A synthetic morphogen reveals principles of tissue patterning



  • Secreted GFP can replace the morphogen Decapentaplegic to correctly pattern a fly wing by a diffusion-based mechanism

  • Gradient formation requires binding of a morphogen by both signalling and non-signalling receptors

  • Lateral movement within membranes and cellular “hopping” by non-signalling receptors could help to shape and broaden the gradient

Multicellular organisms develop through a sequence of patterning events in which cells adopt distinct cell fates. In many instances, these patterns are established by localised production of a morphogen at a restricted source. The resulting concentration gradient is translated into a pattern by the underlying cells, which activate different sets of genes depending on the morphogen concentration they encounter.

As morphogens are secreted into the extracellular space, they could spread by diffusion alone, but such a simple mechanism is thought to face numerous problems, the most obvious being that a soluble molecule would spread in a multi-directional cloud and be quickly dispersed. Experimental evidence indicates that there are a myriad of interlinked regulatory processes governing morphogen action, and this complexity has made it difficult to answer the fundamental question of how the gradient is set up and how it spreads and works over exactly the right distance.

To overcome this problem, a synthetic approach is helpful. Synthetic reconstitution isolates and identifies the key players, and systems can be constructed in which parameters can be carefully tuned, and their effects modelled and measured with precision. Recently, in an innovative spin on this approach, the groups of Jean-Paul Vincent and Guillaume Salbreux have collaborated to engineer a synthetic morphogen which remarkably, is able to recapitulate the ability of the Drosophila morphogen Decapentaplegic (Dpp) to pattern the fly wing (Stapornwongkul et al, 2020). In the process, they have also resolved the longstanding puzzle of gradient formation in favour of a modified diffusion model.

Before coming to the Crick I was already familiar with, and highly impressed by, the work of Jean-Paul. By bringing different disciplines under the same roof, the Crick allowed me, a theoretical physicist, to come in contact with Jean-Paul, a developmental biologist, and discuss a collaboration. The open structure of the Crick then allowed us to meet regularly, make sure that our PhD students could easily exchange ideas, and to make parallel progress in theory and experiment; which were essential conditions for the success of this exciting project.

Guillaume Salbreux

The Vincent lab uses the wing imaginal discs of Drosophila as a model system to study morphogens (see also Case Study 19). The two wing imaginal discs are single-cell-deep sheets of epithelium that grow and invaginate into simple pouches during larval development, eventually metamorphosing into a wing blade by extension and apposition of the dorsal and ventral surfaces of the pouch. The BMP2/4 homologue Decapentaplegic (Dpp) is expressed in a stripe of cells running down the central anterior-posterior boundary of the disc, from where it disperses into the lateral regions of the tissue to form a gradient on either side of its expression stripe. This Dpp gradient induces target gene expression in the wing pouch, and these genes are crucial for correct positioning of the longitudinal wing veins L2 and L5.

To create a synthetic morphogen, Vincent’s PhD student Kristina Stapornwongkul made SecGFP, a secreted version of green fluorescent protein, and expressed it from a stripe of cells very similar to the Dpp expression domain. The protein was detectable in the cells in which it was made, but not in the basolateral space underlying the wing disc epithelium where the Dpp gradient forms, presumably because it was leaking through the basal lamina and into the hemolymph, the larval bloodstream.

This fate does not befall Dpp due to its interaction with binding partners present in the basolateral space, in the form of its cognate signalling receptors and non-signalling extracellular matrix components. Therefore, Stapornwongkul designed a binding partner for SecGFP by fusing a high affinity anti-GFP nanobody to the transmembrane protein CD8, and expressed it in the posterior compartment of the wing disc, next door to the SecGFP expression stripe. When SecGFP was co-expressed with this artificial receptor, Nb1highCD8, a gradient of GFP fluorescence was readily detectable in the basolateral space.

Figure 1

Establishment of a GFP gradient in the wing imaginal disc.

Establishment of a GFP gradient in the wing imaginal disc. A: geography of the wing imaginal disc, B: SecGFP gradient ±high affinity receptor. C: removal of the non-zero tail by fat body trapping of leaked SecGFP

The GFP gradient differed from what would be expected in a classical diffusion-based model: rather than decaying exponentially towards zero, there was a long nonzero tail in the region farthest from the GFP source. Reasoning that the tail might be caused by GFP diffusing into the hemolymph and then re-entering the wing disc tissue, Stapornwongkul devised a method of sopping up any leaked GFP by trapping it with high affinity binding partners expressed in the fat body, a sprawling organ that lines the body cavity and has multiple contacts with hemolymph. This strategy eliminated the nonzero tail, giving a normal diffusion gradient.

