Menu

Close

The context-dependent, combinatorial logic of BMP signaling

Heidi Klumpe, Matthew A. Langley, James M. Linton, Christina J. Su, Yaron E. Antebi, Michael B. Elowitz

Preprint posted on December 08, 2020 https://www.biorxiv.org/content/10.1101/2020.12.08.416503v1

How do ligand-receptor dependent signalling pathways acquire contextuality and specificity?

Selected by Meng Zhu

Background:

 

Cells within a tissue communicate through signalling molecules. An important first step of this communication is the interaction between secreted signaling molecules, also called ligands, and  receptors on the cell surface. Cell-cell signaling can be classified into groups of ligands and receptors that together activate the same downstream effects. Well-studied examples of these include the Fibroblast Growth Factor (FGF), Wnt, and Bone Morphogenetic Protein (BMP) pathways. These pathways are functionally important in developmental, neural, and physiological contexts. In development, for example, interaction between ligands and receptors of these pathways is key to cell proliferation, growth, and differentiation, and the correct expression of these ligands and receptors in certain spatial and temporal patterns is key to embryonic patterning.

 

A fundamental feature of these signalling pathways is redundancy of signaling components – the genome of a multicellular animal can encode several ligands and receptors belonging to the same signalling pathway, and these members are often co-expressed in a certain combination for the various developmental stages or tissues where the pathway plays a role. As inferred by previous studies, a pathway’s many ligand variants can promiscuously bind nearly any receptor variant to activate the pathway, yet, in certain cell contexts, the ligand variants appear to have non-redundant roles. These properties contribute to the complexity of signalling interactions, but also raise several questions relating to the logic of signalling regulation given a certain ligand-receptor expression pattern. For example, can sequence homology between ligands or receptors determine their functional redundancy? And to what extent can different ligands and receptors replace each other? Addressing these questions can not only advance our understanding of signalling pathway interactions, but also help us to better interpret the expression profile of members of a certain signalling pathway in a specific tissue.

 

 

 

In this preprint, Klumpe et al. focused on the BMP pathway as an example to address these questions of signalling specificity. They used a systematic and quantitative approach to survey how combinations of ligands activate the BMP pathway and how these combinatorial properties vary between cell types. Using these data, they can compare the similarity of each ligand’s signaling activity and ask if cells perceive the same differences between ligands, or not. Finally, they used a mathematical model and simulated receptor-ligand binding to provide putative explanations for the functional similarities between different ligands and receptors.

 

Key findings:

 

To quantitively probe the combinatorial interactions between BMP ligands, the authors constructed a BMP reporter NMuMG cell line by hooking a Smad1/5/8 responsive element (which activates downstream transcription when the upstream BMP pathway is activated), with a fusion Histone 2B (H2B)-mCitrine.  The authors used this reporter line to first determine the individual behavior of 10 BMP ligands when activating BMP receptors expressed in NMuMG cells (ACVR1, ACVR2A, ACVR2B, BMPR1A, and BMPR2), and then moved on to quantify pairwise responses. The parameters that they initially focused on were half-maximal activation (EC50) and Relative Ligand Strength (RLS), which measures pathway activity at saturating concentrations normalized to the activity of the strongest ligand. However, the authors found that the ligands that display similar RLS can display differences in the way they interact with other ligands. To better evaluate the ligand equivalence, the authors defined a new parameter, the so- called interaction coefficient (IC), which compares the differences between the responses to a ligand when combined with itself or with another ligand. Based on the IC regime, the interactions between ligands in a pair can be classified into “saturated additive” (IC= 0), “negative interactions” (IC<0), and “positive synergy” (IC>1) (Figure).

 

The authors found that most pairs fell into the “saturated additive” category, whereas others ranged from antagonism to suppression, two kinds of negative interactions (IC < 0). Based on the interaction pattern, the BMP ligands can be grouped into [BMP2, BMP4], [BMP5, BMP6, BMP7, BMP9], [BMP10], [GDF5, GDF7], and [GDF6] (Figure).

 

 

Figure. The definition of Interaction Coefficient (IC) and pairwise analysis of BMP ligands based on IC.

 

 

 

To determine to what extent the interaction metric identified in the NMuMG cells holds true for other cell types, the authors applied the same approach to assay BMP ligand interactions in another cell type, the mouse embryonic stem cell (mESC). The authors found significant differences in single ligand binding parameters as well as pairwise interactions between different ligands. These results thus suggest that ligand interactions are context-dependent.

 

What leads to the contextuality? The authors hypothesized receptor expression to be a cause, as NMuMG and mESC cells differ significantly in their BMP receptor expression profiles. To address this, they first downregulated ACVR1 in NMuMG cells to make its expression level more similar to that of the mESCs. They found that this allows the narrowing of certain differences in the interaction metric between NMuMG and mESC cells, but not others. The authors then downregulated the two most highly expressed receptors, BMPR1A and BMPR2. Knocking down BMPR2, similar to ACVR1 knockdown, recapitulates a difference between NMuMG and mESC and generates some aspects of ligand equivalence observed in mESC. These results suggest that the expression of receptors can account to a certain extent for the how ligand interactions change between different cell types. Besides downregulation of highly expressed receptors, the authors also performed experiments to ectopically express the two lowest expressed receptors, ACVRL1 and BMPR1B. The authors found that the upregulation of these two receptors allows the removal of many of the suppressive effects of their primary binding ligands. These results suggest that the non-additive interactions, such as antagonism and suppression, may arise from receptor competition.

