Somite pattern regulation in the avian segmental plate mesoderm

During avian gastrulation, strips of apparently unsegmented paraxial mesoderm, called the segmental plates, are formed on either side of the notochord. The cells that constitute these segmental plates are derived from the presumptive somite region of the epiblast (Rosenquist, 1966; Vakaet, 1984). They join the caudal ends of the segmental plates after involuting through the cranial end of the primitive streak (Rosenquist, 1966; Vakaet, 1984; Nicolet, 1970, 1971; Packard, 1986a). While somites segment in succes-sive pairs from the cranial ends of the segmental plates, the morphogenetic movements associated with gastrulation continue to add new cells to the caudal ends of the seg-mental plates at least until the 8-somite stage of develop-ment (Packard, 1986a). These morphogenetic movements and some cell divisions within the segmental plate approx-imately compensate for the cells lost in segmentation, although the segmental plates do vary somewhat in length during much of development (Packard and Jacobson, 1976; Packard, 1980a).

Although the segmental plates of living embryos appear to be unsegmented, a series of experimental studies (Menkes and Sandor, 1969; Christ et al., 1974; Packard and Jacobson, 1976; Packard, 1978; Sandor and Fazakas-Todea, 1980) indicate that the spatiotemporal pattern of somite for-mation reflects a stable covert organization (i.e., a prepat-tern) inherent to the segmental plates and extending along their entire length. Furthermore, morphological studies (Meier, 1979, 1982a,b; Triplett and Meier, 1982; Packard and Meier, 1983) have shown that this prepattern can be distinguished as a series of paired circular domains, termed somitomeres, within the segmental plates. The somitomere pattern represents the segmental pattern that the segmental plate cells are specified to form (Jacobson, 1992). The results of these aforementioned studies clearly demonstrate that, for most of the period of somitogenesis, the segmen-tal plates contain at least some cells that together are devel-opmentally specified to form a defined pattern of about 10 somites in a cranial-to-caudal sequence. These studies also strongly suggest that the commitment of cells to form ele-ments of the pattern is acquired prior to the cells entering the caudal end of the segmental plate. Indeed, reversal of the cranial-caudal orientation of prospective paraxial meso-derm lying caudal to the node, that is, before the cells have entered the segmental plate, leads, hours later, to a caudal-to-cranial sequence of segmentation (Christ et al., 1974; also see review by Davidson, 1988). Furthermore, somito-meres have been observed caudal to the node in the newly formed mesoderm that has not yet entered the segmental plates (Meier and Jacobson, 1982; Packard and Meier, 1983; Jacobson and Meier, 1986).

The experiments reported in this paper were inspired in part by findings that, in certain cases, suggest that the somitic prepattern within the segmental plate is susceptible to permanent modification yet, in other cases, suggest that the original specified pattern may be restored after thorough disruption. Heat-shock treatment of chick embryos on the second day of incubation causes reproducible somite pat-tern anomalies (Primmett et al., 1988; 1989; Veini and Bel-lairs, 1986). Many of these anomalies occur 6 to 7 somites caudal to the last somite pair to form prior to the experi-ment. Since the chick segmental plate contains 10 to 12 prospective somites (Packard and Jacobson, 1976; Packard, 1978), the segmentation anomalies must have arisen by modification of the specified somite pattern under circum-stances that did not allow restoration of the proper pattern. However, Menkes and coworkers (Menkes and Miclea, 1962; Menkes and Sandor, 1969) have shown most impres-sively that somite formation continues normally (after a brief delay) even when the organization of the segmental plate has been thoroughly disrupted by cutting the tissue, along with the adherent ectoderm and endoderm, into many pieces and mixing the fragments. Also, Sandor (1972) excised areas of future intersomitic furrows from the seg-mental plates of explanted chicken embryos and found that the somite pattern still formed normally. In these instances, unlike the heat-shock experiments, a mechanism for rapidly restoring the pattern must have functioned. Our experiments were designed to explore further the capacities of the embryo for modification and reconstitution of the segmen-tal pattern. Some of these results have been published in an abbreviated form (Packard, 1986b).