Having established that a single high affinity extracellular binding partner was sufficient to reveal a gradient of SecGFP, Stapornwongkul then asked what happened if Nb1highCD8 was replaced in the posterior compartment with a low affinity GFP binding partner, NblowCD8, which mimicked the binding affinity of the extracellular matrix for a secreted protein. The SecGFP and NblowCD8 combination was insufficient to generate a gradient, although there were low levels of visible GFP fluorescence, showing that NblowCD8 was able to trap extracellular GFP.

Marc de Gennes, a PhD student in the Salbreux lab, formalised the role of extracellular binders and hemolymph leakage in GFP gradient formation by devising a mathematical model taking into account not only diffusion and leakage, but also degradation due to endocytosis of receptor-bound GFP.


The model nicely recapitulated Stapornwongkul’s experimental data, and also had predictive value, correctly anticipating the gradient shapes resulting from experiments manipulating either SecGFP or Nb1highCD8 expression: overexpressing SecGFP resulted in gradient extension, as well as a flattening near the source due to saturation, and overexpression of Nb1highCD8 led to a steepening of the gradient, and a reduction in the nonzero tail, likely due to increased degradation due to endocytosis.

With this information in hand, Stapornwongkul next tested if it was possible to engineer the Dpp signalling pathway to make it responsive to SecGFP. Normally, dimeric Dpp tetramerises upon binding to its receptors Thickveins (Tkv) and Punt (Put), and this leads to phosphorylation and activation of the downstream effector Mad, and transcriptional repression of the brinker (brk) gene. The resulting inverse Brk gradient controls the boundaries of expression of target genes such as spalt (sal) and optomotor-blind (omb), which in turn cause patterned down-regulation of DSRF, required for vein fate specification.

To make Tkv and Put responsive to SecGFP rather than Dpp, the receptors were fused to different anti-GFP nanobodies that recognise non-overlapping epitopes on GFP, generating Nb2highTkv and Nb1highPut. In tissue culture, SecGFP was able to activate the modified receptors, generating good levels of phospho-Mad (pMad). This system was then moved into a Drosophila strain carrying a conditional Dpp allele that allowed SecGFP to be flipped in and Dpp flipped out in the wing pouch.

Rescue of growth and patterning by GFP

Rescue of growth and patterning by GFP

In the absence of Dpp, SecGFP could partially rescue wing development, but only in the presence of the modified Nb2highTkv and Nb1highPut receptors. However, patterning was clearly abnormal, and this was reflected at the molecular level: the whole wing pouch was peppered with low level phosphorylation of Mad, suggesting excessive background signalling activity. As a result, brk was repressed over a wider range than normal, and therefore expression of the Sal, Omb and DSRF proteins did not recapitulate the wild-type situation. Among other defects, vein tissue was found throughout the wings, and vein L5 was too thick.

The group reasoned that background signalling could be due to SecGFP leakage into the hemolymph, as in their first experiments. Indeed, doubling the expression of the Nb2highTkv and Nb1highPut receptors produced a sharply peaked pMad gradient and no background pMad speckling. However, due to the rapid drop-off of the activity gradient, the brk repression domain was too narrow, as were the resulting expression domains of the target genes omb and sal.

These data, combined with the gradient models derived by the Salbreux lab, suggest that at low receptor density, receptor activation is too low to trigger target gene activation; at intermediate density, leakage into the hemolymph occurs, triggering target gene activation too far from the source; and at high receptor density, hemolymph leakage drops, but the gradient scale shortens due to increased endocytosis and degradation in the tissue. In short, varying the density of the high affinity signalling receptors is not sufficient to create a broad gradient.

As mentioned above, morphogens also interact with components of the extracellular matrix, and glypicans, a family of GPI-anchored proteoglycans extensively decorated with heparan sulphate chains, have previously been shown to bind morphogens via electrostatic interactions. To mimic binding to these non-signalling receptors, a low affinity anti-GFP nanobody was fused to a GPI anchor and expressed in flies in the same region as the signalling receptors. Remarkably, in the presence of the two different types of artificial receptor, SecGFP could induce a morphogenetic gradient very similar to that generated by Dpp, restoring wing-disc patterning to near wild-type quality.

How can non-signalling receptors have such a significant effect? Salbreux’s PhD student Luca Cocconi ran simulations of multiple possibilities, first showing that addition of static membrane-tethered non-signalling receptors does not extend the morphogen gradient, as might be expected if their role was to slow SecGFP endocytosis and degradation. Rather, modelling indicated the importance of GPI-anchored proteins’ ability to move laterally within membranes, and perhaps hop between cells; in simulations lacking this diffusion component, the gradient was actually shortened by the presence of non-signalling receptors. Hopping has not been directly observed, but is feasible, and the Vincent lab are currently looking for proof of its existence. 

These remarkable experiments show that a diffusion-based artificial morphogen gradient can pattern a tissue in the presence of signalling receptors and low-affinity non-signalling receptors. The results are not only significant for developmental biology, but also open the way to using mathematical modelling and synthetic biology to design ways of regenerating injured or diseased tissues.