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

How morphogen gradients control patterning and growth in developing tissues remains largely unknown due to lack of tools manipulating morphogen gradients. Here, we generate two membrane-tethered protein binders that manipulate different aspects of Decapentaplegic (Dpp), a morphogen required for overall patterning and growth of the Drosophila wing. One is “HA trap” based on a single-chain variable fragment (scFv) against the HA tag that traps HA-Dpp to mainly block its dispersal, the other is “Dpp trap” based on a Designed Ankyrin Repeat Protein (DARPin) against Dpp that traps Dpp to block both its dispersal and signaling. Using these tools, we found that, while posterior patterning and growth require Dpp dispersal, anterior patterning and growth largely proceed without Dpp dispersal. We show that dpp transcriptional refinement from an initially uniform to a localized expression and persistent signaling in transient dpp source cells render the anterior compartment robust against the absence of Dpp dispersal. Furthermore, despite a critical requirement of dpp for the overall wing growth, neither Dpp dispersal nor direct signaling is critical for lateral wing growth after wing pouch specification. These results challenge the long-standing dogma that Dpp dispersal is strictly required to control and coordinate overall wing patterning and growth.

A fundamental question in developmental biology is how proteins work together to orchestrate developmental processes. Forward and reverse genetic approaches based on mutants and RNAi, together with biochemical analyses, provide insights into how proteins function. However, interpretational gaps often remain between the mutant phenotypes and the underlying mechanisms.
Recently, small, high-affinity protein binders, such as nanobodies, single-chain variable fragments (scFvs), Designed Ankyrin Repeat Proteins (DARPins), and others, have emerged as versatile tools to fill this gap. By fusing these protein binders to wellcharacterized protein domains and expressing the fusion proteins in vivo, protein function can be directly manipulated in a predicted manner [1][2][3][4][5] . For example, when a protein functions with multiple parameters, protein binder tools targeting each or a subset of these parameters could help to dissect the requirement of each parameter. However, it remains challenging to design and customize such protein binder tools.
A class of molecules that exert its function with multiple parameters are morphogens, secreted molecules that disperse from a localized source and regulate target gene expression in a concentration-dependent manner [6][7][8][9] . A morphogen gradient is characterized by its parameters such as rates of secretion, diffusion, and degradation 10 . Temporal dynamics of a morphogen gradient also impact cell fates decisions 11 . Despite a variety of parameters involved, morphogen dispersal is generally thought to be critical for morphogen function based on severe morphogen mutant phenotypes and long-range action of morphogens to reorganize patterning and growth.
However, a recent study challenged this basic assumption for the case of the Wingless (Wg) morphogen, the main Wnt in Drosophila, by showing that a membrane-tethered non-diffusible form of Wg can replace the endogenous Wg without strongly affecting appendage development 12 . Although the precise contribution of Wg dispersal requires further investigations [13][14][15][16] , the study raises the question of how important morphogen dispersal is for tissue patterning and growth in general.
In contrast to Wg, Decapentaplegic (dpp), the vertebrate BMP2/ 4 homolog, is thought to act as a bona fide morphogen in the Drosophila prospective wing. Dpp disperses from a narrow anterior stripe of cells along the anterior−posterior (A−P) compartment boundary to establish a characteristic morphogen gradient in both compartments ( Fig. 1a) 17,18 . How the Dpp dispersal-mediated morphogen gradient achieves and coordinates overall wing patterning and growth has served as a paradigm to study morphogens 19 . However, despite intensive studies, it remains controversial how Dpp/BMP disperses 17,20-23 , controls growth [24][25][26][27][28][29][30][31][32][33] , and coordinates patterning and growth (i.e. scaling) 32,[34][35][36] . Regardless of the actual mechanisms, all the studies are based on the assumption that Dpp dispersal from the anterior stripe of cells controls overall wing patterning and growth, in line with the severe dpp mutant phenotypes.
To directly manipulate dispersal of Dpp, we recently generated morphotrap, a membrane-tethered anti-GFP nanobody, to trap GFP-tagged Dpp and thereby manipulate its dispersal 37 . Using morphotrap, the authors showed that a substantial amount of GFP-Dpp secreted from the anterior stripe of cells can reach the peripheral wing disc and that blocking GFP-Dpp dispersal from the source cells causes severe adult wing patterning and growth defects 37 . These results support the critical role of Dpp dispersal for overall wing patterning and growth 37 . However, the application of morphotrap was limited to rescue conditions by overexpression of GFP-Dpp, due to the lack of an endogenous GFPdpp allele.
In this study, we first generated a platform to manipulate the dpp locus and inserted a tag into dpp in order to investigate the precise requirement of the endogenous Dpp morphogen gradient for wing patterning and growth. We found that while a HA-dpp allele was functional, a GFP-dpp allele was not, thus limiting morphotrap application. To manipulate the endogenous Dpp morphogen gradient, we then generated two protein binder tools analogous to morphotrap. One is HA trap based on anti-HA scFv that traps HA-Dpp through the HA tag to mainly block Dpp dispersal, the other is Dpp trap based on anti-Dpp DARPin that directly binds to Dpp to block Dpp dispersal and signaling in the source cells. Thus, these tools allowed us to distinguish the requirements of Dpp dispersal and cell-autonomous signaling in the source cells for wing pouch growth and patterning.
