but the bands form from the single looped band of the auricularia by fragmentation and rearrangement (Bury, 1895), a remarkable process that is as little understood as de novo band formation in crinoids. In neither case do we know either the morphogenetic mechanism or the means of specifying pattern.

Any pattern-forming system capable of producing so distinctive a pattern is of interest in its own right, but stripe formation is of special interest because of its importance in establishing the basic body plan of insects, currently a subject of intense interest among developmental biologists (e.g. Hafen, Kuroiwa & Gehring, 1984; Martinez-Arias & Lawrence, 1985). And, while plausible morphogenetic models have been proposed that could generate stripes, the majority, including most diffusion-based models, either do not produce stripes or do so only under special circumstances (Kauffman, 1981; Lacalli, 1981). Stripe-forming systems thus offer a relatively stringent test for any patterning model and a possible means of distinguishing between competing explanations of pattern formation.

This paper examines normal development of the striped pattern of ciliary bands in the doliolaria stage of Florometra serratissima. Previous work on this species (Mladenov, 1981; Mladenov & Chia, 1983) shows it to be a useful one for embryological investigation and experiment. Specimens can be collected in good condition from shallow water and the spawning period is unusually long, lasting essentially all year, with reduced activity in winter months. Our results show that the initial pattern in Florometra does not consist of uniform stripes, but is converted secondarily to stripes from a pattern much closer to the predictions of morphogenetic theory. The initial pattern also has features in common with the sinusoidal bands of other echinoderm larvae, and could conceivably be a phylogenetic remnant explainable in evolutionary rather than mechanistic terms. We discuss both possibilities.

Specimens of Florometra serratissima were collected by diving at Bamfield Inlet and Rainy Bay, Barkeley Sound, BC, where they occur on rocky reefs from 18–20 m to deep water. As reported by Mladenov (1981), a given population seems always to have individuals in reproductive condition, and these can be identified by their swollen genital pinnules. Collections were made in April-June of 1984 and 85. The animals were kept in tanks of running sea water at the Bamfield Marine Station. Virtually all adult males have mature spermatozoa. These become active on release from the pinnule, done simply by removing the pinnule with forceps and cutting it open in a dish of filtered sea water. Obtaining ripe ova is more difficult. Maturation occurs as the oocytes are expelled from the germinal epithelium into the pinnule lumen. Ripe females most often contain oocytes rather than mature ova when collected, but will usually ovulate within a few days. The ova can then be dissected from the pinnule and fertilized. It is, however, difficult to obtain more than a few hundreds of eggs in this fashion at any one time. Ova are easily distinguished from oocytes by their smooth, spherical shape, the absence of a capsular chorion and germinal vesicle, and the fact that they do not adhere in clumps. Once collected, the ova were washed several times in filtered sea water and fertilized with drops of sperm suspension. Embryos were raised in small dishes at 10·5°C with daily changes of sea water. Cultures were checked for percentage fertilization and normal development at the 8-cell stage and during gastrulation, and those containing any abnormal embryos were discarded. Obtaining uniform cultures showing optimal development in this species requires some care.

Electron microscopy

Fixation for both transmission (TEM) and scanning electron microscopy (SEM) was by a variant of the semisimultaneous method (Lacalli, 1981a): larvae in sea water were fixed by addition of an equal volume of 2 % glutaraldehyde in 0·2M-cacodylate buffer with 6 % sucrose. After approximately 1 min an additional volume of 4 % OsO4 was added, and specimens were left in this mixed fixative for Ih. They were then washed, partially dehydrated, and stored in 70 % ethanol. Specimens for TEM were later rehydrated, stained overnight in aqueous 2 % uranyl acetate at 60°C, washed, dehydrated, embedded in Spurr’s resin, sectioned, and photographed without further staining. Midsagittal sections through whole larvae collected on slotted grids were used for routine ultrastructural examination. Serial series taken tangentially at the larval surface were used to measure band and interband cell surface areas in late stage larvae.

For SEM, specimens in acetone were dried by freeze drying, standard critical-point methods, or by acetone evaporation. The last is a modification of critical-point drying in which some acetone is allowed to remain in the pressure chamber. This dissolves in the critical phase at the critical point, condenses and rewets the specimen as CO2 is vented, and then evaporates. The rewetting and evaporation cycle is not as damaging as freeze drying, which causes extensive cracking, but is a convenient way of collapsing surface microvilli and cilia to expose the projecting knobs that support each cilium (cf. Figs 6, 7). Preparations of this type were used for all surface area calculations and for the cell counts used to generate area per cell data and cell numbers (Figs 10,11) except where otherwise specified. Division rates (Fig. 11C) are calculated from the slope between pairs of points (at times 1 and 2) from the data on cell number (n) in Fig. 11B as In (n2/n1) ln 2.