Fertile White Leghorn chicken eggs and fertile Japanese quail eggs were obtained from the Poultry Science Division of Cornell Uni-versity, Ithaca, New York. The eggs were stored in a humidified atmosphere at 9°C until needed. After incubation at 38°C in an egg incubator for about 48 hours, each egg was opened into a bowl of Howard’s (1953) saline and the embryos were removed from their yolks, washed briefly in calcium- and magnesium-free Tyrode’s (1910) solution and then placed at room temperature in about 5 ml of calcium- and magnesium-free Tyrode’s solution containing 1% trypsin (hog pancreas, ICN Pharmaceuticals, Inc.) and 0.1% ethylenediaminetetraacetic acid (EDTA). The embryos remained in this solution for 5 to 15 minutes; the time varied with the embryo’s age and the amount of yolk present. When the embryo’s neural tube appeared wrinkled or open, the embryo was transferred to a 35 mm plastic culture plate containing a nutritive agar substratum and about 1 ml of fresh, chicken, whole-egg supernatant (Packard and Jacobson, 1976). The excess supernatant was drawn off and microsurgery was performed with electrolyti-cally sharpened tungsten wire needles. The ectoderm overlying the segmental plate was cut into a flap that was folded laterally to expose the segmental plate. The exposed segmental plate was carefully separated from the neural tube and surrounding meso-derm and then it was gently peeled from the underlying endo-derm. The node region of the embryo, including all three germ layers, was completely removed to prevent continued formation of segmental plate mesoderm (Packard and Jacobson, 1976; Packard, 1980a,b). Grafting of segmental plates was accomplished by transferring the segmental plate to the host embryo’s dish with fine watchmaker’s forceps. The graft was then inserted into the host’s removal site in such a way as to maintain the graft’s orig-inal dorsal-ventral and cranial-caudal orientation and the ectoder-mal flap was then closed over the graft.

The embryos were cultured in vitro at 38°C in a humidified atmosphere of 95% O₂ and 5% CO₂ for 16 to 20 hours (Packard and Jacobson, 1976); the duration of the culture period was chosen on the basis of previous work showing that explants containing segmental plates complete somite formation within 15 hours and begin to lose somites after 20 hours (Packard and Jacobson, 1976). The embryos were then fixed in ethanol:acetic acid (4:1), dehy-drated, embedded in paraffin, cut coronally into 10 μm serial sec-tions, stained according to the Feulgen-Rossenbeck technique (1924), counterstained with fast green, and camera lucida tracings of the neural tube and somites were made from appropriate sec-tions of each embryo. The tracings were digitized with an acoustic tablet and the orthostereoscopic reconstructions were created using custom software (Falen and Packard, 1982).

In some experiments, an impermeable barrier was placed either medial or lateral to the segmental plate. In these experiments, chicken embryos were isolated, treated with enzyme and placed in culture plates as described above. The right segmental plate of each embryo was removed, cut into 15 to 20 pieces and the pieces were randomly reinserted into the removal site. The ectodermal flap was then replaced. A piece of tantalum foil was then cut so that its length matched that of the segmental plate. The foil was held in the desired position (either medial or lateral to the seg-mental plate pieces) and its edge was forced down through all three germ layers of the embryo and into the agar substratum. Finally, the node region was removed and the embryo was cul-tured and fixed as described above. After fixation, the foil was gently removed so as to avoid damage to the embryo. The embryo was then processed for histological examination as described above, except that the sections, which contained no quail cells, were stained with hematoxylin and eosin. Control experiments to test for any toxic effects of the foil were performed by inserting a tantalum foil barrier between the neural tube and the undisturbed right segmental plate of embryos prepared for culture but not treated with enzyme.

A series of control experiments was performed to estimate the effect of the experimental manipulations on the number of somites formed by grafted segmental plates. Chicken or quail embryos with 8 to 21 pairs of somites were treated with trypsin, and the right segmental plate of each embryo was removed and then replaced immediately as indicated in Fig. 1. The embryos were then cultured, fixed and ana-lyzed histologically. A typical result from one of these experiments is shown in Fig. 2. A total of 73 embryos oper-ated on in this way survived the culture period. Of these embryos, 42 were damaged during processing or showed obvious evidence that all or part of the manipulated seg-mental plate had been lost. These embryos were not included in the following analysis. As expected, when the numbers of somites formed by the right and left segmental plates of each embryo were counted from the stereoscopic reconstructions, a high correlation between the figures was revealed (r=0.904, n=31, P<0.001, Fig. 3); no species-or age-related differences were noted. To facilitate compari-son of results in this and other experiments, an ‘index of inequality’ was calculated by averaging for all experiments the absolute difference between the two sides in each embryo; the higher the value of the index, the greater was the average difference between the sides. In the case of the control experiments, the index (X ± s.d.) was 0.68 ± 0.91 somites. Since previous studies had shown that segmental plates removed simultaneously from the same embryo tended to form the same number of somites (Packard, 1980a, Packard and Meier, 1983) and since a similar result was found in these control experiments, it was concluded that the manipulation of the segmental plates resulted in at most only a small change in the number of somites formed by them.

