TL;DR
Gene duplication is a fundamental evolutionary process that contributes to genomic robustness. This study uses high-content single-cell imaging in yeast to reveal how proteins change their location and abundance when their paralogs are deleted, uncovering mechanisms of gene retention. The findings suggest that protein dependency on paralogs is more common than compensation, and that network interactions and alternative protein isoforms play key roles in this evolutionary dance.
What Problem Does This Paper Address?
The retention of duplicated genes (paralogs) after whole-genome duplication (WGD) is a major driver of evolutionary innovation, but the mechanisms by which these paralogs are maintained are not fully understood. While gene duplication can lead to functional divergence, it's unclear how proteins adapt to the loss of their paralogs. Previous research has explored changes in protein and RNA levels, and protein-protein interactions, but the dynamic changes in protein localization and abundance at the single-cell level in response to paralog deletion have been largely unexplored. This study aims to fill this gap by systematically investigating how proteins redistribute within cells when their paralog is absent, providing insights into the forces that govern paralog retention.
Methods
The researchers employed a high-content single-cell imaging approach in yeast. They constructed a large collection of yeast strains (328 strains) representing 82 WGD paralog pairs. For each pair, they created strains where one paralog was tagged with GFP and the other was either wild-type or deleted. These strains were then imaged using automated high-content microscopy.
A key methodological innovation was the use of a deep neural network (ResNet-18) for image analysis. This network performed single-cell segmentation and extracted 128-dimensional features representing protein localization and abundance. This allowed for a quantitative assessment of protein dynamics.
To quantify the response of a protein to its paralog's deletion, they calculated a "redistribution score," which is the Euclidean distance between the feature centroids of the protein in wild-type versus paralog-deleted backgrounds. This score captures combined changes in localization and abundance. They also separately quantified relative protein abundance changes.
Further analysis involved visual inspection for specific relocalization events and the investigation of non-canonical protein isoforms using translation initiation site (TIS) profiling data. Network analysis (protein-protein interactions and genetic interactions) was used to correlate redistribution with functional similarity and network properties.
Datasets
- Yeast strains: 328 yeast strains were constructed, representing 82 WGD paralog pairs. These strains were derived from the O'Shea collection and the GFP collection.
- Protein abundance data: Wild-type protein abundance values were compared to reference values from Ho, Baryshnikova, and Brown (2018).
- Protein-protein interaction (PPI) data: Downloaded from BioGRID (Oughtred et al. 2021).
- Genetic interaction data: Obtained from a previous study (Costanzo et al. 2016).
- Translation Initiation Site (TIS) profiling data: From Eisenberg et al. (2020).
Key Insights and Ideas
- Dependency is more common than compensation: When a paralog is lost, proteins are more likely to show "dependency" (requiring their paralog for normal localization/abundance) than "compensation" (adapting to fill the paralog's role). This suggests that paralogs often function as a unit rather than independently.
- Network context matters: Paralogs that exhibit significant redistribution are more likely to be closely connected in protein-protein interaction (PPI) and genetic interaction networks. This highlights the importance of functional redundancy and network rewiring in driving these dynamic responses.
- Non-canonical protein isoforms as a compensatory mechanism: The study identified a novel mechanism for compensation: the translation of alternative protein isoforms from non-canonical start sites. This can lead to altered localization and abundance, helping to buffer the loss of a paralog.
- Asymmetric divergence: The observation of asymmetric responses (one paralog responding, not the other) suggests that paralogs often diverge asymmetrically, leading to specialized roles or dependencies.
- Ribosomes are not homogenous: The study found that paralogs within the ribosomal subunits showed opposite responses, suggesting that even within large complexes, paralogs can have distinct roles and contribute to robustness in a non-uniform way.
Key Results
- Out of 164 proteins analyzed, 32 showed significant redistribution in response to paralog deletion.
- 30 proteins exhibited significant changes in protein abundance upon paralog deletion.
- Relocalization was visually confirmed in 9 proteins from 7 paralog pairs.
- Paralogs with high redistribution scores were more likely to have negative genetic interactions with their paralog and be closely located in PPI and genetic interaction networks.
- The Cue1-Cue4 paralog pair demonstrated compensatory relocalization of Cue4 to the cytoplasm and an increase in its abundance, linked to the translation of a non-canonical isoform, when Cue1 was deleted.
Limitations and Future Work
- Environmental conditions: The study was conducted under standard laboratory conditions. Future work could explore how paralog responses vary under different environmental stresses, revealing environment-specific paralog dynamics.
- Human cells: The study was performed in yeast. Extending this approach to human cells, where paralogs are more prevalent, could uncover novel mechanisms of paralog retention and potential therapeutic targets.
- "Ambiguous" annotations: Some paralog pairs were selected based on "ambiguous" annotations or distinct localization patterns. While this allowed for focused investigation, it might mean other paralogs with subtler differences were not captured.
- Non-canonical isoforms: While the study identified a non-canonical isoform for Cue4, further systematic investigation into the prevalence and functional impact of such isoforms across other paralogs is warranted.
- Therapeutic potential: The authors suggest that understanding compensatory paralog mechanisms could lead to novel cancer interventions by targeting paralog redistribution to selectively kill tumor cells. This is a promising future direction.
Summary updated