Evidence for short-time divergence and long-time conservation of tissue-specific expression after gene duplication (original) (raw)

Journal Article

,

Jaime Huerta-Cepas is a post-doctoral researcher at the comparative genomics group at CRG, his research focuses in large-scale phylogenetic analyses.

Search for other works by this author on:

,

Joaquin Dopazo is head of the Bioinformatics department at the CIPF, and leads the Functional Genomics group there.

Search for other works by this author on:

,

Martijn A. Huynen leads the Comparative Genomics group at the CMBI and the NCMLS at the Radboud University Nijmegen Medical Centre.

Search for other works by this author on:

Toni Gabaldón leads the Comparative Genomics group at the Bioinformatics and Genomics program in the CRG. He is associate professor of bioinformatics at the UPF. His research topics include comparative genomics and phylogenomics.

Search for other works by this author on:

Received:

19 January 2011

Revision received:

22 March 2011

• PDF
• Split View
• • Article contents
  • Figures & tables
  • Video
  • Audio
  • Supplementary Data
• Cite

Cite

Jaime Huerta-Cepas, Joaquín Dopazo, Martijn A. Huynen, Toni Gabaldón, Evidence for short-time divergence and long-time conservation of tissue-specific expression after gene duplication, Briefings in Bioinformatics, Volume 12, Issue 5, September 2011, Pages 442–448, https://doi.org/10.1093/bib/bbr022
Close

Navbar Search Filter Mobile Enter search term Search

Abstract

Gene duplication is one of the main mechanisms by which genomes can acquire novel functions. It has been proposed that the retention of gene duplicates can be associated to processes of tissue expression divergence. These models predict that acquisition of divergent expression patterns should be acquired shortly after the duplication, and that larger divergence in tissue expression would be expected for paralogs, as compared to orthologs of a similar age. Many studies have shown that gene duplicates tend to have divergent expression patterns and that gene family expansions are associated with high levels of tissue specificity. However, the timeframe in which these processes occur have rarely been investigated in detail, particularly in vertebrates, and most analyses do not include direct comparisons of orthologs as a baseline for the expected levels of tissue specificity in absence of duplications. To assess the specific contribution of duplications to expression divergence, we combine here phylogenetic analyses and expression data from human and mouse. In particular, we study differences in spatial expression among human–mouse paralogs, specifically duplicated after the radiation of mammals, and compare them to pairs of orthologs in the same species. Our results show that gene duplication leads to increased levels of tissue specificity and that this tends to occur promptly after the duplication event.

INTRODUCTION

Sequencing projects have revealed that the presence of duplicated genes is rampant, especially in eukaryotic genomes. These duplicated genes have likely contributed to the acquisition of new functions by their corresponding genomes, as proposed initially by Ohno [1]. The duplication of a gene initially gives rise to a state of genetic redundancy, which can be suppressed by the pseudogenization and eventual elimination of one of the duplicates. Alternatively, during the genetic redundancy phase, accumulating differences in the duplicated sequences may lead to changes in their function, which in turn may render both copies indispensable. Neo-functionalization models imply the acquisition by one of the duplicates of a novel function, which was not present in the common ancestor. In contrast, sub-functionalization models assume that each duplicate will retain only part of the functionality of the ancestor. In any case, the result of both processes is the same: there will be a selective pressure to retain both duplicated copies [2–4]. Similar to changes that have occurred at the sequence level, variations in temporal or spatial gene expression patterns can be regarded as reflections of function changes and thus be analyzed under the same neo- or sub-functionalization models. Thus, differences between tissue expression patterns of duplicated genes, in which each copy becomes expressed on a specific set of tissues, would be expected to play a role in the eventual retention of paralogous copies. Many recent studies have focused on testing some of the predictions of sub- and neo-functionalization models using expression data. For instance, if the acquisition of complementary expression patterns between duplicates facilitates their retention, a prediction of both models is that each round of duplication will result in a reduction of expression breadth (i.e. the total number of tissues in which a gene is expressed). As a result, expression breadth per gene is likely to be narrower in larger families. Consistent with this prediction, a negative correlation of expression breadth and family size has been recently reported in several studies [5, 6]. Similarly, it has been observed that gene duplication is associated to higher expression diversity in yeast, fly and Arabidopsis [7, 8]. Finally, the number of duplications undergone by genes seems to influence the levels of expression divergence [9, 10].