The first experimental question asked was: can a host embryo alter the number of somites that a grafted segmen-tal plate will form? Right segmental plates were removed from trypsinized quail embryos that possessed 5 to 23 pairs of somites. The segmental plates were then grafted in place of the host’s right segmental plate into trypsinized host chicken embryos containing 9 to 24 pairs of somites (Figs 4, 5). Both donor and host embryos were cultured for about 18 hours and then fixed and sectioned; reconstructions were made as described above. Since previous studies showed that segmental plates removed simultaneously from the same embryo tend to form the same number of somites (Packard, 1980a, 1986b), the number of somites that the grafted segmental plate was originally specified to form could be determined by counting the number of somites ultimately formed by the donor embryo’s left, unoperated, segmental plate. The number of somites actually formed by the graft was determined by counting the number of quail somites in the chicken host. A total of 174 experiments were performed. Thirty embryos survived the culture period and processing for histological study and were judged suit-able for analysis. Several cases were eliminated from the study when the host embryos were found to have no somite mesoderm or quail cells on their operated sides. In these cases, the grafts had apparently been lost from the embryos. Reconstructions from a typical experiment are shown in Fig. 6. This experiment differed from the previous control experiment only in that the excised segmental plate was grafted into another embryo rather than being returned to the donor. Therefore, the number of somites formed by the donor’s left segmental plate was compared first with the number of somites formed by the graft (Fig. 7).

Comparison of the data in Figs 3 and 7 shows that, while removing a segmental plate from an embryo and returning it to its original location resulted in little change in the number of somites that it subsequently formed, removal of a segmental plate from an embryo and placing it into another embryo significantly altered the number of somites that it would form. This point was confirmed by both the much lower correlation between the number of somites formed by the two donor segmental plates (r=0.305, n=30, P< 0.10) and the significantly higher index of inequality for this comparison (2.00 ± 2.24 somites, t=2.43, df=59, P<0.02) than in the control experiments.

In order to see if the host embryos had influenced the grafts to form numbers of somites more compatible with the host’s somite pattern, the number of somites formed by the host’s unoperated left segmental plate was compared with the number of somites formed by the graft. As shown in Fig. 8, the correlation between these numbers was sig-nificant (r=0.666, n=30, P<0.001). It was concluded that the host embryos demonstrated a significant ability to alter the number of somites formed by the grafts.

This change of somite number was even more apparent when the data were viewed in such a way that the differ-ences in the number of somites that the donor and host embryos were originally specified to make could be taken into account. The algebraic difference between the number of somites required to maintain the symmetry of the host’s somite pattern (indicated by the number of somites formed on the host’s unoperated side) and the number of somites that the graft was originally specified to make (prospective fate; indicated by the number of somites formed by the donor’s unoperated side) was defined as the ‘disharmony’ between host and graft under those experimental conditions. This figure for disharmony was then compared with the ‘response’ of the graft to the experiment. Response was determined by subtracting the number of somites that the graft was originally specified to make (again indicated by the number of somites formed by the donor’s unoperated side) from the number of somites that the graft actually formed. For example, if the host’s unoperated segmental plate formed 12 somites and the donor’s unoperated seg-mental plate formed 8 somites, the disharmony between host and graft would be +4; that is, the graft would have to form 4 somites, in addition to those it was originally specified to form, in order to maintain the host’s somite pattern. If the graft actually formed 10 somites, the response would be +2; that is, it formed two additional somites. Fig. 9 illustrates the relationship between disharmony and the response of the graft. The correlation between these two measures was statistically significant (r=0.755, n=30, P<0.001), and although nine grafts failed to respond, only 4 of the 30 grafts responded in an unexpected direction; that is, when the disharmony in a given experiment was positive or negative, the response of the graft tended to be, respectively, positive or negative. While most of the grafts did respond, their response was imperfect; the number of somites that they made was often more than or less than the number one might have predicted from the disharmony.

