Communication codes in developmental signaling pathways

Abstract

A handful of core intercellular signaling pathways play pivotal roles in a broad variety of developmental processes. It has remained puzzling how so few pathways can provide the precision and specificity of cell-cell communication required for multicellular development. Solving this requires us to quantitatively understand how developmentally relevant signaling information is actively sensed, transformed and spatially distributed by signaling pathways. Recently, single cell analysis and cell-based reconstitution, among other approaches, have begun to reveal the 'communication codes' through which information is represented in the identities, concentrations, combinations and dynamics of extracellular ligands. They have also revealed how signaling pathways decipher these features and control the spatial distribution of signaling in multicellular contexts. Here, we review recent work reporting the discovery and analysis of communication codes and discuss their implications for diverse developmental processes.

Keywords: Communication codes; Pathway architecture; Signal processing.

Fig. 1.

Developmental signaling pathway architectures sense, process and control ligands in space and time. (A) Major developmental signaling pathways use diverse architectures to control cell-cell communication. In these pathways, ligand-receptor interactions activate intracellular effectors, which then regulate target gene expression. Intracellular signaling activity also induces a myriad of feedback loops to further modulate the signal processing capability of the pathway. TF, transcription factor. (B) At the single cell level, pathways can sense the molecular identity and concentration of individual ligands, the relative concentrations of multiple ligands (combinations) or the temporal dynamics of ligand concentrations. (C) At the tissue level, signaling pathways can actively modulate the distribution of extracellular ligands and intracellular signal perception in space and time. This spatio-temporal control occurs through direct ligand-receptor interactions, secreted inhibitors or modulators, and feedback loops (arrows inside each cell). The ruler and clock represent the spatial and temporal scales of signaling activity in morphogen gradients. Understanding the relationships between pathway architecture (A) and signal processing (B,C) is a fundamental challenge.

Fig. 2.

Discriminating ligand identity. (A) Many signaling pathways share a convergent structure, with ligands interacting with receptors in a many-to-many fashion and information ‘funneling’ down to control a smaller number of intracellular effectors. However, the shared effector can discriminate ligand-receptor identity to induce differential gene expression programs. (B) The Notch pathway ligands Dll1 and Dll4 both bind to the Notch1 receptor but activate the downstream effector NICD with different dynamics. Dll1 induces pulsatile responses, which preferentially activate the transcriptional target Hes1, whereas Dll4 induces sustained responses, which are required for activating Hey1. (C) Different RTK pathways share a common set of intracellular signal transduction components, including PI3K/AKT and RAS/ERK, but induce different cellular responses. For example, both EREG and EGF share the same receptor EGFR, but EREG promotes differentiation (Diff) whereas EGF promotes proliferation (Prolif) in MCF-7 cells. Based on population-level analysis, EREG elicits sustained phosphorylation of ERK (pERK) whereas EGF induces transient pERK (Freed et al., 2017). Similarly, EGF/NGF treatment of PC12 cells induces transient/sustained pERK that correlates with proliferation/differentiation (Marshall, 1995). (D) Single cell analysis has revealed a heterogeneous response of PC12 cells to NGF treatment. At the single cell level, these cell fate decisions depend on both pERK and phosphorylated AKT (pAKT), with a curved boundary separating proliferation and differentiation (left). Feedback loops within the NGF pathway (right) maintain the distribution of pAKT and pERK activities within the cell population close to the boundary. Adapted from Chen et al. with permission (Chen et al., 2012).

Fig. 3.

Different strategies for sensing ligand concentration. (A) In the simplest ‘amplitude modulation’ systems, ligand concentrations are encoded in the concentrations of intracellular effectors. (B) The TGF-β pathway can encode ligand concentration in the fold change, rather than the absolute level, of its effector nuclear Smad3 (S). At the single cell level, cells display highly heterogeneous levels of nuclear Smad3 both before and after TGF-β treatment (top panels), and the distribution of absolute Smad3 level overlaps significantly between different concentration groups (second row). However, fold changes of nuclear Smad3 (post-stimulus divided by its pre-stimulus level in the same cell) show less heterogeneity and better separated distributions between different concentration groups (third row). Furthermore, target gene expression correlates better with the fold change, [S(t_end)/S(t₀)] than with the absolute level, S(t_end), of nuclear Smad3 (bottom row). Dots represent the number of the same transcripts in individual cells treated with either low (gray) or high (blue) concentrations of TGF-β. (C) EGF induces coherent oscillations of pERK shuttling between the nucleus and cytoplasm in certain cellular contexts, and the concentration of EGF regulates the frequency of these oscillations. (D) In developing mouse and chick embryos, SHH secreted from the ventral side of the neural tube forms a concentration gradient. Cells at different positions encode different SHH concentrations into Gli activity profiles with distinct amplitude and duration. Gli activity is then decoded by the downstream fate-decision circuits, composed of multiple transcription factors. A two-dimensional map based on both the amplitude and duration of Gli activity determines the cell fate output, and thus the spatial domains of distinct progenitor cell types. D, dorsal; V, ventral. Adapted from Briscoe & Small with permission (Briscoe and Small, 2015).

