ABSTRACT
The hypothesis that parasegments are a fundamental unit of design in the Drosophila embryo was put forward three years ago (Martinez-Arias & Lawrence, Nature 313, 639-642, 1985). Here, I summarize new evidence in favour of the hypothesis.
Introduction
“… the limited information we have at the moment points to parasegments and compartments as being first in development and foremost in design.” Martinez-Arias & Lawrence, 1985.
In 1980, Morata & Kerridge were studying clones of the homeotic gene, Ultrabithorax (Ubx), in the thoracic segments of Drosophila and found the unexpected - so unexpected that I unwisely advised Gines Morata to quit the bithorax complex and take up skiing. They found that the Ubx gene was required not only in the T3 segment (as expected from Lewis’ model, 1978) but also in the entire posterior compartment of T2. There was no requirement for Ubx+ in anterior T2. Thus Ubx function was not delimited by a border of a segment but by a border within a segment (Morata & Kerridge, 1981). Similarly, Minana & Garcia-Bellido (1982) found that the posterior border of Ubx requirement in the adult was not between segments but within the first abdominal segment, A1.
Struhl (1984) showed that the bithorax complex (BX-C) could be separated into pieces that could still work independently. Each piece determined the development of a set of compartments that were out-of-register with segments. He therefore proposed that the bithorax complex recognized metameres that were not segments: ‘the realms of action of at least some BX-C genes are demarcated not by segment boundaries, but rather by anteroposterior compartment boundaries within segments’. From an independent study of larval phenotypes, Hayes et al. (1984) drew a similar conclusion; the two papers were received within one day of each other! Subsequently, we named these out-of-register metameres parasegments: in the ectoderm, a parasegment consists of a posterior compartment (P) of one segment and an anterior compartment (A) of the next (Martinez-Arias & Lawrence, 1985). We developed and generalized the idea, advancing the thesis that the parasegment, and not the segment, might be the fundamental metamere, both in the ectoderm and in the mesoderm. Our hypothesis was built on incomplete evidence, but more is now available and it is timely to look again at the status of the parasegment.
The aim of developmental genetics is to understand how genetic information is read out as function and form. While much is known about the link between genes and function we are largely ignorant of how genes determine form. Intuition about animal design has proved wayward and, therefore, the information used must be as objective as possible: there are at least three questions relevant to parasegments that yield usefully objective answers. (1) What are the earliest signs of metamerization in the embryo? (2) What are the units of cell lineage during development, and where are their boundaries? (3) In what domains are key genes expressed and required?
(1) What are the first signs of metamerization in the embryo?
The order and arrangement of metameres is ultimately dependent on the mother who lays down the genetic information to build the axes of the embryo (Nüsslein-Volhard et al. 1987). Soon, zygotic genes become active and, amongst these, the gap genes are the first to be expressed (Nüsslein-Volhard & Wieschaus, 1980; Knipple et al. 1985; Tautz et al. 1987). Each gap gene probably defines a zone in the egg and each zone spans several metameres. For example, RNA from the Krüppel gene is expressed in an area that extends from about the anterior border of parasegment 4 back some 3 parasegments (Jäckle et al. 1986). Double labelling with antibodies shows that, in the early gastrula, the bell-shaped curve of Krüppel expression (Gaul & Jackie, 1987) extends from about parasegment 4–7 inclusive; the middle of the curve being near to the anterior border of parasegment 6 (Gaul, unpublished data). It is not easy to see how such a broad zone could lead directly to metamerization, it is more likely the metameres are defined by other genes that act later on. Likely candidates are the pair-rule genes.
One of the earliest acting pair-rule genes is even-skipped (eve) (Nüsslein-Volhard & Wieschaus, 1980; Harding et al. 1986; Macdonald et al. 1986). The stripes of eve appear in midblastoderm and become more defined as blastoderm progresses. Antibody studies show the sharpening of the stripe is achieved mainly at the anterior edge where a wiggly but stable boundary forms (Frasch & Levine, 1987; Lawrence et al. 1987). This sharp line at the anterior edge of the eve stripe delimits the anterior borders of the odd-numbered parasegments; just anterior to the line each cell will belong to, say, parasegment 4, while just posterior each cell will belong to parasegment 5. Similarly, the anterior edge of the fushi taraza (ftz) stripes delimits the anterior edge of the even-numbered parasegments (Lawrence et al. 1987; Carroll et al. 1988). Neither eve nor ftz have definable or stable posterior boundaries, and it therefore seems unlikely that these boundaries could have any important function (Lawrence, 1987). Since expression of the engrailed gene, which defines posterior compartments (Morata & Lawrence, 1975; Kornberg et al. 1985), is subsequent to, and dependent on, eve and ftz expression (Carroll & Scott, 1986; Harding et al. 1986; Howard & Ingham, 1986; Macdonald et al. 1986), it is clear that parasegments are defined before compartments and segments.
