In early embryos, a concentration gradient of the Bicoid protein affects pattern formation. Studies of living embryos reveal a surprising level of accuracy in the Bicoid gradient. But is it accurate enough?
A central idea in developmental biology is Lewis Wolpert's theory of positional information1. This states that a substance present in a concentration gradient induces different developmental fates in cells when present at different concentrations. The first such morphogenetic gradient to be identified was that of the gene transcription factor Bicoid in embryos of the fruitfly Drosophila melanogaster2,3. This protein is distributed with an exponential profile, with its concentration decreasing towards the posterior pole of the embryo. Although the importance of the Bicoid gradient in specifying cellular fates was established, quantitative puzzles remained. These problems have now been largely solved by Gregor and colleagues4,5 in two papers in Cell.
A previous study6 had shown that the Bicoid concentration gradient varied far more widely between embryos than did the expression of the hunchback (hb) gene, which is used as a readout of the effect of Bicoid concentration. This and other studies, however, were performed in fixed tissue, where it is impossible to determine absolute protein concentrations or to follow changes in gene expression over time.
Gregor et al.4 tagged Bicoid with enhanced green fluorescent protein (eGFP), which allowed them to directly observe its gradient in live embryos. For this, the authors constructed a genetic line of fruitflies in which the bicoid gene (bcd) was replaced by a functional bcd–egfp fusion gene. They then monitored the gene's protein product by time-lapse microscopy during the blastoderm stage of early embryonic development.
In early Drosophila development, the embryo is a syncytium — it consists of a mass of cytoplasm, with nuclei that are not separated by cell membranes. The nuclei undergo a series of 13 rapid divisions, with the blastoderm forming at about division 10. The authors found that it is at division 9 — before blastoderm formation — that Bicoid–eGFP is first detected. As it is a DNA-binding protein, Bicoid is localized in the nucleus. But as nuclei lose their envelopes during each division, internally stored Bicoid is released into the cytoplasm.
Gregor and colleagues show that there is a remarkable constancy in nuclear Bicoid concentration between nuclear divisions, with peak concentrations varying by less than 10% at any given anterior–posterior position within the embryo. Evolution has thus provided a startlingly precise mechanism for preserving positional information in nuclei, in the face of repeated dissolutions and restorations of their membranes.
The authors also found that a photo-bleached nucleus regains its fluorescence within a few minutes, indicating that nuclear Bicoid is in tight equilibrium with its cytoplasmic pool. They established a simple model of diffusion-limited active transport and used it to predict the value of a diffusion constant for Bicoid in the cytoplasm. In observing the recovery of fluorescence in the photo-bleached region of the cytoplasm, they confirmed this prediction.
The Bicoid gradient was thought to be exceptional in terms of regulating early patterns of Drosophila gene expression because its distribution can be determined using a diffusion equation that assumes there is a constant source of Bicoid at the anterior pole and first-order decay of diffusing Bicoid7,8. But Gregor and colleagues' results introduce serious complications in this picture.
The authors found that the Bicoid gradient reaches a steady state in 90 minutes. This implies, using the diffusion equation, that the diffusion coefficient is more than 2 µm2 per second; however, photo-bleaching experiments indicate a value of only 0.35 µm2 per second. It may be that diffusivity varies on different timescales, or perhaps is affected by the nuclei themselves. Although the authors provide some evidence for the second possibility by observing that the Bicoid gradient is altered in unfertilized eggs, a true solution to this problem must await further experiments.
In their second paper, Gregor et al.5 investigate just how precise the readout of the Bicoid effect can be. They find that, by the end of the blastoderm stage, certain gene-expression patterns are specified to a resolution of one nucleus, which corresponds to a 10% difference in Bicoid concentration between adjacent nuclei. Can such small differences be perceived by the embryo?
To be detected by the cell, Bicoid molecules must diffuse to a cellular receptor. The theory of this diffusive process was worked out for bacterial chemotaxis — movement along a chemical concentration gradient — many years ago9, when it was established that the limit of detection depends on the square root of the product of the signalling-molecule concentration, its diffusivity, the size of the receptor, and the time period over which the concentration is averaged. This relation is completely general in physical details, as long as the processes involved do not dissipate energy. Thus, the accuracy of detection rises with the square root of the number of molecules sensed.
