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

Gene duplication can impart fragility, not robustness, in the yeast protein interaction network

(Diss et. al., 2017).

yeast cells

Sacccharoyces cerevisiae yeast cells. Figure reproduced from Angelica, 2011.


    Intuitively, gene duplication is thought to provide robustness against deleterious mutations, because a one paralog may be able to compensate for the lost functionality of the other. However, this study supports that duplication may actually cause mutational fragility in some cases, because physical interaction and mutual regulation cause one paralog to be unable to function without the other.
    Researchers identified 56 pairs of paralogous proteins associated with a diverse range of biological functions in a yeast model. They mapped the protein-protein interaction (PPI) network of each protein and measured the intensity of PPI in a wild-type condition and a paralog-deleted condition. If the two paralogs exhibit compensation, PPI would increase in the deleted condition; if they exhibit dependency, the PPI would decrease (Figure 1B). PPI is generally conserved upon paralog deletion, but many cases exhibited significant compensation or deletion (Figure 1C), and compensation and dependency generally did not co-occur (Figure 1D). Most responses to paralog deletion were asymmetric; one duplicate's function was not dependent on or could not be compensated by the other.
    Data did not support paralog up-regulation as a mechanism for deletion compensation, but shift in binding equilibrium was supported as a possible mechanism. In wild-type conditions, the compensating paralog displayed lower PPI intensity compared to its dominant counterpart. Deleting the dominant paralog caused a shift to interaction with the compensating paralog without changing protein or mRNA expression. Tested by overexpressing the compensating paralog, this pattern was observed more for pairs that exhibited compensation (Figure 2E).
    Pairs that exhibited dependency were more likely to form heteromers (Figure 3A), introducing the question of the role of physical interaction of dependent paralogs. The dependent duplicate generally had lower protein abundance than the independent duplicate, especially when the independent duplicate was deleted (Figures 3B and 3C), and deletion of the independent duplicate had a greater impact on fitness. These data support a model of increased physical stability through physical interaction of the paralogs. Investigation of a larger paralog dataset showed that deletion of heteromer-forming paralogs has a larger effect on fitness than nonheteromeric paralogs, implying that they may work as functional units. Further, heteromer forming paralogs sequences and molecular functions are more similar. The data support a possible origin explanation of the duplication of an ancestral homomer (Figure 4A). Ohnologs, or paralogs resulting from whole genome duplication, that form paralogous heteromers are more likely to have a homomer ortholog in an ancestral species (Figure 4C), and ohnologs are more likely to be retained as a pair if they came from an ancestral homomer (Figure 4D). It is possible that dependency stabilizes the dependent paralog, accelerating mutations and evolution (Figure 4E). Overall, the study concludes that while many paralog pairs do exhibit robustness, many actually increase fragility because of paralog dependency.


    The paper was well-written, with convincing evidence, clear presentation of data, and interesting applications of simple methods. I appreciated the inclusion of explanatory figures; a diagram of the processes was much easier to understand than a written account. However, they often presented explanatory figures at the expense of data. Many large conclusions cited supplementary figures, included all of the paralogous heteromers co-function and human cell line data. While the explanatory diagrams made the paper more convenient to read, I would have appreciated seeing more data to support the conclusions in the diagrams.
   A few figures had confusing legends or misleading notes about significance, but they did not hugely distract from the data presentation. For example, the legend in Figure 3C uses a single color for the symbol representing nonsignificance, the same color used for independent pairs. They also only give one data point for independent paralog pairs, and the use of color understates how few of the dependent pairs actually show significant effect. Their conclusion is that independent paralog pairs do not show the same effect as dependent pairs, but this point would have been better supported if they had graphed data for both independent and dependent pairs, and then indicated that more of the dependent pairs showed a significant effect.
    I would suggest further research into which specific proteins are more likely to be compensatory or dependent, beyond ancestral considerations. Are certain gene ontology classes more likely to exhibit compensation? Are there environmental factors that influence whether compensation or dependency occurs?