General observations

The principal events of F. serratissima embryogenesis are summarized in Fig. 1. Ova are opaque, yolky and approx. 210 µm in diameter. Development through gastrulation proceeds within an ornamented fertilization membrane until hatching, which occurs between 35 and 37h. Gastrulation is by invagination, and cilia first appear at the gastrula stage. After hatching there is a period of elongation. Body length increases by 40 % in 15 h to give an ovoid, uniformly ciliated larva by 50h (Figs 4, 5). A section through the epithelium of one such larva, prior to any evidence of band formation, is shown in Fig. 2. The cells are tall, columnar, and apparently of a single, relatively undifferentiated type. Each cell has one cilium, borne on a raised, knob-like projection (Fig. 8). At the cell surface are numerous microvilli supporting a delicate glycocalyx. When prepared for SEM by acetone evaporation, the microvilli, glycocalyx and cilia collapse, leaving the projecting knobs exposed (cf. Figs 4 and 5, 6 and 7). Most of our observations derive from preparations of this type, since they show clearly the distribution of ciliary knobs, and hence of the cells themselves. The underlying arrangement of cells is obscured in preparations that leave the cilia erect and intact.

Fig. 1.

Developmental schedule for F. serratissima at 10·5 °C showing stages of cleavage, gastrulation, the uniformly ciliated hatching larva, and the banded doliolaria.

Fig. 1.

Developmental schedule for F. serratissima at 10·5 °C showing stages of cleavage, gastrulation, the uniformly ciliated hatching larva, and the banded doliolaria.

Fig. 2.

Uniformly ciliated epithelium prior to band formation, from the ventral surface at 58 h. Dividing cell at *. ×3490.

Fig. 2.

Uniformly ciliated epithelium prior to band formation, from the ventral surface at 58 h. Dividing cell at *. ×3490.

Fig. 3.

Section through a ciliary band showing (arrow) its surface extent. Mucus cell at *. ×3510.

Fig. 3.

Section through a ciliary band showing (arrow) its surface extent. Mucus cell at *. ×3510.

Fig. 4.

Figs 48. Larvae prior to band formation. 50 h larva prepared by critical-point drying. Cilia are intact and erect. ×370.

Fig. 4.

Figs 48. Larvae prior to band formation. 50 h larva prepared by critical-point drying. Cilia are intact and erect. ×370.

Fig. 5.

50 h larva prepared by acetone evaporation. Shows collapsed cilia and exposed ciliary knobs. ×440.

Fig. 5.

50 h larva prepared by acetone evaporation. Shows collapsed cilia and exposed ciliary knobs. ×440.

Fig. 6.

Surface detail of a critical-point-dried specimen. ×5580.

Fig. 6.

Surface detail of a critical-point-dried specimen. ×5580.

Fig. 7.

Surface detail of an acetone-evaporated specimen. ×6890.

Fig. 7.

Surface detail of an acetone-evaporated specimen. ×6890.

Fig. 8.

Section through a ciliary knob showing the basal body and rootlet system, the accessory centriole (arrow), and microvilli supporting a thin glycocalyx (gc). The anterior end of the larva is to the left in this figure. Accessory centrioles are located on the posterior (downstroke) side of the rootlets in all cells examined except those of the apical tuft, where they lie on the opposite, i.e. the medial or anterior, side, x 18 250.

Fig. 8.

Section through a ciliary knob showing the basal body and rootlet system, the accessory centriole (arrow), and microvilli supporting a thin glycocalyx (gc). The anterior end of the larva is to the left in this figure. Accessory centrioles are located on the posterior (downstroke) side of the rootlets in all cells examined except those of the apical tuft, where they lie on the opposite, i.e. the medial or anterior, side, x 18 250.

The ciliary bands form between approx. 55 and 75 h, with all elements of the pattern being identifiable and reasonably well defined by 70 h. Once formed, the bands become progressively narrower, and interband cells lose both their cilia and ciliary knobs. Fig. 9 shows the completed pattern at 100 h. The larvae have an apical tuft and four circumferential bands, by convention numbered 14 from the anterior end. These are regular and evenly spaced on the dorsal surface. On the ventral surface, the pattern is interrupted by the vestibular invagination. Band 1 passes above this, arching slightly as it does so. Band 2 is diverted along the sides of the vestibule to pass below it. In some larvae, bands 2 and 3 fuse at this point. Around the vestibule is a roughly circular region, an oral or vestibular field, that is uniformly ciliated. The only other major pattern elements are the small axial bands (e.g. between bands 3 and 4 in the larvae shown in Fig. 9) that connect the circumferential bands in some larvae. As described below, these are remnants of the initial stages of pattern formation, and vary in their occurrence from larva to larva.

Fig. 9.

