Hox Control of Organ Size by Regulation of Morphogen Production and Mobility

Abstract

Selector genes modify developmental pathways to sculpt animal body parts. Although body parts differ in size, the ways in which selector genes create size differences are unknown. We have studied how the Drosophila Hox gene Ultrabithorax (Ubx) limits the size of the haltere, which, by the end of larval development, has ∼fivefold fewer cells than the wing. We find that Ubx controls haltere size by restricting both the transcription and the mobility of the morphogen Decapentaplegic (Dpp). Ubx restricts Dpp’s distribution in the haltere by increasing the amounts of the Dpp receptor, thickveins. Because morphogens control tissue growth in many contexts, these findings provide a potentially general mechanism for how selector genes modify organ sizes.

Changes in body part sizes have been critical for diversification and specialization of animal species during evolution. The beaks of Darwin’s finches provide a famous example for how adaptation can produce variations in size and shape that allowed these birds to take advantage of specialized ecological niches and food supplies (1). Sizes also vary between homologous structures within an individual. For example, vertebrate digits and ribs vary in size, likely due to the activities of selector genes such as the Hox genes (2-4). Although the control of organ growth by selector genes is likely to be common in animal development (2, 5, 6), little is known about the mechanisms underlying this control.

The two flight appendages of Drosophila melanogaster, the wing and the haltere, provide a classic example of serially homologous structures of different sizes (Fig. 1A). Halteres, appendages used for balance during flight, are thought to have been modified from full-sized hindwings during the evolution of two-winged flies from their four-winged ancestors (7, 8). All aspects of haltere development that distinguish it from a wing, including its reduced size, are under the control of the Hox gene Ultra-bithorax (Ubx),which is expressed in all haltere imaginal disc cells but not in wing imaginal disc cells (8, 9) (Fig. 1B). At all stages of development, haltere and wing primordia (imaginal discs) are different sizes. In the embryo, the wing primordium has about twice as many cells as the haltere primordium (7, 10). By the end of larval development, the wing disc has ∼five times more cells (∼50,000) than the haltere disc (∼10,000) (11) [Fig. 1B and Supporting Online Material (SOM) Text]. The wing and haltere appendages will form from the pouch region of these mature discs (fig. S1). The final step that contributes to wing and haltere size differences occurs during metamorphosis, when wing, but not haltere, cells flatten, thus increasing the surface area of the final appendage (12).

Fig. 1.

Reduced Dpp production and transduction in the haltere. (A) Wild-type adult wing and haltere (arrow). (B) Third instar wing (W), haltere (H), and T3 leg (L) imaginal discs stained for Ubx (green) and a ubiquitous nuclear protein (blue). Ubx is present in all haltere disc cells but not in wing disc cells. (C) Removing Ubx activity (lack of GFP) from more than 50% of a haltere disc during larval development using the M⁺(Minute) technique (13) (SOM Text) increased its size [compare with discs in (B) and (F)]. (D)Isolated Ubx^- clones (black, -/-) were not larger than Ubx⁺ twin spots (bright white, +/+)ina Ubx heterozygous haltere (gray, -/+). (E) Ubx mutant:twin spot and neutral:twin spot clone size ratios. Error bars indicate SEM. (F to H) Wild-type wing and haltere discs stained for dpp-lacZ and P-Mad patterns. In the haltere, dpp-lacZ was reduced (arrowheads) and overlapped with a compacted P-Mad gradient (arrows).

Nonautonomous control of haltere size by Ubx

To confirm that Ubx has a postembryonic role in limiting the size of the haltere disc, we generated Ubx^- clones midway through larval development (13). Haltere discs—bearing large Ubx^- clones generated at this time become much larger than wild-type discs (Fig. 1C and SOM Text). Ubx could limit haltere size cell-autonomously by, for example, slowing the cell cycle of haltere cells relative to wing cells. We tested this by comparing the sizes of isolated Ubx^- clones in the haltere with those of their simultaneously generated wild-type twin clones. Contrary to the prediction of a cell-autonomous function for Ubx in size control, Ubx mutant clones did not grow larger than their twins (Fig. 1, D and E), a result that is consistent with earlier experiments suggesting that wing and haltere cells have similar mitotic rates during development (14). Hence, Ubx limits the size of the haltere during larval development by modifying pathways that control organ growth cell-nonautonomously.