Another aspect that has been addressed but for which some controversy remains is that of the timeframe in which gene duplication and subsequent expression divergence occur. In this respect, many studies have investigated the relationship between expression divergence and the age of paralogs, usually estimated from differences in neutral sites (Ks). While several authors have found a significant correlation between Ks and expression divergence [11, 12], others have reported a lack of correlation [13–16]. Such discrepancies may be explained by the use of different methodologies to measure expression divergence, and a lack of sufficient resolution of Ks to estimate the age of duplications [17, 18]. Finally, as noted recently [19], a common drawback of most studies on gene duplicates is that their design does not allow for direct comparison with the situation among orthologs. Indeed, although several studies have addressed the comparison between the expression of orthologs and paralogs at large scale [6, 20–25], they use paralogs from the same species (i.e. intra-specific) that result from duplications at various evolutionary time points, whereas the orthologous genes compared derive from a single speciation event. In that situation, the expression divergence found between paralogs would be the composite result of the effect of divergence time and, if any, of the specific effect of the duplication. Thus, to properly assess the specific effect of duplications it is necessary to control for the time of divergence of genes in each type of homologous relationship.

Here, we exploit the availability of the human phylome [26] to investigate the relationship of tissue specific divergence between paralogs and orthologs, and the relative timeframe of the process of expression divergence. In particular, we compare the levels of tissue expression divergence between human–mouse inter-species paralogs and orthologs of a similar age. From our analyses, we conclude that gene duplication is specifically associated with higher levels of tissue expression divergence and that a significant part of this divergence was acquired shortly after gene duplication.

MATERIALS AND METHODS

Gene expression data sets

Three expression datasets from two independent sources were used to perform all analyses: (i) SymAtlas Absence/Presence calls data set: Expression data from the SymAtlas project for human and mouse species [27]. This data set is based on 69 human tissues and 60 mouse tissues, of which 29 human–mouse tissue pairs are considered homologous. We used the two-samples Affymetrix HG-U133A probeset with MAS5 normalization for human and the GNF1M data set with MAS5 normalization for mouse. To determine whether a gene was present (expressed) or absent (not expressed) in a given tissue we used the Presence (P) and Absence (A) calls, respectively. We assumed marginal expression (M) to be Absences. A gene was considered to be present in a tissue if presence calls were reported for the two replicas included in the probeset. Probes mapping to several genes were removed from the analysis. This dataset is referred to in the manuscript as SymAtlas A/P calls data set. (ii) SymAtlas expression levels data set: from the SymAtlas database, we also used the two-samples Affymetrix HG-U133A probeset for human and the GNF1M dataset for mouse, both under GCRMA normalization and without A/P call analysis. These datasets are referred to as SymAtlas levels in figures and tables. (iii) Bgee Absence/Presence calls data set: expression data for human and mouse genes were retrieved from the microarray expression information stored in the Bgee database [28]. This public resource combines normalized data from different microarray experiments and provides low and high confidence labels for each tissue expression record. This database maintains a manually curated ontology of tissue homologies across species. In our analyses, only high-confidence data were used for all analyses. Expression was based on 71 and 98 tissue categories for human and mouse, respectively. The combined human–mouse data set, referred to as Bgee A/P calls in the manuscript, was based in 44 homologous tissues.

Detecting and dating duplication events

Phylogenetic analyses were carried out to detect and date duplication events leading to human gene paralogs. We used the complete set of human gene phylogenies [26], publicly available from the phylomeDB database [29]. Phylogenetic trees were scanned using the species-overlap approach described in [26] to detect all orthology and paralogy relationships. In brief, this algorithm assumes that a tree node represents a duplication event if it splits two lineages (partitions) with at least one species in common. In addition, duplication events were dated accordingly to the set of species that underwent such event [18]. For this, given the 39 species included in the phylogenetic tree used, we defined 10 evolutionary categories, namely: hominids, primates, mammals, tetrapodes, vertebrates, chordates, metazoa, opisthokonts and basal eukaryotes. Each duplication was assigned to one category by evaluating the source species of the resulting paralogs. Thus, a duplication event involving only genes from mammal species would be assigned to ‘mammals’, meaning that the duplication occurred, at least, before the radiation of mammalian organisms. The software ETE [30] was used to detect and date all duplication and speciation nodes in phylogenetic trees.