Figure Summaries:

yeast cells

Figure 1: Figure 1A shows the possible fates of a pair of paralogs; they can either become functionally independent, retain some redundant elements, or become functionally dependent on each other. Duplication of a gene also affects its protein-protein interaction network, as the paralogs can independently lose or gain ability to interact with proteins. As seen in Figure 1B, when one paralog is deleted, compensatory activity presents as an increase in PPI intensity and dependent activity presents as a decrease in PPI. PPI was measured using a protein-fragment complementation assay; fragments of an enzyme necessary for growth in a restrictive medium are fused to proteins so protein interaction intensity can be measured as a function of colony growth. Figure 1C graphs PPI intensity in wild-type vs. paralog-deleted conditions. Interaction scores were strongly correlated, indicating that generally paralog deletion does not significantly affect PPI network, but several genes deviated from the best fit line, indicating significant compensatory or dependent relationships. Figure 1D shows that compensation and dependency typically did not occur in the same paralog pair.

yeast cells

Figure 2: Only a few cases showed a significant increase in protein levels via flow cytometry in the paralog-deleted condition (Figure 2A), which does not support paralog up-regulation as a mechanism. Figure 2B shows an alternative mechanism for compensation, where the paralogs are mutually exclusive; one paralog shows stronger interaction in the wild-type case because it is more abundant or has a higher affinity. If it is deleted, the other paralog is able to compensate. They tested this theory buy overexpressing the compensating paralog, expecting a decrease in the interaction of the originally dominant paralog. Figure 2C shows a case where this hypothesis was true; the interaction score of the protein in red decreased when its compensating paralog was overexpressed. Figure 2D shows a case where this hypothesis was not true; the interaction score of the protein in red was not affected by overexpressing of its paralog. Overall, paralogs pairs classified as "compensating" were more likely to show decreased PPI upon overexpression of the compensating paralog (Figure 2E)

yeast cells

Figure 3: Figure 3 examines dependent paralog pairs. Figure 3A shows that dependent paralog pairs are significantly more likely to be heteromers, or two proteins that physically interact. Figure 3B shows that in asymmetrical dependent pairs, where one protein can function without the other but not vice versa, the independent protein tends to have a higher abundance. The dependent protein abundance tends to lower further when its independent counterpart is deleted, as measured by flow cytometry and/or Western blots (Figure 3C). Figure 3D shows how different types of gene duplication affect how deletion of one paralog impacts fitness. Deletion of heteromer that originated from small-scale duplications was significantly more detrimental than deletion of singletons, or a non-duplicated gene, while deletion of ohnologs was less detrimental than deletion of singletons. In both cases, deletion of paralogs that formed heteromers had a larger negative impact on fitness than their non-heteromer counterparts, which supports that heteromers are working as functional units; if one is deleted, the other is unable to function.

yeast cells

Figure 4: Figure 4A illustrates three possible dimer configurations after whole genome duplication. Figure 4B illustrates the method of determining the ancestral origin of a current paralogous heteromer resulting from whole genome duplication (ohnomer), comparing the current paralog pair to its ortholog in a related species that did not undergo whole genome duplication. This method revealed that paralogous heteromers were more likely to have a homomer ortholog (Figure 4C), supporting the mechanism outlined in Figure 4A. Additionally, ohnologs that originated from a protein that formed homodimers are more likely to retain both paralogs (Figure 4D), further supporting functional dependence of this class of proteins. In this configuration, the dependent duplicate tends to accumulate more nonsynonymous mutations than its independent counterpart (Figure 4E), suggesting that the presence of an independent paralog allows for greater mutational freedom. Figure 4F summarizes the entire paper: paralogs pairs that originated from ancestral homologs become functionally dependent heteromers after duplication, causing loss of function if one paralog is deleted.

All figures reproduced from Diss et. al., 2017.


1. Diss, Guillaume, Isabelle Gagnon-Arsenault, Anne-Marie Dion-Coté, Hélène Vignaud, Diana I. Ascencio, Caroline M. Berger, and Christian R. Landry. "Gene duplication can impart fragility, not robustness, in the yeast protein interaction network." Science 355, (February 10, 2017): 630-644.

Genomics Page
Biology Home Page
Hartlee's Home Page

Email Questions or Comments:

© Copyright 2017 Department of Biology, Davidson College, Davidson, NC 28035