Ultrabithorax expression patterns in the embryonic and postembryonic ventral nervous system
Metamorphosis in Drosophila brings about a profound change in body form. Although the thoracic and abdominal segments have relatively similar morphologies in the larval body plan, they then diverge dramatically in the adult body plan. Within the adult thorax, there are additional segmental specializations to accommodate segment-constant features (legs) and segment-variable features (wings and halteres). These segmental specializations are sculpted by the Hox genes, with Ubx being the major gene effecting differences within the thorax. The difference in the complexity of the body of the larva versus the adult is paralleled by a difference in the complexity of Ubx expression during embryonic and postembryonic development of the nervous system.
At hatching, Ubx expression is observed in most neurons in parasegments 5 and 6, with expression in the latter being the stronger  (Figure 1). This expression pattern appears to be stable in the embryonic-born neurons throughout larval growth. We have found Ubx expression in the lineages of adult-specific neurons to be quite heterogeneous. With the exception of anterior expression in the median lineage, and extended posterior expression in lineage 9, Ubx expression is still confined to the lineages in parasegments 5 and 6 (posterior T2 through anterior A1). Each lineage, though, develops as an autonomous unit, and each adopts a characteristic pattern of Ubx expression, with the postembryonic neurons in a cluster being either all Ubx+, all Ubx-, or mixed. In the case of mixed expression, Ubx expression is typically restricted to one hemilineage or the other, although it may be found in either the NotchON or NotchOFF set of siblings (Figure 2).
For a given lineage or hemilineage, expression in parasegment 6 was higher than in parasegment 5 (Figures 3B,C, 4B,C, 5B,C, 6B,C, 8B,C and 9I,J). More extreme segmental differences in expression were seen for hemilineages 9A and 17A, in which there was no expression in parasegment 5 but strong expression in parasegment 6 (Figure 2). There were no cases of ‘flip-flopping”, in which Ubx was expressed in one hemilineage in parasegment 5 but in the other hemilineage in parasegment 6. In most cases, the different levels of expression we observed caused segment-specific differences in neuron survival and/or morphology (for example, hemilineages 1A, 6B, 12A and 11B).
The expression of Ubx was highly correlated with whether or not a given hemilineage contributed to a segment-constant (leg neuropil) versus a segment-variable (flight neuropil) portion of the CNS. Most of the hemilineages or lineages that contribute to the leg neuropils were Ubx- (3A, 4, 8, 9, 13, 14, 15, 16, 20, 21, 22 and 24), the major exception being hemilineage 1B, in which Ubx causes the death of inappropriate leg interneurons in segment A1. Hemilineages that contribute to the dorsal flight neuropil, by contrast, were typically Ubx+ (0A, 3B, 6, 7B, 11B, 12A and 19B).
Multiple roles for Ubx in neuronal production and differentiation
In the postembryonic nervous system, positional information conferred by Ubx has dramatically different consequences depending on hemilineage. As might be inferred from the many cases in which neurons that are present in anterior thoracic segments are absent from T3 or A1, Ubx expression promotes programmed cell death in numerous hemilineages including 1A, 6B, 9A, 12A and 19A. By contrast, for hemilineages 19B and 11B, neurons are missing from an anterior segment (T1), and in these cases Ubx is required for hemilineage survival (Figures 5 and 9). Lineage 19 strikingly shows this dichotomy of context dependence since its B (NotchOFF) sibling requires Ubx for survival, whereas its A (NotchON) sibling is killed by Ubx expression. Besides being involved in the selective death or survival of hemilineages, Ubx also regulates the segment-specific survival of whole lineages, as in the case of lineage 11 in T3 and lineage 10 in segment A1 (Figure 9), although, as discussed earlier, we cannot be certain whether this is executed at the level of the postmitotic neurons or the NBs themselves. While the latter two lineages make use of Ubx expression to remove lineages from inappropriate segments, many of the leg-related lineages have shut off Ubx expression to insure the survival of their neurons in the normal Ubx domain of expression. Therefore, we see that the ectopic expression of Ubx in these lineages results in their death, independent of segmental location (Figures 2, 3J,K and 8M-P).
