An axial domain of HOM/Hox gene expression is formed by morphogenetic alignment of independently specified cell lineages in the leech Helobdella

The HOM/Hox genes play a major role in the A-P patterning of a wide variety of animal species (for review, see McGinnis and Krumlauf, 1992). These closely related homeobox genes are organized into genomic clusters, and comparative studies show that the identity of the individual genes and their chromosomal arrangement are for the most part conserved across diverse phyla. In animals with an A-P body axis, the individual genes within the HOM/Hox cluster are expressed in a staggered array of overlapping axial domains, and differing combinations of these gene products are thought to encode axial position by bringing about the diversification of otherwise homologous cells and tissues.

In this paper we address the mechanism by which HOM/Hox gene expression is coordinately regulated in the different cell lineages that contribute to axis formation. In Drosophila, this coordination arises because the homeotic or HOM genes and their immediate upstream regulators are expressed in transverse bands, such that cells located at the same longitudinal coordinate of the embryo receive comparable A-P patterning information (Akam, 1987), even though they become differentially committed to mesoderm, peripheral ectoderm or central nervous system (CNS) along the dorsoventral axis (Nüsslein-Volhard, 1991). However, in many other animal species the axial patterning of different cell lineages is complicated by the fact that the germ layers undergo extensive AP rearrangments during gastrulation. In the vertebrates, some authors have proposed that axial patterns of Hox gene expression originate in one germ layer or tissue, and are locally imprinted to other tissues after gastrulation is complete (DeRobertis et al., 1989; Frohman et al., 1990). However, the mesoderm and neurectoderm of the Xenopus embryo need not come into alignment in order to manifest a normal A-P sequence of gene expression domains (Doniach, 1992; Ruiz i Altaba, 1992), suggesting that axial pattern is specified in these two germ layers independently of spatial proximity between tissues expressing the same positional value.

We here examine the role of morphogenetic alignment in the HOM/Hox gene expression of an annelid, the glossiphoniid leech Helobdella, whose embryonic segmentation arises within the context of an invariant cell lineage (Stent et al., 1992). A major advantage of this organism is that lineally identified segment founder cells can be experimentally diverted into the formation of ectopic body segments (Shankland, 1984; Martindale and Shankland, 1990) so as to test their commitment to a particular segment identity or value of axial differentiation. Fig. 1 provides an overview of the developmental events leading to axis formation in this organism. The early cleavages generate a set of 5 bilaterally paired stem cells or teloblasts, which together generate the segmental tissues. The M teloblast gives rise to the somatic and visceral mesoderm, while the N, O, P and Q teloblasts give rise to distinct portions of the ectoderm. Each teloblast undergoes an iterated, stem cell sequence of highly asymmetric divisions, and the smaller daughter cells - known as primary blast cells - serve as segmental founder cells for their particular teloblast lineage, i.e. blast cells generated by the same teloblast give rise to phenotypically similar descendant clones situated in different body segments (Weisblat and Shankland, 1985).

There is a predictable relationship between a blast cell’s birth rank in the stem cell lineage of its parent teloblast, and the segmental location and developmental identity of its descendant clone. Blast cells depart from the parent teloblast in a linear array, with the firstborn blast cell contributing its descendant clone to the anteriormost body segment, and subsequent blast cells contributing to progressively more posterior segments. However, the different teloblast lineages are not in segmental register with one another at the earliest stages of blast cell formation, and for many blast cells there is a period of several days before the descendant clone comes together with other blast cell clones that are fated to occupy the same body segment of the adult leech (Weisblat and Shankland, 1985; Lans et al., 1993).

Several HOM/Hox genes have been isolated from Helobdella (Shankland et al., 1991; Nardelli-Haefliger and Shankland, 1992) and another leech species (Wysocka-Diller et al., 1989; Aisemberg et al., 1993; Aisemberg and Macagno, 1994). One of these genes, Lox2, appears to be the leech orthologue of the Drosophila HOM gene Ultrabithorax, and in situ hybridization reveals that Lox2 RNA is expressed in a restricted axial domain of the embryo’s germinal plate (Wysocka-Diller et al., 1989; Nardelli-Haefliger and Shankland, 1992). This Lox2 domain has a sharp anterior boundary in the sixth midbody segment (M6), and cell lineage studies indicate that the different teloblast lineages share this common boundary of gene expression (Nardelli-Haefliger and Shankland, 1992).