Here, we show, using these tools, that while posterior patterning and growth require Dpp dispersal, anterior patterning and growth largely proceed without Dpp dispersal but require cell-autonomous Dpp signaling in the source cells. We show that dpp transcriptional refinement from an initially uniform to a localized expression and persistent signaling in transient dpp source cells allow relatively normal anterior patterning and growth despite the absence of Dpp dispersal. Furthermore, despite a critical requirement of dpp for overall wing growth, we also find that neither Dpp dispersal nor direct signaling is critical for the lateral wing pouch to grow once the wing pouch is defined. These results challenge the long-standing dogma that Dpp dispersal controls overall wing patterning and growth and call for a revision of how Dpp controls and coordinates wing patterning and growth.

Results
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 expression 38 . 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 phenotypes 39 (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 approach 25 .
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 scFv 40 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 cells 41 . 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 regions 37 . 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.DsRed 45 (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 Generation of a functional HA-dpp allele. a A schematic view of the wing disc and the adult wing. b A schematic view of a platform manipulating endogenous dpp locus. c Adult wing of a homozygous HA-dpp fly. d Conventional α-HA staining of HA-dpp homozygous wing disc. e Extracellular α-HA staining (Ex HA) and α-pMad staining of HA-dpp homozygous wing disc. Average fluorescence intensity profile of extracellular HA and pMad staining of HA-dpp/HA-dpp wing disc (n = 5). Data are presented as mean ± SD. Scale bar 50 μm. 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 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-26726-6 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 pouch 48 , 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 eliminated 48 .
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 alleles 52 . 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 allele 24 , 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 disc 53 . 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 . b−e Extracellular α-Ollas staining (Ex Ollas), HA trap (mCherry), and merge of control Ollas-HA-dpp/+ disc (b), and of Ollas-HA-dpp/+ ptc > HA trap disc (d). c, e Average fluorescence intensity profile of extracellular α-Ollas staining of (b) and (d) respectively. Ollas-HA-dpp, ptc > + disc (control) (n = 6) (c), and Ollas-HA-dpp, ptc > HA trap disc (n = 13) (e). Data are presented as mean ± SD. f−k Εxtracellular α-Ollas staining, HA trap (mCherry), and merge of Ollas-HA-dpp/+ disc with an anterior clone of cells expressing HA trap (f), of Ollas-HA-dpp/+ disc with a posterior clone of cells expressing HA trap (h), and of Ollas-HA-dpp/+ disc with HA trap expression using ptc-Gal4 and clones of cells expressing HA trap in both compartments (j). g, i, k Quantification of extracellular α-Ollas staining and HA trap (mCherry) of (f), (h), (j), respectively. Arrows indicate clones of cells expressing HA trap where quantification was performed. l Extracellular α-Ollas staining, HA trap (mCherry), pMad, and merge of Ollas-HA-dpp/+ wing disc with clones of cells expressing 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  . c−f Average fluorescence intensity profile of α-pMad (c), α-Brk (d), α-Sal (e), α-Omb (f) staining in (a, b). Data are presented as mean ± SD. g Comparison of compartment size of HA-dpp/HA-dpp, ptc > + control wing pouch (n = 35) and HA-dpp/HA-dpp, ptc > HA trap wing pouch (n = 37). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for the comparison of the A compartment (p = 0.0002) and for comparison of the P compartment (p < 0.0001). ***p < 0.001, ****p < 0.0001. h, i Adult wing of HA-dpp/HA-dpp, ptc > + (control) (h) and HA-dpp/HA-dpp, ptc > HA trap (i). j Comparison of compartment size of (h) and (i). HA-dpp/HA-dpp, ptc > + control adult wing (n = 12) and HA-dpp/HA-dpp, ptc > HA trap adult wing (n = 16). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001) and for comparison of the P compartment (p < 0.0001). ****p < 0.0001. k, l α-pMad, α-Brk, α-Sal, α-Omb, and HA trap (mCherry) (inset) of HA-dpp/HA-dpp, nub > + control wing disc (k) and HA-dpp/HA-dpp, nub > HA trap wing disc (l). m−p Average fluorescence intensity profile of α-pMad (m), α-Brk (n), α-Sal (o), α-Omb (p) staining in (k, l). Data are presented as mean ± SD. q Comparison of compartment size of HA-dpp/HA-dpp, nub > + control wing pouch (n = 33) and HA-dpp/HA-dpp, nub > HA trap wing pouch (n = 38). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001) and for comparison of the P compartment (p < 0.0001). ****p < 0.0001. r, s Adult wing of HA-dpp/HA-dpp, nub > + (control) (r) and HA-dpp/HA-dpp, nub > HA trap (s). t Comparison of compartment size of (r) and (s). HA-dpp/HA-dpp, nub > + control adult wing (n = 11) and HA-dpp/HA-dpp, nub > HA trap adult wing (n = 12). