Somitogenesis in Amphibian Embryos

J.\Embryol. exp. Morph. Vol. 51, pp. 21-50, 1979 27 Printed in Great Britain @? Company of Biologists Limited 1979

Somitogenesis in amphibian embryos

I. Experimental evidence for an interaction between two temporal factors in the specification of somite pattern

By MURRAY PEARSON1 AND TOM ELSDALE2 From the MRC Unit of Clinical and Population Cytogenetics, Western General Hospital, Edinburgh

SUMMARY Somitogenesis is described in two species of anuran amphibians, Xenopus laevis and Rana temporaria, in which the cellular mechanics of somite formation are distinctly different. Heat shocks are employed to demonstrate a wave of cellular change which precedes somite formation down the body axis. This prior wave is shown to be kinematic. It is not a propagated wave. It is a consequence of the temporal activities of the cells laid out in space, but there is no evidence that these activities depend upon an interpretation of their position. Heat shocks cause characteristic segmental abnormalities over a zone of somites which is formed several hours after the shock. Evidence from double heat shock experiments suggests that the pattern of abnormality is the result of (i) a disturbance of co-ordination between pre-somitic cells, and (ii) the time available to those cells for recovery before they are recruited into a segmental pre-pattern at the time of passage of the prior wave. It is a temporal co- ordination that is disturbed and subsequently recovered following a heat shock. This temporal co-ordination of pre-somitic cells does not depend upon position along the axis. The evidence for two physiologically independent temporal patterns of cellular processes, which interact to specify the segmental pattern of somites (their size, shape and number), gives experimental support for the theoretical account of somitogenesis proposed by Cooke & Zeeman (1976).

INTRODUCTION Segmentation is characteristic of the development of many animal forms and a basic principle of their morphogenesis. Somitogenesis in the vertebrate embryo is a striking example. The various aspects of somitogenesis in amphibians have been discussed by Deuchar & Burgess (1967) and Cooke & Zeeman (1976): the craniocaudal sequence; the constancy of the pattern within a species; the co- ordination with the whole body pattern; the precision and versatility of the control mechanisms that regulate the number and size of somites, not only under normal conditions, but also abnormal regimes of ploidy (Hamilton, 1969) and experimental manipulation (Cooke, 1975). In our approach to the problem of somitogenesis we deal first with temporal 1 Author's address: 30 Albion Hill, Brighton, Sussex, U.K. 2 Author's address: MRC Unit of Clinical and Population Cytogenetics, Western General Hospital, Crewe Road, Edinburgh, EH4 2XU, Scotland. 28 M. PEARSON AND T. ELSDALE aspects of the process because only by doing so shall we gain insight into the essential continuity of the whole process as well as the mechanism of dis- continuity between somites. Segmentation begins just behind the head and sweeps down the axis to the end of the tail; we refer to this as a wave of somite formation. (When we refer to a wavefront we mean the movement of the frontier, for instance in this case the frontier between segmented and not yet segmented tissue.) In temporal terms, the problem of the specification of segmental pattern is to understand how the period taken by the wave of somite formation to traverse the axis is specified, and how it is partitioned into just so many of the shorter intervals between the formation of one somite and the next. We have previously demonstrated (Elsdale, Pearson & Whitehead, 1976) how we can identify a crucial change in the properties of the pre-somitic tissue occurring some hours before participation in somite formation. The basic observation was that heat shocks given just before or during somitogenesis lead to abnormalities in the segmental pattern. Heat shock does not disturb already formed somites, and for a while afterwards new somites continue to form as if nothing had happened. After a standard number of normal somites have been made, however, abnormal somites begin to appear. Abnormal segmentation continues for a longer or shorter time depending on the severity and duration of the shock; eventually there is a return to normal segmentation. The anterior part of the abnormal zone is the most disturbed; and the last normal somite is immediately followed by maximally disturbed tissue to give a sharp anterior boundary. Across this boundary, tissue immune to heat shock lies immediately adjacent to tissue maximally disturbed. This means that tissue disturbable by heat shock earlier in its maturation suddenly becomes insensitive in the run up to somite formation. Abnormal zones commencing further and further caudally are produced by shocks to progressively older embryos, from which it follows that the frontier between immune and disturbable tissue moves as a wavefront down the axis ahead of visible segmentation. The wave of somite formation is preceded by a prior wave of change which confers immunity to heat shock. It is not clear, however, whether the normal segmental pattern is specified at this time of acquisition by cells of an immunity to heat shocks, or whether cells then merely become insensitive to any disruption of a segmental prepattern which has already been set. In this paper we present observations and experiments to show that the prior wave of cellular change is concomitant with segmental specification. Several observations support the inference that although heat shock is used to demonstrate the existence of the prior wave, this wave itself is not directly affected by heat shock. There is another component in the specification of segmental pattern. The demonstration of another component which is sensitive to heat shock comes from analysis of the abnormal zone following a single or two successive shocks, and leads to the conclusion that heat shock disturbs a co-ordination of the paraxial mesoderm cells established before the arrival of the wave. Somitogenesis in amphibian embryos. I 29 The results are interpreted according to a scheme for the specification of the somite pattern which involves the coupling of both components, a wave of cellular change by which cells are committed to subsequent somitogenesis, and a co-ordinated periodicity, in the establishment of a segmental prepattern: there is a wave of somite determination. Finally a model coupling suggested by Cooke & Zeeman (1976) is discussed.

MATERIALS AND METHODS Rearing Xenopus embryos came from the Genetics Department, University of Edin- burgh, or from Dr C. Ford, School of Biological Sciences, University of Sussex, from spawnings induced by the injection of chorionic gonadotrophin. Embryos were reared in dechlorinated tapwater, sorted into batches of the same stage during blastula, and again at gastrula stages. Synchronous groups were reared at room temperature or controlled temperatures. Observations on somites were carried out after stripping the epidermis (Elsdale et al. 1976). Clutches of Rana embryos were collected from natural habitats in Perthshire. Synchronous batches were reared in pondwater.