Ubx regulation of Dpp signaling

In the fly wing, Decapentaplegic (Dpp) [a long-range morphogen of the bone morphogenetic protein (BMP) family] has been shown to promote growth (15-17). In both the wing and the haltere, Dpp is produced and secreted from a specialized stripe of cells called the AP organizer, which is induced by the juxtaposition of anterior (A) and posterior (P) compartments, two groups of cells that have separate cell lineages (18). The AP organizer is a stripe of A cells that are instructed to synthesize Dpp by the short-range morphogen Hedgehog (Hh) secreted from adjacent P compartment cells (18-22). Dpp has a positive role in appendage growth. When more Dpp is supplied to the wing disc,either ectopically or within the AP organizer, more cells are incorporated into the developing wing field (22-24). Conversely, mutations that reduce the amount of Dpp lead to smaller wings (fig. S3) (25).

A comparison of the expression patterns of Dpp pathway components in the wing and the haltere demonstrates that Ubx is modifying this pathway (Fig. 1, F to H, fig. S1, and SOM Text). Compared with the wing, the stripe of dpp expression in the haltere was reduced in both its width and intensity, as reported by a lacZ insertion into the dpp locus (dpp-lacZ). There was also a difference in the profile of Dpp pathway activation, as visualized by an antibody that detects P-Mad, the activated form of the Dpp pathway transcription factor Mothers against Dpp (Mad). In the wing, P-Mad staining was low in the cells that transcribe dpp (Fig. 1, F to H, and fig. S1) (26). Immediately anterior and posterior to this activity trough, P-Mad labeling peaked in intensity and then gradually decayed further from the Dpp source, revealing a bimodal activity gradient. In contrast, in the haltere intense P-Mad staining was detected only in a single stripe of cells that overlaps with Dpp-producing cells of the AP organizer (Fig. 1, F to H, and fig. S1).

Because of the coincidence between dpp transcription and peak P-Mad staining in the haltere, we hypothesized that Dpp might be less able to move from haltere cells that secrete this ligand. We tested this idea by generating clones of cells in both wing and haltere discs in which the actin5c promoter drove the expression of a green fluorescent protein (GFP)—tagged version of Dpp (Dpp:GFP) (13, 27, 28). By using an extracellular staining protocol to analyze simultaneously generated clones (29), we observed Dpp:GFP and P-Mad much further from producing cells in the wing than in the haltere (Fig. 2, A to D). These observations strongly suggest that, compared with the wing, Dpp’s mobility—and consequently the range of Dpp pathway activation—is reduced in the haltere.

Fig. 2.

Reduced Dpp mobility in the haltere. (A to D) Simultaneously generated actin5c promoter flp-out clones expressing UAS-dpp::gfp and UAS-GFP in the wing and haltere stained for extracellular GFP (red and white) and P-Mad (blue). The green channel shows GFP auto-fluorescence and marks the clone. The extracellular Dpp::GFP pattern closely correlates with the P-Mad pattern. In (C) and (D), enlarged images of the regions boxed in (A) are shown. (E and F) Overexpressing dpp with ptc-Gal4 (visualized with UAS-GFP>) increased the scale and intensity of P-Mad staining in the wing and the haltere, but the patterns remained qualitatively similar to those of wild type.