Tissue expression complementarity

TEC between the expression profiles of two paralogous genes i and j was calculated by measuring the relative number of tissues in which only one set but not the other was expressed over the total number of tissues in which each gene is expressed:

where d i is the number of tissues in which gene i is specifically expressed (i.e. gene j is not expressed) and tj is the total number of tissues in which gene j is expressed. For one-to-one paralogy relationships, the expression profiles of the two sets were the same as the expression of the two paralogous genes, respectively. TEC in the SymAtlas levels dataset was calculated as follows:

Being Ei,n the level of expression of gene i in tissue n, and N the total number of tissues considered. Note that by always comparing pairs of genes from the same family, this methodology is assumed to be free from functional biases.

Relative contribution of pre- and post-speciation divergence between paralogs

In order to approximate the fraction of TEC acquired before the speciation, we combined the expression patterns of orthologous genes to represent only the tissues in which both orthologs are expressed (a proxy to expression acquired prior to speciation). Next, we calculated the TEC between these two grouped patterns and divided this value by the maximum TEC value observed between inter-species paralogs, a measure of the total divergence observed. Mathematically, this fraction (F) can be represented as follows:

where humanA, humanB, mouseA and mouseB, represent expression patterns of genes related according to Figure 1, and A and B are, respectively, the intersection of expression patterns of A (humanA,mouseA), and B (humanB,mouseB) pairs of orthologs.

Figure 1:

Diagram of the evolutionary relationships considered for analyses. (A and B) Represent classic human–mouse orthology, human–human and mouse–mouse paralogy relationships, respectively. (C) Represents the inter-species paralogy relationship existing between human and mouse sequences. In order to avoid species specific effects, inter-species paralogy relationships were used when comparisons between orthologs and paralogs were required.

RESULTS AND DISCUSSION

Duplication events are specifically associated to higher levels of tissue expression divergence

In order to properly control for the effect of divergence time in our comparisons, we aimed at compiling a set of paralogs and orthologs of a similar divergence time. Moreover, to avoid possible biases derived from comparing intra-species paralogies against inter-species orthology relationships, we limited our analyses to human–mouse paralogous and orthologous pairs (Figure 1). For this, we scanned the complete set of 20 535 human gene phylogenies provided by the human phylome [26, 29] to detect human–mouse one-to-one paralogs originated from duplications occurring after the radiation of mammals but preceding the speciation of primates and rodents. These were compared to human–mouse one-to-one orthologs in terms of their patterns of tissue expression. Expression data for all genes used in the analyses was retrieved from Bgee [28] and SymAtlas [27], using only the reduced set of human–mouse homologous tissues (see ‘Materials and methods’ section). Expression divergence was calculated as the level of tissue expression complementarity (TEC) between two sets of paralogous genes. In brief, TEC is defined as the number of mutually exclusive tissues in which the two genes are expressed divided by the combined expression breadth of the genes resulting from the duplication, so that greater values of TEC indicate higher levels of tissue complementarity (see ‘Materials and methods’ section). We expect this measure to account better for transcriptional differences among distant genes as compared to the traditional correlation coefficient distance, which has been pointed out to present problems [31].

Our results (Figure 2) show that the distribution of TEC between inter-species paralogs is significantly more biased towards higher values than that of human–mouse orthologs in all the expression data sets tested (Two tailed Kolmogorov–Smirnov test, K–S test, P = 2.74E-3, P = 1.82E-21, P = 5.57E-5 for the SymAtlas A/P calls, Bgee A/P calls and SymAtlas levels data sets, respectively). To discard possible biases due to the effect of lineage-specific losses, we repeated the experiment considering only those cases in which the two paralogous copies had been retained in both human and mouse genomes. The results of this restricted dataset are similar to those shown in Supplementary Figure S2. This difference between orthologs and paralogs could be explained by a selective constraint to retain complementary expression patterns in duplicated genes as predicted by the sub- and neo-functionalization models. However, human–mouse paralogs duplicated prior to the radiation of mammals still have had more time to diverge than human–mouse orthologs: the timeframe between the moment of the gene duplication and that of the speciation event of the human and mouse lineages. Thus, unequal divergence times could be regarded as the possible source of the expression differences encountered. In order to test for this possibility, we measured the correlation between the levels of TEC and the lengths of the branches going from the duplication event to the speciation event of the human/mouse lineages (a proxy to the time span between the duplication and speciation events considered), and found that it is not significant (R = 0.05, P = 60). This suggests that the observed differences between orthologous and paralogous pairs are not due to differences in divergence time but rather are specific for the duplication event. Nevertheless, a partial contribution of unequal divergence times to time the effect cannot be completely ruled out.