Heat shocks After removal of their outer jelly coats, embryos were pipetted into bottles containing 150 ml of water and standing in a 37 °C water bath. After the measured period, usually 15 min for Xenopus and 8 min for Rana, they were pipetted back into an excess of water at room temperature or some other controlled temperature.

Microscopy Material fixed in Smith's or Bouin's fixative was dehydrated, embedded in paraffin wax and sectioned at 5-6 ^m. The sections were lightly stained with haemalum for histological examination. For scanning electron microscopy, specimens were prepared by the method described by Bard, Hay & Mellor (1975) and examined in a Cambridge Stereo- scan SI 80.

Bisection o/Rana embryos Embryos were cut in two with tungsten needles in agar dishes containing a solution of 65 % Dulbecco's medium. After 1 h the fragments were cleaned of dead cells and the medium drawn off to be replaced by a 5 % Dulbecco solution. Anterior halves were stripped for scoring at a relatively early stage; the tail halves were cultured longer, until segmentation in controls was complete, and then fixed and stripped.

3 EMB 51 30 M. PEARSON AND T. ELSDALE

RESULTS 1. Normal somitogenesis Somite formation in Xenopus is shown in Fig. 1. Paraxial mesoderm cells first elongate perpendicular to the notochord; then a bundle of the most anterior cells turns through 90° to give a somite, with myoblasts now lying parallel to the notochord. After several hours these myoblasts differentiate as uninucleate muscle fibres stretching from end to end of the somite. The process has been fully described by Hamilton (1969). In Rana the mechanics of somite formation are different. The somites form before cell elongation, and in a manner comparable to many other vertebrates where each somite forms initially as a rosette (Fig. 3). The adhesion between cells of a segmenting somite confers a roughly radial organization on the group, and de-adhesion from the non-segmenting cells posteriorly creates an intersomitic fissure (Fig. 2). Later, myoblasts elongate within the somite and fuse to give multinucleate muscle fibres (Figs. 4 and 5). Counts for the total number of somites formed in Rana vary from 39 to 45, but it is unclear how great is the genuine variation between embryos because boundaries between the small, last formed somites are very difficult to see once myoblasts fuse, and yet before fusion one cannot be certain that the final somite has formed. In Xenopus some 46 somites form to the end of the tail proper; thereafter very small somites are added to the curious tail process which grows throughout larval life.

FIGURES 1-5 Fig. 1. Somite formation at ca. 26-somite stage in Xenopus laevis. A horizontal section showing each segmental group of cells turning through 90°. At this level, each somite (23-26) comprises 4-5 myotomal cells in width, ps, Pre-somitic mesoderm; n, notochord; nt, neural tube. Anterior is to the right. Fig. 2. Somite formation in Rana temporaria. A horizontal section through 15- somite embryo. Caudally (left) pre-somitic cells are unsegmented. Further forward, somites have been formed by rounding up of the segmenting cells to present in section a vague rosette profile. Somites at bottom show this better than at top, the latter are cut slightly oblique, nt, Neural tube; ps, pre-somitic mesoderm; 12-15, ordinal somite number, counted antero-posteriorly from first post-otic somite. Fig. 3. Scanning electron micrograph of Rana neurula showing a partially formed segmental furrow. Unlike Xenopus cells at the segmenting stage are not elongated. Segmentation begins with a furrow dorsally (arrow). Fig. 4. Rana. A horizontal section through a 15-somite embryo showing the anterior first three somites. This is the same embryo as in Fig. 2, but anteriorly at this stage the myoblasts have elongated and fused to give multinucleate cells which differentiate as muscle fibres, n, Notochord; v, otic vesicle. Fig. 5. Scanning electron micrograph of Rana embryo showing elongation and fusion of myoblasts. Elongation begins six or seven somites anterior to the most recently formed, and fusion in somites more than ten anterior to the most recent. Such fusion is evident in the most anterior somites {ant) shown here. Somitogenesis in amphibian embryos. I 31

Figs. 1-5. For legends see facing page.

3-2 32 M. PEARSON AND T. ELSDALE

Xenopus: rate of somite formation

(a) 44 -|

40 •

I 20

.16

10 20 30 40 50 60 70 80 90 100 Time (h) Fig. 6a. For legend see facing page.

In spite of the different methods of forming somites, in both species the effects of heat shocks on the pattern of somitogenesis are very similar.

2. Characterization of the prior wave (a) The course of the prior wave in relation to the wave of somite formation A graph of somite formation against time is simply constructed from counts of the number of somites in embryos sampled successively from a synchronous batch (Figs. 6 and 7). In both species the first several somites form in rapid succession, thereafter a slower rate of formation is constant to the end of the measuring period, to the 37-somite stage in Rana, the 46-somite stage in Xenopus. Hamilton (1969) computed just such a linear relation (r = 0-995) for Xenopus from the data in Nieuwkoop & Faber's (1956) Normal Table. Comparable data on the prior wave of cellular change derive from counts of the total number of normal somites anterior to the abnormal zone developed in Somitogenesis in amphibian embryos. I 33

