Asymmetric requirement of Dpp/BMP morphogen dispersal in the Drosophila wing disc

Generation of a functional HA-dpp allele

To manipulate the endogenous Dpp morphogen gradient, we utilized a MiMIC transposon inserted in the dpp locus (dpp^MI03752), which allows to replace the sequence between the two attP sites in the transposon with any sequence inserted between two inverted attB sites upon integrase expression³⁸. A genomic fragment containing sequences encoding a tagged version of dpp followed by an FRT and a marker was first inserted into the locus (Fig. 1b), then the endogenous dpp exon was removed upon FLP/FRT recombination to keep only the tagged dpp exon (Fig. 1b). Using this strategy, we inserted different tags into the dpp locus and found that while a GFP-dpp allele was homozygous lethal during early embryogenesis, a HA-dpp allele was functional without obvious phenotypes³⁹ (Fig. 1c, see below “Methods”). Immunostainings for the HA-tag including permeabilization steps showed HA-Dpp expression in an anterior stripe of cells along the A−P compartment boundary in the late third instar wing disc (Fig. 1d). In contrast, immunostainings for the HA-tag without permeabilization, which allows antibodies to access only the extracellular antigens, revealed that a shallow extracellular HA-Dpp gradient overlapped with the gradient of phosphorylated Mad (pMad), a downstream transcription factor of Dpp signaling (Fig. 1e). Similar HA tag knock-in dpp alleles have recently been generated by a CRISPR approach²⁵.

Generation and characterization of HA trap

Since we could not apply morphotrap due to the lethality of the GFP-dpp allele, we generated an HA trap, analogous to morphotrap. HA trap consists of an anti-HA scFv⁴⁰ fused to the transmembrane domain of CD8 and mCherry (Fig. 2a). HA trap expression in the anterior stripe of cells of wild-type wing discs using ptc-Gal4 did not interfere with Dpp signaling in the wing disc or patterning and growth of the adult wing (Supplementary Fig. 1). Thus, HA trap is inert in the absence of an HA-tagged protein. While we attempted to visualize extracellular HA-Dpp distribution upon HA trap expression, we noticed that the HA tag can no longer be used for immunostaining when bound to HA trap. We therefore additionally inserted an Ollas tag to generate a functional Ollas-HA-dpp allele in order to visualize the extracellular Dpp distribution using the antibody against the Ollas tag. The extracellular Ollas-HA-Dpp gradient was similar to the extracellular HA-Dpp gradient (Fig. 2b, c).

To test if HA trap can efficiently trap Ollas-HA-Dpp in the Dpp-producing cells, HA trap was expressed in the anterior stripe of cells using ptc-Gal4 in Ollas-HA-Dpp heterozygous wing discs, since ptc-Gal4 expression largely overlaps with dpp– producing cells⁴¹. Under this condition, extracellular immunostainings for the Ollas-tag revealed that Ollas-HA-Dpp accumulated on the anterior stripe of cells, and that the extracellular gradient was abolished (Fig. 2d, e). To test if HA trap can trap Ollas-HA-Dpp outside the anterior stripe of cells, clones of cells expressing Gal4 were randomly induced by heat-shock inducible FLP to express HA trap under UAS control. We found that Ollas-HA-Dpp accumulated in clones of cells expressing HA trap induced outside the main dpp source cells in both compartments (Fig. 2f−i, arrow). If HA trap can efficiently trap Ollas-HA-Dpp in the source cells, the clonal Ollas-HA-Dpp accumulation should be blocked upon HA trap expression in the source cells. Indeed, we found that clonal Ollas-HA-Dpp accumulation in both compartments was drastically reduced upon HA trap expression using ptc-Gal4 (Fig. 2j−k, arrow), indicating that the HA trap can block HA-Dpp dispersal efficiently.

It has been shown that overexpression of GFP-Dpp from the anterior stripe cells leads to accumulation of GFP-Dpp in clones of cells expressing morphotrap in the peripheral regions³⁷. In contrast, we found that Ollas-HA-Dpp accumulated in clones of cells expressing HA trap near the source cells but not in the peripheral regions (Fig. 2l, arrowhead). This raises a question whether Dpp can act in the peripheral regions at physiological levels.