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001). Two-sided Mann−Whitney test was used for comparison of the P compartment (p < 0.0001). ****p < 0.0001. Dashed white lines mark the A−P compartment border. Scale bar 50 μm. 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 region 58 . The same study also showed that the medial Sal expression is Dpp signalingdependent 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,   Fig. 4 Lateral wing pouch growth without Dpp signaling. a, b tkv a12 clones (indicated by the absence of GFP signal) induced in HA-dpp/HA-dpp, ptc > HA trap wing discs (a) and in wild-type wing discs (b). c, d tkvHA FO clones (indicated by the absence of α-HA staining) in wild-type wing discs. Clones were induced at 60−72 h AEL (after egg laying) during mid-second to early third instar stages. e−h α-pMad and 5xQE.DsRed (e, g) and α-Brk and 5xQE.DsRed (f, h) of control wing disc (e, f) and 5xQE.DsRed, dpp FO /dpp FO , tubGal80ts, ci > UAS-FLP wing disc (g, h). Crosses were shifted from 18°C to 29°C at 4-day AEL (early second instar). Scale bar 50 μm.  Average fluorescence intensity profile of extracellular α-Ollas staining of Ollas-HA-dpp/+, ptc > + wing disc (control) (n = 4) and Ollas-HA-dpp/+, ptc > Dpp trap wing disc (n = 5). Data are presented as mean ± SD. c, d α-pMad, α-Brk, α-Sal, α-Omb staining, and Dpp trap (mCherry) expression (inset) of HA-dpp/+, ptc > + control wing disc (c) and HA-dpp/+, ptc > Dpp trap wing disc (d). e−h Average fluorescence intensity profile of α-pMad (e), α-Brk (f), α-Sal (g), α-Omb (h) staining in (c, d). Data are presented as mean ± SD. i Comparison of compartment size of HA-dpp/+, ptc > + control wing pouch (n = 44) and HA-dpp/+, ptc > Dpp trap wing pouch (n = 39). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001) and for comparison of the P compartment (p < 0.0001). ****p < 0.0001. j Comparison of normalized compartment size of wing pouch upon HA trap (n = 37) and Dpp trap (n = 39) expression using ptc-Gal4 (the same data set from Figs. 3g and 5i). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001) and for comparison of the P compartment (p < 0.0001). ****p < 0.0001. k, l α-pMad, α-Brk, α-Sal, α-Omb staining, and Dpp trap (mCherry) expression (inset) of HA-dpp/+, nub > + control wing disc (k) and HA-dpp/+, nub > Dpp trap wing disc (l). m−p Average fluorescence intensity profile of α-pMad (m), α-Brk (n), α-Sal (o), α-Omb (p) staining in (k, l). Data are presented as mean ± SD. q Comparison of compartment size of HA-dpp/+, nub > + control wing pouch (n = 28) and HA-dpp/+, nub > Dpp trap wing pouch (n = 47). Data are presented as mean ± SD. Two-sided Mann−Whitney test was used for comparison of the A compartment (p < 0.0001). Two-sided unpaired Student's t test with unequal variance was used for comparison of the P compartment (p < 0.0001). ****p < 0.0001. r, s Adult wing of HA-dpp/+, nub > + control wing disc (r) and HA-dpp/+, nub > Dpp trap wing disc (s). t Comparison of compartment size of (r, s). HA-dpp/+, nub > + control adult wing (n = 20) and HA-dpp/+, nub > Dpp trap adult wing (n = 20). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment (p < 0.0001) and for comparison of the P compartment (p < 0.0001). ****p < 0.0001. u Comparison of normalized compartment size of wing pouch upon HA trap (n = 38) and Dpp trap (n = 47) expression using nub-Gal4 (the same data set from Figs. 3q and 5q). Data are presented as mean ± SD. Two-sided Mann−Whitney test was used for comparison of the A compartment (p < 0.0001). Two-sided unpaired Student's t test with unequal variance was used for comparison of the P compartment (p < 0.0001). ****p < 0.0001. Scale bar 50 μm. 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 proliferation 59 , 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 wg 12 .
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 compartment 60 . 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 compartment 60 (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 proliferation 59 , or earlier dpp expression covers the entire A compartment as in the case of wg 12 . 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 cells 60 (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 compartment 41 , 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 Fig. 6 Rescue of dpp mutant by cell-autonomous Dpp signaling. a, b α-pMad, α-Brk, α-Sal, α-Omb, 5xQE.DsRed, and DSRF staining of a dpp d8 /dpp d12 wing disc and b dpp d8 /dpp d12 , dpp > tkvQD wing disc. Arrows indicate rescued posterior wing pouch identified by 5xQE.DsRed and DSRF staining. c Adult wing of dpp d8 /dpp d12 . d Adult wing of dpp d8 /dpp d12 , dpp > tkvQD. e A schematic view of converting dpp-Gal4 into a LexA driver, which is permanently expressed in lineage of dpp-Gal4 expressing cells. In this experimental setup, the lineage of dpp-Gal4 (including lineage of non-specific dpp-Gal4 expression) will permanently activate TkvQD and thus pMad signaling. f α-pMad, α-Brk, α-Sal, α-Omb staining, and 5xQE.DsRed signal of dpp d8 /dpp d12 , dpp > FLP, act > y + >LexA-LHG, LOP-tkvQD wing disc. Arrow indicates rescued posterior wing pouch identified by 5xQE.DsRed staining. Scale bar 50 μm. ARTICLE 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 shift 62 , 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.

