Embryonic geometry underlies phenotypic variation in decanalized conditions

A. Huang, T. E. Saunders

Preprint posted on March 16, 2019

Gene regulatory networks provide robustness against variation in embryonic geometry

Selected by Sundar Naganathan

Brief background on Drosophila segmentation

Segmentation refers to serial repetition of similar anatomical modules along the body axis, which results in formation of the head, thorax and abdomen in insects. In Drosophila this segmental pattern is laid down during the first 3 hours of development through a complex hierarchical network of regulatory interactions. A maternally provided mix of different mRNAs that have spatially distinct expression patterns provide the initial regulatory inputs for segmentation. For example, Bicoid (Bcd) protein forms an anterior-to-posterior gradient and Nanos (nos) forms a posterior-to-anterior gradient in the embryo. These maternal gradients activate downstream target genes called gap genes in a concentration dependent manner. Gap genes (e.g. hunchback (hb), Krüppel (Kr), knirps (kni), giant (gt)) are expressed in broad, overlapping domains about 10-20 nuclei wide1. Gap genes regulate expression of pair-rule genes (e.g. hairy, eve, runt), which are the first genes to be expressed in spatially periodic patterns (7-8 stripes about 4 nuclei wide) in the Drosophila embryo2. Finally, at the onset of gastrulation, the segment-polarity genes turn on, which are spatially regulated by pair-rule genes. The segment-polarity genes (e.g. en, odd, slp) show expression in 14 narrow stripes (about 1 or 2-nuclei wide), which are ultimately involved in positioning the morphological segment boundaries later in development.

 The question

Most species have abundant genetic variation and experience a range of environmental conditions and yet phenotypic variation is low. This ability of an organism to maintain a stable phenotype irrespective of genetic and environmental variability is termed as canalization. This concept, first developed by C. H. Waddington in 19423, has since been extensively explored mainly with respect to complex gene regulatory networks that have feed-forward and feedback loops, which ultimately provide robustness to the underlying biological processes.

In addition to genetic and environmental variability, are there other sources of variation that canalization works against to enable developmental robustness? To answer this question, Huang and Saunders chose the early Drosophila embryonic segmentation process and generated decanalized conditions to identify sources of variation that leads to inter-individual variability in phenotypes.

Key discovery

The authors used a previously generated bcd knockout (KO) mutant4, where the anterior embryonic cuticle patterning was entirely defective and the posterior patterning defects were more variable than wild type. This phenotypic variability is due to noisy gap gene expression patterns that emerge in the absence of a maternal Bcd gradient. Similar phenotypic variation was observed in embryos grown from a single pair of bcdKO parents suggesting that environmental conditions and genetic background contributes little to inter-individual variation in the conducted experiments. Interestingly, the variation in the number of segment polarity gene stripes in bcdKO embryos (performed for the gene engrailed, en) was highly correlated with embryonic length along the anterior-posterior (AP) axis and increased variation in gap gene expression patterns was observed with increasing embryonic length. This suggests that embryonic geometry could be a source of variation leading to inter-individual phenotypic variation.

Does a change in embryonic geometry alone alter embryonic pattern formation? To answer this, the authors targeted atypical cadherin Fat2 (fat2RNAi) in the follicle cells to generate embryos with significantly shortened embryonic length, though the total embryo volume only reduced by ~8%. These embryos hatched, with healthy larvae. Under these conditions, changes in gap gene expression patterns were observed with these changes correlating with embryonic length. This implied strongly that the segmentation network is highly robust to variations in embryonic geometry.

The authors then overexpressed bcd, a decanalized condition similar to the bcdKO condition, which also led to increased variability in En stripe patterns with decreased embryonic length. Similar phenotypes were also observed in the cuticle patterning. Interestingly, in embryos significantly shorter in AP length than wild type, the A4 cuticle segment was observed to be defective more frequently than other cuticle segments. Based on the expression profiles of the gap genes, the authors propose that the complex gene regulatory network is vulnerable at specific spatial locations in the embryo under decanalized conditions leading to defects in the A4 cuticle segment.

