Gene expression of functionally-related genes coevolves across fungal species: Detecting coevolution of gene expression using phylogenetic comparative methods

- Gene expression of functionally-related genes coevolves across fungal species:.
- detecting coevolution of gene expression using phylogenetic comparative methods.
- We examined the effects of number of protein interactions and gene expression levels on coevolution, finding both factors are overall poor predictors of the strength of coevolution between a protein pair.
- Simulations further demonstrate the potential issues of analyzing gene expression coevolution without accounting for shared ancestry in a standard hypothesis testing framework..
- Conclusions: Our work highlights potential benefits of using PCMs to detect gene expression coevolution from.
- genes which show similar gene expression patterns across conditions are hypothe- sized to be functionally-related [2–5].
- Most of the previous work examining coevolution of gene expression relied upon the Codon Adaptation Index (CAI) [10] as a proxy for gene expression.
- This makes detecting signals from empirical measures of gene expression, such as from RNA-Seq or mass spectrometry data, particularly useful for many species where codon usage metrics are a poor proxy for gene expression.
- Recent work concluded comparative analysis of gene expression data across species can be confounded by the phylogeny, leading potentially to incorrect inferences [16]..
- Previous work examining coevolution of gene expression did not directly account for the phylogeny when esti- mating correlation coefficients of gene expression across species, which is thought to reflect the strength of coevo- lution between gene pairs.
- Many PCMs have been developed for studying the evolution of gene expression, although this work has not focused on detecting coevolution of gene expression Bedford et al.
- Much of this work relies on model- ing gene expression evolution as an Ornstein-Uhlenbeck (OU) process [28, 29].
- We find physically-interacting proteins show, on aver- age, stronger gene expression coevolution than randomly- generated pairs of proteins using the multivariate BM approach.
- We also find phylogenetically-uncorrected correlations tend to inflate estimates of gene expression coevolution.
- As expected, we find protein pairs with stronger evidence of functional-relatedness show stronger coevolution at the gene expression level.
- We also find gene expression level and the number of pro- tein interactions, which are considered good predictors of evolutionary rate of a gene [33], are poor predictors of the strength of coevolution between protein pairs.
- Con- sistent with previous results, we also find coevolution of gene expression is an overall weak predictor of protein sequence coevolution..
- Overall, the normalized gene expression data are moderately to strongly correlated between all species (Additional file 1, Figure S1).
- Clearly, species which are more closely-related tend to show stronger correlations between normalized gene expression values, consistent with expectations.
- Interacting proteins demonstrate clear coevolution of gene expression.
- This indicates gene expression within tightly- linked groups of physically-interacting proteins show greater signals of coevolution than between proteins which spuriously interact..
- Gene expression evolution was modeled as a multivari- ate Brownian Motion (BM) process using the R package mvMORPH [37] in order to estimate coevolution of gene expression between pairs of proteins.
- Binding proteins have a mean phylogenetically-corrected correlation of ρ ¯ C = 0.45, which is significantly different from the expected value of 0.0 if there was no coevolution of gene expression (One- sample t-test, 95% CI p <.
- 10 −37 , see Methods), indicating interac- tions which are more likely to be true and conserved show stronger coevolution of gene expression (Fig.
- with stronger coevolution of gene expression between the interacting pairs.
- Simulations were performed to confirm potential prob- lems with the use of non-phylogenetic methods for com- paring gene expression across species (see Additional file 1).
- 10 −37 ) between the STRING score and phylogenetically-corrected correlation ρ C indicates more confident and/or conserved interactions tend to have higher ρ C , indicating stronger coevolution at the gene expression level.
- Comparison of 4 methods for detecting coevolution of gene expression using data simulated under Brownian Motion.
- Gene expression and number of interactions are poor predictors of coevolution of gene expression.
- It is well-established both gene expression and loca- tion in a protein-protein interaction network significantly impact the evolutionary behavior of a protein [38–42]..
- One might expect an imbalance in the number of pro- teins involved in a greater number of interactions or more highly expressed interactions to have a more negative impact on fitness, leading to greater constraints on the evolution of gene expression.