Discussion
It has long been thought that Dpp dispersal from the anterior stripe of cells generates the morphogen gradient in both compartments to control overall wing patterning and growth mainly based on the complete loss of wing tissue in dpp disc mutant alleles. Here, we generated two protein binder tools, namely HA trap and Dpp trap, to manipulate distinct parameters of the Dpp morphogen to determine the requirement of Dpp dispersal and cell-autonomous signaling in the source cells. We show that, although endogenous Dpp indeed generates a gradient in both compartments, requirement of Dpp dispersal for wing patterning and growth is much less than previously thought (Fig. 9).

New protein binder tools manipulating distinct aspects of Dpp.
Although nanobodies against GFP have been used most intensively in the field 63,64 , fusion to GFP could affect protein functions, as is the case for GFP-Dpp. To bypass this, we generated HA trap, which is analogous to morphotrap and provides an alternative way to trap secreted proteins. Although HA trap can trap HA-Dpp as efficient as morphotrap, we found several differences between the two. First, while trapping GFP-Dpp by morphotrap in the source cells activates Dpp signaling in at least one cell row in the P compartment 37 , trapping HA-Dpp by HA trap did not (Fig. 3b). We think that the difference is not because GFP-Dpp trapped by morphotrap activates Dpp signaling in trans, since clonal accumulation of GFP-Dpp by morphotrap in the P compartment failed to do so 37 . Second, while morphotrap could trap GFP-Dpp even in the peripheral regions 37 , HA trap did not (Fig. 2l). Third, while trapping GFP-Dpp by morphotrap in the source cells induced excessive Dpp signaling in the source cells and caused severe defects in the adult wing 37 , trapping HA- Fig. 8 Transient anterior dpp source outside Sal domain is required for anterior patterning and growth. a−f α-pMad, α-Brk, and α-LacZ (a, d), α-Sal and α-LacZ (b, e), α-Omb and α-LacZ (c, f) staining of dpp FO /+, tubGal80ts, act > stop > lacZ, ptc > FLP control wing disc (a−c) and dpp FO /dpp FO , tubGal80ts, act > stop > lacZ, ptc > FLP wing disc (d−f). Crosses were shifted from 18°C to 29°C at 5-day AEL (mid-second instar). α-LacZ staining marks the region where dpp is removed upon FLP expression. Dashed white lines mark the A−P compartment border. Scale bar 50 μm. g−j α-pMad, α-Brk, and α-LacZ (g, i), α-Omb and α-LacZ (h, j) staining of dpp FO /dpp FO , tubGal80ts, act > stop > lacZ, ptc > FLP wing disc (g, h) and dpp FO /dpp FO , tubGal80ts, act > stop > lacZ, ci > FLP (i, j). The genotypes of the wing discs in (d−f) and in (g, h) are identical. Crosses were shifted from 18°C to 29°C at 5-day AEL (mid-second instar). α-LacZ staining marks the region where dpp is removed upon FLP expression. Dashed white lines mark the A−P compartment border. Scale bar 50 μm. k Comparison of each compartment size of wing discs (g−j). dpp FO /dpp FO , tubGal80ts, act > stop > lacZ, ptc > FLP wing disc (n = 13) and dpp FO /dpp FO , tubGal80ts, act > stop > lacZ, ci > FLP wing disc (n = 32). Data are presented as mean ± SD. Two-sided unpaired Student's t test with unequal variance was used for comparison of the A compartment size (p < 0.0001) and for comparison of the P compartment size (p = 0.0002). ***p < 0.001, ****p < 0.0001.
Dpp by HA trap in the source cells slightly reduced Dpp signaling in the source cells and caused relatively minor defects in the adult wing ( Fig. 3 and Supplementary Fig. 2a−j). We speculate that these differences are in part due to overexpression of GFP-Dpp. Due to excess amount of GFP-Dpp, some GFP-Dpp may leak from the morphotrap to activate Dpp signaling in the P compartment, or may reach to the peripheral regions. Excessive Dpp signaling may cause cell death during pupal stages to cause severe adult wing defects as previously shown 65 . These differences highlight the importance of investigating endogenous protein functions.
In addition to creating, generating, HA trap, we isolated DARPins against Dpp, and generated a Dpp trap analogous to HA trap. Interestingly, while HA trap blocks mainly Dpp dispersal (Fig. 3), we found that Dpp trap blocks Dpp dispersal and cell-autonomous signaling in the source cells (Fig. 5). We speculate that HA trap binds to the HA tag and thereby allows Dpp to bind to its receptors, while Dpp trap directly binds to Dpp to block its interaction with the receptors. Regardless of the actual mechanisms underlying this difference, these tools allowed us to dissect the requirements of dispersal and cell-autonomous signaling in the source cells for wing pouch growth and patterning. Relatively mild phenotypes by HA trap and severe phenotypes by Dpp trap indicate a minor and asymmetric role of Dpp dispersal and a critical role of cell-autonomous Dpp signaling in the source cells, respectively. Furthermore, these results also suggest that the severe dpp mutant phenotypes do not reflect the role of Dpp dispersal alone, but reflect the role of both Dpp dispersal and cell-autonomous Dpp signaling in the source cells, with more contribution from the latter parameter.