Xenopus 32 -

Stages N/F 12 13 14 15 17 18 19 20

10 15 20 25 30 40 50 60 70 80 Hours at 15 °C Fig. 6. (a) The rate of somite formation in Xenopus laevis at different temperatures. After the first five or six somites, all subsequent somites form at a constant rate at a given temperature. For both 16 °C and 20 °C, different symbols indicate different synchronous batches of embryos. For one batch at either temperature, the range of each somite count is shown by the vertical bar. For each count, the mean was taken of at least five embryos (ten files); the two files of any one embryo are not exactly synchronous, (b) The rate of somite formation at 15 °C (•) is plotted together with the segmental rate of progress of the prior wave (#). The position of the prior wavefront at any time is given by the number of normal somites before the abnormal zone caused by heat shock at the time concerned. For embryos shocked before appearance of the first somite, stages according to Nieuwkoop & Faber (1956) were calibrated to the same time scale for 15 CC. The segmental rate of progress of this prior wave shown by heat shock roughly fits a straight line.

embryos following a heat shock at a known time or stage. When the course of the prior wave is thus plotted as presumptive somites against time, this course is found to be linear from the beginning, and runs parallel to the linear part of the curve for somite formation. The segmental rate of progress of the prior wave is constant with time (Figs. 6 and 7). The separation between the two waves can be read from the graph. It is the difference between the number of somites present at the time of shock and the number subsequently counted anterior to the abnormal zone; that is, the number of normal somites formed following shock. The separation is ca. five somites in Xenopus and between three and four somites in Rana, except for the 34 M. PEARSON AND T. ELSDALE Rana

40 T

36 15 °C 32 •

* 24 • P 5 14°C £ 20

IJ 6

12 •

4 - JL /A 10 20 30 40 50 60 70 80 90 Time (h) Fig. 7. The rate of somite formation in Rana temporaria is also constant at different temperatures (#, •, A) after the first four somites. Vertical bars indicate the range of counts; for each count the mean was taken of at least five embryos (ten files). The broken line (— O —) plots the segmental rate of progress of the prior wave in the same batch of embryos at 18 °C in which somite formation is also plotted. The rate is constant throughout. The graph shows the 'lag' following the prior wave before any given somite is formed. first few somites. The constancy of the relation between the two waves after these first few somites suggests the prior wave may be fundamental to the pattern of somites, and somite formation a secondary consequence. In such case, the rapid appearance of the first few somites would reflect their delayed formation after the passage of the prior wave.

(b) The rate of recruitment of pre-somitic cells into somite formation and the prior wave The first 10-15 somites are of roughly equal size, after which they get smaller in both species. This is not solely due to a decrease in cell size; fewer cells are recruited into more caudal somites. To measure this decreasing rate of recruit- Somitogenesis in amphibian embryos. I 35

i i i i i i i i I i i i i i i 2 4 6 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 Somite number (n) Fig. 8. This figure plots the number of cells per somite width against ordinal numbers of the somites in Xenopus. The average value for each ordinal somite (A) is obtained from a number of counts from different embryos in horizontal section at the level of the notochord. Only counts which were unambiguous are included. The inset diagram shows how the count of x cells is related to the mode of somite formation, and to the length of pre-somitic material included in a somite. Three series of embryos are represented, (i) Normal embryos given by (#) with average counts for somites 1-42. Vertical bars show the range of such counts for each ordinal somite, (ii) Embryos given a heat shock at the 13-somite stage which results in an abnormal zone whose average length extends from somite 16 to 20. Counts for somite width are given for these embryos posterior to the abnormal zone (A)- (iii) A similar series shocked at the 17-somite stage whose abnormal zones extend an average length from somite 20 to 24. Counts for somite widths posterior to this abnormality are given by (O)- For normal embryos (#) each average count is the mean of from 4 to 15 different counts. Up to somite 26,10 or more counts contribute to this mean, smaller numbers of counts in more posterior somites. All counts were made before myoblast differentiation. For heat shocked embryos no ranges are given, since no count varied more than in normal controls; each average is the mean of two to eight different counts. Since the segmental rate of progress of the prior wave is a constant, the ordinal somite number is a measure of time, and the curve describes the decreasing rate of recruitment of pre-somitic cells along the axis into this prior wave, the curve of velocity against time. This velocity is no different behind an abnormal zone following heat shock compared with the same position in normal embryos, since somite width is the same for any ordinal somite. raent, advantage is taken of the peculiar mode of somite formation in Xenopus. In this species, the number of cells counted across a somite in horizontal section gives the number of pre-somitic cells along the craniocaudal axis recruited into that somite (see Fig. 8). We thus obtain in Fig. 8 a quantitative estimate of the length of pre-somitic material along the axis incorporated into each somite. Because the segmental rate of passage of the prior wave (number of presumptive somites per unit time) is constant, this same plot of cells per somite width against ordinal sequence of somites can be translated as the rate of recruitment of cells by this wavefront against time. The only assumption here is that cell proliferation in the interval between the prior wave and somite formation is 36 M. PEARSON AND T. ELSDALE

280

240

200

jj 140

10 14 20 24 30 34 40 42 Somite number Fig. 9. The plot of total number of pre-somitic cells along the axis incorporated into somites against ordinal somite number. The total number of pre-somitic cells along the axis incorporated into somites up to somite n is simply the sum of the average somite widths of somites 1 to n. As in Fig. 8, ordinal somite number is a measure of developmental time for the progress of the prior wave. The curve for cell recruitment therefore gives the axial length versus time curve for the prior wave. The wave begins with a constant velocity and slows gradually in the posterior trunk and anterior tail region. negligible: mitoses are not seen at this stage. Figure 8 is thus the second order curve for the prior wave, whose shape in space-time is given in Fig. 9 by plotting the total number of cells along the pre-somitic axial dimension incorporated into somites against ordinal somite sequence

