The beauty of seeing the real thing: BMP heterodimer detection in vivo reveals dimer composition and activity

Abstract

BMP family ligands can direct cells to divide, differentiate, or die, depending on cell context and specific hetero- or homodimer combinations. In this issue of Developmental Cell, Bauer et al. detect endogenous Drosophila ligand dimers in situ and show that BMP dimer composition affects both signaling range and activity.

During development and tissue homeostasis, individual cells are subject to signals from their immediate neighbors and long-range signals secreted by distant cells. Several classical signaling pathways are used repeatedly in different combinations and contexts to induce specific cell fates and pattern tissues via activation of distinct downstream events.^1,2 The bone morphogenetic proteins (BMPs) are part of the TGF-β superfamily and are one such group of signals that can direct cell division, differentiation, or death, depending on the context, during many aspects of development in invertebrates and vertebrates.³ Dysregulation of BMP signaling has been linked to human disease, for instance cancers in which they have been described as both oncogenes and tumor suppressors.⁴

One reason for the wide-ranging effects of BMP signaling in development and disease is the large number of BMP ligands and receptors and the requirement for these molecules to form multimeric complexes, at the level of both the ligands and the receptors.⁵ How each of these complexes exerts different effects is almost impossible to dissect in mammals because of the thousands of possible combinations of ligand and receptor multimers. While computational modeling demonstrates that a single cell can perform different computations based on ligand identities,⁵ it has been impossible to expand this analysis in vivo. In Drosophila there are many fewer BMPs and receptors, making the task of addressing such questions feasible. In an exciting article in this issue of Developmental Cell, Bauer et al.⁶ develop techniques to detect endogenous BMP dimers and determine how each dimer combination can activate signaling, long-range and short-range, through heteromeric receptor complexes.

Compared to mammals, which have 20 BMPs, Drosophila melanogaster has only three: decapentaplegic (dpp), (BMP2/4 ortholog), and screw (scw) and glass bottom boat (gbb), (BMP5/6/7/8 orthologs).³ During signaling, two BMP dimers bind to heterotetrameric receptor complexes comprising two BMP type I receptors (Saxophone and Thickveins) and two BMP type II receptors (Punt and Wishful thinking). These ligand-receptor complexes are able to activate the same downstream signaling events. It has long been possible to examine the expression patterns of BMP signaling components and to analyze gain- and loss-of-function phenotypes. However, BMP localization studies have largely been limited to (over)expressing fluorescently tagged versions that disrupt protein function and ligand ratios.

Using small epitopes, CRISPR technology, and specific protein binders, Bauer and colleagues⁶ identify endogenous ligand dimers, manipulate their behavior, and examine how different dimers behave in the absence of specific receptors. They take advantage of the well-established model of the developing Drosophila wing, the wing imaginal disc (Figure 1A). Within the wing disc, dpp is expressed in a stripe of cells at the antero-posterior (A/P) boundary, whereas gbb is expressed throughout the disc, and scw is not detected. Using endogenously tagged proteins and a specific immunofluorescence protocol, they show that extracellular (secreted) Dpp and Gbb have the same distribution: high in the A/P stripe and a gradient moving away from the expression borders.⁶ Through spatially restricted knockdown of gbb, the authors demonstrate that secreted Gbb originates from the dpp-expressing stripe, and not the rest of the gbb-expressing cells. Further experiments suggest that in the absence of Dpp, Gbb is retained in the endoplasmic reticulum and so can only be secreted as a Dpp-Gbb dimer from within the dpp expression stripe. Using a protein binder “HATrap,” the authors trapped HA-tagged Dpp, showing that Gbb also colocalized on the HATrap-expressing cells. Further, the authors used an elegant approach by expressing not only HA-Gbb but also Gbb-OLLAS (a different epitope tag) to try to detect HA-Gbb-Gbb-OLLAS homodimers with HATrap (Figures 1B-1D). No Gbb homodimers were detected, and, similarly, no Dpp homodimers were found. Using this same technique but in a gbb mutant background, Dpp homodimers were detected, suggesting Dpp can homodimerize in the absence of Gbb, but Dpp-Gbb heterodimers are the main bioactive secreted “BMP ligand” in wild-type wing discs. These results are consistent with the more severe phenotype in dpp rather than gbb mutants.