Stereopairs of 100h larvae, views of the ventral (top, x360) and dorsal (bottom, x330) surfaces. The two depressions on the ventral surface are the apical pit (upper depression) and vestibule.

Fig. 9.

Stereopairs of 100h larvae, views of the ventral (top, x360) and dorsal (bottom, x330) surfaces. The two depressions on the ventral surface are the apical pit (upper depression) and vestibule.

Fig. 10.

Larval dimensions and cell surface area. Points are means and standard error for five to ten specimens except open circles, which are values taken from single specimens. (A) Larval dimensions from SEM. The top curve gives total length for critical-point-dried (c.p.) specimens starting with (a) the diameter of the late gastrula. The bottom two curves give the length and average radius of the cylindrical portion of the larva from acetone evaporated (a.e.) specimens starting with (b) the radius of the late gastrula. Acetone evaporation causes some shrinkage of specimens compared with critical-point drying, typically on the order of 16-18 %. Data for the lower two curves are corrected to a constant shrinkage of 18%. (B) Surface area per cell. Data are obtained from counts of ciliary knobs within a known area of larval surface in SEMs for unbanded larvae and ciliary bands, and from TEM sections tangential to the larval surface for interband cells, which have no knobs.

Fig. 10.

Larval dimensions and cell surface area. Points are means and standard error for five to ten specimens except open circles, which are values taken from single specimens. (A) Larval dimensions from SEM. The top curve gives total length for critical-point-dried (c.p.) specimens starting with (a) the diameter of the late gastrula. The bottom two curves give the length and average radius of the cylindrical portion of the larva from acetone evaporated (a.e.) specimens starting with (b) the radius of the late gastrula. Acetone evaporation causes some shrinkage of specimens compared with critical-point drying, typically on the order of 16-18 %. Data for the lower two curves are corrected to a constant shrinkage of 18%. (B) Surface area per cell. Data are obtained from counts of ciliary knobs within a known area of larval surface in SEMs for unbanded larvae and ciliary bands, and from TEM sections tangential to the larval surface for interband cells, which have no knobs.

Fig. 11.

Ectodermal cell numbers. Data are for the cylindrical portion of a ‘double dorsal’ larva, i.e. a larva with four complete and uniform circumferential bands, calculated from surface areas (uncorrected data of the type shown in Fig. 10 A) and areas per cell (Fig. 10B). For the gastrulation stages, data are counts of single embryos reduced in proportion to the fraction of cylindrical surface in the immediate post-hatching stage. (A) Log10 cell number. (B) Cell number for the total cylinder of ectoderm and for the bands alone. The dashed line is a forward projection of band cell numbers assuming interband cell number is constant at 1500 cells. (C) Instantaneous division rate calculated from the means of cell numbers in (B). Values after 60 h are for band cells alone.

Fig. 11.

Ectodermal cell numbers. Data are for the cylindrical portion of a ‘double dorsal’ larva, i.e. a larva with four complete and uniform circumferential bands, calculated from surface areas (uncorrected data of the type shown in Fig. 10 A) and areas per cell (Fig. 10B). For the gastrulation stages, data are counts of single embryos reduced in proportion to the fraction of cylindrical surface in the immediate post-hatching stage. (A) Log10 cell number. (B) Cell number for the total cylinder of ectoderm and for the bands alone. The dashed line is a forward projection of band cell numbers assuming interband cell number is constant at 1500 cells. (C) Instantaneous division rate calculated from the means of cell numbers in (B). Values after 60 h are for band cells alone.

Band formation results in the conversion of a uniform epithelium (Fig. 2) to one with alternating domains of band and interband (Fig. 3). Band cells are narrower than their epithelial precursors, especially at their apices, but are otherwise similar in general appearance. Interband cells are flatter, lack cilia and ciliary knobs, and have a surface layer of large, clear vesicles. They expand during band formation, while the band cells become compressed to form a narrow, columnar band. The interband also contains mucus cells in various stages of differentiation, which are absent from the bands. A more complete ultrastructural description of the fully developed larva is provided by Chia et al. (1986).

Quantitative data on overall changes in epithelial cell size and number are shown in Figs 10 and 11. Data are comparatively easy to obtain for the dorsal surface of the larva because of its regular pattern. The ventral surface and hemispherical ends are more difficult to deal with. We restrict ourselves here to data on the cylindrical portion of a hypothetical ‘double dorsal’ larva, i.e. a larva with two dorsal surfaces and with both ends removed as in Fig. 12A. We exclude from consideration the approximately 20 h period of band formation because band and interband regions are too poorly defined during this period for a meaningful determination of cell numbers by our methods.

Fig. 12.