(c) The prior wave is kinematic A frontier of change moving across a tissue may reflect one of two quite different processes. In the one case, the moving frontier may depend upon the propagation of a stimulus from cell to cell in the direction of the movement. Alternatively there may be nothing actually spreading across the tissue to prompt the change, the moving frontier reflecting the predetermination of the cells to change one after another in a temporal sequence across the tissue. In this case we have a kinematic wave. A simple experiment distinguishes between the two sorts of wave: if the tissue is cut in two before the passage of the wave, a propagated wave is stopped at the cut, whereas a kinematic wave is not stopped and appears to jump across the cut. Deuchar & Burgess (1967) proved that the wave of somite formation is kinematic by cutting neurulae in two caudal to the last formed somite. We describe a similar experiment proving that the prior wave is also kinematic. Somitogenesis in amphibian embryos. I 37

Table 1. Results of the bisection experiment Rana embryos were cut in two at the stage of first somite formation, into anterior plus posterior halves. In series 1 the bisected embryos were given no heat shock. Of 12 embryos only 2 failed to develop both anterior and posterior halves up to completion of segmentation. Ten embryos were available for complete scoring. In series II embryos were heat shocked, then immediately bisected. In 4 out of 12 embryos both halves did not survive to scoring, and in another 2 embryos the tail half became so bent that it was only possible to score one side. There were thus 14 out of an initially possible 24 files scored. In all but one of these, abnormal segmentation begins in the anterior half embryo. In both series I and 11 the total number of somites formed by the two halves combined is virtually identical to non-bisected control embryos in series III.

Total Abnormal Mean no. of segmentation in A \Apan fircf* oUI111ICo 111 ( "\ IV! Cct 11 -111 o I No. of 37 °C files somites in ant. + post. Ant. Post. abnormal embryos bisected shock recovered ant. half halves half half somite I. 12 _ 20/24 8-5 39-43 (mean 40-5) 11. 12 + 14/24 11-6 37-42 13/14 13/14 6-9 (mean 39-4) III. 6 controls + 12/12 — 39-42 — — 7-4 not bisected

Rana embryos at the neural fold stage, before the formation of the first somites, were cut transversely in two immediately after an 8 min heat shock. At this stage the prospective somitic material is confined to the posterior half of the embryo; the cut therefore was made rather more than half way along the embryos, nearer the posterior end. The two halves were reared separately. Not only was the total number of somites formed by pairs of half embryos normal, but also the sequence of somite formation was normal, and tail segmentation was completed in posterior halves at the same time as in intact controls. The anterior halves developed the beginning of an abnormal zone in all but one case scored; the posterior halves showed a continuation of the abnormal zone (Table 1). Heat shock was used here to indicate, by the ordinal position of the first abnormal somite, the position of the prior wavefront of cellular change at the time of separation. The presence of abnormal somites in the anterior half proves that the advancing wavefront had not yet reached the posterior half. The demonstration of somitogenesis in the posterior half therefore proves the kinematic nature of the prior wave.

(d) The wave of somite formation and the prior wave after heat shock The rate of segmentation behind the abnormal zone is the same as in controls in both species (Fig. 10). Heat shocked embryos are invariably retarded in their development in comparison with controls. Figure 10 shows that the retardation 38 M. PEARSON AND T. ELSDALE

°36 - (fl) (b) 1 32 Xenopus laeris

0 24 .5 20 | 16 1 12 o c o s 4 4 2 _l J L J 1 L 0 4 8 12 16 20 24 28 32 36 40 0 4 8 12 16 20 24 28 32 36 40 Numbers of somites in controls Number of somites in controls Fig. 10. Retardation in somitogenesis caused by heat shocks at different stages in (a) Rana and (n) Xenopus. In both cases the broken line represents the normal rate of somitogenesis in controls. Arrows indicate the stage at which different groups of embryos were heat shocked, (a) In Rana shocks given at three different stages gave an identical result: a temporary arrest which leads to resumption of domito- genesis at the same rate as in controls, but now a constant 4 somites behind the unshocked controls, (b) The same effect is seen in Xenopus. Heat shock does not affect the rate of subsequent segmentation. The vertical bar shows the extent of the abnormally segmenting zone; estimates of somites' worth of material segmented in this abnormal zone are based on the most dorsal intersomitic furrows where disturbance is least. of somitogenesis is entirely due to a transient arrest immediately following shock, an arrest which appears to affect the whole embryonic process and not just segmentation. If the overall rate of passage (i.e. rate of cell recruitment) of the wave of somite formation were affected by shock, then, since the rate of subsequent segmentation remains the same as controls, we should expect to find either larger and concomitantly fewer somites, or smaller and more numerous somites, formed as a result of a heat shock. We have gathered data on the normality of the somite files as a whole in heat shocked embryos, and in particular on whether the return to normal segmentation posterior to the abnormal zone produces the correct number of somites, correctly sized according to their position along the axis, in the correct time. The following observations were made on batches of embryos receiving shocks of intermediate duration (15 min in Xenopus, 10 min in Rana) resulting in abnormal zones about 8 somites long. (i) The number of somites segmented in files, including an abnormal zone, is the same as in controls in both species. These counts, to the end of the tail proper in Xenopus (46 somites) and to the end of the file in Rana, include estimates of the number of somites, or somites' worth of material, in the abnormal zone. In zones about 8 somites long the most dorsal intersomitic Somitogenesis in amphibian embryos. I 39 boundaries are often minimally disturbed and estimates are unlikely to err by more than ± 1 somite. (ii) In Xenopus the width of the somites behind the abnormal zone, measured by cell number, is normal for their ordinal positions along the axis (Fig. 8). (iii) Since the rate of segmentation is the same as in controls following the initial arrest, the correct number of normal sized somites are formed at the correct time along the axis. Heat shock therefore does not alter the rate of passage of the wave of somite formation. Within the abnormal zone itself, shocked embryos show the same antero-posterior sequence of (abnormal) segmentation as in normally segmenting embryos. The separation between this wave of somite formation and the prior wave can be demonstrated in heat shocked embryos just as in previously unshocked embryos (Figs. 6 and 7). Rana embryos were given a second shock at different stages after a first shock at the 1-3 somite stage. In each case the number of normal somites which formed in the interval between this second shock and the beginning of a second abnormal zone did not differ significantly from controls which had been given no previous shock, as in Fig. 7. The prior wave proceeds as usual at the same rate as the wave of somite formation, preceding by about three to four somites.