There are three scenarios that could account for the spatially coordinated expression of Lox2 gene products in the different teloblast lineages. First, the various cell lineages could be specified by a single pattern of positional information subsequent to their morphogenetic alignment. Second, one or more lineages may be specified prior to alignment, and secondarily imprint their axial pattern onto the other, naive lineages via appositional cell interactions. Finally, the teloblast lineages might be specified prior to alignment, and secondarily align their axial patterns during the assembly of the germinal plate. The results presented here indicate that the Lox2 domain is not specified by positional information acting across the germinal plate, but point instead to a secondary alignment of expression domains specified at an earlier stage in development. This conclusion is consonant with previous studies in which the axial patterning of experimentally displaced cell lineages was characterized by patterns of neuropeptide/neurotransmitter expression (Martindale and Shankland, 1990; Shankland and Martindale, 1992) or mesodermal organogenesis (Gleizer and Stent, 1993).

Embryos were obtained from breeding colonies of the closely related leech species Helobdella triserialis and H. robusta (Shankland et al., 1992). The staging system and cell lineage nomenclature are described by Stent et al. (1992). The initially equipotent O/P teloblasts are here designated as either O or P teloblasts in accordance with their distinct developmental fates.

A 764 bp fragment of the Lox2-Hro cDNA (Nt 574-1337; Nardelli-Haefliger and Shankland, 1992) was amplified with the GeneAmp kit (Perkin-Elmer-Cetus) and ligated into the BamHI cloning site of the pET5b plasmid (Novagen). Recombinant plasmid was transformed into E. coli strain BL21, and liquid cultures induced with IPTG to express a fusion protein comprising 11 N-terminal amino-acids from T7 bacteriophage gene 10, a 4 amino-acid (RGSA) linker, and the 238 C-terminal amino-acids of the predicted LOX2 protein. Total bacterial protein was separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and the induced protein band recovered by electroelution. The eluted protein was used to immunize two adult female rabbits (Pocono Rabbit Farm), and antisera tested by immunostaining Helobdella embryos. The two anti-LOX2 sera gave identical patterns of staining, and serum PRF9258 was selected for further studies because of the lower background staining.

Antiserum was affinity-purified to fusion protein conjugated to a glutaraldehyde-activated filter substrate in a Nalgene U12 chromatography unit. 0.5 ml of serum was repeatedly passed through the filter, and the bound protein fraction eluted in 1.0 ml of 0.1 M glycine buffer (pH 2.3). The eluate was neutralized with 0.4 ml of 1 M Tris buffer (pH 8.0), and 0.1% BSA added for storage.

Total bacterial protein was separated by SDS-PAGE and transferred electrophoretically to nylon. Blots were blocked with milk protein, then incubated with a 1:4000 dilution of affinity-purified anti-LOX2 serum, washed, and incubated with a 1:5000 dilution of alkaline phosphatase-conjugated goat anti-rabbit secondary antibody (Boehringer-Mannheim). Enzyme was reacted with 4-Nitro blue tetrazolium chloride and 5-Bromo-4-chloro-3-indolyl-phosphate substrates (Boehringer-Mannheim). The affinity-purified anti-LOX2 serum recognized a cluster of protein bands of expected mobility in both induced and uninduced cultures. Preabsorption with a lysate of uninduced E. coli eliminated staining of all but a single band found only in induced cultures (Fig. 2A).

Helobdella embryos were fixed overnight at 4°C with 4% formaldehyde in Hepes-buffered saline (HBS; pH 7.4), washed in HBS and dissected with tungsten pins. Tissues were blocked with 10% normal goat serum in HBT (HBS + 0.3% Triton X-100), then incubated with anti-LOX2 serum (see below) in HBT. After washing, the tissue was incubated with a 1:600 dilution of horseradish peroxidase (HRP) conjugated goat anti-rabbit serum (Accurate) in HBT. After a final wash, the enzyme was reacted with diaminobenzidine.