The cylindrical surface of the larva slit midventrally and projected flat as in (A) to show band formation at (B) 64 h and (C) 100h. Bands are stippled, the gradations indicate relative cell density. Interbands are unstippled. * in (B) is the site of the vestibular invagination, (ves) in (C).

Fig. 12.

The cylindrical surface of the larva slit midventrally and projected flat as in (A) to show band formation at (B) 64 h and (C) 100h. Bands are stippled, the gradations indicate relative cell density. Interbands are unstippled. * in (B) is the site of the vestibular invagination, (ves) in (C).

Fig. 10A shows changes in total body length, and values for the length and mean radius of the cylindrical portion of the body, which we have used to calculate cylinder surface areas. This last changes relatively little during development, from 4·5×104μm2 just after hatching to 6×104μm2 at the end of band formation. Area per cell (Fig. 10B) stabilizes after cleavage and gastrulation at approx. 15 μm2cell−1. Band formation involves an increase in cell density in the band, with cells typically having surface areas on the order of 5μm2cell−1 by 100 h. The epithelial cells of the interband region increase in surface area; values in Fig. 12B in fact underestimate this increase because the means are for all interband cells, including differentiating mucus cells, whose surface area is much smaller.

Cell number data are summarized in Fig. 11. The developmental events of interest here, i.e. those associated with pattern formation, occur well after the main period of exponential increase in cell number, as shown in Fig. 11A. Ectodermal cell number continues to increase after hatching (Fig. 11B), but does so slowly. Numbers level off at about 80 h and decrease thereafter. Band and interband behave quite differently in this regard, however. Band cell numbers increase throughout the period of our study, but at a declining rate, to plateau at approx. 3500 cells, or 875 cells per band. Later in development, near settlement, there is a period of major loss of cells from the bands. The three interband zones, when first fully defined and visible in SEM preparations, contain on the order of 1500 cells, i.e. 500 cells per interband or 57 % of the number of cells in a typical band. During the next 60 h, the number of interband cells decreases to about half the original number. We believe this is due to the loss of terminally differentiated mucus cells from the interband. There appears to be no significant cell proliferation in the interbands, as we have yet to find a single mitotic figure in sections through them. Assuming interband cells cease dividing at the time these zones are first specified as pattern elements, an estimate of band cell number can be derived by subtraction from total cell numbers for the period of band formation and before. The dashed part of the lower curve in Fig. 11B shows this. If one then knew or could predict the relative proportion of cells assigned to band and interband when pattern first begins to be expressed, our projection would give the time at which this occurred. For four bands alternating with three interbands, each containing an equal number of cells there would be 4/3 as many band as interband cells, which would happen in a pattern beginning its expression at 50 h. This is an oversimplification of the nature of the initial pattern (see below) and assumes further, that cells are not respecified as to type as the pattern develops, an assumption we can as yet neither confirm nor deny.

Values for the cell proliferation rate as a function of time can, in principle, be obtained from cell number data like that in Fig. 11B. The uncertainty associated with our data is too large to give a completely reliable result, but the means for cell number do fall roughly along a smooth curve, and from these we obtain a smooth, monotonically decreasing curve for the division rate (Fig. 11C). This shows division rate dropping rapidly after 40h to values of 0·01–0·02 divisions per hour by 50–55 h. Thereafter division is restricted to cells of the ciliary band, where we also do find mitotic figures (e.g. Fig. 3), and division rate continues a slow decline.

Band formation

The main elements of the ciliary band pattern do not appear simultaneously. There is instead a posterior-to-anterior and dorsal-to-ventral sequence. Times for the first visible sign of the posterior part of the pattern and of the last-formed anterior part are, respectively, for the dorsal surface, 56 and 64h; for the ventral surface, 62 and 70 h. For the pattern to propagate the length of the larva therefore requires, on each side, about 8h, with events on the dorsal surface preceding comparable ventral events by 6h. The time from first appearance of pattern to its completion, in at least rudimentary form, is then 14 h. Somewhat longer is required for the entire pattern to become uniformly well defined, on the order of 20 h in total, and it is this period that is indicated in the figures (Figs 1,10, 11) as the period of ‘band formation’.