3. The hidden effect of a heat shock The previous demonstration that visible segmentation is normal in all respects behind an abnormal zone might suggest that the effect of a heat shock is exhausted with the abnormal zone. Double heat shock experiments however indicated a hidden effect of heat shock that is more enduring. Rana embryos were given two shocks so arranged that a minimum of two normal somites intervened between the end of the first abnormal zone and the beginning of the second. Shocks of 8 min and 9 min were employed. The first was given at the very beginning of somitogenesis, the second around the 9-somite stage. Controls were given only the latter shock. Results are set out in Table 2. The interesting result is that the abnormal zone resulting from a second shock is shorter than expected, and the severity of the abnormality considerably reduced (Fig. 11). A heat shock therefore partially protects against a subsequent shock. 4. The abnormal zone Apart from the abnormal zone, the passage of the wave of somite formation, and the prior wave, is not affected by heat shock. Within the abnormal zone, only segmental pattern is affected; the cells subsequently differentiate normally. (i) The setting of the abnormal zone. Heat shock is followed by a 'lag' during which several somites segment normally before the disturbance is registered. This 'lag' is independent of the duration of the shock, and following 40 M. PEARSON AND T. ELSDALE

Table 2. The effect of a double heat shock on Rana embryos Experimental embryos were given an 8 min shock at the stage of first formed somites, and a second shock during formation of the resultant abnormal zone. This resulted in a second abnormal zone, separated behind the first by at least two normal somites. Controls were shocked at the same stage as the experimental embryos at the time of their second shock; there is in fact a difference of a mean one somite between these two groups at the time of this shock, shown by the first abnormal somite (FAS) for each. The number of abnormal somites (NAS) is larger by three somites in controls. The same effect of protection against a second shock is seen following 9 min shocks. Not only the length of the abnormal zone, but also the severity of abnormality is reduced by the first shock, as shown in Fig. 11 (S.D., standard deviation; n, number of files scored).

Control Experimental (second shock) (double shock) only 8 min shock Mean FAS 13-3 120 S.D. 0-9 0-8 Mean NAS 2-1 5-5 S.D. 1-4 1-9 n 34 27 9 min shock Mean FAS 10-9 10-9 S.D. 0-6 H Mean NAS 4-4 6-9 S.D. 11 2-5 n 15 14 shocks given after the earliest stages of somitogenesis, the number of 'lag' somites is roughly constant (Figs. 6 and 7). (ii) The abnormal zone may extend from 2 to upward of 20 somites' worth of tissue, the length being proportional to the duration of the heat shock (Table 3). The longer the abnormal zone, the more severe is the visible disturbance of segmentation. (iii) Asymmetry. The abrupt transition from the last normal to the first abnormal somite has been mentioned. The disturbance is greatest anteriorly and then gradually returns to normal posteriorly. (iv) The essential disorder within the abnormal zone. The effect in Xenopus was described by Elsdale et al. (1976): the elongation, alignment, and rotation of cells occurs just as in normal segmentation, but instead of regular segmental blocks, small irregular associations of cells create a chaotic myotome in which regular segmental boundaries are lost (Fig. 12). Because in Rana cellular elongation is deferred until some time after segmentation, the picture seen on stripped Rana embryos is a more direct reflexion of the disturbance caused by heat shocks. Somitogenesis in amphibian embryos. I 41

Fig. 11. A comparison of the abnormality induced by a 9 min shock given to Rana embryos of the same stage (a) after a previous shock, and (b) after no previous shock. In (a) both the length of the abnormal zone and the severity of the disturbance are considerably reduced by the effect of the earlier shock. 42 M. PEARSON AND T. ELSDALE

Table 3. The correlation between length of the abnormal zone, measured by the number of abnormal somites (NAS) and the duration of the shock Rana embryos from the same ovulation were given 6, 8, or 12 min heat shocks (37 °C) at the 9-10 somite stage. Essentially identical results were obtained with embryos at other stages during trunk segmentation; in later tail stages a 12 min shock produced abnormalities to the end of the tail, and shorter shocks corresponsingly produced longer abnormal zones than in the trunk region. The same relation between duration of shock and length of the abnormal zone was described in Xenopus by Elsdale et al. (1976).

Duration of heat shock

6 min 8 min 12 min Mean NAS 2-5 4-5 9-4 S.D. 1-3 1-5 2-2 // 19 39 26

The anterior abnormal zone, especially following severe shock, often presents a solid block of apparently unsegmented tissue equivalent to two or more somites' length. Detailed examination reveals that this superficial appearance is not due to a failure of segmentation but, on the contrary, to an extreme frag- mentation of the impulse to segment, giving rise to a stippled appearance, the result of condensation of cells into small, irregular groups (Fig. 11 b). Posteriorly this fine stippling resolves into a chaotic segmentation characterized by hap- hazard and numerous short furrows and indentations. This appearance in turn simplifies posteriorly, and with the appearance of clear intersomitic boundaries and the disappearance of supernumerary furrows the normal segmental pattern is re-established (Fig. 13). In normal somitogenesis there appears to be a change in intercellular adhesion at the time of segmentation; looser cell contacts are succeeded by a rather tighter adhesion pattern (Fig. 14). Bellairs, Curtis & Sanders (1978) have demonstrated in the chick that a wave of increased cellular adhesiveness leads to somite formation.) In Rana embryos following heat shock, cells undergo the same change within the abnormal zone as in normally segmenting embryos, but instead of forming large, somitic groups, the pre-somitic con- tinuum breaks up into small irregular groupings. The essential disorder observed here is an excessive segmentation, extreme anteriorly, declining posteriorly, which reflects a disorganization of the normal pattern of cellular condensation. There is no evidence of cell death. Somitogenesis in amphibian embryos. I 43