Anti-LOX2 serum gave low levels of non-specific nuclear staining in all tissues examined, and samples used for histochemistry were further purified by preabsorption against another leech homeoprotein, LOX1 (Aisemberg and Macagno, 1994). For preabsorption, the affinity-purified anti-LOX2 serum was diluted 1:50 in HBT, passed repeatedly over a chromatography unit containing the LOX1 protein, and stabilized with 0.1% BSA.

Individual teloblasts were pressure injected with 50 mg/ml fluorescent dextran lineage tracer and 20 mg/ml Fast Green FCF in 0.2 M KCl. Rhodamine dextran amine (RDA) was synthesized as described by Stuart et al. (1990) using a succinimidyl ester of tetramethylrhodamine (Molecular Probes). Fluorescein dextran-amine (FDA) was purchased from Molecular Probes. The fluorescently labeled embryos were fixed and cleared in buffered glycerol containing 4% n-propyl gallate to retard photobleaching, and examined in optical section on a Bio-Rad MRC-600 confocal fluorescence microscope. Confocal images were colored using Adobe Photoshop 2.5 software. HRP reaction product was concurrently visualized with Nomarski optics, photographed with Ektachrome 160T and digitized on a Microtek ScanMaker 1850S.

Lox2 RNA was transcribed and capped with the mMessage mMachine kit (Ambion) from a pET-11c plasmid (Novagen) containing the coding region of the Lox2-Hro cDNA. Approximately 10 pl of a solution of 0.2 mg/ml RNA and 2 mg/ml Fast Green FCF in 0.2 M KCl was pressure-injected into the N teloblast. Injected embryos were raised for 24 hours, fixed, dissected free of the vitelline membrane and stained with anti-LOX2. Immunoreactivity was visualized with a 1:200 dilution of fluorescein-conjugated goat anti-rabbit secondary antibody (Vector), and embryos examined on the confocal fluorescence microscope inside a capillary tube permitting rotation beneath the objective.

Individual teloblast lineages were displaced along the A-P axis as previously described (Shankland, 1984). The targeted lineage was labeled with FDA by microinjection of the parent teloblast, and on the following day the 2-3 anteriormost FDA-labeled blast cells were photoablated by mercury arc illumination as they entered into the germinal band. In most embryos the trailing fragment of the lesioned blast cell chain slipped posteriorly, so that its descendant tissues came to lie out of segmental register with the other teloblast lineages.

Slippage was measured by a procedure in which the teloblast of interest and its contralateral homologue were injected with RDA lineage tracer immediately following the photoablation (Shankland, 1984). In unlesioned control embryos, the anterior boundary of RDA labeling on the right and left sides of the embryo came to lie in the same or adjacent segments, whereas in lesioned embryos the much larger disparity between right and left RDA boundaries served as a measure of axial displacement. As expected, FDA-labeled blast cell clones were observed anterior to the slipped RDA boundary in all embryos examined.

The 238 C-terminal amino-acids of the predicted LOX2 protein, including the homeodomain, were expressed in E. coli as part of a fusion protein (M_r = 28×10³). Bacterial induction yielded a single prominent protein band which migrated on SDS-PAGE at approximately 33×10³. Rabbit anti-LOX2 serum PRF9258 was raised against the isolated fusion protein, and recognizes a single protein band of the correct mobility on immunoblots of induced bacterial lysogens (Fig. 2A).

To demonstrate further that the antiserum recognizes the protein product of the cloned cDNA, we transcribed the Lox2Hro cDNA in vitro, injected capped transcripts into single teloblasts of stage 6 Helobdella embryos, then fixed the embryos 1-2 days after injection and immunostained for LOX2 expression. Anti-LOX2 immunoreactivity was observed in the nuclei - and to a lesser extent cytoplasm - of the injected teloblast and its subsequent blast cell progeny, but not in other, uninjected cell lineages (Fig. 2B). Antiserum PRF9258 is therefore capable of staining the LOX2 protein in histological preparations.

We examined the developmental expression of LOX2 protein by staining dissected embryos with an affinity-purified antiLOX2 serum preabsorbed to reduce homeodomain cross-reactivity (see Materials and Methods). In general, the distribution of anti-LOX2 immunoreactivity is very similar in pattern and time-course to the expression of Lox2 RNA detected by in situ hybridization (Nardelli-Haefliger and Shankland, 1992). Identical staining patterns were obtained with both Helobdella species examined, and results obtained from these two species will here be presented without distinction. No specific staining was obtained with either preimmune serum or secondary antibodies alone.