The reason for this incomplete or imperfect response of the grafts to the host embryos is not clear. We wondered whether this phenomenon might be due to the fact that the cells in the cranial portion of the grafted segmental plate must have formed somites within a few hours after the surgery. It seemed possible that there was insufficient time for the host embryo to alter the number of somites formed by the cranial portion of the graft or that the pattern near the cranial end of the segmental plate was more stably spec-ified than that in the more caudal parts of the segmental plate and hence not susceptible to host influence. To test this possibility, we performed experiments in which the caudal half of the right segmental plate was removed from two quail embryos. These two caudal segmental plate halves were then grafted together in place of the right seg-mental plate of a chicken embryo (Fig. 10). Thus, an amount of segmental plate mesoderm equivalent to one seg-mental plate was grafted, as in the previous experiments, but in this case all of the grafted mesoderm originated from the caudal halves of segmental plates. Would the number of somites formed by caudal segmental plate grafts more closely match the number of somites formed by the unop-erated segmental plate of the host chicken embryo? Eight of the twenty experiments thus performed resulted in sec-tions suitable for reconstructive analysis. At the time of surgery, the chicken embryo hosts possessed from 12 to 22 somite pairs, while the quail donor embryos possessed 12 to 23 somite pairs. The result of the experiment is seen most clearly when the data are expressed as disharmony versus response, as in the previous experiment (Fig. 11). In this case, the value of the disharmony for each experiment was determined by subtracting the number of somites formed by the unoperated side of the host embryo from the total number of somites missing from the operated sides of the two donor embryos (the number of missing somites was judged by subtracting the number of somites formed on the right [operated] side from the number of somites formed on the left [unoperated] side). A response pattern very sim-ilar to that resulting from the grafting of intact segmental plates was observed. The correlation coefficient was 0.814, n=8, P<0.01. Thus, it is clear from these experiments that caudal segmental plate mesoderm is not more responsive to the influences of the host embryo than the cranial seg-mental plate mesoderm.

It also seemed possible that the variable response of the grafts to the host environment might, in part, be due to the interaction between putative pattern-specifying signals emanating from the host and the specified somite (somito-mere) pattern of the graft. Therefore, a fourth series of experiments was carried out in which the excised quail seg-mental plate was bisected lengthwise, cut into about 20 pieces (range=14 to 25 pieces), and the pieces randomly inserted into the host embryo (Fig. 12). These experiments were identical to the preceding intact segmental plate graft-ing experiments except that the specified somite pattern of each graft had been disrupted by cutting the grafts into pieces and scrambling the positions of the pieces. Seg-mental plates treated in this way will be referred to as having been ‘disrupted.’ It was asked whether the change in somite number would still occur and, if so, whether dis-ruption of the graft’s specified somite pattern would permit the host embryo to regulate the grafted mesoderm to form a number of somites more similar to that formed by the host. A total of 39 such experiments were performed. The embryos in 17 of these experiments survived the culture period and processing for histological study and were judged suitable for analysis. A few experiments were excluded because there was either no somite mesoderm or obvious gaps in the somite mesoderm on the host embryo’s operated side. In these cases, some or all of the grafted seg-mental plate mesoderm had apparently been lost. When the numbers of somites formed by the donor embryos were compared with the numbers of somites formed by the dis-rupted and grafted quail segmental plates (Fig. 13), it was evident that there was no significant correlation between these figures (r=0.182, n=17, P<0.72). In order to see whether the grafts had formed numbers of somites more similar to the numbers formed by the hosts, the number of somites formed by the disrupted quail segmental plate grafts was compared with the number of somites formed by the host chick embryos (Fig. 14). There was a weak correla-tion between these numbers (r=0.400, n=17, P<0.03). However, once again, when these same data were viewed with respect to the previously described values of dishar-mony and response for each experiment (Fig. 15), it was clearly evident that a significant relationship between these variables was present (r=0.850, n=17, P<0.001). Thus, the disruption of the graft’s segmental pattern did not appear to affect significantly the graft’s responsiveness to the host’s influences.

The orthostereoscopic reconstructions used in this study not only aided in the accurate counting of somites, they also made it possible to compare the somite patterns observed in the experiments. Analysis of the numbers of somites formed in the above experiments suggested that the grafted segmental plates responded to the host embryo’s environ-ment by tending to form a number of somites different from the number of somites that they were originally specified to form. Study of the tissue reconstructions also suggested that there was a tendency for the positions of the inter-somitic interfaces between the graft-derived somites to occur at nearly the same axial level in the host embryo as the intersomitic interfaces of the host’s segmental pattern. For example, when the cranial-to-caudal lengths of the somites formed by the donor embryo were much longer or shorter than those formed by the host embryo, the somites formed by the graft tended to be of a size that was appro-priate for the host’s segmental pattern. An example of this phenomenon is shown in Fig. 16. Fig. 17 shows another interesting example of this phenomenon. This figure is a reconstruction of a chick host embryo fixed 18 hours after the transfer of a disrupted quail segmental plate graft. The host’s unoperated left segmental plate formed 4 somites. The donor embryo (not shown) formed 7 somites on its unoperated side and the graft also formed 7 somites (4 pairs of host somites are shown cranial to the cranial extent of the graft). In terms of somite number, therefore, it would appear that the graft did not respond to the host’s environ-ment. However, when the reconstruction is carefully stud-ied, it can be seen that opposite the first somite formed by the host following surgery, the graft formed 2 somites that occupied almost the same position in the pattern as the missing host somite would have occupied. The graft formed 1 somite opposite the host’s second somite and 3 small somites opposite the host’s third somite. Thus, while there was no change in the number of graft somites, there was an apparent adjustment of the length and location of the graft somites in such a way as to preserve better the sym-metry of the host’s somite pattern. Although such adjust-ment was not always as clear as in this example, we judged it to be demonstrated at least to some extent in 21 of the 30 intact graft experiments. Therefore, the host embryos may have a more extensive ability to alter the graft somite pattern than was suggested by the study of somite numbers alone.