Fig. 12 summarizes our findings on the pattern itself, its shape and sequence of development. Larvae in stages of band formation are shown in Figs 1316. Larvae at 78 h with complete patterns but as yet no significant band narrowing are shown in Figs 1721 (band width = 20–25μm). Figs 22-24 show 100 h larvae with narrower bands (band width = 10–12 μm) and a pattern comparable to that shown diagrammatically in Fig. 12C. The first sign of band formation, visible at 56h (Fig. 14) and well defined by 60–64 h (Fig. 16) is a cross-shaped domain of closely compressed cells occupying the posterior half of the dorsal surface. The lower arms of this cross extend around the larva to form band 4, which is complete on the ventral side by 62 h, and segregates a posterior cap of enlarged cells. By 60–62 h, the upper arms of the cross, the rudiments of band 3, connect by lateral loops to the rudiment of band 2 (Figs 13,15). Together these elements of the pattern define a roughly ovoid interband domain in the middle of the dorsal surface. When band 1 appears, a mid-dorsal connection is evident between it and band 2. This is variable between larvae, however, and is never so well defined as the connection between bands 3 and 4. In most larvae it has the appearance of two facing loops, each connecting bands 1 and 2, that just meet at the midline, as shown in Fig. 12B. In others, bands 1 and 2 approach each other medially, but form no clear connection. Pattern development is somewhat different on the ventral surface because the vestibule and its ciliary field must be accommodated in the pattern. The ventral portions of bands 3 and 4 are formed as on the dorsal side from the upper and lower arms, respectively, of a cross-shaped domain (Fig. 20). With band 3, the medial portion develops first, and only later connects to the already welldefined lateral loops that join the dorsal parts of bands 2 and 3 (Fig. 21). As the vestibular invagination appears, a zone of uniform ciliation develops surrounding it.that is not linked in any obvious way with other pattern elements. The ventral portion of band 1 is completed above this, roughly halfway between the vestibule and the apical pit. The ventral portion of band 2 develops along the lower margin of the vestibular field, and separate from it. This last element of the pattern is always narrower than the other bands and is variable from larva to larva. In some it passes as a complete band beneath the vestibule, but in others it fuses with band 3 or simply terminates as shown in Fig. 12C. Examples are shown in Figs 2124. As development proceeds, the interband domains, at first ovoid, expand laterally to become more rectangular. The axial connections between bands are reduced until, eventually, continuous interband stripes are formed, though in most larvae some remnant of the axial connections will persist. Examples visible in the figures include: the posterior mid-dorsal connection (Fig. 9 bottom), the posterior midventral one (Figs 9 top, 22), and several lateral ones (Figs 23, 24).

Fig. 13.

Figs 1316. Initial stages of band formation, all oriented with the anterior pole up. Side view at 62h, dorsal surface to the right. Note lateral loop connecting bands 2 and 3. x360.

Fig. 13.

Figs 1316. Initial stages of band formation, all oriented with the anterior pole up. Side view at 62h, dorsal surface to the right. Note lateral loop connecting bands 2 and 3. x360.

Fig. 14.

Dorsal view at 62 h showing an early stage in the development of the posterior cross. x380.

Fig. 14.

Dorsal view at 62 h showing an early stage in the development of the posterior cross. x380.

Fig. 15.

Side view of band 4 and the lateral loop (arrow) at 64 h, dorsal surface to the right. x485.

Fig. 15.

Side view of band 4 and the lateral loop (arrow) at 64 h, dorsal surface to the right. x485.

Fig. 16.

Dorsal view showing the posterior cross at 64h. x540.

Fig. 16.

Dorsal view showing the posterior cross at 64h. x540.

Fig. 17.

Figs 1721. Larvae at 78 h. All pattern elements are complete and well defined. The anterior pole is on the right. Ventral surface, x330.

Fig. 17.

Figs 1721. Larvae at 78 h. All pattern elements are complete and well defined. The anterior pole is on the right. Ventral surface, x330.

Fig. 18.

Side view, ventral surface up. x315.

Fig. 18.

Side view, ventral surface up. x315.

Fig. 19.

Dorsal surface. x340.

Fig. 19.

Dorsal surface. x340.

Fig. 20.

Detail of the ventral posterior cross. x580.

Fig. 20.

Detail of the ventral posterior cross. x580.

Fig. 21.

Oblique side view. Shows bands 2 and 3 joining beneath the vestibule (ves) and the small axial remnant of the lateral loop (arrow). x590.

Fig. 21.

Oblique side view. Shows bands 2 and 3 joining beneath the vestibule (ves) and the small axial remnant of the lateral loop (arrow). x590.

Fig. 22.

Figs 2224. Details of 100 h larvae showing the pattern in its fully developed state. Ventral surface showing the vestibular field and incomplete fusion, below the vestibule, of bands 2 and 3. Shows also a remnant of the axial connection between bands 3 and 4 (arrow). x615.

Fig. 22.

Figs 2224. Details of 100 h larvae showing the pattern in its fully developed state. Ventral surface showing the vestibular field and incomplete fusion, below the vestibule, of bands 2 and 3. Shows also a remnant of the axial connection between bands 3 and 4 (arrow). x615.

Fig. 23.

Oblique side view of the larva in Fig. 22 showing again the partial fusion between bands 2 and 3. x450.

Fig. 23.

Oblique side view of the larva in Fig. 22 showing again the partial fusion between bands 2 and 3. x450.

Fig. 24.