Figure 1. BMP dimer distribution in wing imaginal discs.

(A) BMP signaling activity in wing imaginal discs. Dpp (blue) is expressed in a stripe at the A/P border of the wing disc, whereas Gbb (red) is expressed throughout the disc (inset: wing disc schematic; gray line indicates A/P cross-section where activity and protein gradients are analyzed). Gradient schematics for Dpp-Gbb heterodimers (purple) and Dpp-Dpp homodimers (blue) are indicated with Y axis showing protein levels; X axis reflects extend in the A/P-axis (gray line in inset). Note that in wild-type, BMP activity is a gradient from its peak at A/P border to the edges of the disc (purple, Dpp-Gbb dimers), whereas in gbb mutants, the gradient is reduced to the center of the disc with a sharp cutoff (Dpp-Dpp dimers). See also (E).

(B) Dpp and Gbb heterodimerize in the endoplasmic reticulum and are secreted. Using HATrap (gray) in cells expressing Dpp and Gbb, Dpp-HA is trapped and retains Gbb-OLLAS but does not retain Dpp-OLLAS. Similarly, Gbb-HA does not retain Gbb-OLLAS.

(C and D) In gbb mutants, Dpp-Dpp homodimers form and can be captured by HATrap (C), whereas in cells that do not express Dpp, Gbb-Gbb homodimers are not secreted (D).

(E) In wild-type cells, Dpp-Gbb heterodimers are secreted and can reach distant cells, whereas in gbb mutants, Dpp-Dpp homodimers are secreted but only travel short distances.

With these techniques established, the authors then asked how dimer composition, Dpp-Gbb or Dpp-Dpp, affects gradient distribution and signaling activation. Analysis of the Dpp gradient showed a much sharper cutoff, closer to the A/P boundary, in gbb mutants than in wild-type (comparing Dpp-Dpp homodimers to Dpp-Gbb heterodimers). Therefore, Dpp binding to Gbb is essential to shape the long-range signaling gradient (Figure 1E). Dpp homodimers were also less efficient at activating downstream signaling. The authors postulated that ligand-receptor interaction might be affected by dimer composition. Through genetic manipulation, they assessed whether loss of type I receptor Saxophone or type II receptor Punt affected Dpp-Gbb and Dpp-Dpp dimers differently. Interestingly, while Saxophone is required for optimal Dpp-Gbb signaling, it inhibits Dpp-Dpp activity. In contrast, Punt is necessary for peak Dpp-Gbb signaling and for any Dpp-Dpp signaling activity at all. Thus, in wild-type, Dpp-Gbb are the only dimers that are secreted, and they require heterotrimeric receptor complexes containing both type I receptors, Thickveins and Saxophone, and the type II receptor Punt for peak signaling activity. This analysis of BMP signaling is a great advance for the field given the capability of dissecting the specific requirements for ligands and receptors in such fine mechanistic detail in vivo.

Overall, this approach allows investigation of endogenous BMP signaling in other contexts, such as dorsal patterning of the Drosophila embryo,⁷ to test whether Dpp hetero- and homodimer properties are conserved. Screw is expressed in the embryo, and, unlike Gbb, Scw-Scw homodimers are secreted and functional.⁷ One aspect that is not addressed here are proteins that regulate extracellular distribution of BMPs during development, such as Sog and Tsg,⁷ and how this might relate to gradient formation in the wing disc. Although there are 20 vertebrate BMPs, not all are expressed in the same tissue at the same time. It is exciting to think this technique could be used where there are a limited number of active BMPs to investigate early patterning in the fish or frog embryo.^3,8

HATrap is part of a growing repertoire of genetically encoded binders that can be used to manipulate protein localization and function in vivo. Trapping extracellular proteins, altering the intra-cellular localization of proteins, or even targeting them for temporally and spatially restricted degradation are methods that open up new avenues for experimental design.⁹ The technique used here depends on small epitopes, which could also be applied to other signaling pathways with secreted ligands, such as Wnts and other growth factor families.