Ultrabithorax confers spatial identity in a context-specific manner in the Drosophila postembryonic ventral nervous system

Background In holometabolous insects such as Drosophila melanogaster, neuroblasts produce an initial population of diverse neurons during embryogenesis and a much larger set of adult-specific neurons during larval life. In the ventral CNS, many of these secondary neuronal lineages differ significantly from one body segment to another, suggesting a role for anteroposterior patterning genes. Results Here we systematically characterize the expression pattern and function of the Hox gene Ultrabithorax (Ubx) in all 25 postembryonic lineages. We find that Ubx is expressed in a segment-, lineage-, and hemilineage-specific manner in the thoracic and anterior abdominal segments. When Ubx is removed from neuroblasts via mitotic recombination, neurons in these segments exhibit the morphologies and survival patterns of their anterior thoracic counterparts. Conversely, when Ubx is ectopically expressed in anterior thoracic segments, neurons exhibit complementary posterior transformation phenotypes. Conclusion Our findings demonstrate that Ubx plays a critical role in conferring segment-appropriate morphology and survival on individual neurons in the adult-specific ventral CNS. Moreover, while always conferring spatial identity in some sense, Ubx has been co-opted during evolution for distinct and even opposite functions in different neuronal hemilineages.


Background
The insect ventral CNS, like the body as a whole, is built on a plan of repeating segmental units that then undergo regional specialization. The neurons of a segmental unit arise from a stereotyped two-dimensional array of 30 uniquely identifiable neural stem cells (neuroblasts, NB) per hemisegment [1][2][3]. These NBs undergo repeated asymmetric divisions, thereby producing a series of ganglion mother cells, GMCs [4], each of which divides to produce a pair of postmitotic daughters [5,6]. These daughters then acquire distinct fates via Notch signaling [7,8]. In insects with complete metamorphosis, like Drosophila melanogaster, the NBs typically have an initial burst of proliferation to generate the neurons of the larval CNS and then later a subset feature an extended proliferative period during larval life, producing most of the neurons of the adult CNS [9,10]. During the postembryonic neurogenic phase, Notch signaling between sibling cells produces two morphologically distinguishable cell types that accumulate to form two distinct hemilineages, one of which may be eliminated by programmed cell death [10].
In the embryo, the NB arrays are almost identical between thoracic and abdominal neuromeres [11], although there are some regional differences in the neurons produced by thoracic versus abdominal homologs [12][13][14]. During the postembryonic neurogenic phase, however, there are dramatic differences between the numbers of thoracic versus abdominal NBs [15]. Within the thorax particular lineages exhibit segment-specific differences in their cellular composition [10].
Given their roles in anteroposterior patterning of the embryonic CNS (reviewed in [16]), the Hox genes are excellent candidates for conferring segmental identity in the postembryonic nervous system. For example, in late stage embryos, Abdominal-A (Abd-A) represses proliferation of many NBs in the abdomen [17], and a burst of abd-A expression causes the apoptosis of persistent abdominal lineages during the third instar [17,18]. Also, Ultrabithorax (Ubx) represses the formation of leg neuropils in the first abdominal segment (A1) [19], and in Ubxanimals, thoracic-specific NBs are retained in the A1 neuromere during the postembryonic neurogenic period [20].
The development of methods to label and manipulate NB lineages [21] has allowed the detailed characterization of the postembryonic lineages that generate the adultspecific neurons [9,10]. Using these methods we find that Ubx is expressed in a segment-, lineage-, and siblingspecific manner that correlates with morphological differences observed in different segments for particular lineages. Moreover, removal of Ubx from a lineage via the MARCM (mosaic analysis with a repressible cell marker) method results in anterior transformation of its morphology and survival pattern, whereas ectopic expression of Ubx results in posterior transformation. Interestingly, Ubx can promote survival, death, and/or segment-specific changes in neurite morphology, depending on the hemilineage. Taken together, these data demonstrate that Ubx has been co-opted during evolution to regulate the segmental identity of secondary neurons in a contextdependent manner during development.