This picture may not be complete but it does suggest that parasegments are the first metameres to be defined in development. The simplest hypothesis that will now suffice is that ftz and eve draw the parasegmental boundaries so that each cell is allocated with respect to them.
Some genes upstream of eve and ftz are probably concerned in placing the stripes but not, specifically, in allocating cells to parasegments. Apart from the maternal effect and gap genes, which act much earlier (reviewed in Nüsslein-Volhard et al. 1987), there are two pair-rule genes that are upstream to ftz and eve (runt and hairy, Carroll & Scott, 1986; Howard & Ingham, 1986; Frasch & Levine, 1987; Ingham & Gergen, this volume). Both genes are expressed in stripes which are out of phase with each other (Ingham et al. 1985a; Gergen & Butler, 1988; Ingham & Gergen, this volume). However,’we do not know that the wild-type functions of these genes are; I would expect them to be expressed in stripes that are different in character to ftz and eve, for example, they might lack sharp boundaries.
The definition of parasegments at the genetic level has anatomical consequences. Grooves appear in the extended germband and these coincide with parasegmental boundaries (Martinez-Arias & Lawrence, 1985; Ingham et al. 1985b). In the mesoderm, deeper grooves are formed and these also demarcate the parasegments (Lawrence & Martinez-Arias, 1985). These grooves are the first visible signs of metamerization in both Drosophila and other insects.
(2) What are the units of cell lineage during development and where are their boundaries?
Much of this material has been reviewed before and needs little reiteration. Clonal analysis has shown that early in the embryo, that is by one cell division after blastoderm, the epidermal cells are divided into A and P polyclones that will generate the A and P compartments of the adult (Garcia Bellido et al. 1973; Crick & Lawrence, 1975; Lawrence, 1981). This means that by then, which is probably the newly extended germband stage when epidermal cell divisions restart (Campos-Ortega & Hartenstein, 1985), all the precursors of adult epidermal cells are allocated so that both parasegmental (P/A) and segmental (A/P) borders are defined. The expression of engrailed in the extended germband shows clear and sharp stripes which suggests that all cells of the ectoderm (precursors of the larval and adult epidermis and nervous system) are allocated to either A or P compartments by that stage.
Cell lineage experiments cannot prove that para-segments are established first and then subdivided into compartments, because both steps occur in the absence of cell division. However, the cell lineage data are fully consistent with that sequence of events.
Clonal analysis shows that the mesoderm is divided into metameres which each generate a precisely defined set of adult muscles that span the segment. Clones of cells in the mesoderm do not cross from one set of muscles to another, even if they are founded-soon after blastoderm (Lawrence, 1982). The existence of these metameres, and their boundaries, was confirmed by nuclear transplantation (Lawrence & Johnston, 1986); they probably derive from the embryonic parasegments (Martinez-Arias & Lawrence, 1985). However, the mesoderm differs fundamentally from the ectoderm; mesodermal parasegments are not subdivided into anterior and posterior compartments (Lawrence, 1982). Moreover, although the engrailed gene is transiently expressed in mesodermal cells (Kornberg et al. 1985), it is probably not required there (Lawrence & Johnston, 1984).
When marked mesodermal cells are transplanted from one embryo to another they can colonize muscles in more than one larval segment, and this has been interpreted as evidence against my hypothesis that the mesoderm is subdivided into single metameres (Beer et al. 1987). However, it is doubtful that transplanted cells, when separated from their neigh-bours and implanted into a (necessarily) wounded region of a host embryo, would behave as they do in situ. The simplest viable hypothesis therefore is that the mesoderm, both somatic and visceral, is subdivided into parasegments by the eve, ftz and other genes in exactly the same way as the ectoderm. Indeed, these subdivisions may be independent of, and may pre-exist, the definition of the mesoderm.
(3) In what domains are key genes expressed and required?