We also see that Ubx expression can regulate segment-specific morphology without affecting cell survival. For example, the median lineage 0A neurons normally project to the pI commissure and elaborate their processes in T1 (Figure 6A), but they project past that point to the aI commissure in segments T2 to A1 (Figure 6B-D). These differences persist when cell death is blocked (Figure 6E-H). Therefore, the loss or gain of Ubx function is able to alter axon guidance and target recognition, presumably due to segment-inappropriate expression of signal transduction pathway components.
Besides lineage- and hemilineage-restricted patterns, we saw examples such as lineages 8, 15 and 23 in which Ubx expression was confined to only two to three cells in the cluster. Because our analysis was at the cluster level, we could not determine whether the loss of Ubx resulted in the death of this small number of cells. Also, the low-level Ubx expression in these cells might direct later events that occur as the neurons mature during metamorphosis.
We conclude that the effects of Ubx expression are not universal for secondary lineages but instead are lineage- and even hemilineage-dependent, implying independent co-option of Ubx by distinct mechanisms of regulation. This ability to act as a micromanager rather than a global switch would also allow Ubx to sculpt numerous species-specific differences in nervous system development during insect evolution without disruption of the largely conserved neuroblast array [23, 24].
Candidate mechanisms for context-dependent Ubx function
In the postembryonic CNS, Ubx carries out such diverse downstream functions as promoting NB or neuron death, promoting cell survival, and influencing axon guidance. These context-specific responses could be mediated through a number of different mechanisms: for example, expression levels governing threshold-dependent processes, alternative splicing, and the presence of specific cofactors and/or collaborators, any of which could influence DNA binding specificity and/or activation versus repression of gene targets.
Levels of Ubx expression are known to be important to developmental patterning. For example, in the Drosophila leg, gene dosage contributes to species-specific bristle patterns . Moreover, low levels of Ubx are sufficient to repress an eighth bristle row on the posterior femur in T2 and T3, but higher levels are required for the repression of trichomes , suggesting that different levels of Ubx are required for distinct functions during development.
Similar differential responses to different levels of Ubx also appear to be in play for the postembryonic lineages since we see several cases in which neurons express Ubx at low levels in one segment but die in response to higher levels in the next. The best example is the 12A hemilineage, which makes the 12id and 12im axon bundles in T1 and the 12id (but not always 12im) bundle in T2, and dies in T3. These differences are associated with three levels of Ubx expression in this hemilineage: none in T1, intermediate levels in T2, and high levels in T3. Importantly, when the 12A neurons in T3 are rescued by dronc mutants, they produce only the 12id bundle, but when rescued by removal of Ubx, both the 12id and the 12im branches are produced. This suggests that low levels of Ubx prevent formation of the 12im branch, but high levels cause cell death.
The six Ubx isoforms are believed to have distinct roles in regulating target gene expression in different tissues during development [27, 28], but it has long been standard practice to use a single isoform for overexpression studies. While our loss of function experiments were carried out using a null allele of Ubx, eliminating all possible isoforms, the gain of function experiments used a transgene constructed from the Ia isoform . We nonetheless observed overexpression phenotypes that appeared to be the opposite of those from loss of function experiments, suggesting that Ia can substitute for the isoforms normally expressed in the CNS. This is consistent with reports that expressing any isoform ectopically at high enough levels can compensate for the lack of the normal one . Nonetheless, it remains an untested possibility that different isoforms of Ubx are responsible for distinct functions in the postembryonic lineages.
A final possibility is that different Hox cofactors and/or collaborators are present in the cells of different NB lineages. Cofactors of the Pbx/Meis family (Extradenticle and Homothorax in Drosophila) are TALE homeodomain proteins that bind DNA cooperatively with Hox genes to increase target specificity in vivo, reviewed in . The homeodomain protein Engrailed, which is differentially expressed between the siblings of at least some lineages (JWT, unpublished work), has also been shown to be a Hox cofactor and appears to be involved specifically in target gene repression via recruitment of the co-repressor Groucho [32, 33]. In addition, other transcription factors such as Teashirt and Sloppy paired appear to function as Hox collaborators at Hox-targeted cis regulatory elements [32, 34, 35].