Comparative study of human mitochondrial proteome reveals extensive protein subcellular relocalization after gene duplications

Abstract

Background: Gene and genome duplication is the principle creative force in evolution. Recently, protein subcellular relocalization, or neolocalization was proposed as one of the mechanisms responsible for the retention of duplicated genes. This hypothesis received support from the analysis of yeast genomes, but has not been tested thoroughly on animal genomes. In order to evaluate the importance of subcellular relocalizations for retention of duplicated genes in animal genomes, we systematically analyzed nuclear encoded mitochondrial proteins in the human genome by reconstructing phylogenies of mitochondrial multigene families.

Results: The 456 human mitochondrial proteins selected for this study were clustered into 305 gene families including 92 multigene families. Among the multigene families, 59 (64%) consisted of both mitochondrial and cytosolic (non-mitochondrial) proteins (mt-cy families) while the remaining 33 (36%) were composed of mitochondrial proteins (mt-mt families). Phylogenetic analyses of mt-cy families revealed three different scenarios of their neolocalization following gene duplication: 1) relocalization from mitochondria to cytosol, 2) from cytosol to mitochondria and 3) multiple subcellular relocalizations. The neolocalizations were most commonly enabled by the gain or loss of N-terminal mitochondrial targeting signals. The majority of detected subcellular relocalization events occurred early in animal evolution, preceding the evolution of tetrapods. Mt-mt protein families showed a somewhat different pattern, where gene duplication occurred more evenly in time. However, for both types of protein families, most duplication events appear to roughly coincide with two rounds of genome duplications early in vertebrate evolution. Finally, we evaluated the effects of inaccurate and incomplete annotation of mitochondrial proteins and found that our conclusion of the importance of subcellular relocalization after gene duplication on the genomic scale was robust to potential gene misannotation.

Conclusion: Our results suggest that protein subcellular relocalization is an important mechanism for the retention and gain of function of duplicated genes in animal genome evolution.

Figure 1

Maximum likelihood phylogeny of the arginase family. Numbers indicate bootstrap support based on 100 replicates. ARG1: Type I arginases; ARG2: Type II arginases. Colored boxes indicate annotated and/or predicted subcellular locations of the proteins: cytoplasm (yellow) and mitochondria (green). There is no subcellular information for the proteins in Anopheles gambiae and Nematostella vectensis.

Figure 2

Maximum likelihood phylogeny of the DNA topoisomerase typeIB family. Numbers indicate bootstrap support based on 100 replicates. TOP1: DNA topoisomerase 1; TOP1MT: mitochondrial DNA topoisomerase 1. Colored boxes indicate annotated and/or predicted subcellular locations of the proteins: nucleus/cytoplasm (blue) and mitochondria (green). There is no subcellular information for the proteins in Anopheles gambiae and Nematostella vectensis.

Figure 3

Evolutionary rates in mitochondrial vs. non-mitochondrial proteins. (A) A schematic phylogeny of a mt-cy two gene family with gene duplication occurred in the vertebrate lineage. Branch lengths before the divergence between fish and tetrapods are marked as a and b for mitochondrial and cytosolic proteins, respectively. The corresponding average branch lengths after this divergence are marked as a' and b'. (B) The ratios of branch lengths for mitochondrial vs. nuclear paralogs (a/b, a'/b', and (a+a')/(b+b')) were calculated on the maximum likelihood topologies as illustrated in (A) with the exception of the TST family, for which the divergence between birds (chicken) and mammals was used. TOP1MT, TST, SHMT2 and CDS2 families had undergone relocalization from cytosol to mitochondria, while the remaining 6 families had the opposite direction of relocalizations.

Figure 4

Maximum likelihood phylogeny of the class I sirtuin family. Numbers indicate bootstrap support based on 100 replicates. SIRT1: Sirtuin 1; SIRT2: Sirtuin 2; SIRT3: Sirtuin 3. Colored boxes indicate annotated and/or predicted subcellular locations of the proteins: nucleus (purple), cytoplasm (yellow) and mitochondria (green). SIRT3 in Rattus norvegicus and Mus musculus lost the mitochondrial N-terminal targeting signal and thus were retained in the cytoplasm.

Figure 5

The presence of mitochondrial N-terminal targeting signals (A) and mitochondrial Pfam domains (B) for human mitochondrial (mt) and non-mitochondrial (nonmt) proteins. N-terminal mitochondrial targeting signals were inferred for proteins in mt-cy families based on targetP predictions [38]. Mitochondrial Pfam domains refer to those domains that were found only in eukaryotic (excluding human) mitochondrial proteins [40].

Figure 6

Maestro score distributions for human mitochondrial and non-mitochondrial proteins. Maestro scoring system incorporates eight genomic-scale data sets (targeting sequence prediction, protein domain enrichment, presence of cis-regulatory motifs, yeast homology, ancestry, tandem-mass spectrometry, coexpression and transcriptional induction during mitochondrial biogenesis) for predicting nuclear encoded mitochondrial proteins [40]. The cutoff score of 5.65 is indicated as the vertical bar.