Results
Overview of Ubx expression in the larval nervous system As initially described by White and Wilcox [22], the major domain of Ubx expression in the embryonic CNS is parasegment 6 ( Figure 1A), with weaker expression in parasegment 5 (posterior T2 and anterior T3) and an isolated cluster of neurons in the midline of parasegment 4. Posterior to A1, Ubx expression is weak and spotty but still occurs in some neurons through A7. Within parasegment 6, the great majority of the neurons show strong Ubx expression ( Figure 1A').
By the end of the last (third) larval stage, the larval neurons have been joined by clusters of secondary neurons. The former are in a compact layer next to the neuropil, whereas the latter are in superficial clusters that extend from the larval neuron layer to the surface of the CNS. The larval neurons show the same pattern of Ubx expression as seen at hatching ( Figure 1B' , C). Ubx expression in the postembryonic lineages is also mostly confined to parasegments 5 and 6, with that expression in the latter being stronger. However, unlike in the larval neurons, Ubx expression in the clusters of postembryonic-born neurons was quite heterogeneous, even in parasegment 6. The NBs and GMCs did not express Ubx, but Ubx expression within the associated cluster of postembryonic neurons varied from cluster to cluster ( Figure 1B,B'), suggesting a lineage-based regulation. The T2 lineages that exhibit any Ubx expression are 0, 3, 11, 12, and 19, all of which are in the engrailed domain (JWT & D.W. Williams, unpublished work) and, thus, in the anterior portion of parasegment 5. These lineages express much higher levels of Ubx in T3 but fail to express it in A1 (parasegment 7). Ubx expression in the postembryonic lineages is summarized in Figure 2, and examples of expression patterns for the positive lineages are given in the following figures.
Lineages that were Ubx positive typically had all Ubx+ cells or roughly equal numbers of Ubx+ and Ubx-neurons. For the latter cases, our Ubx manipulations described below argue that one sibling from the GMC division becomes Ubx+ and the other Ubx-, thereby resulting in roughly equal numbers of the two expression types. There were a few lineages in which Ubx expression appeared not to be divided along hemilineage lines.
Lineage 12 in segment T3 and lineage 1 in A1 both had a few Ubx+ cells apically, near the NB and GMCs. However, as shown below, in both lineages Ubx expression is responsible for the death of the neurons of one hemilineage, and the cells that we observed were the newly-born cells that had not yet initiated programmed cell death. Expression patterns that were clearly not hemilineage related were seen for the largely negative lineages 8, 15, and 23. Each had one to a few weakly Ubx+ neurons in T3, but our Ubx manipulations did not reveal a role for this expression.
Ubx regulates segment-specific neuronal programmed cell death of particular hemilineages Lineage 1 Lineage 1 provides a striking example of segmentspecific survival in the secondary lineages. The neurons in the 1A hemilineage form the contralateral (1c) axon bundle that projects across the anterior ventral (aV) commissure to the contralateral leg neuropil, and those Circles show the relative size and position of the segmental lineages. The two halves of each circle represent the A (Notch ON ) and B (Notch OFF ) hemilineages. Neurons in hemilineages with dashed borders die soon after their birth. (A) Summary of expression of Ubx in wild-type (WT) clones and in clones in which cell death is blocked by mutation of the dronc gene. Ubx expression typically showed a hemilineage restriction and was weak to moderate (pink) or strong (red) depending on parasegment (PS). The Ubx expression in hemilineages that normally died was revealed in droncclones (*). Red ? = no Ubx expression data. (B,C) The effects of loss of Ubx (Ubx-) or ectopic expression of Ubx (Ubx+) in MARCM clones. Only the lineages that were changed by a given treatment are numbered. The changes include the survival or death of hemilineages (solid versus dashed outlines) and alteration in projection patterns (Star). A1: first abdominal segment; ND: no data; PS4-6: fourth to sixth parasegments; T1-T3: first to third thoracic segments; Ubx: Ultrabithorax; WT: wild-type.
We examined the Ubx expression pattern in wild-type (WT) and droncclones. In WT clones in T1 (n = 3/3, data not shown) and T2 (n = 7/7, Figure 3A'), all lineage 1 cells were Ubx-, while approximately half of the cells in T3 clones were weakly positive for Ubx (n = 3/3, Figure 3B'). In A1, all or most cells were Ubx-, with a few cells near the NB occasionally observed to be strongly positive (n = 3/5, Figure 3C'). For droncclones, in which programmed cell death was blocked, the A1 cluster was enlarged and showed a robust 1c bundle consistent with the survival of the 1A sibling (n = 11/12, 2.91 ± 0.56 μm, Figure 3F). Approximately half of the cells in the enlarged cluster were strongly positive for Ubx (n = 9/9, Figure 3F'), suggesting that the 1A sibling in A1 expresses a high level of Ubx prior to dying.

