This web page was produced as an assignment for an undergraduate course at Davidson College.

Ingrid Wagnon's Genomics Home Page

Assignment 2: The impact of gene duplication on the protein interaction network of S. cerevisiae

The figures used on this webpage are from the original paper "Gene duplication can impart fragility, not robustness, in the yeast protein" by G. Diss et al. published in Science (2017).

        Summary

After the duplication of the genome, the two duplicates can evolve in different way. They can remain the same, gain or lose some functions. Thus they might become dependent from each other or they can be compensatory (if some functions still overlapping).

In this paper, they choose to work on a yeast model (S. cerevisiae) to measure the protein interaction network of 56 pairs of paralogous proteins. They used protein complementation assay (PCA) to measure the formation of proteins complexes, the size of the colony being proportional to the protein-protein interaction (PPI) score. They work with wild type cells (WT) or cells missing one paralog.

Overall the protein interaction network was maintain even in the cells lacking one paralog. However they observed that sometimes it was not the case and there was a phenomenon of dependency or compensation. In general pairs were asymetric, that means that dependency and compensation were rarely co-occuring within the same pair.

They focused on the mechanism of compensation. Two paralogs can interact with the same target but in general, one paralog is going to be favoured compared to the other. In general they found that compensation do not result in the increasing of the expression of the remaining copy, but more in a shift of the reaction. Indeed, when one paralog is deleted, the other has no competitor and can interact with its target (even if he wasn't the favoured one originally).

After that, they wanted to know more about the physical interaction of the paralogs and its role in dependency. Looking at the distribution of heteromers, they found that a lot of them are part of a dependent pair. They also noticed that within a pair, the dependent duplicates has a lower protein abundance. So one duplicates is less stable than the other.

This previous observation lead them to think from an evolutionary point of view. It seems that the paralogous heteromers would (in majority) came from an ancestral homomer (self interacting protein) and that's probably why this paralogs would develop a dependency. The dependent duplicate appears to evolve faster than its stable copy.

The authors of the paper want to highlight the contradiction in the effects caused by the duplication. The presence of two identical copies or compensating paralogs promotes the robustness of the species which can survive even with a mutation occurring in one of its copy. But a lot of paralogs also develop a dependency and can now be heteromers and need to interact physically with each other to be completely functional. The species is thus more sensible to mutations, and instead of increasing robustness, the duplication has increased fragility.


        My Opinion

I think that what was the most interesting in this paper was their position concerning the effect of duplication. We are used to look at the bright side of the duplication and how it is promoting the resistance and the refinement of the protein interaction network of species, but we are rarely looking at how it can also be a source of fragility.

They used a lot of schemes that were really helpful in the understanding of the paper (concerning the evolution of the paralogs after duplication). Overall, I found that the figures and their legends were clear, except for figure 2.C. and D. and figure 3.C. and D. that I thought were pretty challenging (especially figure 3) and were lacking information both in the figure, the legends and the text (for figure 3).

I also found that some words that they used would have deserved more explanations, so we could understand the paper more in details (like the words ohnologs or small-scale duplicate).

I think that we knew (more or less) that duplication had some bad side and that, logically, the specialization of the duplicates, their dependency and their heteromerization was a source of fragility. But I think that maybe it was important than someone highlighted it, so we can keep it in mind.

Evolution come with bright side but also with bad ones and we should be aware of it to keep a skeptical and objective point of view.


        Figure description


        Figure 1

In panel A and B, they are using schemes to explain how two duplicates of a same gene can evolve. So, after duplication, we can have two identical copies of the gene or some modifications can appear (gain or loss of function). The schema in panel A gives us three outcomes: 1. The duplicates have lost and gain functions in a way that they are now functionally independent; 2. The duplicates have lost and gain function but they share some redundant element(s); 3. The duplicates have developed an functional dependency to each other. In panel B we are looking at the evolution of the duplicates from a protein-protein interaction (PPI) network perspective. After its duplication, the network goes through gain and loss of function of the two paralogs. After the deletion of one paralog, we can distinguish two outcome: The lost of one protein can be compensated by the other paralog (and we have an increase in the protein-protein interaction), or the network cannot be restored because the paralogs are dependent.

Researchers wanted to study this phenomenon and choose the yeast model.