As in Fig. 23, another larva showing (arrow) the axial remnant of the lateral loop. x440.

Fig. 24.

As in Fig. 23, another larva showing (arrow) the axial remnant of the lateral loop. x440.

The principal new finding reported here is that the regular, striped pattern characteristic of the doliolaria does not develop directly. It instead arises secondarily from an initial pattern that has some stripe-like elements but is not, strictly speaking, striped. The interband zones, when they first appear, are arranged as an approximately close-packed array of circular or oval domains separated by and surrounded by belts of developing band of remarkably uniform width. Subsequent refinement of this initial pattern produces, in the end, regular stripes. The interband domains enlarge, due in part simply to the enlargement of the cells contained in them. They also change shape, becoming more rectangular as the bands progressively align with the transverse axis of the larva and axial connections between them disappear. From our data it is not clear whether these later changes are simply a continuation of the original pattern-forming process or depend upon some secondary, quite separate mechanism. Assuming the initial, visible events of pattern formation reflect an underlying morphogenetic prepattern imposed on the system as a whole, late pattern events could be due to a change in the prepattern that causes cells to be respecified as to type, i.e. band cells could be converted to interband and vice versa. Band-interband boundaries could then move without the cells themselves being required to move. Our data do not allow a choice between this and the alternative, that domain boundaries change only as the cells in each of the original domains rearrange themselves in relation to each other and to cells in neighbouring domains. This would presumably involve local cell-cell interactions of some type, and might at first seem unlikely because it requires that large-scale pattern features like the precisely aligned band-interband boundaries be generated and maintained by interactions of a strictly local nature. Some means would be required of making information on the system as a whole, i.e. on the larval axis, available to individual cells. In fact the cells do have an internal reference system that could do this. The ciliary rootlets and, presumably, the cytoskeletal elements attached to them, are aligned with the larval axis. This should, in principle, permit cells to distinguish internally between anterior and posterior and, in terms of axes, between axial and transverse. Whether such information is utilized in morphogenesis remains to be seen.

Since this account is restricted to visible events of pattern expression, it provides no direct evidence as to the nature, site or time of formation of the underlying prepattern, if such exists. We have suggested above, on the basis of cell counts, that 50 h is a reasonable time for what must be one of the earliest events of pattern expression, cell specification, to occur in the ectoderm, but the prepattern responsible could have formed long before this. Further, rather than being intrinsic to the ectoderm, pattern formation could involve the underlying tissues of the embryo as well. We are currently examining this question experimentally, but assume here, for the sake of discussion, that pattern and prepattern are sufficiently similar that our interpretation of the former is relevant also to the latter.

The initial pattern in Florometra is, as patterns go, both bizarre and unexpected. It is a surprisingly complex pattern rather than a simple one. We consider two possible explanations for this: (1) that the pattern is a phylogenetic remnant reflecting the arrangement of ciliary bands found in other echinoderm larvae, which are complicated, and (2) that it arises as a morphogenetic necessity because limitations inherent in the patterning mechanism prevent the direct formation of the deceptively ‘simple’ pattern of uniform stripes.

The pattern is a phylogenetic remnant

Free-swimming, feeding larvae occur in all echinoderm classes except the Crinoidea, and in all of these the principal ciliary band is circumoral. In the sea urchin pluteus the band is initially circular as shown in Fig. 25A (Czihak, 1962), but becomes convoluted as the larva grows. The bipinnaria larva of asteroids and the holothurian auricularia both have convoluted bands with large anterior and posterior loops extending along their lateral surfaces (Fig. 25D). These apparently develop by expansion of the oral field at its anterior-lateral and posterior-lateral margins as shown schematically in Fig. 25B. A continuation of this process, whether by localized proliferation or cell rearrangement, would eventually generate two zones of mid-dorsal band fusion as shown in Fig. 25C, essentially the pattern we find in Florometra. Fusions of this type do in fact occur in asteroid and holothurian larvae, resulting in the segregation of portions of the band system as separate preoral and anal loops. Near metamorphosis, the convoluted band of the holothurian auricularia further rearranges, and a banded doliolaria is formed (Bury, 1895). According to Grave (1903), there is evidence for similar rearrangements late in larval development in the other echinoderm classes. The ciliary band may thus pass through a similar sequence of pattern stages in all feeding larvae regardless of class, changing from a simple circumoral band, to a convoluted band, to a convoluted band with mid-dorsal fusions, to a stage with multiple bands. The last two stages would correspond to what we see in the crinoid as, respectively, the initial and final pattern. If crinoids once had a feeding larva that has now been lost, the developmental sequence by which it develops might now be truncated or compressed to eliminate the early events associated with the feeding stage, so that only the late events remain or, at least, are visibly in evidence. This would in part provide a rationale for explaining the appearance, initially, of a complex rather than a simple pattern. Alternatively, a banded larva like the doliolaria could be the basic larval type for echinoderms, so that feeding larvae would have had to have evolved from this. If an initial stage like that described here for the crinoid is also a basic, primitive feature, it is relatively easy to see how a convoluted feeding band could have arisen secondarily simply by doing the reverse of whatever converts (B) to (C) in Fig. 25, though this does not explain the initial pattern itself. Our work does not resolve these issues, but does show clearly the need for a careful reexamination of the process of band rearrangement in other echinoderm classes, particularly in the holothurians.