Loss and gain of function experiments confirmed that
Ubx regulates the survival of lineage 1A neurons. Ubxclones in A1 exhibited a robust 1c bundle as well as the expected 1i bundle (n = 23/24, 3.53 ± 0.96 μm, Figure 3I). Interestingly, the ectopic 1c bundle hooked upwards towards T3, rather than taking the expected trajectory towards the posterior part of the segment. A similar phenotype was also seen in dronc-clones for lineage 1 in segment A1. This altered pathway may be due to a lack of leg neuropil target cues in abdominal segments. The misexpression of Ubx in UAS-Ubx clones apparently led to the death of both hemilineages of lineage 1 neurons, regardless of segment. Only a few thoracic lineage 1 clones were observed, and those had few cells and very thin, faint projections, most likely indicating that the neurons were dying (n = 11/12, axon bundle diameter = 1.80 ± 0.27 μm, Figure 3J, K). Thus, a high level of Ubx expression can result in the death of the neurons of both hemilineages, although only hemilineage 1A neurons normally express it. Also, although the hemilineage 1A neurons in T3 normally express a moderate level of Ubx, they die in response to the high levels in the MARCM clones. Therefore, the ability of Ubx to cause the death of these neurons appears to be concentration dependent.
In Ubxclones, the 0A interneurons in T2 (n = 5/5, Figure 6J) and T3 (n = 7/7, Figure 6K) adopted a T1-like morphology, projecting diffusely over the pI commissure. Conversely, with ectopic expression of Ubx, the 0A interneurons in T1 adopted a posterior morphology, now projecting to the aI commissure (n = 13/15, Figure 6M). Taken together, these data demonstrate that Ubx acts to specify the segment-appropriate axon morphology of lineage 0 neurons in the thoracic neuromeres. Consistent with their lack of Ubx expression in WT or droncclones, the axon morphology of A1 clones were unaffected by Ubx manipulation (n = 3/3, Figure 6L).
Ubxclones in T3 looked normal (n = 9/13, Figure 7G), but those in A1 were larger and exhibited a much more robust and diffuse 9i process as compared with WT (n = 4/5, Figure 7H). Ubxclones in A2 either featured more robust 9i projections traveling along with 9c (n = 8/14) or an additional, more dorsal contralateral projection similar to those seen in dronc -clones (n = 7/14, Figure 7I). For thoracic UAS-Ubx MARCM clones, the 9i bundle was either absent (n = 2/15 for T3, not shown) or remained closely associated with the 9c bundle and lacked the characteristic medial "hook" (n = 11/15 for T3, Figure 7J). These results suggest that for lineage 9, Ubx regulates both cell survival and axon guidance in A1 and A2.

UAS-Ubx in A1
Misexpression of Ubx caused the death of both 12A and 12B siblings in all three thoracic segments, as evidenced by dramatically thinned and/or absent projections (T1: n = 17/17, Figure 8M; T2: n = 4/4, Figure 8N; T3: n = 8/8, Figure 8O) and no 12i bifurcation. The characteristic position of the lineage 12 bundle in the neurotactin scaffold relative to lineages 3 and 6 permitted unequivocal identification of this bundle even in the absence of CD8-GFPlabeled processes (Figures 8M' , N' , O'). Where the clone should reside, we often saw a small cluster of cells with truncated or no projections. In such preparations the neurotactin-positive bundles (12i and 12c) were missing, confirming that these cells did not survive. These data strongly suggest that a high level of Ubx expression promotes the death of both lineage 12 siblings, while an intermediate level of Ubx permits survival but controls the segment-specific bifurcation of the 12i bundle.
In addition to these three examples, we also found a possible role for Ubx in determining the axon projections of lineages 3 and 7 ( Figure 2). Overexpression of Ubx in lineage 3 resulted in survival of only the 3id sibling in the thoracic segments (n = 28/43), and the terminal elaborations normally found in T1 were missing (n = 43/43). Overexpression of Ubx in lineage 7 resulted in the axon bundle turning posteriorly instead of anteriorly (n = 16/51) or failing to turn (n = 32/51) after crossing the midline in all three thoracic segments (data not shown). However, given that there was no abnormal phenotype in Ubxclones for either lineage, we cannot conclude definitively that Ubx normally regulates axon guidance in these lineages.