 

To better understand the ligand equivalence, the author generated a more global ligand equivalence picture by combining datasets  for different cell types and clustered the BMP receptors as 5 equivalent categories [BMP2, BMP4], [BMP5, BMP6, BMP7], [BMP9], [BMP10], and [GDF5, GDF6, GDF7]. By analysing the protein sequence, the authors found that, surprisingly, the ligand sequence similarity does not fully predict functional similarity in a single receptor context, though it did match the global functional similarity across all cell lines. Then what can account for the functional similarity, or ligand equivalence in general, and how it changes across contexts? The authors turned to a mathematical approach to simulate receptor-ligand binding scenarios. The model takes into account primarily the affinities to form a trimeric ligand-receptor complex and the limitation of receptor amount to generate ligand competition, and the level of activity for each type of complex. The parameters of the model were fitted using experimental measurements coming from individual and pairwise analysis performed above.

 

With best fit parameters, the model can recapitulate many of the experimental observations on differential behaviours of ligand groups, suggesting that these experimentally observed behaviours can be explained by a model of affinity-based competition. Furthermore, and more importantly, the model predicted the generation of high activity and low activity complexes between different combinations of BMP receptors and ligands. The low ligand activity is associated not with the formation of fewer complexes, but with the formation of many low activity complexes. The model also suggested that the change in BMP signalling by the change of single-receptor expression can result from direct or indirect effects as the receptor participates in the formation of a broad range of complexes. Finally, the model predicted that the addition of a second ligand can result in a shift in the activity of the first ligand, as the ligands influence each other’s ability to form complexes from shared components.

 

Overall, the study suggests several contextual mechanisms of which ligand equivalence to be controlled: 1) the expression of receptors, such as when BMP10 can only replace BMP9 in the presence of ACVRL1; 2) the composition of co-expressing ligands, such as when BMP4 and BMP7 can only replace each other in the absence of antagonising ligands such as GDFs. In general, this work addressed the mechanism of contextuality and is insightful for the studies of systems in which multi-component signalling pathways participate.

 

What I like about this preprint:

 

FGF, Wnt, and BMP pathways are the key players in a broad range of developmental contexts. They come on stage at different space and time to execute distinct functions. How these distinct functions can arise by the same category of signalling pathway is an important topic that I, as a developmental biologist, very much wish to understand.   This work provides a systematic investigation to this matter and provides profound insights into this question. A take-home message of this work is that the identity of a ligand and receptor that is present in a certain context is crucial, and thus one cannot simply use the number of ligands or receptors in a certain signalling pathway category as a readout for the level of activity for this pathway.

 

 

Posted on: 11th February 2021

doi: https://doi.org/10.1242/prelights.27355

Read preprint (No Ratings Yet)




Author's response

Heidi Klumpe shared

  1. The fact that protein sequence can poorly explain the function similarity is intriguing. Do the authors have ideas of any other factors at the biochemical level can account for ligand equivalence?

 

We hypothesize that sequence is a poor predictor of functional similarity in a certain cell context for two reasons. First, functional similarity ultimately depends on certain emergent parameters, such as the affinities of the ligands for the receptors and the activities of the resulting ligand-receptor complexes (as we explored in our model). Protein sequence is only a proxy for those parameters, as evidenced by the difficulty of building models that predict affinity from sequence. Second, the full ligand sequence contains a lot of information that a cell will “overlook,” because the cell’s receptors will only “see” the parts of ligand sequence they are sensitive to. More precisely, receptors do not contact every amino acid on the ligand when they are bound, so not every amino acid difference can create an affinity or activity difference that ultimately manifests in a unique response.

 

 

  1. Do the authors think the ability to activate certain downstream signalling cascade (such as Smad1/3/5) can also account for differential ligand behaviour (which is not considered as a primary factor in the current work)?

 

There are many interesting dimensions to this question! First, one factor that we consider in the model, but which was harder to probe experimentally, is the activity of each signaling complex (i.e. how well each ligand-receptor complex activates the downstream pathway). Other work on BMP signalling shows that ligands and receptors can bind each other and form “non-signalling complexes.” Based on our model fitting, we hypothesize that such less active complexes are key to producing many of the counter-intuitive, non-additive responses we observed. So an important difference between ligands is not only their affinity for the receptors, but how much Smad they can phosphorylate with those receptors!

However, as your question also points out, BMP ligands can produce many different outputs. The primary output is the phosphorylation of three transcription factors: Smad1, Smad5, and Smad8 (often written as Smad1/5/8). This is what we measured (indirectly) in the preprint. However, there is some evidence that BMP ligands can activate other Type I receptors usually considered reserved for a related set of proteins, like TGF-βs. These other Type I receptors phosphorylate Smad2 and Smad3, which are related to Smad1/5/8 but produce different downstream effects. Moreover, BMP ligands can activate other non-canonical outputs. These include PI3K, Erk, and MKK6, though the mechanism is not well understood.

Since producing these additional outputs makes use of the same pathway components (e.g. shared Type II receptors), ligand outputs measured in the “Smad1/5/8” channel could depend on what happens in these other output channels. In this case, the unique responses generated by some ligands in Smad1/5/8 could be a result of their ability to activate other outputs. However, the existence of other outputs suggests a further and intriguing possibility. BMP ligands which activate Smad1/5/8 in the same way are not guaranteed to activate other outputs in the same way. Therefore, there may exist other ligand differences that we have not yet detected!

 

 

 

 

 

 

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here
Close