Fig. 25.

Band patterns in echinoderm larvae: a generalized scheme for generating complex patterns from simple ones. (A) A simple, circumoral band as seen in early sea urchin development. (B) A convoluted band generated by localized proliferation or cell rearrangement (at arrows), similar in general form to the bipinnaria/auriculariatype larva shown in (D). Continued growth of the convolutions would give mid-dorsal fusions and a pattern much like that seen in Florometra ((C), larva rotated to show the dorsal surface).

Fig. 25.

Band patterns in echinoderm larvae: a generalized scheme for generating complex patterns from simple ones. (A) A simple, circumoral band as seen in early sea urchin development. (B) A convoluted band generated by localized proliferation or cell rearrangement (at arrows), similar in general form to the bipinnaria/auriculariatype larva shown in (D). Continued growth of the convolutions would give mid-dorsal fusions and a pattern much like that seen in Florometra ((C), larva rotated to show the dorsal surface).

The pattern is a morphogenetic necessity

As an alternative we suggest that the initial pattern, whether derived from ancestral larval forms or not, is necessary because the pattern-forming mechanism is incapable of producing a regular, precisely ordered series of stripes directly. Stripes could in principle be produced in an epithelium by various means, for example, by cell counting or by the propagation of chains of inductive events or wave-like signals from one pole of the embryo (Cooke, 1981; Meinhardt, 1982), but there is as yet no evidence that phenomena of this type are important in echinoderm development. A somewhat more promising class of pattern-forming mechanisms are those involving isotropic diffusion of morphogenetically active substances between cells. Most of the activator-inhibitor models described by Meinhardt (1982) would, for example, be of this type. Among such models are a few that will produce stripes, but most do not. The problem, in general terms, is that the equations derived from these, in their linear form, generally admit a number of wave patterns as solutions. While some of these are striped or stripe-like, there will also be non-striped patterns of similar scale, and all compete for domination of the system as a whole as the pattern develops. The outcome is particularly difficult to control in a two-dimensional system as in the present case, where pattern develops on the surface of an ellipsoid. A glance through a catalogue of Chladni figures (e.g. Waller, 1961) gives some idea of just how many competing patterns one can have. In physical terms, a good analogy is the raucous sound of a cymbal, a typical badly controlled two-dimensional system with many competing vibrational modes, in contrast with the pure sound of a one-dimensional violin string (Harrison, 1982). Among the patterns possible in two dimensions are some that at least approximate the initial pattern seen in Florometra. Two of these, taken from Waller’s catalogue, are shown as Fig. 26A and B. They can be compared with our schematic version of the real pattern, shown as Fig. 26C.

Fig. 26.

(A,B) Standing-wave patterns on an ellipse, taken from Waller (1961). Lines represent nodes, i.e. equilibrium concentration values in the case of a morphogenetic model, and oppositely shaded domains show regions of displacement, in opposite direction, from equilibrium. (A) Pattern with four transverse nodes. (B) Pattern with four transverse and one radial node. Unshaded regions approximate in position the interband domains in the Florometra initial pattern, schematized in (C) for an ellipse.

Fig. 26.

(A,B) Standing-wave patterns on an ellipse, taken from Waller (1961). Lines represent nodes, i.e. equilibrium concentration values in the case of a morphogenetic model, and oppositely shaded domains show regions of displacement, in opposite direction, from equilibrium. (A) Pattern with four transverse nodes. (B) Pattern with four transverse and one radial node. Unshaded regions approximate in position the interband domains in the Florometra initial pattern, schematized in (C) for an ellipse.

From among the various competing linear patterns, most realistic non-linear models will select and amplify one or a few patterns to generate a quite specific final result. Meinhardt’s model for mutual activation of locally exclusive states is an example that produces stripes, but it is much more common to obtain patterns consisting of peaks of activated territory surrounded by valleys of unactivated territory. Patterns of this type with the peaks in close-packed array are characteristic of models of the substrate depletion type (Lacalli, 1981b; Meinhardt, 1982), a large and important class of diffusion-based models that includes, for example, the Brusselator and its variants. Such models are good candidates for explaining patterns like that in Florometra if one can equate the activated peaks generated by the model with the initial interband domains, and the valleys with developing bands. This is to some extent justified by the fact that it is the interbands, not the bands, i.e. the territory we identify as being activated, that is ultrastructurally more differentiated in terms of its divergence from the embryonic cell type, and it is the interband cells that first cease division.