Asymmetric patterning and growth defects by HA trap

After we validated that HA trap can efficiently block Dpp dispersal, we then expressed HA trap using different Gal4 driver lines in HA-dpp homozygous wing discs to address the requirement of Dpp dispersal. Normally, Dpp binds to the Dpp receptors Thickveins (Tkv) and Punt, inducing a pMad gradient and an inverse gradient of Brk, a transcription repressor repressed by Dpp signaling. The two opposite gradients regulate growth and patterning (nested target gene expression, such as sal, and omb) to define adult wing vein positions (such as L2 and L5) (Fig. 3a)^{10,19,42,43,44}.

Upon HA trap expression in the anterior stripe of cells using ptc-Gal4, pMad, Sal, and Omb expression were undetectable in the P compartment and Brk was also upregulated in the P compartment (Fig. 3b−f), indicating that HA trap efficiently blocked HA-Dpp dispersal from source cells and interfered with patterning. The posterior wing pouch growth was also affected as revealed by the expression of an intervein marker DSRF and a wing pouch marker 5xQE.DsRed⁴⁵ (Fig. 3b arrow, 3g). Interestingly, although 5xQE.DsRed contains five copies of the 806 bp Quadrant Enhancer (QE) of the wing master gene vg containing a Mad binding site and is therefore thought to be directly regulated by Dpp signaling^46,47, 5xQE.DsRed remained expressed in the P compartment without detectable Dpp signaling (Fig. 3b, arrow). In the A compartment, pMad was slightly reduced in the anterior medial region (Fig. 3b, c), probably because HA trap partially blocked Dpp signaling upon binding to HA-Dpp (Fig. 2l, arrow). Nevertheless, the anterior Brk gradient was not strongly affected (Fig. 3b, d). Although maximum intensity of Sal or Omb was reduced, nested expression of Sal and Omb was maintained in the A compartment and the anterior growth defects was milder than the posterior growth defects (Fig. 3b, e, f, g). Consistent with these phenotypes in the wing discs, while posterior patterning and growth were severely affected, anterior patterning and growth were relatively normal in the resulting adult wings (Fig. 3h−j).

Similar asymmetric defects in patterning and growth were observed upon HA trap expression in the region covering the entire wing pouch using nub–Gal4 (Fig. 3k−t) or in the entire anterior compartment using ci-Gal4 (Supplementary Fig. 2a−j). Furthermore, even when HA trap was expressed using both nub-Gal4 and ptc–Gal4, the resulting phenotypes were not enhanced (Supplementary Fig. 3). To test whether the posterior growth defects upon HA trap expression is caused by cell death, Caspase-3 was analyzed. We found that Caspase-3 was not upregulated upon HA trap expression, and blocking apoptosis by apoptosis inhibitor p35 did not rescue these growth defects upon HA trap expression (Supplementary Fig. 4). Thus, the posterior growth defects upon HA trap expression is not caused by cell death. Taken together, these results suggest that, while critical for posterior patterning and growth, Dpp dispersal is largely dispensable for anterior patterning and growth.

Lateral wing pouch growth without Dpp signaling

A critical role of Dpp dispersal for posterior patterning and growth is consistent with a role of Dpp as a morphogen. However, the overall phenotypes caused by HA trap was surprisingly mild when compared to the phenotypes seen in dpp mutants (see below). Given the requirement of Dpp signaling for cell proliferation and survival in the entire wing pouch⁴⁸, it was surprising that about 40% of the posterior wing pouch was able to grow and differentiate into adult wing tissue without detectable Dpp signaling (Fig. 3).

We therefore tested whether the posterior growth and 5xQE.DsRed expression seen upon HA trap expression is caused by low levels of HA-Dpp leaking from the HA trap expressed in the source. In this case, the posterior growth and 5xQE.DsRed expression seen upon HA trap expression should be dependent on tkv, an essential receptor for Dpp signaling. To test this, mutant clones of tkv^a12 (characterized as a null allele^49,50) were induced in wing discs expressing HA trap with ptc-Gal4 between mid-second and beginning of third instar stages and analyzed in the late third instar stage. We found that tkv^a12 clones often survived and expressed the 5xQE.DsRed reporter in the anterior lateral regions as well as in the entire posterior region. We also noticed that tkv^a12 clones survived and expressed the 5xQE.DsRed reporter even next to the source cells in the P compartment (Fig. 4a). These results indicate that the lateral growth and 5xQE.DsRed expression seen upon HA trap expression is independent of Dpp signaling, and not caused by a leakage of HA-Dpp from the HA trap, even if such leakage would occur.