The demonstration that host embryos can alter the somite pattern of grafted segmental plates raises the question of what regions of the host embryos are active in this alter-ation. We decided to test whether tissues either medial or lateral to the segmental plate are important to the reestab-lishment of the somite pattern in disrupted segmental plates. Impermeable tantalum foil barriers were inserted either medial or lateral to the pieces of a disrupted segmental plate and, after the same culture period used in the previous experiments, the resulting somite pattern was observed.

First, a control experiment was performed to test for any possible toxic effects of the tantalum foil barrier. Four chicken embryos having 11 to 19 pairs of somites were placed on the agar culture substratum and a piece of foil, equivalent in length to the segmental plate, was placed between the neural tube and the right segmental plate. The foil was pushed entirely through the embryo and into the agar (Fig. 18). After 16 to 20 hours in culture, each of the embryos had formed symmetrical somite pairs through the level of the foil (Fig. 19). The foil did not appear to inter-fere with somite formation.

In the second series of experiments, chicken embryos possessing approximately 10 to 20 somite pairs were treated with trypsin, placed on the agar culture medium, and their right segmental plates were removed. The segmental plates were then cut into 15 to 20 pieces and the pieces were ran-domly reinserted into the removal site. The tantalum foil barrier was then placed immediately lateral to the pieces (Fig. 20). Following the culture period, the embryos were fixed, the foil was removed and the embryos were processed for histological study. Seven embryos were judged suitable for study. In two of the experiments, there was no somite mesoderm on the embryo’s right side (the side with the foil barrier). Apparently, the pieces of segmental plate had been lost during the culture period. Somites that appeared to be normal were observed medial to the barrier in the remain-ing five experiments (Fig. 21). However, in each of the experiments, there were fewer somites on the operated side (Table 1). The smaller number of somites on the operated side was always associated with one or more obvious gaps in the row of somites. This observation suggested that one or more of the segmental plate pieces had been lost.

In the final series of barrier experiments, chicken embryos were treated exactly as for the previous experi-ments, except that the foil barrier was placed between the neural tube and the disrupted segmental plate (Fig. 22). Six-teen embryos survived the culture period and histological processing. The mean number of somites formed on the unoperated side of these embryos was 8.5, while the mean number formed on the operated side was 1.1 (Table 2). Nine of the sixteen embryos formed no somites on the operated side. When paraffin sections of these embryos were exam-ined in the light microscope, a dense mass of mesoderm was seen lateral to the barrier, but medial to the kidney and lateral plate mesoderm (Fig. 23). This dense mesoderm was not continuous along the length of the barrier in any of these embryos. Rather, there were gaps in the mesoderm, much like the gaps seen in the row of somites that formed medial to the barrier in the previous experimental series. Again, these gaps probably represented the loss of some of the segmental plate pieces; the tantulum foil likely inter-feres with the healing of the grafted pieces into the host site. Of the remaining seven embryos, each exhibited 1 to 4 somites in the dense mesoderm lateral to the barrier. In all but one of these embryos, the somites that formed on the operated side were either at the cranial end or the caudal end of the barrier. The embryo that was the exception formed 11 somites on the unoperated side and 4 small somites on the operated side. These 4 somites were found at the level of the 6th and 7th somites on the unoperated side. All but 2 of the somites observed lateral to the bar-rier were much smaller than the somites on the embryo’s contralateral side. Some of these small somites had enlarged myocoels with a very short somite epithelium so that the somites took on the appearance of hollow vesicles. It was clear that, once the prepattern of segmentation in the seg-mental plate had been destroyed by ‘disrupting’ the seg-mental plate, the presence of a tantalum foil barrier between the pieces of the segmental plate and the midline tissues of the host embryo largely inhibited the formation of somites.