Asymmetric requirement of Dpp dispersal. The relatively minor requirement of Dpp dispersal for the anterior compartment is reminiscent of the minor requirement of dispersal of Wg 12 . Wg dispersal is largely dispensable for the wing growth, likely due to the early uniform wg transcription in the entire wing pouch and a memory of the earlier signaling 12 . In contrast, the requirement of Dpp dispersal is asymmetric along the A−P axis due to the early uniform dpp transcription in the A compartment. In both cases, transcriptional refinement of each morphogen and persistent signaling by transient morphogen expression appear to be a key for robustness against the absence of dispersal of each morphogen.
Although not identified, a dpp source outside the anterior stripe of cells has previously been implicated to control the entire wing pouch growth based on the minor growth defects by removal of dpp from the anterior stripe of cells using dpp-Gal4 and severe growth defects by removal of dpp from the A compartment using ci-Gal4 24 . However, we and others showed that removal of dpp using dpp-Gal4 was imprecise and inefficient, and that dpp derived from the anterior stripe of cells is indeed critical for wing growth 25,41,61 . Thus, the presence of a dpp source outside the anterior stripe of cells has been questioned. Our results suggest that such a dpp source indeed exists and contributes to the anterior patterning and growth, but not to the growth of the entire wing pouch (Figs. 7, 8).
It remains unknown how transient dpp expression can maintain Dpp signaling in the anterior lateral region (Fig. 8). One possibility is that feedback factors control the duration of Dpp signaling. Various feedback factors have been shown to regulate Dpp signaling. For example, the Drosophila tumor Fig. 9 Asymmetric requirement of Dpp dispersal for wing patterning and growth, and a refined growth equalization model. Based on complete lack of wing pouch in dpp disc alleles, it has long been thought that Dpp dispersal from the anterior stripe of cells controls the overall patterning and growth of the Drosophila wing pouch. In contrast, the present study shows that while critical for the posterior patterning and growth, Dpp dispersal is largely dispensable for the anterior patterning and growth. The asymmetric requirement of Dpp dispersal is in part due to dpp transcriptional refinement from an initially uniform to a localized expression and persistent signaling in transient dpp source cells. Furthermore, despite a critical requirement of Dpp signaling for the medial wing pouch growth, Dpp signaling is dispensable for the lateral wing pouch growth after wing pouch specification. We propose a refined growth equalization model, in which Dpp signaling removes a growth repressor Brk to allow medial regions to grow, while Dpp signaling sets Brk expression to the lateral region to repress the lateral growth with higher growth rate to equalize growth rates. In this refined model, both the Dpp signaling-dependent medial region and -independent lateral region exist within the wing pouch. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-26726-6 ARTICLE NATURE COMMUNICATIONS | (2021)12:6435 | https://doi.org/10.1038/s41467-021-26726-6 | www.nature.com/naturecommunications necrosis factor α homolog Eiger and a secreted BMP-binding protein Crossveinless-2 (Cv-2) are positively regulated by Dpp signaling to either positively and negatively influence Dpp signaling in the early embryo, respectively 66 . While JNK is not activated by Eiger in the wild-type wing pouch 67 , Cv-2 acts as a positive feedback factor to specify the posterior crossvein during pupal stages 68 . Whether these two factors act during wing pouch patterning and growth remains to be addressed. Pentagon (Pent) is a secreted feedback factor repressed by Dpp signaling to positively regulate Dpp signaling in the wing disc 69 . Pent is produced in the lateral region of the wing pouch and regulates Dpp signaling and proliferation there. Thus, in addition to its role for scaling, Pent may control the duration of Dpp signaling in the lateral region. Another possibility is that, since it takes quite some time (20−24 h at 29°C) to eliminate Dpp protein upon excision of dpp 24,41 , relatively stable dpp mRNA and/or protein after the termination of transient dpp transcription may contribute to Brk repression (~28 h at 29°C).
Growth without Dpp dispersal and signaling. Despite a critical requirement of dpp for the entire wing pouch growth, our results uncover Dpp dispersal-and signaling-independent lateral wing pouch growth (Figs. 3, 4, 6 and Supplementary Figs. 5, 6) The presence of Dpp signaling-independent growth appears inconsistent with the fact that the wing pouch is completely lost in dpp mutants (Fig. 6a, c and Supplementary Fig. 7b). This could be in part due to a failure of the initial specification of the wing pouch in dpp disc alleles 52 . Indeed, despite severe growth defects, part of the wing pouch could still grow upon removal of dpp after wing pouch specification (Fig. 4e−h) and upon removal of tkv after wing pouch specification ( Supplementary Fig. 5).
Consistent with the presence of Dpp signaling-independent lateral growth, it has been shown that lateral wing fates are less sensitive than medial wing fates in various dpp mutant alleles 70 . However, this appears counter-intuitive since the lateral region, where morphogen level are low, is expected to be more sensitive than the medial region to a reduction of morphogen levels. It has been proposed that another BMP-type ligand, Glass bottom boat (Gbb), which is expressed ubiquitously in the wing pouch, contributes to lateral cell fates 70 . However, since Gbb signaling is also mediated by Tkv 71 but the lateral growth is independent of Tkv (Figs. 3, 4, 6 and Supplementary Figs. 5, 6), we think that the lateral region develops independent of direct Dpp and Gbb signaling. Thus, the lateral region is less sensitive than the medial region in various dpp mutant alleles probably because the lateral region can develop independent of Dpp signaling. What regulates the Dpp signaling-independent lateral wing pouch growth? Given that 5xQE.DsRed is also dependent on Wg 45,54 , Wg may regulate 5xQE.DsRed expression and growth in the absence of Dpp signaling.