To test whether Dpp signaling-independent growth occurs also during normal development, tkv^a12 clones were induced in the wild-type wing disc during mid-second and early third instar stage. We found that tkv^a12 clones were eliminated from the medial regions but often survived and expressed the 5xQE.DsRed reporter in the lateral wing pouch (Fig. 4b). Since tkv^a12 may not be a complete null allele, we then inserted an FRT cassette in the tkv locus and generated a tkv flip-out allele (tkvHA^FO) to induce FLP/FRT-mediated excision of tkv. By generating tkv null clones upon heat-shock inducible FLP expression, we confirmed that tkv null clones often survived and expressed the 5xQE.DsRed reporter in the lateral wing pouch (Fig. 4c, d, arrow). We also found that, while most often eliminated, medial tkv null clones survived and expressed 5xQE.DsRed in rare cases (Fig. 4d), indicating that Dpp signaling is dispensable for 5xQE.DsRed expression also in the medial region, but medial cells lacking Dpp signaling are normally eliminated⁴⁸.

How can Dpp signaling-independent wing pouch growth and 5xQE.DsRed expression be reconciled with a critical role of Dpp signaling for the entire wing pouch growth?⁴⁸ First, tkv clones generated in the developing wing pouch have been shown to be eliminated by apoptosis or extrusion and do not survive in the adult wing^48,51. However, tkv clones survive better in the P compartment where Dpp signaling is blocked by HA trap (Fig. 4a) and in the lateral region of wild-type wing disc where Dpp signaling is generally low (Fig. 4b−d). This raises a possibility that tkv clones are eliminated when surrounded by wild-type cells, even if tkv clones could grow and survive to a certain extent. Second, wing pouch and 5xQE.DsRed expression were completely lost in dpp mutants (see below). It has been shown that initial wing pouch specification is mediated by Dpp derived from the peripodial membrane, which covers the developing wing pouch, and this early dpp expression in the peripodial membrane is lost in dpp disc alleles⁵². Thus, wing pouch and 5xQE.DsRed expression could be lost in dpp disc alleles due to failure of initial specification of the wing disc and subsequent elimination of cells.

To minimize these potential problems, we applied Gal80ts to conditionally remove dpp from the entire A compartment using ci–Gal4. At the permissive temperature of 18 °C, Gal80ts actively represses Gal4 activity. At restrictive temperature of 29 °C, Gal80ts can no longer block Gal4 activity; thus, Gal4 can be temporally activated using temperature shifts. Upon FLP expression, dpp was removed by FLP/FRT-mediated excision via dpp^FO allele²⁴, in which an FRT cassette was inserted into the dpp locus. To remove dpp from the beginning of second instar stage when the wing pouch is specified, the larvae were raised at 18 °C for 4 days and then shifted to 29 °C. By removing dpp from the entire A compartment using ci–Gal4 under this condition, we found that 5xQE.DsRed remained expressed despite severe growth defects in the late third instar stage (Fig. 4e−h). Similarly, genetic removal of tkv via tkvHA^FO from the A compartment using ci–Gal4 or from the P compartment using hh–Gal4 from the second instar stage revealed that, despite severe growth defects, 5xQE.DsRed remained expressed in each compartment lacking tkv (Supplementary Fig. 5). Surprisingly, similar results were obtained even when tkv was removed from the entire P compartment using hh–Gal4 from the embryonic stages without Gal80ts (Supplementary Fig. 6). These results further support the presence of Dpp signaling-independent 5xQE.DsRed expression and wing pouch growth.