The work presented here was designed to investigate the related but distinct capacities of the segmental plate to undergo modification or restoration of the segmental pat-tern. As described in the Introduction, we were intrigued by the apparent contradiction between evidence indicating that the segmental pattern is already firmly established within the segmental plate and evidence indicating that the pattern can either be modified by heat shock without restoration of the original pattern or reestablished follow-ing surgical disruption. By taking advantage of the known bilateral symmetry of the segmental pattern in the two seg-mental plates of any given embryo, we were able to devise experiments that revealed the ability of the avian embryo to modify the segmental pattern.

The control experiments showed that the surgical manip-ulations required for segmental plate grafting did not sig-nificantly alter the number of somites formed by a seg-mental plate. Because the correlation between the numbers of somites formed on the unoperated and operated sides of these embryos was so high, we expected to be able to detect a change of two or more in the number of somites formed by segmental plates grafted homotopically. It is interesting to compare these control data with those published in figure 8 of Packard (1980a). At that time, both segmental plates were removed from Japanese quail embryos as separate tissue explants that included the lateral plate mesoderm, ectoderm and endoderm, along with various combinations of neural tube and notochord. The explants were cultured for 14 to 20 hours and the number of somites formed in the left explant from each embryo was compared with the number formed in the right explant. The correlation between these numbers was significant (r=0.657; P<0.001; n=90), and was taken to indicate that segmental plates removed at the same time from the same embryo form nearly the same number of somites. The major difference seen between the present data and those published previ-ously is the small number of somites in some of the recent graft experiments. The only explanation we have for these low somite counts is that the cuts made at the caudal end of the segmental plate must have tended to be more cranial than similar cuts made about ten years earlier. This would have led to smaller numbers of somites while maintaining the bilateral symmetry. It is important to note that while the mean somite number was lower in the recent series, the slope of the linear regression is remarkably similar in both sets of data, which suggests that the left-right symmetry in the two groups of experiments was the same. Therefore, the enzyme treatments and surgical manipulations used in the present experiments did not introduce dissimilarity between the operated and unoperated segmental plates.

There is compelling evidence that at least some of the cells along the entire length of the avian segmental plate are committed to forming a particular somite pattern. (1) Segmental plates from avian embryos are able to form somites in the absence of further contact with the tissues that normally surround them (Sandor and Amels, 1970; 1971; Packard and Jacobson, 1976; Bellairs and Veini, 1980; Sandor and Fazakas-Todea, 1980). In addition, when avian segmental plates are cultured, each forms somites along its entire length. These somites are usually 10 to 12 in number (Packard and Jacobson, 1976; Packard, 1978; 1980a). (2) Although there is some variation in the number of somites formed by segmental plates removed from dif-ferent embryos, when both the right and left segmental plates are removed from a given embryo, they are very likely to form the same or nearly the same number of somites after a suitable time in culture (Packard, 1980a; 1986b). (3) If, during early somitogenesis, the node, prim-itive streak and the immediately adjacent region of the embryo are removed, the embryo will produce about 10 additional pairs of somites and then somite formation stops (Packard and Jacobson, 1976). Removal of the node and streak apparently halts the formation of new segmental plate, so that somite formation continues only until the somite prepattern present in the segmental plates at the time of the operation is expressed. (4) When segmental plates are transected, each resulting piece is able to form the por-tion of the somite pattern that it would have formed if the cuts had not been made, since the total number of somites made by the pieces is still 10 to 12 (Packard, 1978). Sim-ilarly, when segmental plates are bisected lengthwise, they form double files of small somites, with the reduplicated somites usually being of equal number (Menkes and Miclea, 1962). (5) When the segmental plate is excised and reim-planted so that the cranial-caudal axis is reversed, it forms somites in a caudal-to-cranial sequence, that is, in the sequence consistent with its original orientation (Menkes et al., 1968; Menkes and Sandor, 1969; Christ et al., 1974). (6) When a segmental plate is observed with stereo scan-ning electron microscopy, about 10 ‘somitomeres’ can be seen arranged in tandem along its entire length (Meier, 1979, 1982a,b; Triplett and Meier, 1982). These regions of cells, with their processes arranged in a circular pattern, have been shown to be present in equal numbers in the two segmental plates of a given embryo (Packard and Meier, 1983). Furthermore, in explanted segmental plates, somite formation proceeds at the expense of preexisting somito-meres; that is, somitomeres normally become somites (Packard and Meier, 1983).

