Summary Morphogen gradients specify cell fates during development, with a classic example being the bone morphogenetic protein (BMP) gradient’s conserved role in embryonic dorsal-ventral axis patterning. Here, we elucidate how the BMP gradient is interpreted in the Drosophila embryo by combining live imaging with computational modeling to infer transcriptional burst parameters at single-cell resolution. By comparing burst kinetics in cells receiving different levels of BMP signaling, we show that BMP signaling controls burst frequency by regulating the promoter activation rate. We provide evidence that the promoter activation rate is influenced by both enhancer and promoter sequences, whereas Pol II loading rate is primarily modulated by the enhancer. Consistent with BMP-dependent regulation of burst frequency, the numbers of BMP target gene transcripts per cell are graded across their expression domains. We suggest that graded mRNA output is a general feature of morphogen gradient interpretation and discuss how this can impact on cell-fate decisions.

The Hunchback (Hb) transcription factor is critical for anterior-posterior patterning of the Drosophila embryo. Despite the maternal hb mRNA acting as a paradigm for translational regulation, due to its repression in the posterior of the embryo, little is known about the translatability of zygotically transcribed hb mRNAs. Here we adapt the SunTag system, developed for imaging translation at single mRNA resolution in tissue culture cells, to the Drosophila embryo to study the translation dynamics of zygotic hb mRNAs. Using single-molecule imaging in fixed and live embryos, we provide evidence for translational repression of zygotic SunTag-hb mRNAs. While the proportion of SunTag-hb mRNAs translated is initially uniform, translation declines from the anterior over time until it becomes restricted to a posterior band in the expression domain. We discuss how regulated hb mRNA translation may help establish the sharp Hb expression boundary, which is a model for precision and noise during developmental patterning. Overall, our data show how use of the SunTag method on fixed and live embryos is a powerful combination for elucidating spatiotemporal regulation of mRNA translation in Drosophila.

Morphogen gradients specify cell fates during development, with a classic example being the BMP gradient's conserved role in embryonic dorsal-ventral axis patterning. Here we use quantitative imaging and computational modelling to determine how the BMP gradient is interpreted at singlecell resolution in the Drosophila embryo. We show that BMP signalling levels are decoded by modulating promoter occupancy, the time the promoter is active, predominantly through regulating the promoter activation rate. As a result, graded mRNA numbers are detected for BMP target genes in cells across their expression domains. Introducing a heterologous promoter into a BMP target gene changes burst amplitude but not promoter occupancy suggesting that, while the promoter sequence controls amplitude, occupancy depends on the amount of BMP signal decoded by the enhancer. We provide evidence that graded mRNA output is a general feature of morphogen gradient interpretation and discuss how this can impact on cell fate decisions.

Concepts from dynamical systems theory, including multi-stability, oscillations, robustness and stochasticity, are critical for understanding gene regulation during cell fate decisions, inflammation and stem cell heterogeneity. However, the prevalence of the structures within gene networks that drive these dynamical behaviours, such as autoregulation or feedback by microRnAs, is unknown. We integrate transcription factor binding site (tfBS) and microRnA target data to generate a gene interaction network across 28 human tissues. This network was analysed for motifs capable of driving dynamical gene expression, including oscillations. Identified autoregulatory motifs involve 56% of transcription factors (tfs) studied. tfs that autoregulate have more interactions with microRnAs than non-autoregulatory genes and 89% of autoregulatory TFs were found in dual feedback motifs with a microRnA. Both autoregulatory and dual feedback motifs were enriched in the network. tfs that autoregulate were highly conserved between tissues. Dual feedback motifs with microRnAs were also conserved between tissues, but less so, and TFs regulate different combinations of microRNAs in a tissue-dependent manner. The study of these motifs highlights ever more genes that have complex regulatory dynamics. These data provide a resource for the identification of TFs which regulate the dynamical properties of human gene expression. Cell fate changes are a key feature of development, regeneration and cancer, and are often thought of as a "landscape" that cells move through 1,2. Cell fate changes are driven by changes in gene expression: turning genes on or off, or changing their levels above or below a threshold where a cell fate change occurs. "Omic" technologies have been successful in cataloguing changes in gene expression during cell fate transitions. Many computational tools have been developed for the ordering of gene expression changes in pseudotime, delineating cell fate bifurcation points and linking genes into networks 3-5. However, while we have a good understanding of the fates/states that cells transition through and their order in time/space, the mechanisms that allow cells to move through the fate/ state landscape are not well understood. Gene regulatory networks are maps of interactions between different transcription factors (TFs), cofactors, and the genes or transcripts they target 6. Networks are commonly represented diagrammatically as graphs of the connecting components, such as TFs and their targets. Network motifs are small repeating patterns found within larger networks 6. Modelling of networks in this manner allows us to develop an understanding of how components interact and what behaviours they may generate 6-9. Although it is clear that gene interactions are dynamic and change over time, current approaches in many biological studies focus on the qualitative analysis of genes or simple interactions between gene pairs: in short, how the perturbation of one gene affects the expression of another. However, gene expression is...

the autoregulatory loops (M1) are present more often than expected by chance, this could lead to all motifs appearing to be enriched. We therefore used two remaining 24% were detected in four or more tissues (Figure.5A).Where TFs have binding site data in more than one tissue, we asked whether their connections were conserved between those tissues. To this end, we investigated the conservation of M1 and M2 type motifs between tissues. M1 autoregulatory motifs are well conserved between different tissue types: the autoregulatory ability of over half of all TFs was conserved in 100% of tissues where data exist ( Figure.5B).This suggests that their biological function may be linked to their ability to autoregulate. The presence of M2 motifs was only slightly less well conserved between tissues (Figure.5C). When M2 motifs are not conserved between tissues, the most significant factor is loss of autoregulation -19% of all M2 motifs lose their autoregulatory signal between tissues, whereas only 1.3% of motifs are not conserved because they are missing TF regulation of the microRNA. However, even when the presence of the M2 motif is conserved, the identity of the components of Affiliations ContributionsSGJ and NP supervised this study. TGM, SGJ and NP conceived and designed this study. All computational and bioinformatics experiments and subsequent analysis were conducted by TGM. Ethics approval and consent to participateNot applicable. Consent for publicationNot applicable.