How is 5xQE.DsRed expression regulated if QE is not directly regulated by Dpp signaling? While 5xQE.DsRed expression is completely lost in dpp mutants, we found that 5xQE.DsRed reporter expression was rescued in dpp, brk double mutant wing discs (Supplementary Fig. 7), indicating that 5xQE.DsRed expression is largely induced by repressing brk, similar to the regulation of other dpp target genes. Indeed, QE has been shown to be activated in brk mutant clones in the wing disc⁵³. However, this notion appears inconsistent with the fact that 5xQE.DsRed expression was not repressed in the region where Brk is high in various conditions, in which Dpp signaling is compromised (Figs. 3b, 4h and Supplementary Fig. 5dʹ, hʹ). We noticed that the observed high Brk levels upon Dpp trapping were comparable to the Brk level in the lateral region of the control wing disc (Fig. 3d, n), and Brk and 5xQE.DsRed were co-expressed in the lateral region of the control wing disc (Fig. 4f and Supplementary Fig. 5bʹ fʹ). Thus, we speculate that Brk is not sufficient to repress 5xQE.DsRed expression at physiological levels in lateral regions and that there are additional inductive inputs such as Wg^45,54.

Severe patterning and growth defects by Dpp trap

Even if the lateral wing pouch region can grow independent of Dpp signaling after wing pouch specification (Fig. 4e−h), this growth cannot account for the overall minor growth phenotypes caused by HA trap (Fig. 3 and Supplementary Fig. 2a−j). How can relatively normal patterning and growth be achieved without Dpp dispersal? Since pMad was completely lost in dpp mutants (Fig. 4g) but remained active in the source cells upon HA trap expression (Fig. 3 and Supplementary Fig. 2a−j), we asked whether Dpp signaling in the source cells could account for the minor phenotypes caused by HA trap.

To test this, we selected DARPins, protein binders based on ankyrin repeats^55,56,57, that bind to the mature Dpp ligand and block Dpp signaling. For each of the 36 candidates obtained from the in vitro screening, we generated a Dpp trap by fusing the anti-Dpp DARPin to the transmembrane domain of CD8 and mCherry (Fig. 5a). By expressing each trap in the wing disc, we identified one Dpp trap (containing DARPin 1242_F1), which efficiently blocked Dpp dispersal (Fig. 5b) and signaling (Fig. 5c, d). We found that the expression of the Dpp trap using ptc-Gal4 (Fig. 5c−i), nub-Gal4 (Fig. 5k−t), and ci-Gal4 (Supplementary Fig. 2k−t) caused severe signaling defects as well as patterning and growth defects, similar to dpp mutants (Fig. 4g, h). Adult wings expressing Dpp trap using nub-Gal4 were recovered and also showed severe patterning and growth defects comparable to dpp mutants (Fig. 5s). Although Caspase-3 was upregulated upon Dpp trap expression (Supplementary Fig. 4a, c, d), the growth defects were not rescued by p35 (Supplementary Fig. 4h−j), indicating that apoptosis was not the main cause of growth defects caused by Dpp trap. Furthermore, these severe phenotypes were not due to a common scaffold effect of DARPins, since one of the traps (containing DARPin 1240_C9) that failed to trap Dpp did not interfere with pMad accumulation in the wing disc or patterning and growth of the adult wing when expressed using ptc–Gal4 (Supplementary Fig. 8).

We note that upon Dpp trap expression, Sal expression was lost from the medial region but appeared to be upregulated in the lateral region (Fig. 5d, l and Supplementary Fig. 2o, arrow). It has previously been shown that Sal is expressed not only in the medial region but also in the lateral region⁵⁸. The same study also showed that the medial Sal expression is Dpp signaling-dependent but lateral Sal expression is Brk-dependent. Thus, upregulation of Brk upon Dpp trap expression could cause the lateral Sal upregulation. However, when we focused on the peripheral region of the control wing disc (basal confocal section), we noticed that the lateral Sal expression of the control wing disc was actually comparable to that of the wing disc expressing Dpp trap (Supplementary Fig. 9a, b). Thus, when we focused on the medial Sal expression (apical confocal section), the lateral Sal expression of the control wing disc was simply missed due to the tissue architecture. Consistently, when dpp was removed from the entire A compartment using ci-Gal4 from mid-second instar stage, Sal expression was lost from the medial region but not significantly upregulated in the lateral region (Supplementary Fig. 9c−e).

