The hourglass model of evolutionary conservation during embryogenesis extends to developmental enhancers with signatures of positive selection

Introduction

Since the 19th century there have been many attempts to find a pattern when comparing morphological trajectories of different species during development. In the 1990s the Hourglass (HG) model was proposed (Duboule 1994; Raff 1996). According to the HG model we should find, when comparing the morphology of animals in the same phylum, high variance in early and late development but low variance in mid-development. The conserved mid-developmental stage is usually referred to as the phylotipic stage or phylotypic period.

Although initially proposed based on morphology, the HG model has received support from many comparative gene expression studies. These studies found that the transcriptome of different species (e.g. Drosophila flies) is more similar in mid embryonic development (Kalinka 2010). Some recent analyses have found that the protein coding sequence of genes expressed in the phylotypic stage show relatively low positive selection and strong puryfing selection (Liu 2018 , Coronado Zamora 2019). However, it is still unknown if selection signatures on regulatory regions would follow this pattern.

Liu and collaborators tackle this question using DNAase-seq to identify stage specific enhancers throughout the embryonic development of the fruitfly, investigating their conservation and looking for signatures of positive selection.

About the preprint

The authors performed DNase-seq to identify active regulatory elements across five matched embryonic development stages in two Drosophila species: D. melanogaster and D. virilis. Importantly, the five stages (TP1-TP5) include early, mid and late development, with stage TP3 within the phylotypic period.

Using a measure of enhancer conservation they found that TP3 (phylotypic period) has a higher proportion of conserved enhancers (Figure 1C). A high conservation in the sequence of enhancers can be, however, caused by strong purifying selection (in which most mutations are eliminated due to deleterious effects) or by a relatively weaker positive selection.

 

 

To test the latter scenario, they used a new approach that detects signatures of positive selection on genomic regulatory regions (Liu 2020). This approach is based on a Machine Learning algorithm that estimates the effect of nucleotide substitutions on chromatin stability (Gandhi 2014). More specifically, it consists of a gapped k-mer support vector machine (gkm-SVM) model. Support vector machine models are supervised learning models with associated learning algorithms that analyse data used for classification and regression analysis. Given a set of training examples, each marked as belonging to one or the other of two categories, an SVM training algorithm builds a model that assigns new examples to one category or the other. In this case the positive categories are stage-specific enhancers and the negative set is an equal number of sequences randomly sampled from the genome with similar characteristics (e.g. sequence length, nucleotide composition, etc). This gkm-SVM not only differentiates enhancer sequences from random sequences, but also identifies sequence features within the enhancer elements that determine their stage-specific occupancy.

Testing for positive selection

To identify signatures of positive selection, the authors used for further analyses only stage-specific enhancers with at least two substitutions between D. melanogaster and D. simulans. D. simulans is evolutionary closer to D. melanogaster (2.5 MY divergence) than D. virilis (40 MY divergence), being therefore a better option for this approach.

Using D. yakuba as an outgroup, they inferred the enhancer ancestor sequence and then calculated a deltaSVM score, which estimates the effect of substitutions on the gkm-SVM function score. This means that, if the substitutions that occurred in the D. melanogaster branch after divergence from D.simulans increased the chromatin accessibility of an enhancer, the deltaSVM score of this enhancer would be positive, and if the score is greater than expected by chance, it is assumed to have been positively selected.

The authors found that the phylotypic period (stage TP3) had a much lower proportion of enhancers with evidence of positive selection than the other stages. The pattern of conservation and positive selection on developmental enhancers seems then to support the Hourglass model (Figure 2).

 

Why I chose this preprint

Until now, the Hourglass model has been mostly tested using developmental transcriptomics data. This is the first study that tests the Hourglass model at the regulatory level trying to elucidate the role of developmental enhancers in the evolution of gene expression during embryogenesis. With this study, the authors help filling in a knowledge gap in the HG field.

 

Questions to the authors:

Q1: Instead of “U” shape expected from the HG model, it seems that the positive selection on enhancers shows an “M” shape. What do you think would be the cause for the relatively lower positive selection of the earliest and latest stages you analysed?

Q2: Do you think selective pressures on developmental processes could equally prevent positive selection on enhancers and codifying regions?

Q3: You mention that the higher number of enhancer in the phylotypic period suggests more redundancy and thus higher regulatory robustness in gene expression. Would it be possible to test this experimentally?

References

Duboule D. 1994. Temporal colinearity and the phylotypic progression: a basis for the stability of a vertebrate Bauplan and the evolution of morphologies through heterochrony. Development 1994:135–142.

Raff RA. 1996. The shape of life : genes, development, and the evolution of animal form. University of Chicago Press.

Kalinka AT, Varga KM, Gerrard DT, Preibisch S, Corcoran DL, Jarrells J, Ohler U, Bergman CM, Tomancak P. 2010. Gene expression divergence recapitulates the developmental hourglass model. Nature 468:811–814.

Liu J, Robinson-Rechavi M. 2018a. Developmental Constraints on Genome Evolution in Four Bilaterian Model Species. Genome Biol. Evol. 10:2266–2277.

Coronado-Zamora M, Salvador-Martínez I, Castellano D, Barbadilla A, Salazar-Ciudad I. Adaptation and Conservation throughout the Drosophila melanogaster Life-Cycle. Genome Biol Evol. 2019;11(5):1463-1482. doi:10.1093/gbe/evz086

Liu J, Robinson-Rechavi M. 2020. Robust inference of positive selection on regulatory sequences in human brain. bioRxiv. https://doi.org/10.1101/2020.03.09.984047.

Ghandi M, Lee D, Mohammad-Noori M, Beer MA. 2014. Enhanced Regulatory Sequence Prediction Using Gapped k-mer Features. PLoS Comput. Biol. 10:e1003711.

Tags: evodevo, hourglass

doi: https://doi.org/10.1242/prelights.26321

Read preprint (No Ratings Yet)