Nevertheless, when a segmental plate was placed in another embryo, rather than returned to its host embryo as in the control experiments, the number of somites formed by the graft was significantly different from the number formed by the unoperated segmental plate of the donor. Does such evidence that a segmental plate can be influ-enced to make more or fewer somites than it was originally specified to make conflict with the evidence that the avian segmental plate contains a prepattern of somite segmenta-tion? We would expect that altering the somite pattern would require that the specified segmental pattern and, therefore, the somitomeric prepattern be changed. Whether such a respecification of the prepattern seems plausible depends on how one conceives of somitomeres. One view is that somitomeres are stable, rather rigid structures, as exemplified by Keynes and Stern (1988): “in order to main-tain a fixed arrangement of somitomeres, it is essential that there be little or no cell movement within the segmental plate, or at least that any movement be restricted to a single somitomere.” For these authors, somitomeres seem to be the formal equivalent of somites. If one insists on this view, it is indeed difficult to imagine how somitomeres might be created or destroyed in adjusting the prepattern.

The alternative view, which we favor, is to regard somit-omeres as dynamic structures, recognizable by patterns of cell orientation within a layer of mesodermal cells that is constantly changing because of cell division and cell move-ments (Jacobson and Meier, 1986; Jacobson, 1988; 1992). Anderson and Meier (1981) have shown that, as the chick cranial neural crest migrates over the dorsal surface of the head paraxial mesoderm, the cells of the neural crest take on the characteristic swirling pattern of the underlying mesodermal somitomeres. So, even a sheet of migrating cells can exhibit the somitomere pattern. If these patterns of cell orientation are so readily influenced as to permit a moving sheet of neural crest cells to take on the underly-ing somitomere pattern, one can imagine that the cells of a grafted segmental plate could adjust their cellular orienta-tions in response to host influences to reflect more closely the original segmental pattern of a host embryo. Recalling that the somitomere pattern ultimately becomes transformed into the somite pattern (Packard and Meier, 1983), we suggest that the putative adjustment of the somitomere pat-tern would reflect a respecification of the segmental pattern which, in turn, would lead to the altered numbers of somites in our grafted segmental plates. In any case, it is clear that, to use the terminology of Slack (1991, pp. 18-33), the spec-ified segmental pattern in the avian segmental plate is not determined, but is capable of regulation when the segmen-tal plate is grafted homotopically into another embryo.

The numerical evidence for the response of grafts to the host embryo environment was confirmed and amplified by analysis of the orthostereoscopic reconstructions. It was apparent from comparing the somites formed in the grafts with those in the donor and host embryos that the location and size of somites could be adjusted by the host embryo. The size of the graft somites tended to be similar to that of the host somites. When the graft made more somites than the host, two to three small somites were often seen in the space normally occupied by one somite. The impression that we gained from studying the reconstructions was that an intersomitic interface tended to form on the operated side opposite each intersomitic interface on the unoperated side, while additional intersomitic interfaces on the operated side could occur at any level. Even the occasional graft that showed no response at all in terms of the number of somites it formed could show clear evidence of a response in the reconstructions. Thus, the ability of the grafts to respond to the influences of the host was even more pronounced than was suggested by the numerical evidence alone.

What components of the host embryos bring about or influence the respecification of the graft segmental pattern? We attempted to address this question by placing an imper-meable barrier either medial or lateral to disrupted seg-mental plates. While somite formation was unimpeded when the barrier was lateral to the pieces of segmental plate, at most only a few somites formed when the barrier was placed medial to the grafted fragments. We consid-ered the possibility that when barriers were placed medial to the disrupted segmental plates, blood from the dorsal aorta was unable to reach the segmental plate cells. Impaired blood supply might account for the failure of somite formation. However, when barriers were placed medial to the segmental plates in unoperated chick embryos, somites formed in a normal fashion. These results, therefore, suggest that contact with tissues near the midline of the embryo is required for disrupted segmental plates to form somites.