Figure 2:

Orthologs versus paralogs TEC. Empirical cumulative distribution function of TEC among human–mouse orthologs (black line) and among human–mouse inter-species paralogs (red line), using three different expression data sets: SymAtlas presence calls (A), Bgee presence calls (B), and SymAtlas expression levels (C), with paralogs showing distributions more biased towards higher levels of TEC. For a comparison with TEC levels obtained using intra-specific paralogs see Supplementary Figure S1.

A significant part of the tissue expression divergence between paralogs was gained shortly after the duplication event

That paralogs have higher levels of tissue expression divergence relative to orthologs of a similar age, does not necessarily imply that this level of divergence was acquired shortly after the duplication. Alternative scenarios may involve a slow but continuous accumulation in tissue expression patterns since the time of duplication. To test this, we used the set of duplications that have two representatives in both human and mouse genomes. In this case, the comparison of tissue expression patterns between intra- and inter-species paralogs can provide information on whether the observed level of complementarity was gained before or after the speciation event. If a significant part of the complementarity was gained prior to speciation (Figure 3), one would expect that the complementarity observed between human–human paralogs resembles to that observed between mouse–mouse paralogs that emerged from the same duplication. In the opposite scenario, the acquisition of complementarity after the speciation event, one would expect independence between the complementarities that were acquired in the human and mouse lineages. To test this, we searched for duplications that occurred in the mammals, before the split of human and mouse lineages and resulting in two human genes and two mouse genes. We then compared, for each of these duplications, the TEC level between human-human paralogous sequences with that of mouse–mouse paralogs. We observed a significant positive correlation (Pearson R = 0.34, P = 3.85E-09, SymAtlas A/P calls data set) between the two measures. Additional tests support that this correlation is due to a very similar tissue complementarity pattern between human and mouse paralogs. Indeed, we found that, when human–human TEC and mouse–mouse TEC are high, there is an increase in (i) the similarity of expression patterns of orthologous genes (Pearson correlation R = 0.17, P = 3.85E-05 SymAtlas A/P calls data set) and (ii) the level of TEC between inter-species paralogs (Pearson correlation R = 0.66, P = 5.30E-73 SymAtlas A/P calls). Taken together, these three correlations suggest that the expression complementarity between the considered human and mouse duplicates was mostly acquired prior to the human–mouse speciation event (Figure 3), and tends to be conserved in evolution. We believe that alternative explanations than similarity by common ancestry are highly unlikely, since they would imply widespread convergent evolution, only for orthologous genes, in the absence of any apparent selective force. Note that this finding is consistent with the idea that gene retention was favoured by the acquisition of divergent tissue expression patterns. Similar results were obtained from the two other data sets considered, indicating that these results are robust to changes in the type of measurement or the set of expression data considered (Supplementary Table S1). It is difficult to quantify the fraction of the differences among paralogs, which was gained before the speciation, especially in the absence of known ancestral expression profiles. Nevertheless, we can obtain a rough estimate by assuming that tissue expression shared by a given pair of orthologs, but not their paralogs, was acquired prior to speciation (see ‘Material and methods’ section). Such measure indicates that roughly 80% (Bgee data set, 72% for SymAtlas) of the total complementary expression observed was likely acquired prior to the speciation event.

Figure 3:

Relative timing of the acquisition of tissue expression divergence. A possible timeline for the acquisition of TEC between mammal-specific duplicates: as indicated in the example, tissue specificity would be mostly gained by paralogs prior to the speciation of human and mouse species, thus being inherited by both lineages.