Panel C is showing us the interaction score in wild type yeast (x-axis) and in the yeast in which they deleted one paralog (y-axis). The interaction score is defined in function of the size of the yeast colony. In fact they did a protein complementation assay (PCA) in which the quantity of protein complex formed is correlated with the growth of the colony. If y<x significantly, we have a situation of dependency. If y>x significantly, we have a situation of compensation. In the majority of case, the PPI wasn't pertubated (In grey).

In panel D they are looking at the perturbation score in the different pair of paralogs. What we can see is that it’s rare that paralogs are concerned by both compensation and dependency.

 


        Figure 2

In a first time, researchers were interested in the phenomenon of compensation and wanted to know if it was the result of an increasing of the transcription of the remaining copy. In panel A, they used flow cytometry to measure the abundance of the proteins (tagged with green fluorescent protein) in the wild type yeast or the one with one paralog deleted. Except for four proteins, there was no significant change in the fluorescence between the two conditions (the points on the figure are following the diagonal). That means that the compensation by the augmentation of the production of the protein of the remaining paralog doesn’t seem to be a common mechanism.

To better understand the phenomenon of compensation, they overexpressed one of the paralog in the yeast and measured the interaction score (panel C and D). They hypothesized that the compensation effect could happen thanks to the fact that the products of the remaining paralog can now bind to their targets (which that they weren’t able to do before because the products of the other paralogs were “favored”). They found that in some cases, paralogs can mutually exclude each other as shown in C, where the overexpression of PMP3 leads to a decrease in the interaction score between Snap4p and Yeh1p. But in some other case the overexpression of one paralog doesn’t have a significant effect on the PPI (as shown in D for the binding of Yap1802p and Vma8p under the overexpression of YAP1801).

They compared compensated and non compensated PPI and noticed that non compensated PPI are more likely to be affected by the overexpression of one paralog (panel E).

 


        Figure 3

The researchers were interested to know the effect of the physical interaction between paralogs (to know if there was some kind of regulation).

In panel A they looked at the number of heterodimers and non heterodimers for the dependent and independent pairs. They found that there is more heterodimers in dependent pairs and there is not a lot of them in independent pairs.

In panel B they looked at asymmetric pairs (only one paralog is dependent from the other) and measured the abundance of the dependent (y-axis) and independent duplicates (x-axis). We can see that the independent duplicates tend to be more abundant than the dependent one.

Continuing on this idea they used flow cytometry to measure the fluorescence (and so the abundance) of GFP-tagged protein in wild type yeast and yeast missing one paralog. They distinguished the dependent and independent paralogs and found that dependent paralogs are the only one with significantly less fluorescence (compared to wild type).

Looking at different classes of genes and using protein complementation assay, they found that the deletion is more deleterious for heteromer small-scale duplicates (SSDs) than for singletons; and more deleterious for singleton than for heteromer ohnologs.

They argued that these results are coherent with the model in which the protein produced by a paralog are stabilized thanks to the second paralog (dependent relationship).



        Figure 4

They wanted to understand the origin of heterologous paralogs and had the intuition that they might come from an homomer ancestor that had evolve after the whole genome duplication (panel A). So they chose to study the model of S. pombe which is a species close to S. cerevisiae ancestor (didn’t go through whole genome duplication) (panel B).

When they compared S. pombe and S. cerevisiae orthologs, they found that around 19% of paralogous heteromers in S. cerevisiae had an homomer orthologs in S. pombe (panel C). And when they looked at different post whole duplication species with both ohnologs, they found that it is more likely that they came from a ancestral homomer (than a non-homomer) (panel D).

They compared the ratio of non synonymous over synonymous substitution for dependent and independent duplicates and found that the latter tend to be bigger for dependent duplicates (panel E). The dependent duplicates evolve faster than the independent one (it contain more non synonymous mutations).

In this figure, they are arguing that the duplication of the genome that leads the homomers to evolve into paralogous heteromers (which are dependent from each other) increases the fragility of species. Indeed, the loss of one duplicates has consequence on the other and the PPI network in which they are both involved. However the presence of paralogous heteromers  also promotes a faster evolution of the dependent duplicate.


Diss G, Gagnon-Arsenault I, Dion-Coté A-M, Vignaud H, Ascencio D, Berger C, Landry C. 2017. Gene duplication can impart fragility, not robustness, in the yeast protein interaction network. Science (335): 630-634.
http://www.bio.davidson.edu/genomics/2017/2017_gene_dup.pdf


Return to Home

 

Genomics Page
Biology Home Page

Email Questions or Comments: inwagnon@davidson.edu


© Copyright 2016 Department of Biology, Davidson College, Davidson, NC 28035