By comparing the phenotypes caused by Dpp trap and HA trap, we noticed that the phenotypes caused by Dpp trap were much stronger than those caused by HA trap (Figs. 3, 5 and Supplementary Fig. 2). Indeed, comparison of each compartment size when expressing HA trap and Dpp trap using different Gal4 driver lines also showed that each compartment size was smaller upon Dpp trap expression than upon HA trap expression (Fig. 5j, u and Supplementary Fig. 2u). To test if the difference could be due to more efficient blocking of Dpp dispersal by Dpp trap than by HA trap, each trap was expressed in the anterior stripe of cells using ptc–Gal4 and posterior pMad signal was analyzed, since the extracellular staining was not sensitive enough to detect significant differences in leakage (Figs. 2e and 5b), and the posterior pMad activation would reflect the amount of leaked Dpp since the two traps were specifically expressed in the anterior stripe of cells. We found that HA trap blocked posterior pMad signal more efficiently than Dpp trap (Supplementary Fig. 10), indicating that Dpp trap actually blocks Dpp dispersal less efficiently than HA trap. Thus, the severe phenotypes caused by Dpp trap are likely because Dpp trap blocks Dpp signaling more efficiently than HA trap.

Interestingly, despite the slight leakage of Dpp from the Dpp trap (Supplementary Fig. 10), anterior Dpp trap expression caused more severe posterior growth defects than HA trap (Fig. 5j and Supplementary Fig. 2u), indicating that anterior Dpp signaling is non-autonomously required for the posterior growth. We note that, even though the anterior Dpp signaling was eliminated, genetic removal of tkv from the A compartment using ci-Gal4 did not interfere with posterior growth as severe as Dpp trap (Supplementary Fig. 5c, d), probably because Dpp secreted from the A compartment can disperse to control posterior growth. Taken together, these results suggest that Dpp signaling in the source cells is required for a majority of patterning and growth seen upon HA trap expression.

Rescue of dpp mutants by cell-autonomous Dpp signaling

Our results so far suggest that, while the requirement for Dpp dispersal is relatively minor and asymmetric, Dpp signaling in the source cells is critically required for the majority of patterning and growth seen upon blocking Dpp dispersal. This raises the question of how cell-autonomous Dpp signaling in the source cells can control patterning and growth outside the anterior stripe of cells, the main dpp source cells. There are a couple of possible scenarios; if dpp expression is successively restricted to the anterior stripe of cells during development, the anterior stripe of cells may retain and deliver the earlier Dpp signaling to the peripheral region after they leave from the anterior stripe of cells via proliferation⁵⁹, or downstream factor(s) of Dpp signaling in the anterior stripe of cells may act non-autonomously to control patterning and growth outside the stripe of cells. Alternatively, dpp expression may not be restricted to the anterior stripe of cells in the early stages, similar to what has been shown in the case of wg¹².

Before we address these possibilities, we first asked how important cell-autonomous Dpp signaling in the source cells is for wing pouch patterning and growth. If the relatively mild phenotypes caused by HA trap are due to cell-autonomous Dpp signaling in the source cells, a constitutively active version of Tkv (TkvQD)⁵⁰ expressed in the anterior stripe of cells using dpp-Gal4 should rescue severe patterning and growth defects in dpp mutants (Fig. 6a, c) to an extent mimicking the phenotypes caused by HA trap (Fig. 3). Indeed, under this condition, pMad activation in the anterior stripe of cells rescued nested Sal and Omb expression in the A compartment (Fig. 6b). Interestingly, growth, but not patterning, was also partially rescued in the P compartment as indicated by DSRF and 5xQE.DsRed expression (Fig. 6b, arrow), thus indeed mimicking phenotypes caused by HA trap (Fig. 3). Unfortunately, the resulting adult flies were not recovered at 25 °C. However, when the temperature was shifted from 25 °C to 18 °C during mid- to late-third instar stages in order to reduce Gal4 activity during pupal stages, rare survivors were recovered from the pupal cases or managed to hatch although they died shortly after hatching. In such survivors, although the anterior wing veins tends to be affected, probably due to continuous TkvQD expression during pupal stages even in the lower temperature, the anterior growth was rescued more than the posterior growth, similar to phenotypes caused by HA trap (Fig. 6d). These results suggest that the phenotypes caused by HA trap expression largely depend on cell-autonomous Dpp signaling in the source cells.

