Studies in fruitflies support the idea that regulatory regions of genes control development by acting as molecular 'computers', calculating cell fate according to the combined effects of several signalling pathways.
Multicellular organisms are made up of a huge number of cell types, specialized to perform particular functions. During development, the fate of a cell is determined by outside signals that ultimately lead to specific changes in gene expression inside the cell. But, compared to the number of cell types, the number of signalling pathways is tiny, so each must be quite general. How, then, are these signals translated into specific patterns of gene expression? Three groups, writing in Cell1,2,3, address this problem in different developmental contexts. Their results strengthen a view of gene regulation4 as the deciphering of a combinatorial code, in which the quality and quantity of signals are directly sensed and computed by regulatory elements in the target gene.
Developmental decisions can be seen as a set of logical operations that integrate the effects of short- and long-range signals through the action of transcription factors4 (proteins that help to transcribe genes into messenger RNAs). Take, for example, the case of muscle formation in the fruitfly Drosophila melanogaster, where several types of signalling pathway result in the gene even-skipped being turned on in just one cell of a cluster of mesodermal cells (Fig. 1a). Mesoderm is the tissue from which muscle develops, and cells that express even-skipped go on to form particular types of heart and body-wall muscles. What is the molecular 'computer' that calculates the effects of the signalling pathways on the even-skipped gene? Halfon et al. 1 propose an answer, in the shape of an 'enhancer' — a regulatory region of the gene.
The authors1 find that this particular enhancer region is needed to ensure the expression of even-skipped in the requisite cell from a cluster of mesodermal cells. Within the enhancer region, the authors identify sites that bind regulatory factors positioned near the ends of receptor-tyrosine-kinase signalling pathways. They also find regions that bind to factors from the Wingless and Decapentaplegic signalling systems. Finally, they find binding sites for Twist, a protein specific to undifferentiated mesodermal cells, and for Tinman, a marker of dorsal mesodermal cells.
By altering these binding sites, or by changing the combination of regulatory factors that are expressed, Halfon et al. show that all of these signalling pathways, and all of the binding sites, are needed to achieve proper expression of even-skipped (Fig. 1b). Here, then, the combinatorial code involves the Wingless, Decapentaplegic and receptor-tyrosine-kinase signalling pathways, as well the tissue-specific expression of Twist and Tinman. The code is deciphered by the enhancer of the even-skipped gene.
In the Drosophila eye, cone cells, pigment cells and some types of photoreceptor cell develop from a common pool of precursor cells in response to different signals. The precursor cells use receptor-tyrosine-kinase pathways, sometimes in the context of signalling through the Notch pathway, to make developmental decisions. Flores et al.2 and Xu et al.3 show that the enhancer regions of target genes are again at the heart of these decisions.
Cone cells form from a cluster of 'equivalent' precursor cells that are responsive to the Drosophila epidermal-growth-factor receptor (DER; a receptor tyrosine kinase), and that express the transcription factor Lozenge. Signalling through the Notch pathway in some of these cells results in the expression of the transcription factor D-Pax2, and a fate as a cone cell. Flores et al.2 show that the combinatorial code for expression of D-Pax2 is deciphered by the enhancer region of the D-Pax2 gene, and involves Lozenge and proteins from the Notch and DER pathways.
Finally, Xu et al.3 study the specialization of so-called R7 photoreceptor cells, which involves changes in the levels of the prospero gene in a group of the precursor cells. The authors identify an enhancer in the prospero gene that ensures low levels of prospero expression, a feature of the early stages of R7-cell specialization. This enhancer also responds to DER signalling, and has binding sites for Lozenge. In other words, the combinatorial code that produces these early, low levels of prospero expression is deciphered at the prospero enhancer region.
Of the cells that express Lozenge, respond to DER and express low levels of prospero, those that pick up the Notch signal activate expression of D-Pax2 and become cone cells. However, those cells in which the Sevenless protein, a receptor tyrosine kinase, is activated, but in which Notch signalling is absent, express high levels of prospero — a feature of the late stages of R7-cell specialization — and become R7 cells3. Flores et al.2 confirm this: according to this model, Notch signalling in 'presumptive' R7 cells should convert them into cone cells, and this is what happens.
So it seems that the transcription factors that define domains in a tissue — such as Twist in undifferentiated mesoderm, or Lozenge in some precursor cells in the eye —are continually required in a given cell lineage. Some signalling pathways, such as those involving Wingless and Decapentaplegic, may be similarly required. They provide a background, or context, that is probably continuously sensed by the target gene. The 'final' choice of fate is determined by more general inductive signals, such as regulated receptor-tyrosine-kinase signalling acting within this background. The combinatorial logic allows the repeated use of a small number of elements to give a unique output. This way of controlling gene expression requires that all types of signal act directly at the target.
The next level of complexity is presumably provided by the existence of autoregulatory and feedback circuits that are kicked off by the protein product of the target gene. It will be interesting to find out how stable such circuits are. And a cell's response to signals is not limited to modifying gene expression. It will also modify proteins, for example, or move proteins around the cell, and these processes will add further layers of complexity.
Earlier analysis of gene regulation in sea urchins4 showed the possibility of modelling gene regulation as logic circuits, and revealed the computational power of enhancers. The latest three papers1,2,3 provide much-needed data to allow further development of such models. And as more gene circuits are unravelled, we may well find that the formal logic underlying their regulation has much to contribute to our understanding of other complex circuits, such as the neuronal circuits seen in the brain.
Halfon, M. S. et al. Cell 103, 63–74 (2000).
Flores, G. V. et al. Cell 103, 75–85 (2000).
Xu, C., Kauffmann, R. C., Zhang, J., Kladny, S. & Carthew, R. W. Cell 103, 86–94 (2000).
Yuh, C.-H, Bolouri, H. & Davidson, E. H. Science 279, 1896– 1902 (1998).
Carmena, A. et al. Genes Dev. 12, 3910– 3922 (1998).
About this article
Seminars in Cell & Developmental Biology (2017)
Synergistic regulatory networks mediated by microRNAs and transcription factors under drought, heat and salt stresses in Oryza Sativa spp.
Transcriptional profiling of Arabidopsis root hairs and pollen defines an apical cell growth signature
BMC Plant Biology (2014)
COPS: Detecting Co-Occurrence and Spatial Arrangement of Transcription Factor Binding Motifs in Genome-Wide Datasets
PLoS ONE (2012)