- However, we find both the number of interactions and the gene expression to be weak predictors of the strength of coevolution of gene expression.
- indicating protein pairs involved in more interactions tend to show stronger constraint on the evolution of gene expression..
- Surprisingly, the mean ancestral gene expression esti- mates are negatively correlated with the phylogenetically- corrected correlations ρ C , with ρ S = −0.09 (Fig.
- Given phylogenetically-corrected correlations ρ C corre- late with the number of interactions and mean ancestral gene expression, differences between the binding and control groups in terms of number of interactions and gene expression could introduce small biases when com- paring the ρ C distributions.
- The average mean ancestral gene expression estimate distributions for the binding and control group are extremely similar (0.414 vs.
- This makes dif- ferences in the gene expression distributions an unlikely source of bias when comparing the binding and con- trol groups.
- 4 Effects of number of interactions and gene expression on strength of coevolution.
- The relationship of a the mean degree (average number of interactions between a protein pair) and b mean ancestral gene expression estimate with the phylogenetically-corrected correlation ρ C for the binding group.
- 0.09 (p for mean ancestral gene expression.
- This suggests both metrics are poor predictors of the strength of coevolution of gene expression between protein pairs.
- Despite this, the overall inter- pretation is the same: interacting proteins show greater coevolution at the gene expression level than randomly generated pairs of proteins..
- Coevolution of gene expression weakly reflects coevolution of protein sequences.
- tion at the gene expression level based on CAI [7, 8]..
- We also found a significant correlation between our phylogenetically-corrected correlation ρ C and the mea- sure of gene expression coevolution from Clark et.
- time, making CAI a reliable proxy for gene expression in these cases..
- A broad-scale analysis based on the Covariance Ratio test [34, 35] found coevolution of gene expression was stronger within groups of tightly-linked protein interactions compared to coevolution between proteins with weaker or no interactions (Covariance Score.
- Consistent with this, we find physically-interacting proteins show a clear signal of gene expression coevolution compared to randomly-generated pairs of proteins, with mean phylogenetically-corrected correlations ρ ¯ C of 0.45 vs.
- We also find the number of protein-protein interactions a pro- tein is involved in and its gene expression level – two common metrics known to affect the evolution of protein sequence – are overall weak predictors of gene expression coevolution.
- Sur- prisingly, highly expressed protein pairs actually tended to show weaker coevolution of gene expression (weighted Spearman rank correlation ρ S.
- We also find an overall weak correlation between gene expression coevo- lution and protein sequence coevolution (weighted Spear- man rank correlation ρ S = 0.10), consistent with previous work [7, 8].
- [7] and our mea- sure of gene expression coevolution based on empiri- cal RNA-Seq data (weighted Spearman Rank correlation ρ S = 0.22).
- CAI and similar codon usage metrics often show moderate to strong correlations with empir- ical gene expression estimates .
- It is worth noting that our estimates of gene expression coevolution and the esti- mates from [7] do not come from the same 18 species..
- for gene expression based on codon usage, reflect the evo- lutionary average expression level for a given gene (assum- ing strength of selection on codon usage scales with gene expression), but this may not reflect expression of a gene for a given experimental treatment .
- Furthermore, using multivariate PCMs allows for the treatment of gene expression measured under various conditions as separate traits [1]..
- Future work should focus on the examination of coevolution of gene expression using the OU model.
- While other methods for nor- malizing RNA-Seq measurements for across species exist, our results indicate transformation to the standard log- normal was suitable for the purpose of determining if functionally-related genes show stronger coevolution of gene expression than randomly-generated pairs.
- Despite this, we were still able to pick up evolutionary signals indi- cating coevolution of gene expression.
- However, analyses attempting to make more pre- cise conclusions about the evolution or coevolution of gene expression should ideally use measurements pro- duced under better controlled conditions.
- Gene expression data.
- Gene expression levels were estimated from publicly available RNA-Seq datasets taken from SRA using the pseudo-alignment tool, Salmon [61].
- where X is the gene expression vector for a species), con- sistent with the transformation used by [19].
- Analysis of gene expression data.
- Coevolution of gene expression was broadly examined using the Covariance Ratio test implemented in geo- morph [34–36].
