Evolutionary Model of Protein Families by Means Of
Total Page:16
File Type:pdf, Size:1020Kb
bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. EvoProDom: Evolutionary model of protein families by means of translocations of protein domains Gon Carmi1 † Alessandro Gorohovski 1 † Milana Frenkel-Morgenstern 1,* 1Cancer Genomics and BioComputing of Complex Diseases Lab, The Azrieli Faculty of Medicine, Bar-Ilan University, 8 Henrietta Szold St, Safed 13195, ISRAEL *Corresponding Author: e-mail: [email protected] † These authors contributed equally to this work. bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 1 Abstract Here, we developed a novel “evolution of protein domains” (EvoProDom) model for evolution of proteins, which was based on “mix and merge” of protein domains. We collected and integrated genomic and proteome data for 109 organisms. These data include protein domain content and orthologous protein families. In EvoProDom, we defined evolutionary events, such as translocations, as reciprocal exchanges of protein domains between orthologous proteins of different organisms. We found that protein domains, which frequently appear in translocation events, were enriched in trans-splicing events, i.e., producing novel transcripts fused from two distinct genes. We presented in EvoProDom, a general method to obtain protein domain content and orthologous protein annotation, by predicting these data from protein sequences using the Pfam search tool and KoFamKOALA, respectively. This method can be implemented in other research such as proteomics, protein design and host-virus interactions. Keywords: protein evolution, protein domains, translocations. bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 2 1 Introduction 2 Proteins are built from a set of domains that correspond to conserved regions with 3 distinct functional and structural characteristics (1). In accordance with the domain- 4 oriented view of proteins, domains group together to form domain architectures (DAs), 5 i.e., ordered sequences of domains. Some domains participate in specific DAs, while 6 others are found in diverse DAs. This later property is termed “domain promiscuity” or 7 “domain mobility”. Analyses of domain promiscuity can reveal the mechanisms by which 8 domains are gained or lost (2). Marsh and Teichmann in 2010 (1) described five 9 mechanisms by which proteins gain domains: (i) gene fusion, the fusion of a pair of 10 adjacent genes via alternative splicing in non-coding intergenic regions; (ii) exon 11 extension, the expansion of exon regions into adjacent introns, which encode a new 12 domain; (iii) exon recombination, direct merging of two exons from two different genes; 13 (iv) intron recombination or exon shuffling, insertion of an exon into an intron of a 14 different gene; and (v) retroposition, a sequence located within one gene, which is 15 transposed into a different gene along with a flanking genetic sequence. The mechanism 16 that added a given domain to a protein can be identified from the properties of the gained 17 domain, e.g., the position of the domain in a protein sequence and the number of exons. 18 For example, multi-exon domain gain at the c-termini is due to gene fusion. Additionally, 19 during metazoan evolution, new physical protein-protein interactions (PPIs) emerged 20 consequent to exon shuffling of domains that mediated the interaction (3). An additional 21 work, by Bornberg-Bauer and Mar Albà in 2013 (4), refined and expanded these 22 mechanisms and introduced new concepts such as intrinsically disordered regions, 23 implied links between the emergence of de-novo domains, and de-novo genes (4). bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 3 24 We developed a novel “evolution of protein domains” (EvoProDom) model for the 25 evolution of proteins, which was based on “mix and merge” of protein domains. We 26 collected and integrated genomic and proteome data for 109 organisms. These data 27 included protein domain content and orthologous protein content. In EvoProDom, we 28 defined evolutionary events, such as translocations, as the reciprocal exchanges of protein 29 domains between orthologous proteins of different organisms. We found that protein 30 domains, which frequently appear in translocation events, were enriched in trans-splicing 31 events, i.e., producing transcripts as a slippage of two distinct genes (5). We presented in 32 EvoProDom, a general method to obtain protein domain content and orthologous protein 33 annotation, by predicting these data from protein sequences using the Pfam search tool 34 (6, 7) and KoFamKOALA (8), respectively. This method can be implemented in other 35 research fields such as proteomics (9), protein design (10) and host-virus interactions 36 (11). 