Anti-LOX2 immunoreactivity first becomes detectable during embryonic stage 9, and by the middle of stage 10, the Helobdella embryo has developed a stable and segmentally restricted pattern of expression that persists into the adult.

Immunoreactivity shows a nuclear localization befitting a transcription factor, with the highest level of expression observed in neurons of the segmental ganglia of the CNS (Fig. 3). The expression domain has a sharp anterior boundary in ganglion M6 (Fig. 3A,B), but the posterior boundary is not as readily defined (Fig. 3C). At the beginning of embryonic stage 11, the most posterior neuronal staining was typically observed in a bilateral pair of cells located at the lateral edge of the ganglion as far posterior as segments M19/M20.

The body wall and digestive tract also express a segmentally repeated pattern of anti-LOX2 immunoreactivity whose anterior boundary falls in segment M6 (Fig. 3A). Immunoreactive nuclei were observed in the musculature, nephridia, and other, uncharacterized mesodermal tissues. However, we failed to discern anti-LOX2 immunostaining of identified peripheral sensory neurons, epidermis or endoderm. In addition to segmentally repeated structures, we also found high levels of LOX2 expression in a cluster of nuclei associated with the developing reproductive organs in segment M6 (Fig. 3A).

The distribution of LOX2 protein was examined in central neurons from each of the five teloblast lineages. Individual teloblasts were injected with RDA prior to blast cell formation, and the injected embryos raised to the beginning of stage 11. The CNS was stained for LOX2 using HRP immunochemistry, and neurons containing the fluorescent lineage tracer were individually examined for HRP reaction product. Largely symmetrical patterns of neuronal expression were observed in lineages descended from homologous right and left teloblasts.

The O teloblast gives rise to approximately 40 neurons/hemiganglion, distributed in three major clusters (Weisblat and Shankland, 1985). Individual neurons within a cluster often exhibit widely differing levels of anti-LOX2 immunoreactivity, and comparison of homologous clusters in different ganglia revealed distinct, segment-specific patterns of expression.

The posteroventral (PV) cluster of O-derived neurons is situated immediately lateral to the posterior connective. The PV neurons do not express LOX2 in M5 or more anterior ganglia, but in M6 and immediately posterior ganglia we routinely encountered 3-4 immunoreactive PV neurons (Fig. 4A-F). However, the PV neurons in ganglion M6 express LOX2 at a much higher level than their homologues in any other ganglion (compare Fig. 4D and 4F). Immunoreactive PV neurons were found as far posteriorly as ganglion M16, but not in more posterior segments.

The crescent (CR) cluster of O-derived neurons extends from the ganglion’s lateral edge to its ventral midline. The CR neurons do not express LOX2 in M6 or more anterior ganglia, but several CR neurons express detectable levels of anti-LOX2 immunoreactivity in more posterior ganglia (Fig. 4G-L). In ganglion M7, a single large cell - designated as neuron OM7-situated near the medial edge of the CR cluster, expresses a distinctively high level of anti-LOX2 immunoreactivity (Fig. 4J). This distinctive pattern of LOX2 expression was not observed in any other segment (Fig. 3B), and lineal homologues of the OM7 neuron in other ganglia must either degenerate or express a significantly lower concentration of LOX2 protein (Fig. 4L).

Neurons of the O-derived anterodorsal cluster also express anti-LOX2 immunoreactivity in a restricted segmental domain, with anterior boundary in ganglion M7. This component of the expression pattern was not characterized in further detail.

The N teloblast gives rise to approximately 150 neurons distributed in a complex pattern throughout the segmental hemiganglion (Kramer and Weisblat, 1985). Differing numbers of N-derived neurons express anti-LOX2 immunoreactivity in particular midbody segments. The most anterior staining was observed in 2-3 neurons situated dorsal to the O-derived PV neuron cluster of ganglion M6 (Fig. 5A,B). (This portion of the N lineage Lox2 expression pattern was not previously detected by in situ hybridization). There is a stepwise increase in the number of immunoreactive N-derived neurons in ganglia M7 and M8 (Fig. 5C,D), with a similarly large fraction of the ganglionic neurons expressing LOX2 over the segmental domain M8-M14. The number and intensity of immunoreactive neurons decreases in yet more posterior ganglia, and we were unable to detect any immunoreactive N-derived neurons posterior to ganglion M17.