In addition to the Dpp signaling-independent lateral growth, requirement of anterior Dpp signaling for posterior growth (Fig. 5j and Supplementary Fig. 2u) and rescue of posterior growth in dpp mutants by anterior Dpp signaling (Fig. 6) indicates that anterior Dpp signaling is non-autonomously involved in posterior lateral growth. We note that similar rescue of posterior growth in dpp mutant by anterior Dpp signaling has previously been recognized, but the rescued posterior growth was interpreted as growth of the hinge region, without immunostainings for relevant markers 59 . It remains unknown how anterior Dpp signaling contributes to posterior growth. One possibility is that factors from the A compartment may act non-autonomously to promote posterior growth. Such factors include, but are not limited to, direct downstream factors of Dpp signaling. Given that 5xQE.DsRed is dependent on Wg 45,54 , Wg derived from the rescued A compartment may regulate 5xQE.DsRed expression and growth in the P compartment. Alternatively, mechanical forces may be involved in the non-autonomous growth. It has been proposed that growth factors such as Dpp induce medial growth and subsequently stretch the peripheral regions to induce lateral growth. As the wing disc grows, the peripheral regions in turn compresses the medial region of the wing disc to inhibit the growth of the medial region [72][73][74][75][76] . Thus, the growth of the A compartment may stretch the P compartment cells to stimulate their proliferation. It has also been shown that juxtaposition of cells with different Dpp signaling level can induce proliferation non-autonomously but the growth is transient 26,27 . Therefore, we think it is unlikely that the difference of Dpp signaling levels between two compartments can induce sustained growth.
A refined growth model. The presence of Dpp signalingindependent lateral wing pouch growth is at odds with all the growth models assuming that Dpp dispersal directly controls overall wing patterning and growth based on the complete loss of wing tissue in dpp mutants [24][25][26][27][28][29][30][31][32][33] (Fig. 9). For example, no wing pouch growth is expected without Dpp signaling due to a lack of either a temporal increase of Dpp signaling (temporal model) 32 , a detectable Dpp signal (threshold model) 25,61,77 , or a slope of Dpp signaling activity (gradient model) 26,27 .
It has recently been proposed that Wg and Dpp control wing pouch size by two distinct mechanisms 33,45,54,78 . One is an intracellular mechanism, in which Vg controls its own expression through QE in response to Wg and Dpp; the other one is an intercellular mechanism, in which Vg-positive wing pouch cells send a feed-forward signal to induce QE-dependent vg expression in the neighboring pre-wing cells to recruit them into the wing pouch in response to Dpp and Wg. The dispersal of each morphogen is critical for the two mechanisms in the model 33,45,54,78 . In the genetic setup used, in which morphogens and vg expression are eliminated, the two mechanisms appear to recapitulate the dynamics of vg expression seen during normal wing disc development, in which vg expression is gradually expanded in the wing pouch area. However, although sufficient, these mechanisms do not necessarily account for vg expression under physiological conditions. Indeed, we found Dpp dispersal and/or signaling-independent vg expression in various conditions. Among the models, the presence of Dpp signaling-independent lateral wing pouch growth appears most consistent with the growth equalization model, in which Dpp signaling removes Brk to allow medial regions to grow, while Dpp signaling limits Brk expression to lateral regions to suppress Dpp signalingindependent lateral growth with higher proliferation nature to equalize the non-uniform growth 22,28 (Fig. 9). However, the identity of medial and lateral regions remained undefined in this model. Given the complete lack of wing pouch in dpp mutants, it is tempting to speculate that the Dpp signaling-dependent medial region corresponds to the entire wing pouch region, and the Dpp signaling-independent lateral region corresponds to the hinge region located next to the wing pouch region. Indeed, using morphotrap, such a Dpp signaling-insensitive posterior growth has previously been observed and interpreted as the growth of the hinge region due to severe adult wing defects 37 . However, in contradiction to this interpretation, which would predict overgrowth of the hinge region in brk mutant, a massive overgrowth of wing pouch region was observed in brk mutant (Supplementary Fig. 7). Based on our results that the lateral wing pouch regions can grow independent of direct Dpp signaling (Figs. 3,4,6 and Supplementary Figs. 5,6), we therefore suggest to refine the growth equalization model and propose that both Dpp signalingdependent medial and Dpp signaling-independent lateral regions are located within the wing pouch (Fig. 9). The permissive role of Dpp signaling in modulating a non-uniform growth potential within the wing pouch raises questions about what kind of instructive signals control proliferation and growth, and how the non-uniform growth potential emerges within the wing pouch independent of the Dpp/Brk system 44 .
In summary, our approach applying customized protein binder tools to manipulate distinct parameters of Dpp challenges the long-standing dogma that Dpp dispersal controls overall wing patterning and growth. Given that the tools developed in this study are easily applicable in other tissues, it would be interesting to investigate the precise requirement of Dpp dispersal and signaling in other tissues or between different organs.

Methods
Data reporting. No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.  Table 1.