37 38 Materials and Methods 39 The EvoProDom model was based on the data of full genome and annotated proteomes. 40 In addition, the model utilized orthologous protein annotation and protein domain 41 content. Orthologous protein groups were used to group proteins (Refseqs) from 42 different organisms, thereby linking protein domain changes among orthologous proteins 43 with their corresponding groups of organisms. Orthologous proteins were realized as 44 Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs, namely, KEGG ortholog 45 (KO) (12, 13). Protein domain content was realized as Pfam domains, and this content 46 was associated with proteins. Accordingly, proteins were modeled as a group or list of bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 4 47 Pfam domains, and orthologous proteins were modeled as a group of proteins with the 48 same KO number. Both Pfam domains and KO assignments to proteins were predicted 49 from protein sequences alone, using the Pfam search tool (6, 7) and KoFamKOALA (8), 50 respectively. By applying these protein sequence-based methods to obtain protein domain 51 content and orthologous protein annotation, new organisms were easily added to 52 EvoProDom. Moreover, this method is unique to EvoProDom and is applicable to other 53 protein domain -based applications such as proteomics (9), protein design (10) and host- 54 virus interactions (11). 55 Statistical analysis was performed using R (R: A language and environment for 56 statistical, 3.3.2,2016). 57 Data resources 58 In total, the EvoProDom model was tested on a collection of 109 organisms with full 59 genome and annotated proteomes (Entrez/NCBI (14)). These organisms were grouped 60 according to six categories: (i) 15 fish; (ii) 4 subterranean, 8 fossorial and 21 61 aboveground animals (15, 16); (iii) 65 organisms with known PPIs (BioGrid version 62 3.5.173, (17, 18)); (vi) 17 organisms with HiC datasets; (v) 4 cats; and (iv) 15 pathogenic 63 organisms (19). Organisms with HiC datasets were obtained by searching ‘HiC’ in the 64 NCBI GEO database (Table 1). 65 Orthologous protein annotation 66 Orthologous annotation was based on Kyoto Encyclopedia of Genes and Genomes 67 (KEGG) orthologs, namely, KEGG ortholog (KO) (12, 13). Proteins were assigned to 68 KO groups, by utilizing KoFamKOALA, a Hidden Markov Model (HMM) profile based 69 search tool (8). Since this tool assigned proteins to KO groups based on protein bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 5 70 sequences, in-house script was written to automatically assign proteins to KO groups. In 71 this study, only proteins with KO annotation were used for analysis. Additionally, 72 organism code was generated by selecting 3-4 letters from an organism name in 73 uppercase format (lower case code represents organisms from the KEGG database) 74 (Table 1). 75 Protein domain detection 76 Pfam (release 32.0) domains were predicted from protein sequences, using a dedicated 77 HMM-based search tool (6, 7). Therefore, in-house script was written and protein domain 78 content was predicted from protein sequences. Additionally, Pfam domains were 79 members of super families (clan, pfam nomenclature), which we used for our 80 classification. These data were supplemented to the protein domain content. 81 EvoProDomDB 82 Genomic and proteomic data, along with orthologous protein and protein domain content 83 data, were related by shared data. Thus, a relational database, EvoProDomDB, was 84 written in MySQL on MariaDB (10.0.26) for the efficient search engine. The 85 EvoProDom model was implemented and tested on EvoProDomDB. EvoProDomDB was 86 organized with orthologous protein and protein content for the 1,835,600 protein products 87 that were distributed among 23,147 KO groups, containing 17,929 unique Pfam domains. 88 Pfam domains were distributed among 629 super families, and EvoProDomDB integrated 89 data for 109 organisms from diverse taxa. EvoProDomDB was built from six relational 90 tables with common features, e.g., organism id and other features, as shown in the 91 scheme (Figure 1). Relational tables, taxonomy, ko_annotation, clan_domain, bioRxiv preprint doi: https://doi.org/10.1101/2020.02.23.961532; this version posted February 25, 2020.