We did not attempt to identify individual N-derived neurons solely on the basis of LOX2 expression. Several neurons within the N lineage can be identified by anti-serotonin immunostaining (Stuart et al., 1989), but double-staining did not reveal any detectable anti-LOX2 immunoreactivity in the Retzius neuron, and barely detectable expression in the lateral serotonin neurons.

The three remaining teloblast lineages displayed many of the same general features of LOX2 expression. The M, P and Q lineages all express LOX2 within a restricted segmental domain, with the anteriormost neuronal expression in ganglion M7. The posterior boundaries were not as sharply defined, although we did ascertain that two lateral immunoreactive neurons seen in the more posterior midbody ganglia are separately descended from the P and Q teloblasts.

As seen with the O lineage, individual neurons show distinctive patterns of segmental expression. For example, the Q lineage includes several anteroventral cluster neurons (Weisblat and Shankland, 1985) that express LOX2 intensely in ganglion M7 but at a much lower level in more posterior ganglia. In contrast, the Q-derived qz1 neuron displays a high level of anti-LOX2 immunoreactivity throughout the entire expression domain.

To ascertain the relative contributions of cell lineage and position in defining the LOX2 expression domain, we characterized the pattern of LOX2 expression in 21 Helobdella embryos whose right O lineage had been ‘slipped’ posteriorly so that its constituent blast cell clones were frameshifted out of register with the other teloblast lineages (Shankland, 1984; Martindale and Shankland, 1990). The developmental fate of the slipped lineage was followed by means of FDA lineage tracer that had been previously injected into the parent teloblast, and the magnitude of the displacement calculated by labeling cells with a series of secondary injections (Shankland, 1984).

Posterior slippage of the right O lineage altered the segmental domain of LOX2 expression. These alterations appeared to be specific to the neuronal derivatives of the slipped cell lineage, since the other teloblast lineages displayed essentially normal patterns of LOX2 expression. A normal pattern of LOX2 expression was observed in 2 control embryos that received the same array of lineage tracer injections but were not induced to undergo slippage.

O-derived neurons whose precursors have been displaced into the normal LOX2 domain from anterior segments do not express anti-LOX2 immunoreactivity, even though they have taken the place of segmental homologues, which normally do so. There were 6 experimental embryos in which the PV cluster neurons, normally destined to occupy ganglia M4 and/or M5, had taken part in the formation of ganglia within the normal LOX2 domain, and these neurons consistently failed to express detectable LOX2 protein (Fig. 6A,B). This lack of expression was not a general characteristic of the manipulated cell lineage, since PV neurons normally destined for M6 or more posterior ganglia do express anti-LOX2 immunoreactivity when slipped into more posterior segments. It should be noted, however, that PV neurons destined for M6 did not show their normally high levels of LOX2 expression when displaced into more posterior ganglia, suggesting that LOX2 expression may be enhanced for the PV neurons situated in ganglion M6 by inductive interactions with some segment-specific environmental feature.

We also observed ectopic LOX2 expression by neurons displaced outside the normal expression domain. Posterior slippage of the right O lineage resulted in the anomalous expression of anti-LOX2 immunoreactivity by O-derived neurons on the right side of the posterior midbody ganglia (Fig. 6C), suggesting that neurons fated by lineage to express the LOX2 protein will do so even if forced to take part in the formation of ectopic ganglia where this gene is not normally expressed.

Slippage of the O lineage brought about a predictable repositioning of the highly immunoreactive OM7 neuron. The OM7 neuron on the unperturbed left side of the CNS was consistently situated in ganglion M7, and in 20 out of 21 experimental embryos the slipped, right side O lineage gave rise to a single OM7 neuron, which was variably located 3-8 segments more posteriorly (Fig. 6D). (The remaining embryo did not have a neuron on the right side of the CNS that exhibited the OM7 phenotype, and lineage tracer analysis revealed that the blast cell that would normally have given rise to this particular neuron had been ablated during the slippage procedure.) For each experimental embryo, the repositioning of the right OM7 neuron was found to correlate closely (0.2±1.0 segment; mean ± s.d.) with the number of segments the O lineage had been displaced. This correlation suggests that slippage simply frameshifted the O lineage pattern of LOX2 expression relative to the other axial tissues, without altering the segment identity of its constitutent blast cell clones.