How can cell-autonomous Dpp signaling in the source cells control posterior growth if dpp expression is restricted to the A compartment? One trivial possibility is that the posterior growth was induced by non-specific dpp-Gal4 expression in the P compartment. To test this, dpp-Gal4 was converted into a ubiquitous LexA driver to express TkvQD permanently in lineage of dpp-Gal4 (Fig. 6e). In this setup, pMad was constitutively activated in all the cells where dpp–Gal4 has been expressed, including cells expressing dpp–Gal4 non-specifically. Under this condition, pMad, Sal, and Omb were uniformly upregulated in the A compartment, and Brk was completely lost in the entire A compartment (Fig. 6f), indicating that dpp–Gal4 has been expressed in the entire A compartment⁶⁰. In contrast, pMad was not activated in the P compartment, but 5xQE.DsRed was still induced in the P compartment (Fig. 6f, arrow), indicating that non-autonomous posterior growth control by anterior Dpp signaling is permissive rather than instructive (see “Discussion”).

Initial uniform dpp transcription in the anterior compartment

Next, we asked how cell-autonomous Dpp signaling in the source cells can control anterior patterning and growth. The uniform lineage of Dpp-producing cells in the A compartment⁶⁰ (Fig. 6f) suggests two possibilities; either dpp expression is always restricted to the anterior stripe of cells but the lineage of these cells can cover the lateral region via proliferation⁵⁹, or earlier dpp expression covers the entire A compartment as in the case of wg¹². Since the existing dpp-Gal4 line is derived from a fragment of the dpp disc enhancer inserted outside the dpp locus, we first generated an endogenous dpp-Gal4 line using our platform (Fig. 7a). We traced its lineage with G-TRACE analysis, in which RFP expression labels the real-time Gal4-expressing cells and GFP expression labels the entire lineage of the Gal4-expressing cells⁶⁰ (Fig. 7b). We found that the lineage of Dpp-producing cells indeed covers the entire A compartment (Fig. 7c).

To distinguish between the two possibilities mentioned above, we then generated a dpp transcription reporter line by inserting a destabilized GFP (half-life <2 h) into the dpp locus (Fig. 7d). Consistent with the latter possibility, we found that this dpp transcription reporter was uniformly expressed in the entire A compartment until the early third instar stage (Fig. 7e, eʹ, f, fʹ) and refined to an anterior stripe expression during much of the third instar stage (Fig. 7g, gʹ, h, hʹ). To directly follow dpp transcription, we also performed smFISH using RNAscope technology to visualize dpp transcripts in situ. Consistent with the dynamic expression of the dpp transcription reporter, we found uniform anterior dpp transcription until the early third instar stage (Fig. 7i, iʹ, j, jʹ) and an anterior stripe of dpp transcription in the later stages (Fig. 7k, kʹ, l, lʹ). Despite the initial broad anterior dpp expression, we found that pMad signal is low in the middle of the wing disc and graded toward the lateral regions, similar to the pMad gradient in the later stages (Supplementary Fig. 11).

Transient dpp source outside Sal domain is required for anterior patterning and growth

The earlier anterior dpp source outside the anterior stripe of cells could provide a local dpp source to control anterior patterning and growth when Dpp dispersal is blocked. However, since ptc-Gal4 is also initially expressed in the entire A compartment⁴¹, the relatively minor defects by HA trap could be due to the perdurance of Dpp signaling via artificially stabilized HA-Dpp by HA trap. To avoid HA trap expression in the entire A compartment, we applied tubGal80ts to express HA trap using ptc–Gal4 at defined time points in the A compartment. To do so, the larvae were raised at 18 °C until a temperature shift to 29 °C to induce Gal4 expression (Supplementary Fig. 12a). Upon HA trap expression from the mid-second instar stage, the lineage of ptc–Gal4 covered at most the anterior Sal domain (see below), which corresponds to the region between L2 and L4 in the adult wing. We found that the later the temperature shift, the milder the posterior growth defects (Supplementary Fig. 12a, b). In contrast, the A compartment size (between L1 and L4) remained rather normal, independent of the timing of the temperature shift (Supplementary Fig. 12a, c). Interestingly, the size of peripheral regions (between L1 and L2) and the specification of L2 were not affected, independent of the timing of the temperature shift (Supplementary Fig. 12a, d). These results are consistent with a role of an anterior lateral dpp source for patterning and/or growth in the anterior lateral regions.