We also tested whether the decreased production of Dpp in the haltere contributes to the different pattern of pathway activation observed in this tissue compared with the wing. This is unlikely because, even in haltere discs that overexpress Dpp in its normal expression domain, peak P-Mad staining was still observed close to Dpp-expressing cells (Fig. 2, E and F) (13). Despite increased dpp expression, no P-Mad activity trough was observed in these haltere discs. Further, although they become larger, these discs remained smaller than wild-type wing discs. We conclude that the decreased Dpp production in the haltere contributes to its reduced growth, but there must be mechanisms that also limit the extent of Dpp pathway activation, even in the presence of increased Dpp production.

One way in which Dpp’s activation profile can be modified is by varying the production of the type I Dpp receptor, Thickveins (Tkv) (26, 30). In the wing, tkv expression is low within and around the source of Dpp, resulting in low Dpp signal transduction in these cells and robust Dpp diffusion (26, 30, 31) (Fig. 3, A and B, and fig. S1). Low tkv expression in the medial wing is due to repression by both Hh and Dpp (26, 30). Accordingly, tkv expression is highest in lateral regions of the wing disc, where Hh and Dpp signaling are low. In contrast to the wing, tkv transcription and protein levels were high in all cells of the haltere (Fig. 3, A and B). Thus, the more restricted Dpp mobility and P-Mad pattern in the haltere may result from a failure to repress tkv medially. To test this idea, we supplied all cells of the wing disc with uniform UAS-tkv⁺ expression, to mimic the haltere pattern (Fig. 3C) (13). The resulting P-Mad pattern in these wing discs was very similar to that found in the wild-type haltere: The P-Mad trough was gone, and the activity gradient was compacted into a single stripe that coincides with Dpp-producing cells. Conversely, lowering the amount of Tkv in the haltere by expressing an RNA interference (RNAi) hairpin construct directed against tkv (UAS-tkvRNAi) in Dpp-producing cells induced a bimodal pattern of P-Mad staining similar to that of the wild-type wing disc (Fig. 3, D to F) (13). Thus, different amounts of Tkv result in qualitative differences in the P-Mad profiles of the wing and the haltere.

Fig. 3.

Tkv production influences Dpp signaling and appendage size. (A) tkv-lacZ expression was high throughout the haltere, whereas in the wing it was low medially and high laterally. (B) Tkv antibody staining showed a pattern similar to that of the tkv-lacZ enhancer trap, with less resolution. (C) Driving uniform UAS-tkv⁺ expression in the wing using tubGal4 compacted the Dpp activity gradient and created haltere-like P-Mad staining pattern (arrow). (D to F) Expressing UAS-tkvRNAi in the haltere using ptc-Gal4 (visualized with UAS-GFP) reduced Tkv staining [yellow arrow in (D) and (E)] and resulted in a bimodal P-Mad staining pattern [yellow arrowheads in (F), which shows a magnification of the region boxed in (D)]. (G) Adult wings uniformly expressing UAS-tkv⁺ using tubGal4 were ∼30% smaller than control wings. (H) Quantification of wing size reduction caused by uniform UAS-tkv⁺ expression (orange) compared to controls (green). Error bars indicate SEM. (I) Adult halteres uniformly expressing UAS-tkvRNAi using vg-tubGal4 were up to 60% larger than control halteres. (J) Quantification of haltere size increase caused by uniform tkvRNAi expression (orange) compared to controls (green). The average increase seen is 46%.