A lineage-specific displacement of the LOX2 expression pattern was likewise observed following experimental slippage of the right or left N lineage. The N lineage was slipped 1-3 segments posteriorly in 24 embryos, and in all cases there was a clear reduction in the number of anti-LOX2 immunoreactive neurons on the operated side of ganglia M7 and M8 (Fig. 6E). These ganglia contain a full complement of N-derived neurons as indicated by the distribution of FDA lineage tracer, and thus it would appear that N-derived neurons normally destined for more anterior ganglia do not undertake LOX2 expression when transplanted into the normal expression domain. The slipped lineage did express LOX2 protein in more posterior ganglia, and the anterior boundary of that staining correlated well with the measured displacement.

We did not observe any clear alteration in the posterior boundary of LOX2 expression following N slippage. However, it was not clear that such a shift would have been detected. The N lineage expresses very low levels of LOX2 protein in the posterior midbody ganglia (see above), and the small degree of slippage obtained in these experiments is probably insufficient to produce an obvious change in the staining of those ganglia.

Unilateral slippage of the N lineage did not have any consistent effect on the pattern of LOX2 expression in the other teloblast lineages, but some anomalies were observed. In 4 of the 24 N-slipped embryos, the CNS displayed a pattern of LOX2 expression that would be expected if the ipsilateral O lineage had been shifted anteriorly by 1 segment - i.e. immunoreactive PV neurons in ganglion M5, and the ipsilateral OM7 neuron in ganglion M6. We also encountered a fifth experimental embryo with immunoreactive PV neurons in ganglion M5, but no displacement of the OM7 neuron. All of the remaining embryos that had experienced posterior N slippage showed an outwardly normal pattern of LOX2 expression in the other teloblast lineages, including paired OM7 neurons in ganglion M7 (Fig. 6E).

The O lineage was not lesioned in these experiments and contained no lineage tracer, but several lines of evidence suggest that the observed anomalies in LOX2 expression arose through a misalignment of the O teloblast lineage relative to the segmental body plan as a whole. First, the inconstant nature of this phenomenon is characteristic of a variable morphogenetic defect. The O lineage is immediately adjacent to the N lineage at the time of slippage, and it has been shown previously that disruptions of the N lineage can lead to abnormal patterns of ganglion formation (Stuart et al., 1989). Second, it is not readily apparent how a posterior displacement of the N lineage could respecify cell identities within the O lineage in such a way as to produce an anterior shift. One additional point is the fact that we also observed anomalous anterior expression of LOX2 immunoreactivity in 3 of the 70 embryos in which teloblast lineages were labeled without intentional slippage. In one such embryo the O lineage had been labeled, and had undergone a spontaneous break in segment M7 such that the the CR cluster including the OM7 neuron remained in its normal location, and the immunoreactive PV neurons normally destined for ganglion M6 had been displaced anteriorly into ganglion M5.

Single teloblast lineages that are forced to develop several segments out of register with the remaining axial tissues manifest a pattern of axial differentiation that is essentially normal within the context of their own A-P organization, but shifted with respect to the body as a whole. Blast cells displaced into the LOX2 expression domain from more anterior regions generate descendant neurons that behave according to their lineage history, in that they do not express LOX2 protein, and blast cells displaced out of the LOX2 expression domain generate descendant neurons that express LOX2 protein in discordance with their axial position. Moreover, axial displacement of the entire teloblast lineage brings about a precise repositioning of certain readily identifiable neurons or groups of neurons with distinctive patterns of LOX2 expression. Thus, lineally identified neurons can manifest their own characteristic pattern of gene expression even when that expression is inappropriate for the ganglion in which they reside. The most distinctive of these neuronal phenotypes, the OM7 neuron, was never duplicated or missing, as might have been expected if axial displacement had respecified the slipped cell lineage so as to disrupt the normal sequence of segmental identities along its length.