To apply this formula, Gregor et al. measured the absolute concentration of Bicoid by placing embryos expressing Bicoid–eGFP in a bath containing GFP at a known concentration, thus assigning an absolute concentration scale to their measurements. They found that the nuclear concentration of Bicoid was about 8 nanomolar in the centre of the embryo, which amounts to about 700 molecules per nucleus. In this region of the embryo, Bicoid must be averaged over a period of about 2 hours for a 10% difference in concentration to be detectable. The actual timescale is much shorter because boundaries of gene-expression domains form over a period of 7 minutes. These timescales imply a Bicoid discrimination threshold of 20–40% in neighbouring nuclei.
Embryo-to-embryo variations in gene expression also provide information on the actual discrimination threshold. The anomalous positional accuracy described in earlier work6 had two main components. First, when measured as a percentage of egg length, the range in position of Bicoid concentration thresholds for activation of the expression of the hunchback gene (xbcd) was found to be six times that for the hb border it nominally controls (xhb). Moreover, xhb correlated with embryo size, whereas xbcd did not, raising the question of how the Hunchback border scales with the size of the egg.
Gregor et al. present in vivo data from 15 live embryos imaged side by side. These data indicate that, based on absolute distance — rather than percentage of egg length — the range in xbcd is about twice that of xhb (Fig. 1). I think these results indicate that the anomalous positional-accuracy problem still exists, but that the anomaly is smaller than was thought. The small range of xbcd in absolute distance units may mean that the apparently different scaling properties of xbcd and xhb were a fixation artefact, a point that has serious implications for theoreticians.
Gregor and colleagues argue that a 10% change in Bicoid concentration is detectable and that the Bicoid gradient is sufficiently accurate to be used for specifying the position of the border of hb expression in the anterior domain. To make this point, they marshal an intricate set of quantitative arguments that are ultimately unconvincing, because they are based on a picture in which Bicoid is the only input to hb expression. This assumption is demonstrably false. The mean position of xhb is altered in embryos that have mutations in giant and other gap genes — genes involved in allocating domains in the insect embryo. Furthermore, the variance of xhb is doubled when one chromosome arm is removed6. Although such effects are much smaller than is seen for other gap genes, it is dangerous to ignore them in a study that aims to obtain a complete quantitative characterization of the control of hb expression.
The most radical elements of the authors' conclusions are not well supported. But there is no question that this work is a landmark that may prove to be as revolutionary as were the methods for imaging protein and RNA in fixed tissue that were developed 25 years ago. Moreover, these findings confirm that it is unlikely that either experimentalists or theoreticians will run out of fascinating phenomena to investigate in the Drosophila embryo any time soon.
Wolpert, L. J. Theor. Biol. 25, 1–47 (1969).
Driever, W. & Nüsslein-Volhard, C. Cell 54, 83–93 (1988).
Driever, W. & Nüsslein-Volhard, C. Cell 54, 95–104 (1988).
Gregor, T., Wieschaus, E. F., McGregor, A. P., Bialek, W. & Tank, D. W. Cell 130, 141–152 (2007).
Gregor, T., Tank, D. W., Wieschaus, E. F. & Bialek, W. Cell 130, 153–164 (2007).
Houchmandzadeh, B., Wieschaus, E. & Leibler, S. Nature 415, 798–802 (2002).
Houchmandzadeh, B., Wieschaus, E. & Leibler, S. Phys. Rev. E 72, 061920 (2005).
Gregor, T., Bialek, W., de Ruyter van Steveninck, R. R., Tank, D. W. & Wieschaus, E. F. Proc. Natl Acad. Sci. USA 102, 18403–18407 (2005).
Berg, H. C. & Purcell, E. M. Biophys. J. 20, 193–219 (1977).
About this article
Frontiers in Endocrinology (2018)
Scientific Reports (2016)
Current Opinion in Genetics & Development (2012)
Cellular and Molecular Life Sciences (2011)
Gene Expression Noise in Spatial Patterning: hunchback Promoter Structure Affects Noise Amplitude and Distribution in Drosophila Segmentation
PLoS Computational Biology (2011)