tkv expression and appendage size

We hypothesized that the more limited pathway activation in the haltere might contribute to its smaller size. If correct, increasing tkv expression in the wing should reduce its size. Adult wings from flies expressing uniform UAS-tkv⁺ were ∼30% smaller than control wings; however, wing cell size remained the same (Fig.3,G and H, and fig. S2) (13, 30). Similar results were seen in staged imaginal discs and when UAS-tkv⁺ expression was limited to the wing and the haltere (fig. S2). Conversely, reducing Tkv amounts by uniformly expressing UAS-tkvRNAi in wings and halteres increased haltere size by 30 to 60% (Fig. 3, I and J). In a complementary experiment, we reduced tkv transcription in the haltere by expressing a known tkv repressor, master of thickveins (mtv)(32). In this experiment, we measured haltere discs instead of the adult appendage and found, consistently, that the appendage-generating region of these discs increased in size by ∼40% (fig. S2). Thus, different amounts of Tkv not only affect Dpp pathway activation but also affect organ size. The fact that manipulating only Tkv production does not fully transform the sizes of these appendages suggests that additional mechanisms, such as the reduced amounts of dpp transcription and the modulation of other morphogen pathways by Ubx, also contribute to size regulation. Consistently, when Dpp production is decreased in wing discs that uniformly express UAS-tkv⁺, wing size was reduced more than it was by either single manipulation (fig. S3).

Ubx regulation of tkv

We next address how Ubx up-regulates tkv in the haltere. tkv-lacZ expression and amounts of Tkv protein were cell-autonomously reduced in medial Ubx^- clones, whereas lateral Ubx mutant tissue retained high amounts of Tkv (Fig. 4, A to D, and fig. S4). Because tkv is repressed by Dpp and Hh signaling in the wing (26, 30), these results suggest that, in the haltere, these signals are not able to repress tkv. Consistently, activation of the Dpp pathway by expressing a constitutively active form of Tkv (Tkv^QD) resulted in cell-autonomous tkv-lacZ repression in the wing pouch (Fig. 4, E and F), whereas repression is not observed in the corresponding region of the haltere disc (Fig. 4, G and H).

Fig. 4.

Dpp and Ubx collaborate to repress a tkv repressor in the haltere. (A to D) Ubx mutant tissue in the medial haltere (absence of GFP) shows a cell-autonomous reduction in tkv-lacZ and reduced P-Mad staining. High P-Mad and tkv-lacZ staining can be detected in a Ubx⁺ island (yellow arrow) that is separated from Dpp-producing cells by Ubx^- tissue (*). The approximate position of the AP boundary is indicated by a white arrow in (D). (E to H)Clones expressing UAS-tkv^QD (marked with GFP) repress tkv-lacZ in the wing pouch (cyan arrow) but not in the analogous domain of the haltere (yellow arrowheads). (I and J) Wild-type wing and haltere discs stained for mtv-lacZ and P-Mad. mtv-lacZ is strongly expressed in Dpp-producing cells of the wing (cyan arrow) but is repressed in Dpp-producing cells of the haltere (yellow arrow). (K and L) Clones expressing UAS-tkv^QD (marked with GFP) strongly repress mtv-lacZ in the haltere (yellow arrows). Similar clones in the wing repress mtv-lacZ moderately in the P compartment (cyan arrow) and not at all in the A compartment (cyan arrowheads).

In Ubx mosaic haltere discs, we also found that medial Ubx⁺ tissue showed stronger P-Mad staining than Ubx^- tissue at the same distance from the Dpp source (Fig. 4, A to D). We interpret this observation as evidence that Ubx⁺ tissue is more effective at trapping and transducing Dpp than Ubx^- tissue because of higher Tkv production in Ubx⁺ cells.

To further understand the control of tkv by Ubx, we examined the known tkv repressor, mtv (32). In medial wing disc cells, mtv expression is approximately complementary to tkv expression (Fig. 4, I and J, and fig. S1), and mtv^- clones in this region of the wing disc cell autonomously derepressed tkv (fig. S4) (32). In the haltere, very low mtv-lacZ expression was detected in the cells that stained strongly for P-Mad, suggesting that mtv is repressed by Dpp in this appendage (Fig. 4, I and J). Accordingly, strong repression of mtv-lacZ was seen in UAS-tkv^QD-expressing haltere pouch clones, whereas weak or no repression was seen in analogous wing clones (Fig. 4, K and L). We also found that, as expected, Ubx^- clones in the medial haltere cell autonomously derepressed mtv-lacZ (fig. S4).