This independence between the segmental patterning of the displaced teloblast lineage and that of the germinal plate as a whole argues that the normally restricted domain of LOX2 expression cannot be defined by positional cues that act after the teloblast lineages have become aligned within the germinal plate. Alignment occurs relatively late in the development of the leech embryo, and involves extensive longitudinal reorganization of the individual teloblast lineages (Weisblat and Shankland, 1985; Lans et al., 1993). The slippage experiments performed here disrupt the segmental register of the operated cell lineage at a much earlier developmental stage - prior to either neurogenesis or the onset of LOX2 expression - with the result that the slipped blast cells never acquire their normally fated positions in the overall framework of body segmentation. Thus, the axial domain of LOX2 expression within the displaced cell lineage cannot be directly imprinted from the other, unperturbed tissues such as the mesodermal M lineage, nor can it be specified by any set of positional cues acting across the five teloblast lineages coordinately. In a similar vein, the slipped lineage does not itself appear to generate positional cues that have any significant effect on LOX2 expression in the other, unperturbed lineages. The M, P and Q lineages were not displaced in this study, but previous slippage and ablation experiments using other segmental markers suggest that they also manifest a large degree of developmental independence (Gleizer and Stent, 1993; Stuart et al., 1989).

Axial differentiation is a complex, multistep process, and we do not wish to exclude the possibility that segment-specific cell interactions could play some role in the regulation of Lox2 or other leech HOM/Hox genes. The large number of neurons within each lineage and the cellular complexity of the LOX2 expression pattern leave open the possibility that some repositioned neurons may have experienced unnoticed changes in gene expression. Our data indicate that expression pattern is largely determined by factors intrinsic to the teloblast lineage, but we can not rule out position-dependent modifications of the underlying lineage-dependent pattern. For example, the O-derived PV neurons in ganglion M6 do not retain their distinctively high level of LOX2 expression when forced to differentiate in more posterior ganglia, and may normally take part in a segment-specific interaction which enhances their expression of this protein. An analogous situation has been observed in the specification of other aspects of segmental specificity. Early in embryogenesis the N lineage develops an axial pattern that specifies the formation of several neuronal phenotypes which are normally restricted to particular body segments (Martindale and Shankland, 1990; Shankland and Martindale, 1992), but later on the CNS elaborates an additional tier of segmental differences through the interaction of neurons with segment-specific mesodermal targets (Loer et al., 1987), and the competitive interaction of homologue neurons situated in different body segments (Gao and Macagno, 1987; Blair et al., 1990).

Slippage experiments only examine the importance of interactions between the different teloblast lineages, and do not directly address the mechanism by which axial pattern is established within a single lineage. During normal development there is an exact correlation between a blast cell’s lineage history and the segmental location of its descendant clone, and it has been suggested that individual blast cell daughters of a given teloblast already possess distinct identities at the time of their birth (Martindale and Shankland, 1990). While it is still possible that the segment identity of the blast cell could be specified by positional cues encountered subsequent to its birth, the present findings indicate that any such positional cues must arise within or be acting upon each teloblast lineage independently.

In other organisms as well there is a growing body of evidence for the independent specification of axial pattern in different embryonic cell lineages. Inductive cell interactions are of primary importance in the axis formation of vertebrate embryos (Slack and Tannahill, 1992), but even so, different cell lineages can manifest or even acquire their axial patterns independently. In chick, hindbrain rhombomeres retain their own intrinsic pattern of Hox gene expression when transplanted relative to the paraxial mesoderm (Guthrie et al., 1992; Kuratani and Eichele, 1993), indicating that the neural tube is committed to an autonomous pattern of axial differentiation by the stage of transplantation. Even more profound dissociation of mesodermal and ectodermal patterning has been reported in the Xenopus embryo. Neurectoderm and mesoderm are able to generate complex and largely normal axial patterns of molecular differentiation when explants are cultured in an endto-end planar orientation (Doniach, 1992), and similar patterns of differentiation are also observed in embryos in which the normal alignment of these germ layers is prevented by aberrant patterns of gastrulation (Ruiz i Altaba, 1992). The neurectoderm of the Xenopus embryo must interact with the mesoderm to form an axial pattern, but these experiments argue that the two germ layers need not occupy their normally colinear spatial framework in order to develop A-P patterns of positional differentiation.