In view of the results of the barrier experiments, could it be that the neural tube and/or the notochord are seg-mented at the level of the segmental plate? If so, these struc-tures might guide the reconstitution of the segmental pat-tern in the adjacent segmental plate mesoderm. Stern (1990) has suggested that the avian notochord is the archetypal segmented structure in the trunk of vertebrates and that it may imprint segmental information onto the adjacent parax-ial mesoderm. Kimmel et al. (1991) argue strongly against this notion, noting that zebrafish lacking a notochord are segmented. However, even if the paraxial mesoderm can segment in the absence of notochord, it remains possible that axial position is similarly specified in notochord and segmental plate and that this positional information is used to maintain the axial and paraxial tissues in register. With regard to the neural tube, the evidence available at this time suggests that it does not acquire segmental properties until after somite formation in the adjacent mesoderm (Kimmel et al., 1991; Stern et al., 1988, 1991). Indeed, Stern et al. (1991) have shown that segmental lineage restrictions in the chick neural tube are dependent on the presence of somites. Despite the lack of evidence for segmentation of the neural tube and notochord at the level of the segmen-tal plates, it is possible that the presence of the somitomere pattern in the segmental plate could lead to a segmental variation in some molecular entity (e.g., a component of the extracellular matrix) distributed along the sides of the axial structures. Following experimental disruption of the somitomere pattern, these segmental variations might guide the restoration of the segmental pattern.

Our results, then, suggest a possible interaction between axial structures and the presomitic mesoderm in maintain-ing the symmetry of the segmental pattern. It is important to distinguish this proposed interaction from two other types of postulated interaction between axial tissues and paraxial mesoderm. The first of these interactions, whereby axial level is specified within the paraxial mesoderm, is thought to occur earlier during gastrulation. Kessel and Gruss (1991), who analyzed homeotic transformations of verte-brae in mouse embryos treated with retinoic acid during gastrulation, propose the following mechanism for specifi-cation of the embryonic axis: “A retinoic acid signal gen-erated from midline embryonic structures during gastrula-tion is received by ingressing cells, which respond with the sequential activation of more and more Hox genes,” thus defining more and more caudal levels along the embryonic axis. While it is conceivable that retinoic acid from Hensen’s node participates somehow in establishing the somitic prepattern at the same time that it activates Hox genes, the evidence summarized above suggests instead that elements of the segmental prepattern form caudal to Hensen’s node by an independent mechanism and that it is these elements (recognizable as somitomeres) that are receiving and responding to the putative retinoic acid signal. It would be of great interest to examine vertebral structures and patterns of Hox gene expression after disruption of seg-mental plates as described in this paper; it may be that, even though the presomitic pattern is seemingly restored, proper axial specification is not. The second axial-paraxial inter-action may occur at later stages, just before or shortly after somites have formed, and probably plays an important role in the formation of the sclerotome and dermamyotome (Kenny-Mobbs and Thorogood, 1987; Aoyma and Asamoto, 1988), shaping of the somites (Packard and Jacobson, 1979; Drake et al., 1992) and the further devel-opment of the sclerotome after its formation (Packard and Jacobson, 1976; Drake et al., 1992). In the absence of other evidence, this later influence on somite differentiation must be presumed distinct from the proposed function of axial structures in maintaining the symmetry of the somitic prepattern.

We believe that the results of our experiments, when con-sidered along with the literature cited above, are most plau-sibly explained as follows. Although the segmental plate contains a specified segmental pattern, this pattern is not determined and so is subject to regulation. When a seg-mental plate is grafted homotopically into a host embryo, the cells in the graft change their positions and reorient their cell processes in response to influences emanating from tissues near the midline of the host. This reorientation leads to a new specified segmental pattern in the form of a new somitomere pattern, that subsequently is expressed as an altered somite pattern. It is conceivable that pattern dis-ruptions induced by heat shock and other agents (Primmett et al., 1988, 1989) are often repaired by such respecifica-tion so that only the most severe damage is later detected in the somite pattern. Further study of heat-shocked embryos may be very helpful in understanding somite pat-tern regulation.

It is not clear whether the pattern-regulating influences demonstrated here represent a continuation of the processes that originally established the segmental pattern in the seg-mental plate or whether they represent a separate mecha-nism that maintains the specified pattern until it is expressed as somites. The presence of a bilaterally symmetrical seg-mental pattern has great importance in the evolutionary sense, because disruptions or incompatibilities in that pat-tern would lead to problems in locomotion. The continua-tion of pattern-specifying processes or the presence of a corrective process to repair disturbances of the specified segmental pattern would help ensure a symmetrical somite pattern.

The authors are indebted to Marisa Martini and Sandra McGillis for their tireless technical support, to Dr Douglas Robertson for enlightening discussions and assistance with matters statistical, and to Dr Steven Falen for the reconstruction software. The dia-grams were created by Martha Hefner, MS and the data were graphed by Brian Harris, MS This work was supported by the fol-lowing grants from the US National Institutes of Health: HD13396 (D. S. P.) and NS27409 (D. C. T.) and by a grant from the US Department of Agriculture (D. C. T.).