Fig. 4.

Sensing combinations of multiple ligands in the same or orthogonal pathways. (A) Multiple BMP ligands can often bind to the same receptors. Pathway activity in this context can be a complex function of ligand combinations. Cells expressing the same receptors can compute different functions of distinct ligand combinations, including additive, ratiometric or imbalanced relationships between the two ligands (top row). Cells can also change their receptor expression profiles to compute different functions of the same ligand combinations (bottom row). For example, BMP4 and BMP9 exhibit an additive relationship in NMuMG cells, whereas knocking down BMPR2 in NMuMG cells or altering the BMPR expression profile entirely in a different cell type, such as mouse embryonic stem cells (mESC), completely changes the ligand relationship. (B) SHH and BMP set up anti-parallel gradients in the developing neural tube to specify several dorsal (D), intermediate (I) and ventral (V) neural progenitor fate domains. The ligand concentrations of the two orthogonal pathways control the activities of their canonical intracellular effectors, Gli and pSmad, which are subsequently decoded by a gene regulatory network, represented here in its abstract form. Different combinations of BMP and SHH concentrations lead to distinct cell fates: SHH-low/BMP-low produces intermediate fates, whereas SHH-high/BMP-high produces either dorsal or ventral fates in a stochastic manner but not intermediate fates, which suggests cells do not measure the relative level of the two ligands. Different dorsal progenitor fates are indicated by different shades of red, and different ventral progenitor fates are indicated by different shades of blue. Adapted from Zagorski et al. with permission (Zagorski et al., 2017).

Fig. 5.

Perceiving rates of change in ligand concentration. (A) Step-like but sustained TGF-β treatment leads to adaptive dynamics of intracellular signaling activity, which is measured by the level of nuclear Smad4. Sustained TGF-β exposure is less effective than pulsatile TGF-β exposure at inhibiting myoblast differentiation into myotubes, which is induced by a low level of serum in the culture. (B) In the same experimental system, gradually increasing TGF-β concentration activates the signal less strongly than a sudden step-like rise in the ligand concentration. The rate of change in ligand concentration correlates with response amplitude.

Fig. 6.

Reconstituted systems enable quantitative analysis of communication codes in space and time. (A,B) A 2D culture system of PSM cells can be used as an in vitro model for somitogenesis. The activities of multiple pathways, including Notch, Wnt and FGF, oscillate in the PSM. Whereas isolated individual PSM cells exhibit pulses of signaling activation in vitro, populations of densely packed PSM cells display synchronized oscillations in a density-dependent manner (A). The relative phases between Notch and Wnt signal oscillation differ at different locations within the PSM, with an anti-phase relationship in the posterior and an in-phase relationship in the anterior, which triggers segmentation (B). The phasing between two oscillatory signals can therefore encode spatial information. (C) A bottom-up morphogen gradient reconstitution system enables quantitative analysis of the causal relationship between pathway architecture and tissue patterning. By engineering mouse fibroblasts into morphogen-sending and -receiving cells and plating the two populations under defined spatial arrangements, gradients can form within the cell layer in a petri dish. Spatio-temporal dynamics can be quantitatively measured using time-lapse imaging. (D) Unique architectural features of the SHH pathway. PTCH receptor (purple) inhibits downstream signal and transcriptional targets (yellow) in the absence of SHH (blue) (1). SHH-PTCH binding leads to inactivation of PTCH and SHH (2), and thus activation of the downstream targets. Signal activation induces an evolutionarily conserved negative feedback through upregulation of PTCH (3), which both sequesters ligand extracellularly and inhibits signal intracellularly, and therefore is bifunctional (red arrows). (E) Rewiring the SHH pathway to explore different architectures and measuring the resulting gradients revealed different degrees of robustness to variations in SHH production rate: without feedback (minimal), the amplitude (the response in the first cell next to the source) and length of the signaling gradient are sensitive to an increase in the ligand production rate (second row versus first row); with the evolutionarily conserved PTCH feedback (natural), both gradient amplitude and length become more robust (second row versus third row); with intracellular negative feedback (rewired) from a mutant PTCH (orange) that does not bind SHH but suppresses the intracellular signal (Briscoe et al., 2001), gradient amplitude but not lengthscale becomes more robust compared with no feedback (last row versus second row). These results directly link pathway architecture to patterning behavior.

Fig. 7.

Active signal processing allows specificity and precision in cell-cell communication. Developmental signaling pathways can be viewed as programs that control message addressing (‘who can talk to whom’; 1), message content (‘which target program to activate’; 2), and message delivery (‘when and where the information should be received’; 3). The diverse signal processing schemes used by different pathways not only transduce signals but actively modulate them in ways that enable specificity and precision in multicellular development.