These findings raise an important question about the construction of the primary body axis in the leech. If the LOX2 expression domain is independently specified in different teloblast lineages, it is then necessary for those lineages to bring their respective expression domains into alignment, i.e. with the anterior boundaries of expression in segment M6. Alignment of the teloblast lineages occurs several days prior to the earliest detected expression of the LOX2 protein, and is therefore unlikely to depend upon the reciprocal recognition of LOX2-expressing cells. Indeed, one could envision that segment identity does not play a role in the alignment process, since blast cells assemble in a stereotyped fashion beginning with the anteriormost segment and proceeding posteriorly (Weisblat and Shankland, 1985; Lans et al., 1993), and the ability of a slipped teloblast lineage to integrate out of its normal segmental register with the remainder of the germinal plate argues that segment identity mismatches do not prohibit segmental organogenesis.

There is, however, direct experimental evidence that segment identity can influence the assembly process. The mesodermal M teloblast lineage realigns itself with the contralateral mesoderm when displaced by a single segment (Gleizer and Stent, 1993), implying that the constituent blast cell clones can discriminate the segment identity of their neighbors, and can reposition themselves accordingly. The slipped ectodermal lineages examined in the present study did not display any obvious propensity for corrective realignment, suggesting that this may be restricted to mesoderm. Alternatively, it may be that blast cell clones are only able to make corrective realignments following a limited slippage that leaves them in close proximity, and presumably direct physical contact, with their normal segmental neighbors.

The formation of a restricted axial domain of LOX2 expression in the leech embryo clearly depends upon the secondary assembly and alignment of cell lineages that have already undergone an independent axial specification. It is therefore prudent to view the development of a polarized A-P body plan as a process of axis construction, with distinct cellular events underlying the primary specification of cellular identities and the secondary morphogenetic rearrangement or sorting of the specified cells. These two phases are readily separable in the development of the leech embryo, but one can also envision a situation in which cell identity specification and morphogenetic rearrangements are going on concurrently within the same field of cells.

The HOM/Hox genes have been shown to play a functional role in the A-P differentiation of the CNS in insects (Ghysen et al., 1985), vertebrates (Carpenter et al., 1993), and nematodes (Wang et al., 1993). The developmental function of these genes has not yet been tested in the leech, but the observed pattern of LOX2 expression has the potential to encode a complicated multisegmental pattern of neuronal differentiation. Within the LOX2 expression domain, each ganglion is a mosaic, with only a subset of neurons displaying immunologically detectable levels of the protein. Moreover, individual neurons express LOX2 over somewhat different segmental domains, with the result that ganglia located near the anterior and posteior boundaries of that domain manifest distinct cellular patterns of LOX2 expression. This segmental heterogeneity is further complicated by the fact that lineally homologous cells in neighboring ganglia often show quantitatively distinct levels of LOX2 expression. If one assumes that the leech HOM/Hox genes do influence the phenotypic differentiation of the neurons in which they are expressed, this highly diversified segmental pattern of LOX2 expression could provide a detailed template for the cellular diversification of individual ganglia located in the posterior midbody region of the CNS.

Some neurons exhibit patterns of LOX2 expression that are unique to a particular body segment. For example, the PV cluster neurons of ganglion M6 express reproducibly higher levels of LOX2 protein than the PV neurons of any other body segment, and the OM7 neuron in ganglion M7 shows a distinctively high level of LOX2 accumulation not found in any of the neurons located at a similar position in other ganglia. These neurons have not yet been characterized with respect to other aspects of their differentiation, and it will be interesting to learn whether they take on differentiated phenotypes and physiological functions that are unique to those particular body segments.

We are indebted to Andreas Galluser for advice on bacterial protein expression, to Duncan Stuart for advice on fluorescent dextrans, to Gabriel Aisemberg for supply of a LOX1 expression vector, and to Bob Ezzell for his savvy with digitized imaging. We are also grateful to Karl Deisseroth for undertaking pilot experiments on ectopic gene expression, and to Rick Gilles and Michelle Perkins for technical assistance. This work was supported by research grants from the NIH (RO1-HD21735) and March of Dimes (1-FY93-0088). D. N. H. was supported by a postdoctoral grant from the Fonds National Suisse de la Recherche Scientifique, and A. B. was supported in part by the Harvard Mahoney Neuroscience Institute.