Lineage 11
Lineage 11 is normally present in T1 and T2, but not in T3 or in the abdomen [9]. Only the 11A hemilineage producing the 11im bundle is present in T1 (n = 3/3, Figure 9E), while 11id is also present in T2 (n = 7/7, Figure 9F), indicating the additional presence of the 11B hemilineage. Suppression of cell death resulted in the appearance of both hemilineages in T3 (n = 2/3, Figure 9I). Ubx expression in these neurons was weak in T2 (n = 5/6, Figure 9H') but very strong in T3 (n = 3/3, Figure 9I'). The loss of Ubx also resulted in the appearance of lineage 11 clones in T3; however, only 11im was present, with 11id greatly reduced or absent (n = 11/13, Figure 9L). Interestingly, 11id was reduced or absent in T2 Ubxclones as well (n = 12/15, Figure 9K), indicating that a low level of Ubx is required for the survival of the 11B neurons that produce the 11id bundle. We found no lineage 11 clones when we ectopically expressed Ubx in the UAS-Ubx experiments. Based on these data, Ubx is responsible for the lack of lineage 11 neurons in T3 and of lineage 10 neurons in A1. One possibility is that Ubx causes the death of both siblings right after they are born. However, in WT individuals, we have not seen any indications of these lineages in the respective segments, but we typically look at the end of larval life, so the NB and all of its progeny may have died by that time. Alternatively, Ubx expression may cause the death of the NB itself. Previously it was reported that postembryonic NB survival in the ventral CNS is governed by the mutually exclusive expression and antagonistic functions of Antennapedia and Abd-A but not Ubx [17]. The NBs that give rise to lineage 10 in A1 and to lineage 11 in T3 may be the exceptions to this rule.

Discussion
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 [23] (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 adultspecific 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 Notch ON or Notch OFF 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).

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 (Notch OFF ) sibling requires Ubx for survival, whereas its A (Notch ON ) 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 [25]. 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 [26], 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 [29]. 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 [30]. 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 [31]. 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].

Conclusions
We have found the Hox gene Ubx to be a key regulator of anteroposterior patterning in the postembryonic ventral nervous system of Drosophila melanogaster. In the larva, Ubx is not expressed homogenously within its general domain (parasegments 5 and 6), but rather in specific NB lineages and even hemilineages. Ubx is both necessary and sufficient for many of the segmentspecific differences in NB lineage morphology observed in previous studies. Moreover, Ubx acts in a contextdependent manner, promoting programmed cell death, promoting cell survival, and/or regulating axon morphology, depending on the hemilineage. In some hemilineages, the function of Ubx is segment-specific and appears to depend on the level of expression. Thus, Ubx has been co-opted during evolution for multiple roles in sculpting the postembryonic ventral nervous system in a segment-appropriate manner.

Fly stocks
This study employed the MARCM (mosaic analysis with a repressible cell marker) system, in which the FLP/FRT method is used to generate clones of cells lacking GAL80, a suppressor of GAL4 [21]. This allows a reporter gene, UAS-mCD8::GFP, to be driven exclusively in cells that are homozygous mutant for a gene of interest or in cells expressing an additional UAS-transgene of interest.

Generation of MARCM clones
To generateWT control and UAS-Ubx MARCM clones, eggs of the appropriate genotype were collected for 12 h on grape agar plates at 25°C, incubated at 25°C for 12 h, and then heat-shocked at 37.5°C. Treatment was similar for droncclones except that eggs were collected for 24 h and then heat-shocked immediately. To generate Ubx -MARCM clones, eggs were collected for 12 h on grape agar plates at 25°C and then heat-shocked immediately. Following heat shock, embryos or larvae were transferred to instant fly media (Carolina Biological Supply, Burlington, NC, USA) and reared at 29°C to increase expression of the GAL4 C155 driver and visibility of MARCM clones.