Genotypes by figures. Each genotype by figures was provided as a Supplementary
Immunostainings and antibodies. Staged larvae were dissected and transferred directly to cold fixative (4% PFA in PBS) and fixed for 20 min at room temperature. After fixation, discs were extensively washed with PBT (PBS plus 0.3% Triton-X) and blocked in PBT plus 5% normal goat serum (Sigma-Aldrich) for >30 min at 4°C, followed by incubation with primary antibody overnight at 4°C. The next day discs were washed in PBT and then incubated in secondary antibody in PBT plus 5% normal goat serum for 2 h at room temperature on a rotor without light. After another round of washes with PBT, samples were mounted in Vectashield (H-1000, Vector Laboratories) 37 . For extracellular staining, dissected larvae were incubated with primary antibodies in M3 medium for 1 h on ice before fixation to allow antibodies to access only the extracellular antigens. Each fly cross was set up together with a proper control and genotypes were processed in parallel. If the genotype could be distinguished, experimental and control samples were processed in the same tube. To minimize variations, embryos were staged by collecting eggs for 2−4 h. An average intensity image from three sequential images from a representative wing disc is shown for all the experiments. Images of wing discs were obtained using a Leica SP5-II-MATRIX confocal microscope (section thickness 1 μm) and Leica LAS AF (ver. 2.6.0.7266). Images were analyzed using ImageJ RNAscope. smFISH using RNAscope technology with the probe against dpp (Cat No. 896761) has previously been successful to visualize dpp mRNA in the germline stem cell niche 79 . The accession number for the probes against dpp target 682−1673 is NM_057963.5. RNAscope was performed in an Eppendorf tube. Larvae were dissected in PBS and fixed in 4% paraformaldehyde in PBS for 30 min. Fixed larvae were washed with PBS and then with PBT (PBS containing 0.03% Triton X-100). Fixed larvae were dehydrated at RT (room temperature) in a series of 25, 50, 75 and 100% methanol in PBT, and stored overnight in 100% methanol at −20°C. Larvae were rehydrated at RT in a series of 75, 50, 25, 0% methanol in PBT. Larvae were treated with protease using Pretreat 3 (RNAscope H2O2 & Protease Plus Reagents; ACD, 322330) at RT for 5 min. After washing with PBT, hybridization using probes against dpp was performed overnight at 40°C. The following day, larvae were washed with RNAscope wash buffer and re-fixed with 4% paraformaldehyde in PBS for 10 min. After washing with PBS, fluorescent signal was developed using RNAscope Fluorescent Multiplex Reagent Kit according to the manufacturer's instructions. Wing discs were mounted in Vectashield (H-1000, Vector Laboratories) and dissected for imaging. Images of wing discs were obtained using a Leica SP5-II-MATRIX confocal microscope (section thickness 1 μm) and Leica LAS AF (ver. 2.6.0.7266). Images were analyzed using ImageJ (v.2.0.0-rc69/1.52p). Figures were prepared using Omero (ver5.9.1) and Illustrator (24.1.3).

Quantification
Quantification of pMad, Brk, Sal, and Omb. From each z-stack image, signal intensity profile along A/P axis was extracted from average projection of three sequential images using ImageJ (v.2.0.0-rc69/1.52p). Each signal intensity profile collected in Excel (Ver. 16.51) was aligned along A/P compartment boundary (based on anti-Ptc staining) and average signal intensity profile from different samples was generated and plotted by the script (wing_disc-alignment.py). The average intensity profile from control and experimental samples was then compared by the script (wingdisc_comparison.py). Both scripts can be obtained from https://etiennees.github.io/Wing_disc-alignment/. The resulting signal intensity profiles (mean with SD) were generated by Prism (v.8 Generation of HA-dpp and GFP-dpp knock-in allele Cloning of plasmids for injection. A fragment containing multi-cloning sites (MCS) between two inverted attB sites was synthesized and inserted in the pBS (BamHI) vector (from Mario Metzler). A genomic fragment of dpp between dpp MI03752 and PBac{RB}e00178 (about 4.4 kb), as well as an FRT and 3xP3mCherry were inserted in this MCS by standard cloning procedures. A fragment encoding HA tag or GFP was inserted between the XhoI and NheI sites inserted after the last Furin processing site 18 .
Removal of the endogenous dpp exon by FLP/FRT recombination (Fig. 1b (ii)). Males from the above stock were crossed with females of genotype hsFLP; al, Pbac{RB}e00178 /SM6, al, sp and subjected to heat-shock at 37°C for 1 h/day until flies hatch. Pbac{RB}e00178 contains FRT sequence and w+. Upon recombination, the dpp genomic fragment with a tag is followed by FRT and w+. In the case of HA-dpp, males of hsFLP; HA-dpp/al, PBac{RB}e00178 were recovered and crossed with yw; al, b, c, sp/ SM6, al, sp. From this cross, HA-dpp allele with a successful recombination (males of yw; HA-dpp(w+)/SM6, al, sp or yw; HA-dpp(w+)/al, b, c, sp) were screened for w+ and against al. Each male was then individually crossed with virgins of yw; al, b, c, sp/ SM6, al, sp to establish the yw; HA-dpp/SM6, al, sp stock. However, in the case of GFP-dpp, hsFLP; GFP-dpp/al, PBac{RB}e00178 flies were not recovered, indicating that GFP-dpp allele is haploinsufficient. Indeed, insertion of GFP-dpp in the dpp locus and maintenance of the resulting stock (GFPdpp(w-)/Cyo, P23) required P23, a transgene containing a dpp genomic fragment to rescue dpp haploinsufficiency. We thus concluded that GFP-dpp allele is haploinsufficient. Consistent with this, the original stock (GFP-dpp(w−)/Cyo, P23) never became homozygous.