Fig. 12. A light micrograph, horizontal section through Xenopus embryos following a 15 min heat shock and subsequent abnormal segmentation. In the abnormal zones indicated ( ) cells have rotated, but chaotically. Thus there is no proper alignment of myoblast nuclei as in normal somites seen anteriorly in (a). In both (a) and (b) the return to normal segmentation posteriorly is evident. In both, however, one side is seen to re-establish normal segmentation before the other. The two files do show independent variation in the extent of the abnormal zone, but the whole series of sections would need to be read to establish the true extent of an abnormal zone in any embryo. A single section may be misleading, n, Noto- chord; nt, neural tube.

DISCUSSION Heat shocks have been used to demonstrate, firstly, that segmentation can be disturbed, and secondly that there is a wave of change in the presomitic tissue prior to somite formation. After the passage of this wave of change segmental specification is not disturbed by heat shock. Two observations together suggest that this prior wave is a determinant of the succeeding wave of somite formation : (i) the constancy of the developmental relation between the two waves, and (ii) the nature of the change before somite formation, an abrupt change with which future somite boundaries become insensitive to disturbance by shocks. Our hypothesis is that the behavioural changes responsible for the association of cells into segmental groups, for somite formation, are triggered by the passage of the prior wave. 44 M. PEARSON AND T. ELSDALE