CONCLUSIONS

Using a phylogeny-based approach and two independent sources of expression data, we have characterized the relationship between duplication, divergence time and tissue expression divergence. Our results show that paralogs, as compared to orthologs of a similar age, have higher levels of tissue expression divergence. This indicates a specific relationship of the duplication event with levels of tissue expression divergence between paralogs. A possible explanation for this result would be that after duplication, the resulting paralogs would evolve independently in terms of their pattern of tissue expression. This divergence could eventually lead to the situation in which the retention of both genes is favoured simply because only their combined expression pattern will cover all tissues in which the expression of the gene is necessary or favourable. Note that this holds even if both duplicates perform exactly the same function and even if their coding sequences are identical. That is, the mere pattern of tissue expression may constitute a factor for the conservation of duplicated genes. Of note, non-segmental duplications in which one of the duplicated genes is transferred to a completely distinct genomic location may provide the means for a rapid change in expression profiles right upon duplication. Duplicates with identical patterns of tissue expression would rely on other mechanisms for their retention, i.e. change in function or an advantageous effect of dosage increase, but would, in this model have a higher chance of losing one of the copies. This would explain the observed higher degree of complementarity observed between groups of paralogs as compared to orthologs. Another result that supports this idea is that much of the TEC found between human–mouse paralogs was likely acquired prior to the divergence of these two lineages. Altogether, our results provide evidence that gene duplication promotes tissue specificity in the short term, a prediction that emerges from many models of gene duplication. The species investigated in this study, human and mouse, were chosen for the availability of appropriate comparable expression data across tissues. As similar sets become available to a broader and more diverse set of species, it will be important to assess the generalities of our findings. This will be particularly important in plants, where genome duplications have taken place in several lineages [32].

SUPPLEMENTARY DATA

Supplementary data are available online at http://bib.oxfordjournals.org/.

• We find that an increase in tissue specificity in expression profiles promptly follows the duplication of a gene, suggesting a specific effect of the duplication.
• This tissue specificity tends to be conserved after subsequent speciation events so that the complementarity of expression patterns of orthologous genes is smaller than that of paralogs.
• Altogether, our data are in line with predictions of several gene duplication models in which changes promoting the conservation of duplicated genes should shortly follow the duplication event.

FUNDING

Spanish Ministry of Science and Innovation (BFU2009-09168 to T.G.)and (Subprograma Juan de la Cierva, JCI2010-07614 to J.H.C.); European Science Foundation (ESF), Frontiers of Functional Genomics program (to J.H.C.); Project (BIO BIO2008-04212 to J.D.), Spanish Ministry of Science and Innovation and PROMETEO/2010/001 from the GVA-FEDER. The INB and the CIBER de Enfermedades Raras, are initiatives of the ISCIII.

Acknowledgements

The authors want to acknowledge Javier Díez and Frederic Bastian for their assistance with the Bgee dataset.

References

1

,

Evolution by gene duplication

,

1970

New York

Springer

2

et al.

Preservation of duplicate genes by complementary, degenerative mutations

,

Genetics

,

1999

, vol.

151

(pg.

1531

-

45

)

3

The probability of duplicate gene preservation by subfunctionalization

,

Genetics

,

2000

, vol.

154

(pg.

459

-

73

)

4

The evolution of gene duplications: classifying and distinguishing between models

,

Nat Rev Genet

, vol.

11

(pg.

97

-

108

)

5

et al.

Relating tissue specialization to the differentiation of expression of singleton and duplicate mouse proteins

,

Genome Biol

,

2006

, vol.

7

pg.

R89

6

Divergence of spatial gene expression profiles following species-specific gene duplications in human and mouse

,

Genome Res

,

2004

, vol.

14

(pg.

1870

-

9

)

7

et al.

Duplicate genes increase gene expression diversity within and between species

,

Nat Genet

,

2004

, vol.

36

(pg.

577

-

9

)

8

Duplicate genes increase expression diversity in closely related species and allopolyploids

,

Proc Natl Acad Sci USA

,

2009

, vol.

106

(pg.

2295

-

300

)

9

Heterogeneous patterns of gene expression diversification in mammalian gene duplicates

,

Mol Biol Evol

,

2010

, vol.