In the wing, Dpp and mtv are mandatory repressors of tkv: In the absence of either, tkv expression is high. In the haltere in the presence of Ubx, Dpp is a repressor of mtv. Consequently, high levels of these obligate tkv repressors (Dpp signaling and mtv) do not coexist in the haltere, resulting in tkv derepression. Consistent with this model, when we forced mtv expression in the medial haltere, where it coexists with Dpp signaling, it repressed tkv-lacZ (fig. S4). We note, however, that Ubx is likely to control tkv through additional means, because mtv mutant wing clones did not derepress tkv-lacZ expression to haltere levels (fig. S4), and ectopic mtv in the haltere did not repress tkv-lacZ expression to the extent seen in the medial wing (fig. S4).

Control of the relative position of Dpp and Hh signaling by tkv regulation

Because of high Tkv production in the wild-type haltere disc, peak Dpp signal transduction occurs in the AP organizer, the same cells that transduce the Hh signal. Thus, in the haltere, the activity profiles for these two signal transduction pathways coincide with each other (Fig. 1, F to H, and 5I). In contrast, low tkv expression in the wing AP organizer results in two peaks of Dpp signaling that are on either side of Hh-transducing cells. This difference will have important consequences for the expression of genes that are targets of both pathways. For example, dpp is activated by Hh and repressed by Dpp signaling (19-22, 33). In the haltere, these two conflicting inputs occur in the same cells, possibly contributing to reduced dpp expression compared with the wing. Ubx^- clones cell-autonomously up-regulated dpp-lacZ in the haltere (Fig. 5, A and B). To test whether Ubx lowers dpp transcription in part by aligning Dpp and Hh signaling, we expressed uniform UAS-tkv⁺ in the dorsal half of the wing disc. As a result, in this region of the wing disc both signals peaked in the same cells, and dpp-lacZ expression was reduced compared with the ventral half of these wing discs (Fig. 5, C and D, and fig. S5). Conversely, expressing tkvRNAi in dorsal haltere cells increased dpp-lacZ expression (fig. S5). Thus, Ubx reduces dpp transcription in part by changing where peak Dpp signaling occurs in the disc (Fig. 5I). We note that Ubx is likely to reduce dpp expression in additional ways, because increasing tkv expression does not lower dpp-lacZ expression to that observed in wild-type haltere. Nevertheless, varying the relative spatial relationships between signal transduction pathways is a potentially powerful mechanism for modifying the outputs from commonly used pathways. We suggest that selector genes may work through molecules that control ligand distribution to vary the spatial relationships between these and other signal transduction pathways in diverse contexts during development.

Fig. 5.

Contributions of dpp transcription and mobility to growth. (A and B) dpp-lacZ is up-regulated in Ubx mutant haltere tissue (marked by loss of GFP) within the AP organizer. (C and D) UAS-tkv⁺ expression in dorsal cells using ap-Gal4 results in dpp-lacZ down-regulation (thin arrow) compared with ventral cells (thick arrow). (E to G) Examples of +/Ubx^- (E), pbx/Ubx^- (F), and pbx/Ubx^-; actGal4>tkv⁺ (G) haltere discs stained for Nubbin (Nub), a marker of the appendage, and a marker of the AP compartment boundary (yellow line). (H)P:A ratios of the Nub domains of +/Ubx^-, pbx/Ubx^-, and pbx/Ubx^-; actGal4>tkv⁺ haltere + discs. Error bars indicate SEM. (I) Summary of the consequences of different spatial relationships of Dpp and Hh signaling for dpp transcription in the wing and haltere. (J) Summary of how Dpp represses mtv in the presence of Ubx to control tkv expression, Dpp mobility, and growth in the haltere compared with the wing.