- Gene expression evolution was modeled as a multi- variate Brownian Motion process using the R package mvMORPH [37] in order to examine the strength of coevolution between pairs of proteins (as opposed to coevolution within modules).
- reflects the degree to which gene expression estimates are correlated over evolutionary time and can be calculated from the evolutionary rate matrix .
- Under the hypothe- sis that gene expression coevolves between proteins which physically-interact, we expect the mean value of ρ C for the binding group to be significantly different from 0.
- The phylogenetically-corrected correlation ρ C , which reflects the strength of gene expression coevolution.
- We expect stronger coevolution of gene expression between pro- teins which are more functionally-related.
- It is well-established both gene expression and number of interactions in a protein-protein interaction network impact the evolutionary behavior of a protein [38, 45];.
- We hypothesized proteins pairs which are, on average, more highly expressed and involved in more interactions would show stronger coevolution of gene expression.
- the average number of interactions for each protein) and the mean phylogenetically-corrected average gene expression value were calculated.
- The phylogenetically- corrected average gene expression value for a protein is taken as the ancestral state value estimated at the root of the tree by mvMORPH..
- Furthermore, previous studies have examined the rela- tionship between sequence evolution and gene expression evolution [7, 8].
- We compared our estimates of gene expression coevolution to measures of sequence evolu- tion taken from Clark et al.
- also examined gene expression coevolution using the Codon Adaptation Index (CAI), which allowed us to compare our results based on empirical estimates of gene expression with a commonly-used proxy based on codon usage [10]..
- To determine if functional-relatedness, gene expression, number of protein interactions, and sequence coevolu- tion have an impact on the strength of gene expression coevolution, a weighted rank-based (i.e.
- Assessing accuracy of methods for detecting coevolution of gene expression.
- On the other hand, protein pairs from the control set were simulated forcing independent evolution of gene expression (i.e.
- For the PCM approach, protein pairs were consid- ered coevolving if a Likelihood Ratio test (as imple- mented in mvMORPH) comparing the model allow- ing coevolution of gene expression to a null model forcing independent evolution of gene expression had a Benjamini-Hochberg corrected p-value <0.05.
- the control group) as a means of deter- mining statistically significant gene expression coevo- lution using phylogenetically-uncorrected correlations..
- more than 2 genes) showed sig- nificant coevolution of gene expression by comparing the median phylogenetically-uncorrected correlation to the median correlations from 10,000 randomly-generated gene sets.
- Comparing performance of different methods for assessing coevolution of gene expression after removing protein pairs with a STRING Score less than 400..
- Heatmap showing overall similarity between species gene expression estimates.
- Distance between CAI and empirical-based measures of coevolution as function of gene expression.
- Effects of number of interactions and gene expression on strength of coevolution without filtering genes violating BM.
- Phylogenetic Analysis of Gene Expression.
- A relationship between gene expression and protein interactions on the proteome scale: analysis of the bacteriophage T7 and the yeast Saccharomyces cerevisiae.
- Coevolution of gene expression among interacting proteins.
- Comparative expression profiling reveals widespread coordinated evolution of gene expression across eukaryotes.
- Optimization of gene expression by natural selection..
- The evolution of gene expression levels in mammalian organs.
- A Phylogenetic Mixture Model for the Evolution of Gene Expression.
- Pervasive Correlated Evolution in Gene Expression Shapes Cell and Tissue Type Transcriptomes.
- Comparative Methods for the Analysis of Gene-Expression Evolution: An Example Using Yeast.
- Modeling Gene Expression Evolution with an Extended Ornstein–Uhlenbeck Process Accounting for Within- Species Variation.
- Estimating Gene Expression and Codon-Specific Translational Efficiencies, Mutation Biases, and Selection Coefficients from Genomic Data Alone.
- Unexpected correlations between gene expression and codon usage bias from microarray data for the whole.
- Intra and Interspecific Variations of Gene Expression Levels in Yeast Are Largely Neutral: (Nei Lecture, SMBE 2016, Gold Coast)

Xem thử không khả dụng, vui lòng xem tại trang nguồn
hoặc xem Tóm tắt