In summary, we suggest the initial pattern in Florometra provides circumstantial evidence for the operation of a diffusion-based patterning mechanism, and that the pattern is converted secondarily to stripes, possibly by some other, as yet unspecified process. This indirect means of forming stripes may be necessary if the organism possesses no means of forming stripes except through a series of transient intermediate steps. Transient intermediate patterns have now been reported in another stripe-forming system, during segmentation in Drosophila (Hafen et al. 1984; Weir & Kornberg, 1985), and there is some theoretical justification for this expectation if a diffusion-based mechanism is responsible (Kauffman, 1981). Whether there are further similarities between the two systems remains to be seen.

This work was supported by NSERC Canada. We thank the director and staff of the Bamfield Marine Station for their assistance.

Bury
,
H.
(
1888
).
The early stages in the development of Antedon rosacea
.
Phil. Trans. R. Soc. B
179
,
257
301
.
Bury
,
H.
(
1895
).
The metamorphosis of echinoderms
.
Q. JI Microsc. Sci
.
38
,
45
135
.
Chia
,
F. S.
,
Burke
,
R. D.
,
Koss
,
R.
,
Mladenov
,
P. V.
&
Rumrill
,
S. S.
(
1986
).
Fine structure of the doliolaria larva of the feather star Florometra serratissima, with special emphasis on the nervous system
.
J. Morph
, (in press).
Cooke
,
J.
(
1981
).
The problem of periodic pattern in embryos
.
Phil. Trans. R. Soc. B
295
,
509
524
.
Czihak
,
G.
(
1962
).
Entwicklungsphysiologie der Echinodermen
.
Fortschr. Zool
.
14
,
238
267
.
Grave
,
C.
(
1903
).
On the occurrence among echinoderms of larvae with cilia arranged in transverse rings, with a suggestion as to their significance
.
Biol. Bull. mar. biol. Lab., Woods Hole
5
,
169
186
.
Hafen
,
E.
,
Kuroiwa
,
A.
&
Gehring
,
W. J.
(
1984
).
Spatial distribution of transcripts from the segmentation genefushi tarazu during Drosophila embryonic development
.
Cell
37
,
833
841
.
Harrison
,
L. G.
(
1982
).
An overview of kinetic theory in developmental modeling
.
In Developmental Order: Its Origin and Regulation
(ed.
S.
Subtelny
&
P. B.
Green
), pp.
3
33
.
New York
:
Alan R. Liss
.
Holland
,
N. D.
&
Kubota
,
H.
(
1975
).
Correlated scanning and transmission electron microscopy of larvae of the feather star Comanthus japónica
.
Trans. Am. microsc. Soc
.
94
,
58
70
.
Kauffman
,
S. A.
(
1981
).
Pattern formation in the Drosophila embryo
.
Phil. Trans. R. Soc. B
295
,
567
594
.
Lacalli
,
T. C.
(
1981a
).
Structure and development of the apical organ in trochophores of Spirobranchus polycerus, Phyllodoce maculata and Phyllodoce mucosa
.
Proc. R. Soc. B
212
,
381
402
.
Lacalli
,
T. C.
(
1981b
).
Dissipative structures and morphogenetic pattern in unicellular algae
.
Phil. Trans. R. Soc. B
294
,
547
588
.
Martinez-Arias
,
A.
&
Lawrence
,
P. A.
(
1985
).
Parasegments and compartments in the Drosophila embryo
.
Nature
313
,
639
642
.
Meinhardt
,
H.
(
1982
).
Models of Biological Pattern Formation
.
London
:
Academic Press
.
Mladenov
,
P. V.
(
1981
).
Development and reproductive biology of the feather star Florometra serratissima
.
209
pp.
Ph.D. Thesis, University of Alberta
.
Mladenov
,
P. V.
&
Chia
,
F. S.
(
1983
).
Development, settling behaviour, metamorphosis and pentacrinoid feeding and growth of the feather star Florometra serratissima
.
Marine Biology
73
,
309
323
.
Mortensen
,
T.
(
1920
).
Studies in the development of crinoids
.
Carnegie Inst. Publication Series no. 294
.
Waller
,
M. D.
(
1961
).
Chladni Figures: A Study in Symmetry
.
London
:
G. Bell & Sons
.
Weir
,
M. P.
&
Kornberg
,
T.
(
1985
).
Patterns of engrailed and fushi tarazu transcripts reveal novel intermediate stages in Drosophila segmentation
.
Nature
318
,
433
439
.