Fig. 13. Scanning electron micrograph of stripped Rana tail-bud embryo showing the typical effect of an 8 min heat shock during the neurula stage. The abnormal zone is localized, is most severe anteriorly, and then returns gradually to normal segmentation posteriorly. The nature of the abnormality is discussed fully in the text. Note that the dorsal embryo is toward the bottom of the figure, ventral to the top. Anterior to the left. Fig. 14. Newly formed somites in the same embryo as Fig. 13, enlarged to show apparently tighter adhesion in somites than in the posterior, non-segmented pre-somitic mesoderm cells. Somitogenesis in amphibian embryos. I 45 This hypothesis immediately raises two questions: (i) what is the cause of the prior wave? and (ii) is this wave itself part of the mechanism of segmental patterning and its regulation, or is it merely a temporal reflexion of an under- lying spatial pattern ? These two questions are interrelated, and we shall approach them by considering the kinematic nature of the prior wave. A kinematic wave does not depend upon the propagation of a stimulus. Each cell, pre-determined to do something at a particular time, acts autonomously. The smooth passage of a kinematic wavefront implies that the cells are laid out along the axis in the order in which they are set to change; if the order were changed, the passage of the wavefront would be correspondingly erratic. A kinematic wave does not depend on cells knowing where they are along the axis and interpreting their positions, but on cells knowing when independently of their whereabouts. We would expect the wave of somite formation to be kinematic, as Deuchar & Burgess (1967) showed, if it were the consequence of the prior wave. The demonstration that the prior wave is likewise kinematic indicates that it too is the consequence of an earlier primary event. Our experiment tells us only that the paraxial mesoderm is temporally determined as early as the neural fold stage at which embryos were cut in two; in a following paper, however, we shall present additional evidence suggesting temporal determination as early as the beginning of gastrulation. Zeeman (1975) envisages a temporal determination established with the passage of a primary wave across the mid blastula co- incidental with primary mesodermal induction. On this view, the kinematic waves we demonstrate in connexion with somitogenesis are the reflexion of a larger field of temporal determination, established across the marginal zone of the young embryo, responsible for the intrinsic developmental dynamics of the primary organizer. Zeeman's theorem describes the shape of such a wave, arising out of the dynamical organization of the intercellular biochemistry, by a cusp catastrophe. The waves we describe do not fit a cusp catastrophe at their beginning or end (compare Fig. 8 with Fig. 3, Pearson & McLaren, 1977), but in any case the substrate deformations during invagination rule out the likeli- hood of correspondence between such secondary waves and a postulated primary wave. Our results provide no evidence to bear crucially on the theorem. The temporal determination of the prior wave in somitogenesis, already set by early gastrulation (Elsdale & Pearson, 1979), suggests how this wave may be involved in the specification and regulation of the segmental pattern. Cooke (1975) has demonstrated that embryos artificially reduced in size at the early blastula stage subsequently develop a normal body pattern including the normal number of somites appropriately reduced in size and containing fewer than normal cells. The interval between the formation of one somite and the next is not altered and somitogenesis is completed in the same time as in control embryos. This means that the wave of somite formation takes the same time to traverse the shorter axis in reduced embryos as it does to traverse the longer axis 4 EMB 51 46 M. PEARSON AND T. ELSDALE in normal embryos, implying that regulation in the former is achieved by a slower progress of the wave of somite formation. The rate of segmentation is the same in both cases, so that the necessary regulation must result from the timing of cellular activities which lead to their incorporation into the kinematic wave of somite formation. The timing of such cellular activities must be set by an earlier process. If the prior wave is itself determined by some earlier process, it is nonetheless the prior wave which appears immediately to regulate somite formation, since only after its passage is somite specification fixed. This is our second hypothesis: the timing of the cells' commitment to somite formation, which is manifest at a supracellular level as a prior wave along the mesodermal axis, is fundamentally a part of the patterning process. It is the timing of cellular mechanisms rather than an original positioning of cells which underlies segmental pattern. We could envisage two ways by which the segmental pattern might arise with the prior wave and be affected by heat shocks. The segmental period might be inherent in the timing of the cellular mechanisms which lead to the commitment of cells to somite formation, or the period might arise independently of the prior wave to be imposed upon it at the moment of this commitment. The crucial considerations are the following. Firstly, the asymmetry of the abnormal zone: behind the sharp anterior frontier there is a gradual return to normal posteriorly. Secondly, the length of the abnormal zone is proportional to the duration of the shock, so that a shock longer by minutes leads to a longer abnormal zone formed over several hours. We can view the whole somite file, including an abnormal zone, as a record of the developmental history of pre-somitic cells up to somite formation; that is, we can translate from the linear dimension of the file into a time course which has been stamped on the final somite pattern by heat shock. Posteriorly, normal segmentation reflects cells in an early phase of maturation at the time of the shock; further anteriorly the maturing cells register heat shock in a subsequently disturbed segmentation which increases up to the anterior frontier of the abnormal zone. Cells then switch from maximal sensitivity to insensitivity, and the segmental pattern is fixed. Before cells reach this point of commitment, therefore, they may be deranged by heat shock depending on (a) how close they are to the wavefront at the time of shock, and (b) on the duration of the shock. The closer cells are to commitment, the more readily they are induced to abnormal segmentation. There are two possible explanations for this behaviour recorded by abnormal files. (i) The first explanation assumes that the picture provided by the file with an abnormal zone directly reflects the responsiveness - or non-responsiveness - of the cells along the axis at the time of shock. Once disturbed, a cell remains so in the absence of any means of recovery. This explanation however offers no obvious reason for the asymmetry of the abnormal zone beyond ascribing it to Somitogenesis in amphibian embryos. I Al increasing sensitivity. Nor does it predict significantly longer abnormalities for slightly longer shocks unless we postulate a subliminal sensitivity posteriorly which can be brought to the point of disturbance by a longer shock. On this hypothesis we might expect the effects of two shocks to be additive, the subliminal effect of a first shock sensitizing the tissue for a greater response to the second, but this is the opposite of what we find. In fact there is a shorter and less severe abnormality, in comparison with controls, in response to the second shock. This result of the double heat shock experiment indicates a restorative mechanism, such that the effect of a first shock which is not seen in any visible segmental abnormality enables the tissue to recover from the effect of a second shock. (ii) The second alternative explanation depends upon the ability of cells to recover from the effect of heat shock. The pattern of visible abnormality may then be translated into an account of recovery by labile cells which have not yet been fixed in the segmental pattern by the passage of the wave of commitment. On this basis it is the state of the cells at the time of commitment which is alone crucial for orderly somitogenesis. Cells immediately behind the anterior frontier of the abnormality register the most severe disturbance, not because they are most sensitive, but because there is less time for recovery before the final determination of segmental pattern. Because the disturbance is transient, only those cells committed immediately following a shock will stabilize the full disturbance. Cells committed later will have more time to recover. A longer shock causes more severe disturbance; and so recovery takes longer. In this case the abnormal zone extends posteriorly into a region where recovery would have been complete following a shorter shock. When a second shock is delivered, cells which would have recovered from the first shock now recover more rapidly from the second disturbance. Recovery therefore is an active process. The evidence for independent recovery from heat shocks before a final determination of the somite pattern means that segmental specification does not emerge as an inherent pattern in the cellular maturation processes that lead up to the prior wave. Segmental specification is concomitant with the passage of the prior wave, but does not simply arise from it. There is no evidence that the prior wave itself is disturbed by heat shocks. The effect of heat shocks. The visible pattern abnormality is an excessive segmentation, reflecting a tendency of the cells to associate into smaller, more irregular groups than normal. Heat shock has fragmented some principle of intercellular co-ordination established before the passage of the wavefront. It is proposed that heat shock undoes this co-ordination over the whole length of the paraxial mesoderm not yet overtaken by the prior wave. The loss is transient, however, and a slow recovery toward re-establishment of co-ordination ensues. To account for the fixation of the disturbance, and the delayed appearance of abnormality after a shock, we propose that with the passage of the prior wave a segmental pre-pattern is set up on the basis of the degree of intercellular 4-2 48 M. PEARSON AND T. ELSDALE co-ordination pertaining at the time. This scheme (Fig. 15) accounts for all the characteristics of the abnormal zone. Turning to the nature of the second component in pattern specification we consider the basis for intercellular co-ordination. This could in principle be either a temporal or a spatial co-ordination. The latter suggestion however would amount to the conceptual indulgence of a second pre-pattern. Further- more, since the segmental rate of progress of the prior wave is constant throughout (Figs. 6 and 7), a segmental specification at the time of the prior wave must be a constantly periodic process. It is most reasonable therefore to assume a periodicity in the responsible mechanism and to implicate this periodic component in the partitioning of the longer time interval taken by the prior wave to traverse the axis. We propose therefore that each cell of the paraxial mesoderm carries two independent pieces of equipment: first, a pre-set specification of the timing of abrupt change, or a count-down apparatus, impervious to outside influence by heat shocks, and second, a component uniquely susceptible to heat shock and on the basis of which the intercellular co-ordination is established. The hypothesis that this component is a clock synchronizable with similar clocks in other cells could be experimentally tested. While there is evidence that the first component is present already by the beginning of gastrulation, co-ordination is not established until the late gastrula (Elsdale & Pearson, 1979). There is no evidence to suggest that cells are precipitated by heat shock into any abnormal state other than to disturb their temporal co-ordination. It is the synchroniza- tion of a population of cells which is affected, upon which normal segmental patterning depends. Similarly, in connexion with recovery from heat shock, it is a resynchronization. Coupling of the two components: the clock and wavefront model. Our results indicate that crucial for the specification of the somite pattern is a process involving the momentary coupling of both components in the setting up of a segmental pre-pattern. The pre-pattern is established as a wave of somite determination which is, in practice, inseparable from the prior wave. The components we describe are essentially the same as those proposed in a theoretical model published by Cooke & Zeeman (1976), who suggest how the coupling might work. Suppose a cyclic process in each cell having a period equal to the interval between the formation of one somite and the next. The cycles are phase-linked. The cells cycle in unison. Suppose further that there are two parts of the cycle, an ON part and an OFF part, and that only during the former can the cells undergo abrupt change. Cells that complete their 'count- down ' to the abrupt change which commits them to somite formation during the OFF part of the cycle can do nothing and must wait until they re-enter the ON part of the cycle. Hence, as the cells pass into the OFF part of their cycles, the wavefront is halted, and on re-entering the ON part, a batch of waiting cells undergo their abrupt changes synchronously, and are thus committed to define a somite. Somitogenesis in amphibian embryos. I 49

