Protein innovation through template switching in the Saccharomyces cerevisiae lineage

Abstract

DNA polymerase template switching between short, non-identical inverted repeats (IRs) is a genetic mechanism that leads to the homogenization of IR arms and to IR spacer inversion, which cause multinucleotide mutations (MNMs). It is unknown if and how template switching affects gene evolution. In this study, we performed a phylogenetic analysis to determine the effect of template switching between IR arms on coding DNA of Saccharomyces cerevisiae. To achieve this, perfect IRs that co-occurred with MNMs between a strain and its parental node were identified in S. cerevisiae strains. We determined that template switching introduced MNMs into 39 protein-coding genes through S. cerevisiae evolution, resulting in both arm homogenization and inversion of the IR spacer. These events in turn resulted in nonsynonymous substitutions and up to five neighboring amino acid replacements in a single gene. The study demonstrates that template switching is a powerful generator of multiple substitutions within codons. Additionally, some template switching events occurred more than once during S. cerevisiae evolution. Our findings suggest that template switching constitutes a general mutagenic mechanism that results in both nonsynonymous substitutions and parallel evolution, which are traditionally considered as evidence for positive selection, without the need for adaptive explanations.

Figure 1

Template switching causes IR arm homogenization. (a) A recently occurring perfect IR in UTP5 overlapped with a MNM on the branch leading to BC187 (shown as vertical red lines on the branch). Alignment of BC187 (perfect IR) and S288c (imperfect IR) is shown, with arrows representing IR arms, the dotted line representing the IR spacer and asterisks representing matches between strains. A very short IR with an arm length of 4 bp and a spacer of 7 bp appears in S288c. In strain BC187, a longer IR with an arm length of 11 bp was formed. (b–d) Template switching converted an imperfect IR to a perfect IR. (b) Assuming intramolecular template switching, the nascent strand folded on itself and served as a template. (c) Assuming intermolecular template switching, the nascent strand invades the sister chromatid and uses it as a template. (d) The second switch returns the nascent strand into the original template, resulting in a perfect IR, represented by the upper sequence in A. Uppercase letters represent the IR arms, with the same arms colored in the same color. The red dots represent mismatches between the arms. The direction of the replication fork is indicated by an arrow. (e) Six substitutions were induced during the template switching, resulting in the replacement of two amino acids (T38V, S39H).

Figure 2

Template switching events forming an IR in the gene MSH4. (a) The homogenous IR has an arm length of 8 bp (upper sequence). The IR homogenesis was formed by an AA → TT MNM and resulted in two amino acid changes, i.e., L394F, I395F. The spacer is not shown. (b) Placing of template switching events on the tree. Fourteen strains with perfect IR and the TT form are shown in red, strains with the AA form are shown in green and other forms are shown in black. Template switching occurred on three terminal branches in S. cerevisiae lineage (red circles): on the branches leading to YJM269, RedStar, and EC9-8. While not directly estimated in this study, an ancestral sequence reconstruction revealed one extra event on one internal branch. The best maximum likelihood tree reconstructed by RAxML using DNA sequences from 4304 orthologs and the ancestral sequence reconstruction of codons 394–395 performed by FASTML are presented. (see “Methods” section). Bootstrap values are shown on branches. Tree outgroup is not drawn to scale.

Figure 3

Intermolecular template switching causes spacer inversion. (a) A perfect IR in JAY291 on SYG1 overlapping with a MNM located on the spacer. The MNM occurred in the branch leading to JAY291 (shown as vertical lines on the tree). In this case, the sister taxon, represented here by S288c, also has a perfect IR with the same arms. (b) The mechanism that created spacer inversion (AAGTC → GACTT) was intermolecular template switching, with the first switch having occurred during synthesis of the first arm. (c) The second switch returned the nascent strand to the original template, resulting in an inverted spacer of JAY291 compared to S288c. (d) IR spacer inversion resulted in one amino acid replacement (Y46G) in JAY291. Uppercase letters represent the IR arms, with the same arms shown in the same color. The red dots represent mismatches. The direction of the replication fork is indicated by an arrow.

Figure 4

Intermolecular template switching causes arm homogenization and spacer inversion. (a) REG2 has a perfect IR with an arm length of 16 bp in YHM339. The ancestral form represented by S288c has an arm length of 5 bp. The MNM occurred on the IR arm in the branch leading to YHM339 (shown as vertical lines on tree). (b) Intermolecular template switching, in which the first switch occurred during the synthesis of the first arm, was responsible for spacer inversion. (c) The second switch returned the nascent strand to the original template, resulting in an inverted spacer of YHM339 compared to S288c. (d) Seven substitutions were induced by this event (three on the spacer and four on the arm, two of which were part of a continuous MNM) resulting in five modified amino acids (KSELDP → DFEQRR, positions 206–211).

Figure 5

Spacer flip inversion in IR of SPO75. (a) IR with an arm length of 11 bp shows a perfect form in all S. cerevisiae strains. The 2 bp-spacer has either the ancestral form AA (red) or the derived form TT (blue) in positions 1 and 2 of codon 409 coding AAA (K) and TTA (L) respectively. (b) Tree presenting the template switching events. Strains in red have AA on the spacer, strains in blue have TT on the spacer, and strains in gray have indels in this region. S. paradoxus has AA in this locus but does not share the same IR. Three transitions are shown in full circles (two from L into K, shown in red, and one from K into L, shown in blue) on terminal branches. According to ancestral sequence reconstruction, there were also one to two reversal events from K to L on internal branches, one of which has only low support (empty circle). The best maximum likelihood tree reconstructed by RAxML using DNA sequences of 4304 orthologs and the ancestral sequence reconstruction of codon 409 by FASTML, are presented (see “Methods” section). Bootstrap values are shown on branches. Tree outgroup is not drawn to scale.

Figure 6

Effect of template switching on codons and amino acids. (a) Location of substitutions on two neighboring codons (six nucleotide positions). Synonymous substitutions are shown in green, and nonsynonymous substitutions are shown in red. For example, the MNM of gene YGR098C (top) are located in the 3rd position of the first codon and on the 1st position of the second codon. Both substitutions caused the synonymous changes shown in green. Similarly, YIL047C (bottom) has four neighboring substitutions spanning two codons. The first codon has three substitutions that replace the amino acid, shown in red, and the second codon has one synonymous substitution, shown in green. (b) The amino acid changes induced by MNMs in 39 proteins are shown; one event from its type per gene locus. The matrix is colored by Grantham's physicochemical distance table which has values of 5–215, with a mean distance of 100. Green represents similar amino acid pairs, while red represents distant amino acid pairs. Thirteen events had an amino acid with a physicochemical distance of 120 or higher (in bold).