Since 1985 much new data relevant to this question have been accumulated. Broadly, the answer has two parts; first, the initial pattern of expression of the major homeotic genes is strictly parasegmental, and probably remains so in the mesoderm. Second, in the ectoderm, the initial parasegmental pattern becomes more complex and confusing later on (see, for example, Martinez-Arias et al. 1987).
The confusion is partly due to inadequate techniques and analysis, partly due to a genuine increase in the complexity of gene expression. In situ hybridization has, in practice, insufficient resolution to decide the exact pattern of expression. Since sections are used and since cells are, at best, pseudohexagonally packed, locating gene expression is no better than plus or minus one cell (see Jäckle et al. 1986) and, when 35S probes are used, poorer than that. At the blastoderm stage, the metameres average only about 3i cells across (Sullivan, 1987) and it is therefore not surprising that mistakes can be made; for example, when determined by in situ hybridization, the anterior boundary of the third ftz stripe was widely reported as being out of register with the anterior boundary of Ubx expression and of parasegment 6 (Akam & Martinez-Arias, 1985; Akam, 1985; Ingham et al. 1986; Ish-Horowicz et al. 1985) when by anti body staining it is clear that they coincide exactly (Lawrence et al. 1987; Carroll et al. 1988). In mapping expression of a gene, the anatomical landmarks are not sufficiently precise or well-enough observed or agreed to be used at the cellular level. Since a parasegment is a specific set of cells, with no ambiguity, but with wiggly boundaries, gene expression should ideally be examined at cellular resolution. This, for the moment, means double labelling with antibodies, one of which is a marker, such as engrailed, ftz or eve, that defines the parasegmental borders. One particularly useful and persistent marker is βgalactosidase which can be put under the control of ftz or eve 5’ regions (Hiromi et al. 1985; Lawrence et al. 1987).
The best-studied case of a homeotic selector gene is Ubx+. The Ubx protein can first be detected at the beginning of germband extension and is confined then to parasegment 6. Soon after extension of the germband, Ubx expression becomes visible in a complex pattern including parasegment 5 – 13 in the epidermis and in the somatic mesoderm from 6 – 12 inclusive (Akam & Martinez-Arias, 1985; Akam, 1987; see Fig. 1). Expression of Ubx+ in the visceral mesoderm appears later, is confined to parasegment 7 and is regulated somewhat independently from the ectoderm (Bienz et al. 1988). Double-labelling experiments show that the anterior border of Ubx expression in parasegment 6 is exactly coincident with the parasegment border as defined by ftz-figaf as is the anterior border of the patch of Ubx expression exactly coincident with the eve β gal border of para-segment 5. The same coincidence can be seen in the underlying somatic mesoderm at the anterior edge of parasegment 6 (Lawrence & Johnston, unpublished data).
Later expression of Ubx in the larval muscles shows that the muscles of Al (derived from parasegment 6, Martinez-Arias & Lawrence, 1985; Lawrence & Martinez-Arias, 1985) but not T3, all express Ubx at a low level, and those of A2 at a higher level (Hooper, 1986) suggesting that control of Ubx expression is somewhat independent in each parasegment. Experiments mapping the requirement for the three genes of the bithorax complex, Ubx, abd-A and AbdB, in the adult all implicate parasegments not segments (e.g. Casanova et al. 1986).
All in all, the studies of Ubx point to the earliest and most fundamental unit of design being a parasegment. The same conclusion can readily be drawn with respect to other selector genes such as Deformed, Sex combs reduced (see Fig. 2) and Antennapedia (see Akam, 1987 for references). However, because of the lack of double-labelling experiments and changes in the domains of expression during development there is still room for other interpretations.
Conclusion
All three questions give essentially the same answer: (1) The parasegments appear to be the first metameres to be defined as specific sets of cells and it is the grooves between the parasegments that are the earliest visible signs of segmentation in both ectoderm and mesoderm. (2) Experiments on cell lineage suggest that parasegments are defined in both ectoderm and mesoderm and later subdivided into compartments in the ectoderm only. (3) Expression of Ubx+ and other genes shows that the zones of gene expression in the ectoderm, somatic and visceral mesoderm coincide with parasegments suggesting that the parasegment is a basic domain of genetic control. Whether the parasegment is the true archetypal metamere is unproved, comparative molecular genetics will be required to find out. However, given the many lines of evidence in Drosophila and the early origin of parasegments in development, it would be astounding if other insects, and even annelids, were made of fundamentally different units.