The formation of an abnormal zone following lieat shock

ABC D

Somites 12 3 4 5 6 7 8 9 10 1112 13 14 15 16 17 18 1920 2122 23 2425

T2 * * * *

Axis Anterior Posterior Fig. .15. This is a board on which the reader is invited to perform two mobile demonstrations, by placing a ruler along the base of the board and moving it slowly upwards to exit from the top of the figure. (1) The wavefronts of somite determination and somite formation. As the ruler is raised the points of intersection with the lines labelled somite determination and somite formation move along the ruler from left to right simulating the passage of the two wavefronts along the embryonic axis as development proceeds. (2) The creation of an abnormal zone following heat shock. The four superimposed horizontal lines T1-T4 represent four sequential stages in the development of an embryo following heat shock. At Tl six somites have been determined (A-B) and the first somite is about to form. A heat shock at this time disturbs (asterisks) all of the undetermined pre-somitic mesoderm, but determined tissue anterior to B is immune. The first six somites form sequentially as the ruler is moved towards T2. Concomitantly the wavefront of determination advances from B to C across tissue disturbed by heat shock. Immediately behind B the tissue has no time to recover before determination and the maximal disturbance is fixed; by the time the wavefront has reached C however, recovery is pictured as complete. Moving the ruler beyond T2 towards T3 somites 7-12 form; these somites are to a greater (anterior) or lesser (posterior) extent abnormal reflecting the degree of disturbance fixed at the time of determination. Concomitantly the wavefront of determination continues to advance across tissue now fully recovered from the heat shock. Moving upwards from T3 to T4, somite formation returns to normal posterior to the abnormal zone. 50 M. PEARSON AND T. ELSDALE The overall result is that a somite's worth of cells undergo abrupt change together, followed by an interval of time before the next batch changes and so on in discontinuous steps along the body axis. As Cooke & Zeeman see it the cells become ready to change in smooth sequence down the axis and it is due to coupling with the cyclic component that the observed wavefront of somite determination is presumed to move in spurts and pauses, thus setting up the segmental pre-pattern. The coupling is formally analogous to an escapement mechanism in a clock by which a smooth energetic input from the mainspring is transformed into ticks. This model eschews reliance on any form of qualitative differentiation, including positional interpretations, and thereby assumes that somitogenesis is essentially the sequential division of a homogeneous loaf. In the absence of evidence for any qualitative difference among pre-somitic cells, this seems a realistic foundation for any account of somitogenesis. Any explanation of somitogenesis does not principally concern the particular changes of cell surface property nor the mode of cell behaviour, but rather the spatio-temporal pattern of such changes within the morphogenetic cell population. Thus in Xenopus laevis and Rana temporaria the cellular mechanics of somite formation are very different, yet the pattern of segmentation, and the effects of heat shocks, are very similar. The same account suffices for both.

We thank Jonathan Bard for scanning electron micrographs and for criticising the manu- script; also Alison Abbott and Sandy Bruce for assistance. MJP thanks Professor Evans for hospitality at the Western General, MRC for paying train fares and other expenses, Brian Goodwin and Christ Ford for providing facilities at the University of Sussex.

REFERENCES BARD, J. B. L., HAY, E. D. & MELLOR, S. M. (1975). Formation of the endothelium of the avian cornea: a study of cell movement in vivo. Devi Biol. 42, 334-361. BELLAIRS, R., CURTIS, A. S. G. & SANDERS, E. J. (1978). Cell adhesiveness and embryonic differentiation. /. Embryol. exp. Morph. 46, 207-213. COOKE, J. (1975). Control of somite number during morphogenesis of a vertebrate, Xenopus laevis. Nature, Loud. 254, 196-199. COOKE, J. & ZEEMAN, E. C. (1976). A clock and wavefront model for control of the number of repeated structures during animal morphogenesis. /. theor. Biol. 58, 455-476. DEUCHAR, E. M. & BURGESS, A. M. C. (1967). Somite segmentation in amphibian embryos: is there a transmitted control mechanism?/. Embryol. exp. Morph. 17, 349-358. ELSDALE, T. R. & PEARSON, M. J. (1979). /. Embryol. exp. Morph. (in the press). ELSDALE, T. R., PEARSON, M. J. & WHITEHEAD, M. (1976). Abnormalities in somite segmentation following heat shock to Xenopus embryos. /. Embryol. exp. Morph. 35, 625-635. HAMILTON, L. (1969). The formation of somites in Xenopus. J. Embryol. exp. Morph. 22, 253-264. NIEUWKOOP, P. D. & FABER, J. (1956). Normal Tables o/Xenopus laevis (Daudin). Amsterdam: North Holland. PEARSON, M. J. & MCLAREN, D. I. (1977). A criticism of catastrophe modelling of the differentiative process in amphibian development. /. theor. Biol. 69, 721-734. ZEEMAN, E. C. (1975). In Lectures on Mathematics in the Life Sciences, vol. 7. Rhode Island: Am. Math. Soc. (Received 19 June 1978, revised 9 January 1979)