27

(pg.

325

-

35

)

10

Historical constraints on vertebrate genome evolution

,

Genome Biol Evol

,

2010

, vol.

2010

(pg.

13

-

18

)

11

et al.

Rapid divergence in expression between duplicate genes inferred from microarray data

,

Trends Genet: TIG

,

2002

, vol.

18

(pg.

609

-

13

)

12

Divergence in the spatial pattern of gene expression between human duplicate genes

,

Genome Res

,

2003

, vol.

13

(pg.

1638

-

45

)

13

Divergence in Expression between Duplicated Genes in Arabidopsis

,

Mol Biol Evol

,

2007

, vol.

24

(pg.

2298

-

309

)

14

et al.

Transcriptional similarities, dissimilarities, and conservation of cis-elements in duplicated genes of Arabidopsis

,

Plant Physiol

,

2004

, vol.

136

(pg.

3009

-

22

)

15

Asymmetric functional divergence of duplicate genes in yeast

,

Mol Biol Evol

,

2002

, vol.

19

(pg.

1760

-

8

)

16

Evolution of gene sequence and gene expression are not correlated in yeast

,

Trends Genet

,

2008

, vol.

24

(pg.

109

-

13

)

17

Very low gene duplication rate in the yeast genome

,

Science

,

2004

, vol.

306

(pg.

1367

-

70

)

18

Assigning duplication events to relative temporal scales in genome-wide studies

,

Bioinformatics

,

2011

, vol.

27

(pg.

38

-

45

)

19

How confident can we be that orthologs are similar, but paralogs differ?

,

Trends Genet

,

2009

, vol.

25

(pg.

210

-

16

)

20

Evolutionary significance of gene expression divergence

,

Gene

,

2005

, vol.

345

(pg.

119

-

26

)

21

Incongruent expression profiles between human and mouse orthologous genes suggest widespread neutral evolution of transcription control

,

OMICS

,

2004

, vol.

8

(pg.

15

-

24

)

22

et al.

Evolution of stress-regulated gene expression in duplicate genes of Arabidopsis thaliana

,

PLoS Genet

,

2009

, vol.

5

pg.

e1000581

23

Comparative analysis indicates regulatory neofunctionalization of yeast duplicates

,

Genome Biol

,

2007

, vol.

8

pg.

R50

24

et al.

Rapid and asymmetric divergence of duplicate genes in the human gene coexpression network

,

BMC Bioinformatics

,

2006

, vol.

7

pg.

46

25

et al.

Nonrandom divergence of gene expression following gene and genome duplications in the flowering plant Arabidopsis thaliana

,

Genome Biol

,

2006

, vol.

7

pg.

R13

26

et al.

The human phylome

,

Genome Biol

,

2007

, vol.

8

pg.

R109

27

et al.

A gene atlas of the mouse and human protein-encoding transcriptomes

,

Proc Natl Acad Sci USA

,

2004

, vol.

101

(pg.

6062

-

7

)

28

et al.

Bgee: Integrating and Comparing Heterogeneous Transcriptome Data Among Species

,

Data Integration Life Sci

,

2008

, vol.

5109

(pg.

124

-

31

)

29

et al.

PhylomeDB v3.0: an expanding repository of genome-wide collections of trees, alignments and phylogeny-based orthology and paralogy predictions

,

Nucleic Acids Res

,

2011

, vol.

39

(pg.

D556

-

60

)

30

ETE: a python Environment for Tree Exploration

,

BMC Bioinformatics

,

2010

, vol.

11

pg.

24

31

A problem with the correlation coefficient as a measure of gene expression divergence

,

Genetics

,

2009

, vol.

183

(pg.

1597

-

600

)

32

et al.

The flowering world: a tale of duplications

,

Trends Plant Sci

,

2009

, vol.

14

(pg.

680

-

8

)

© The Author 2011. Published by Oxford University Press. For Permissions, please email: journals.permissions@oup.com

Supplementary data

Citations

Views

Altmetric

Metrics

Total Views 1,306

927 Pageviews

379 PDF Downloads

Since 11/1/2016

Citations

43 Web of Science

×

Email alerts

Citing articles via

• Latest

• Most Read

• Most Cited

More from Oxford Academic