To directly test this, we applied Gal80ts to genetically remove dpp via dpp^FO allele upon FLP expression using ptc–Gal4 from the mid-second instar stage. To do so, the larvae were raised at 18 °C for 5 days before a temperature shift to 29 °C and were then dissected 48 h later. In this setup, dpp was removed approximately from the anterior Sal region, where cells in which the FRT cassette was removed were marked by lacZ staining (Fig. 8a−f). Given that it takes about 20 h to eliminate the majority of Dpp protein under this condition^24,41, wing pouches are devoid of the majority of the Dpp protein derived from the anterior stripe of cells for 28 h at 29 °C until they reach the late third instar stage, which corresponds to a lack of Dpp protein secreted from the main source from early third instar stages onward.

Under this setup, we found that pMad, Sal, and Omb were significantly reduced in the P compartment, consistent with the removal of the main dpp source (Fig. 8a−f). In contrast, in the A compartment, low levels of pMad persisted and Brk remained graded with lowest expression outside the lacZ positive region (Fig. 8d, arrow). As a consequence, while Sal was completely lost (Fig. 8e), weak Omb remained expressed in the A compartment (Fig. 8f). Consistent with a critical role of the dpp stripe for wing pouch growth^25,41,61, both anterior and posterior growth were affected (see below, Fig. 8k).

To test if the remaining anterior Dpp signaling activity is due to Dpp produced outside the anterior Sal domain, we then compared removal of dpp from the anterior Sal domain using ptc–Gal4 (the same setup above) with removal of dpp from the entire A compartment using ci–Gal4 from the mid-second instar stage. We found that the anterior weak Dpp signaling and Omb expression seen upon removal of dpp from the anterior Sal domain using ptc–Gal4 (Fig. 8g, h) was completely lost by removing dpp from the entire A compartment using ci–Gal4 (Fig. 8i, j). Furthermore, anterior growth defects upon removal of dpp from the anterior Sal domain was further enhanced upon removal of dpp from the entire A compartment (Fig. 8k), indicating that the anterior dpp source outside the anterior Sal domain is locally required for anterior Dpp signaling and growth.

How can the transient dpp transcription sustain Dpp target gene expression? One possibility is that persistent low pMad levels are continuously required to repress Brk. Alternatively, Brk repression by early Dpp signaling persists in the later stages without continuous Dpp signaling, for example, by epigenetic regulation or via autoregulation. To distinguish between these possibilities, we applied tubGal80ts to genetically remove tkv via tkvHA^FO allele upon FLP expression from the entire A compartment using ci-Gal4 at different time points. To do so, the larvae were raised at 18 °C until a temperature shift to 29 °C to induce Gal4 expression. Consistent with a role of tkv in wing pouch growth, the earlier tkv was removed, the smaller the A compartment was (Supplementary Fig. 13). We found that Brk is largely derepressed as early as 16 h after tkv was removed (Supplementary Fig. 13). By considering perdurance activity of Gal80ts for 6 h after temperature shift⁶², we can estimate that Brk is derepressed within 10 h at 29 °C after Dpp signaling was lost. Given that the transient dpp transcription, which terminated in the early third instar stage, can sustain the anterior pMad signaling and Brk repression at least 28 h at 29 °C after Dpp protein from the main source is eliminated, these results suggest that persistent weak pMad signaling is continuously required to repress Brk.

Taken together, Dpp dispersal-independent anterior patterning and growth can therefore be achieved by a combination of a persistent weak signaling by transient dpp transcription outside the stripe and a stronger signaling by continuous dpp transcription in the anterior stripe of cells.