Evolution of the Highly Repetitive PEVK Region of Titin Across Mammals

Abstract

The protein titin plays a key role in vertebrate muscle where it acts like a giant molecular spring. Despite its importance and conservation over vertebrate evolution, a lack of high quality annotations in non-model species makes comparative evolutionary studies of titin challenging. The PEVK region of titin-named for its high proportion of Pro-Glu-Val-Lys amino acids-is particularly difficult to annotate due to its abundance of alternatively spliced isoforms and short, highly repetitive exons. To understand PEVK evolution across mammals, we developed a bioinformatics tool, PEVK_Finder, to annotate PEVK exons from genomic sequences of titin and applied it to a diverse set of mammals. PEVK_Finder consistently outperforms standard annotation tools across a broad range of conditions and improves annotations of the PEVK region in non-model mammalian species. We find that the PEVK region can be divided into two subregions (PEVK-N, PEVK-C) with distinct patterns of evolutionary constraint and divergence. The bipartite nature of the PEVK region has implications for titin diversification. In the PEVK-N region, certain exons are conserved and may be essential, but natural selection also acts on particular codons. In the PEVK-C, exons are more homogenous and length variation of the PEVK region may provide the raw material for evolutionary adaptation in titin function. The PEVK-C region can be further divided into a highly repetitive region (PEVK-CA) and one that is more variable (PEVK-CB). Taken together, we find that the very complexity that makes titin a challenge for annotation tools may also promote evolutionary adaptation.

Keywords: PEVK region; comparative genomics; gene prediction; molecular evolution; titin.

Figure 1

PEVK Finder tool optimization and evaluation of human TTN. a) Match scores of human PEVK exon sets were generated using different combinations of minimum exon length, PEVK ratio and sliding window length parameter settings. b) PEVK_Finder recovered more PEVK exons than other existing gene prediction tools (GENSCAN, Augustus, FGENESH and geneid). c) PEVK_Finder outperformed GENSCAN at recovering the exon-intron distribution of human TTN PEVK exons identified by cDNA. Vertical lines indicate exon boundaries, and the thickness of the lines is determined by the exon coordinates. Gray circles indicate exons that were missed by either PEVK Finder or GENSCAN, and black asterisks indicate automatically annotated exons that were not annotated by cDNA. In this figure, the first exon indicated by an asterisk is technically outside the pre-defined bounds of the PEVK region. The PEVK ratio scale is given in the figure.

Figure 2

Phylogenetic comparison of PEVK exon structure. PEVK_Finder exon-intron plots were overlaid onto a time-calibrated phylogeny for 39 of 41 mammalian species. The dotted line indicates the approximate boundary of the PEVK-N and PEVK-C segments of the PEVK region, based on the human boundary between the two segments. PEVK ratio scale is the same as in Fig. 1. In this figure, Neomonachus schaunslandi and Tupaia chinensis are represented by their alternative names, Monachus schaunslandi and Tupaia belangeri, respectively.

Figure 3

Exon and nucleotide-level comparisons of PEVK-N and PEVK-C regions. a) A heat map of substitutions among PEVK exons within an individual: I) PEVK-N vs. PEVK-C, II) PEVK-C vs. PEVK-C, III) PEVK-N vs. PEVK-N, and IV) PEVK-C vs. PEVK-N. Dark pink indicates exon pairs with few substitutions, whereas light pink and green indicate exons pairs with many substitutions. In general, PEVK-C exons are highly repetitive and homogeneous, while PEVK-N exons are more variable. b) A reciprocal dot plot of the human PEVK-C nucleotide sequence that shows the repetitive nature of the PEVK-C region in humans (Dotlet, https://dotlet.vital-it.ch). c) Mean pairwise substitutions per exon across all 41 species for each quadrant from Figure 3a. Bars represent mean ± SE d) Mean P,E,V,K amino acids per exon across all 41 mammalian species. There is significantly more glutamate (E), valine (V) and lysine (K) per exon in the PEVK-N region (gray bars) and more proline (P) per exon in the PEVK-C region (black bars) across all 41 species. Bars represent mean ± SE. Asterisks denote significance at P < 0.05. e) Ratio of glutamate (E) per exon in human TTN PEVK. PEVK_Finder confirms that there is relatively more glutamate (E) per exon in the PEVK-N region compared to the PEVK-C region.

Figure 4

Human PEVK_Finder exon-intron plot depicting orthologous exons and codons under significant positive selection. Dark gray bars (labeled with their respective exon numbers) indicate orthologous exons, gray circles indicate exons missed by PEVK Finder, and asterisks indicate codons under selection at varying levels of significance. Asterisks indicate posterior probabilities of * > 0.5; ** > 0.75; *** > 0.95. The probability for only the most significant codon for a given exon is noted. Exons 114, 116, 122, 125, 130, 135, 136, 137, 138, 141, 142, 143, 144, 146, 152 153, and 161 are either constitutively expressed or expressed in >95% of TTN transcripts in skeletal muscle according to Savarese et al. (2018). Details about codons under selection are in Figures S5–S12. As in figure 1c, the first exon in this figure is technically outside the pre-defined bounds of the human PEVK region.