Dpp mobility versus dpp transcription

The finding that increased tkv expression results in decreased dpp transcription reveals an unexpected link between Dpp mobility and Dpp production. Because of this link, the above experiments do not discriminate between growth effects due to differences in Dpp mobility per se as opposed to secondary consequences on Dpp production. To distinguish between these scenarios, we made use of a compartment-specific Ubx regulatory allele, posterior bithorax (pbx), that lacks detectable Ubx in the P compartment when paired with a Ubx null allele but still has normal Ubx expression in the A compartment (fig. S6) (8). Consequently, in pbx/Ubx^- haltere discs, the P compartment increased in size such that the P:A size ratio was 1.45 (Fig. 5, E and F); the P:A ratio of +/Ubx^- haltere discs was ∼0.35 (13). We suggest that Dpp more readily diffuses into and through the P compartments of pbx/Ubx^- discs because of the wing-like expression pattern of tkv and that this wing-like diffusion results in its robust growth.

To test whether differences in Tkv-regulated Dpp diffusion affect tissue growth independently of an effect on Dpp production, we examined the consequences of expressing UAS-tkv⁺ uniformly in pbx/Ubx^- haltere discs. If Tkv’s effect on growth is mediated only by lowering Dpp production, both compartments should be reduced in size and thus maintain the same size ratio. However, if reducing Dpp mobility directly affects growth, the P compartment should be reduced in size more than the A compartment, which, in pbx/Ubx^- discs, already has high tkv expression. We found that expressing uniform tkv⁺in pbx/Ubx^- discs decreased the size of the P compartment more than the A compartment, resulting in a P:A ratio of 0.83 (Fig. 5, E to H). Because uniform tkv⁺ returned the P:A ratio back to the wild-type ratio by ∼56% (from 1.45 to 0.83, whereas +/Ubx^- discs have a P:A ratio of ∼0.35), these results suggest that this single variable is sufficient to provide ∼50% rescue of the size of an otherwise Ubx mutant P compartment.

Discussion

We have investigated the mechanism underlying a classic yet poorly understood phenomenon in biology: how size variations are genetically programmed in animal development. Many experiments show that organ size is not governed by counting cell divisions but instead depends on disc-intrinsic yet cell-nonautonomous mechanisms, possibly relying on morphogen signaling (34). Our results support this idea by showing that alterations in a morphogen gradient contribute to size differences between appendages. In the example investigated here, Ubx limits the size of the haltere by reducing both Dpp production and Dpp mobility. Moreover, both of these effects are due, in part, to higher tkv expression in the medial haltere (Fig. 5, I and J). In many morphogen systems, the receptors themselves have been shown to control the distribution of the ligand and, consequently, pathway activation (30, 35-37). We show that a selector gene exploits this phenomenon to modify organ size.

Although the mechanism by which Dpp controls proliferation is not fully understood, recent results argue that, in the medial wing disc, cells may compare the amount of Dpp transduction with their neighbors, whereas lateral cells proliferate in response to absolute Dpp levels (17). Our results suggest several ways in which the altered Dpp gradient in the haltere could limit its growth. First, proliferation of lateral haltere cells may be limited because they perceive less Dpp. Second, the narrower Dpp gradient results in fewer cells exposed to the gradient in the medial haltere. Another notable difference is that, because there are two peaks of Dpp signaling in the wing but only one in the haltere, the wing has four distinct slopes whereas the haltere has only two. The less complex Dpp activity landscape of the haltere may also contribute to its reduced growth.

On the basis of these results, we suggest that altering the shape and intensity of morphogen gradients may be a general mechanism by which selector genes affect tissue sizes in animal development. Consistent with this view, wingless (wg), another long-range morphogen in the wing, is partially repressed in the haltere (38). Intriguingly, some of the size and shape differences in the beaks of Darwin’s finches are controlled by alterations in the production of the Dpp ortholog BMP4 (39). Our results suggest that differences in the diffusion of this ligand may also contribute to the range of beak morphologies that have evolved in these species.