Construction of α-HA scFv. The accession numbers for VH and VL of anti-HA antibody are VH: LC522514 and VL: LC522515. cDNA of HA scFv was constructed by combining coding sequences of variable regions of the heavy chain (V H : 1−423 of LC522514) and of the light chain (V L : 67−420 of LC522515) cloned from anti-HA hybridoma (clone 12CA5) 40 with a linker sequence (5ʹ-accggtGGC GGAGGCTCTGGCGGAGGAGGTTCCGGCGGAGGTGGAAGCgatatc-3ʹ) in the order of V H -linker-V L . The coding sequence of HA scFv was cloned into pCS2+mcs-2FT-T for FLAG-tagging. Requests for HA scFv should be addressed to Y.M. (mii@nibb.ac.jp). To generate HA trap, the region encoding morphotrap (VHH-GFP4) was replaced with KpnI and SphI sites in pLOTattB-VHH-GFP4:CD8-mChery 37 . A fragment encoding HA scFv was amplified by PCR and then inserted via KpnI and SphI sites by standard cloning procedures.
Selection of Dpp-binding DARPins and generation of Dpp trap. Streptavidinbinding peptide (SBP)-tagged mature C-terminal domain of Dpp was cloned into pRSFDuet vector by a standard cloning. Dpp was overexpressed in E. coli, extracted from inclusion bodies, refolded, and purified by heparin affinity chromatography followed by reverse phase HPLC 80 . To isolate suitable DARPins, SBP-tagged Dpp was immobilized on streptavidin magnetic beads and used as a target for DARPin selections by employing multiple rounds of Ribosome Display 81,82 . Due to the aggregation and precipitation propensity of the purified SBP-Dpp, the refolded dimers previously stored in 6 M urea buffer (6 M urea, 50 mM Tris-HCl, 2 mM EDTA pH8.0, 0.25 M NaCl) were diluted to a concentration of 100−120 µg/ml in the same buffer and subsequently dialyzed against 4 mM HCl at 4°C overnight. To ensure binding of correctly folded Dpp to the beads, this solution was diluted five times in the used selection buffer just prior to bead loading and the start of the ribosome display selection. In each panning round, the target concentration presented on magnetic beads was reduced, while the washing stringency was simultaneously increased to enrich for binders with high affinities 81 . In addition, from the second round onward, a pre-panning against Streptavidin beads was performed prior to the real selection to reduce the amounts of matrix binders. After four rounds of selection, the enriched pool was cloned into an E. coli expression vector, enabling the production of both N-terminally His 8 -and C-terminally FLAG-tagged DARPins. Nearly 400 colonies of transformed E. coli were picked and the encoded binders expressed in small scale. Bacterial crude extracts were subsequently used in enzyme-linked immunosorbent assay (ELISA) screenings, detecting the binding of candidate DARPins to streptavidin-immobilized Dpp, or just streptavidin (indicating background binding) by using a FLAG-tag based detection system. Of those 127 candidate DARPins interacting with streptavidin-immobilized Dpp, 73 (or 57%) specifically bound to Dpp (i.e., having at least threefold higher signal for streptavidin-immobilized Dpp than to streptavidin alone). Thirty-six of these (50%) revealed unique and full-length sequences. To generate Dpp trap, the region encoding morphotrap (VHH-GFP4) was replaced with KpnI and SphI sites in pLOTattB-VHH-GFP4:CD8-mChery 37 . Each fragment encoding a DARPin was amplified by PCR and then inserted via KpnI and SphI sites by standard cloning procedures.
Generation of tkvHA FO (Flip-out) allele. The tkvHA allele was previously described 83 . An FRT cassette was inserted in the re-insertion vector for tkvHA (Genewiz) and re-inserted into the attP site in the tkv locus.
Statistics and reproducibility. All images were obtained from multiple animals (n > 3). The experiments were repeated at least two times independently with similar results. Statistical significance was assessed by Prism (v.8.4.3(471)) based on the normality tests. The observed phenotypes were highly reproducible as indicated by the significance of p values obtained by statistical tests.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data that support all experimental findings of this study are available within the paper and its Supplementary Information files or from the corresponding author S.M. upon request. Raw data necessary to reproduce all statistical analyses and results in the paper are provided in the source data file provided with this paper. The accession numbers for VH and VL of anti-HA antibody are VH: LC522514 and VL: LC522515.
The accession number for the probes against dpp target 682-1673 is NM_057963.5. Source data are provided with this paper.

Code availability
Both wing_disc-alignment.py and wingdisc_comparison.py are available from (https:// etiennees.github.io/Wing_disc-alignment/). Briefly, wing_disc-alignment.py was used to align each signal intensity profile along the A/P compartment boundary and generate average signal intensity profile from different samples. wingdisc_comparison.py was then used to compare the average intensity profile from control and experimental samples.