Production of the anti-Ultrabithorax antiserum
To generate Ubx antibody 7701, we first expressed a fusion-antigen using the cDNA clone RE43738 received through the Drosophila Genomics Resource Center. The sequence chosen for its low complexity and low paralogy to other genomic regions was ATGAACTCGTACTTTG AACAGGCCTCCGGCTTTTATGGCCATCCGCACCAG GCCACCGGAATGGCAATGGGCAGCGGTGGCCAC CACGACCAGACGGCCAGTGCAGCGGCGGCCGCG TACAGAGGATTCCCTCTCTCGCTGGGCATGAGTCC CTATGCCAACCACCATCTGCAGCGCACCACCCAGG ACTCGCCCTACGATGCCAGCATCACGGCCGCCTGC AACAAGATATACGGCGATGGAGCCGGAGCCTACAA ACAGGACTGCCTGAACATCAAGGCGGATGCGGT GAATGGCTACAAAGACATTTGGAACACG. The corresponding protein region is MNSYFEQASGFYGHPHQA TGMAMGSGGHHDQTASAAAAAYRGFPLSLGMSPY ANHHLQRTTQDSPYDASITAACNKIYGDGAGAYKQ DCLNIKADAVNGYKDIWNT, which corresponds to position 1-106 of the Ubx protein.
The complementary DNA sequence was cloned by LIC cloning within the pMCSG19 vector [38] to express and purify the antigen as described. The antigen was injected in rabbits by Millipore using their standard polyclonal protocol. Sera were then used to test antibody specificity by fluorescent immunostaining on Drosophila embryos ( Figure 10).

Immunohistochemistry
Nervous systems were dissected from wandering third instar larvae in PBS (pH 7.2), fixed in 3.7% buffered formaldehyde at room temperature, and then washed in 0.3% PBS-TX (PBS with 0.3% Triton-X100). Fixed samples were blocked in 2% normal donkey serum (Jackson Immunoresearch Laboratories, West Grove, PA, USA) in PBS-TX for 30 min and then incubated with rat anti-mCD8 (Caltag/Invitrogen, Grand Island, NY, USA) at 1:100 and either rabbit anti-Ubx at 1:1000 (7701) (Figure 10; [39]) or mouse monoclonal anti-neurotactin (BP106, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA) at 1:30 for several days at 4°C. After primary antibodies were washed out, tissues were incubated overnight at 4°C in a 1:300 dilution of fluorescein isothiocyanate-conjugated donkey anti-rat IgG and Texas Red-conjugated donkey antimouse IgG (Jackson Immunoresearch Laboratories). After additional washes, tissues were mounted on polylysine coated coverslips, dehydrated through an ethanol series, cleared in xylene, and mounted in DPX (Sigma-Aldrich, St. Louis, MO, USA).

Microscopy and image processing
Fluorescently stained nervous systems were imaged using either a 63× oil objective on a BioRad MRC 1024 confocal microscope or a 40× oil objective on a Leica SP5 Spectral Systems confocal microscope. Z-stacks were collected sequentially with averaging at 0.5 to 1.0 μm intervals.
Raw data stacks were imported into ImageJ (http:// rsbweb.nih.gov/ij/) using a Bio-Formats plug-in (LOCI, University of Wisconsin, Madison, WI, USA) and either merged or projected into three-dimensional representations for analysis. Lineages were identified based on morphology, NB location, and/or projections into the neurotactin scaffold with reference to our published atlas [9,10].
Images are shown as single views of three-dimensionally reconstructed and rotated confocal stacks or as single optical slices, as indicated. Confocal stacks were processed and assembled into figures using ImageJ, Microsoft Powerpoint, and the Adobe Photoshop Suite. Multiple clones are typically labeled in the same sample and often obscure one another in a simple projection. For clarity, individual clones were cropped out in their entirety and adjusted for brightness and contrast. Only the slices featuring the most relevant portion of the neurotactin axon scaffold were projected and shown as landmarks.
For selected lineages, we made two-dimensional z-projections in ImageJ to measure the diameter of the axon bundle that exited the cell cluster. We used the *straight* line-drawing tool in combination with the plot profile macro to measure the diameter of the bundle as it crossed the midline.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions ECM conceived of the study, participated in the design and coordination, carried out the Ubx loss of function experiments, assisted with clonal analysis for all MARCM experiments, created figures, and drafted the manuscript. KED carried out and analyzed the Ubx gain of function experiments, assisted by ARC. DRA, KTR, and MEC carried out the wild-type and dronc -Ubx expression experiments and assisted with data collection and analysis. NN and KPW generated and characterized the 7701 antibody to Ubx, created the relevant figure, and critiqued the manuscript. JWT participated in the design and coordination, carried out the early larval expression experiment, created figures, and helped to draft the manuscript. All authors read and approved the final manuscript.