Why I chose this preprint

Embryogenesis is an intricately woven dance regulated spatiotemporally by complex gene regulatory processes. Decades of work on Drosophila early embryos have uncovered many of the hierarchical regulatory networks that pattern the segmented body plan. This volume of available data allows for investigating questions on robustness of gene regulatory networks and their evolution. By using this information, the work performed in this manuscript shows that gene regulatory networks have likely evolved to also provide robustness against varying embryonic geometry. Importantly, they also show that under decanalized conditions, e.g. when bcd is overexpressed, there are likely to be vulnerable points in the regulatory network leading to morphogenetic defects in precise spatial locations in the embryo.

Open questions

  1. Based on a qualitative description of gene expression profiles, the authors have predicted that there is likely to be a susceptible point in the gene regulatory network leading to biased phenotypes in the A4 cuticle segment. Given the volume of data available at the authors’ disposal, it will be wonderful to build or use previously published mathematical models to quantitatively predict this bias.
  2. The number of En stripes shows positive linear correlation with embryonic length in bcdKO embryos, whereas the number of stripes remains constant in fat2RNAi embryos. Does this mean that the number of stripes is scaled with embryonic length in bcdKO conditions, whereas the width of the stripes is scaled in fat2RNAi conditions ensuring constant number of stripes? What determines these different scaling processes?
  3. Significantly different changes in embryonic length have been achieved mainly through genetically altered conditions (bcdKO, fat2RNAi and bcd overexpression). In this scenario, it is not clear to me the authors’ claim of embryonic geometry as an “additional source of variation” that gene regulatory networks have to buffer against.


  1. Jaeger J., The gap gene network, Cell. Mol. Life Sci., 2011.
  2. Clark E., Dynamic patterning by the Drosophila pair- rule network reconciles long-germ and short- germ segmentation, PLoS Biol., 2017.
  3. Waddington C. H., Canalization of development and the inheritance of acquired characters, Nature, 1942.
  4. Huang A, et al., Decoding temporal interpretation of the morphogen Bicoid in the early Drosophila embryo, eLife, 2017.

Tags: canalization, fly, genetic networks, pattern formation, robustness

Posted on: 6th May 2019

Read preprint (No Ratings Yet)

  • Author's response

    T. E. Saunders shared

    1. This is a very interesting point. We have been using the model published by Bieler et al. (Biophysical Journal 2011) to test the models of gap gene patterning under geometry perturbations. Interestingly, we found that the model was robust to around a 30% reduction in embryo length before significantly departing from the wild type pattern. This contrasts with the Fat2 embryos, that remained precisely patterned even with 40-50% reduction in embryo length. It was hard to test how the model was failing, as values changed significantly around the break point. This tells us that the models are significantly less robust to geometric variation than the wild type system. However, to do a full exploration of the parameter space to uncover the fundamental weaknesses in the model is a project in its own right. I hope our work motivates theoretical biologists to explore the role of geometry more in gene patterning networks.
      As to why A4 is the most sensitive part of the network, the simplest answer may be that this is where the signals from anterior and posterior patterning networks are weakest. Therefore, this stripe may be most sensitive to variation in the network geometry. However, to really get at this requires exploration of the patterning in size reduced embryos combined with mutations in specific elements of the gap genes (e.g. Kruppel). Other groups have noticed that perturbations to the gap genes or the patterning inputs results in increased patterning variability (e.g. Janssens et al. Developmental Biology 2013). It will be interesting to revisit these and other results to test whether embryo size correlates with the observed phenotypes.
    2. In the bcdKO condition, the number of En stripes does scale with the embryo length. In the fat2RNAi conditions, some of the stripes do appear thinner, but is not a clear phenotype – there is quite a lot of variability in the En stripes, as shown in Figure 4D. However, the positioning of the stripes does scale in the fat2RNAi embryos. Scaling of the patterning appears to be via the gap gene network. However, Bcd appears to be a key input into this network to enable the gap gene to scale precisely. However, the precise mechanism of this remains unclear.
    3. When we look at natural variation in egg size in wild type populations, there is still a relatively large change – eggs vary by ±10% in length and volume. So, this is a source of variability that is potentially significant in early development. Second, there is a surprisingly large variation in egg size amongst closely related Drosophila species (Markow et al. Journal of Evolutionary Biology 2009). Drosophila Sechelia lay significantly larger eggs, while Drosophila pseudoobscura lay much smaller eggs, than Drosophila melanogaster. The early patterning networks in these embryos are similar. Therefore, the gap gene network appears to able to adapt to embryo size. It will be very interesting to probe how these different species have adapted the networks to ensure precise patterning with the changing geometries.

    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