Multilevel Gene Expression Changes in Lineages Containing Adaptive Copy Number Variants

Abstract

Copy number variants (CNVs) are an important class of genetic variation that can mediate rapid adaptive evolution. Whereas, CNVs can increase the relative fitness of the organism, they can also incur a cost due to the associated increased gene expression and repetitive DNA. We previously evolved populations of Saccharomyces cerevisiae over hundreds of generations in glutamine-limited (Gln-) chemostats and observed the recurrent evolution of CNVs at the GAP1 locus. To understand the role that gene expression plays in adaptation, both in relation to the adaptation of the organism to the selective condition and as a consequence of the CNV, we measured the transcriptome, translatome, and proteome of 4 strains of evolved yeast, each with a unique CNV, and their ancestor in Gln- chemostats. We find CNV-amplified genes correlate with higher mRNA abundance; however, this effect is reduced at the level of the proteome, consistent with post-transcriptional dosage compensation. By normalizing each level of gene expression by the abundance of the preceding step we were able to identify widespread differences in the efficiency of each level of gene expression. Genes with significantly different translational efficiency were enriched for potential regulatory mechanisms including either upstream open reading frames, RNA-binding sites for Ssd1, or both. Genes with lower protein expression efficiency were enriched for genes encoding proteins in protein complexes. Taken together, our study reveals widespread changes in gene expression at multiple regulatory levels in lineages containing adaptive CNVs highlighting the diverse ways in which genome evolution shapes gene expression.

Keywords: Ssd1; adaptation; chemostat; copy number variation; gene expresssion; uORF.

Fig. 1.

Study design and characterization of CNV strains. a) Strain provenance (Lauer et al. 2018; above the dotted line) and experimental design to assess gene expression effects using RNA-seq, ribosome profiling (Ribo-seq), and TMT-labeled mass spectrometry (this study, beneath dotted line) in glutamine-limited chemostats. b) A schematic showing the GAP1-locus amplified genes in each strain and their copy number. A common core of 17 genes is amplified in every evolved strain.

Fig. 2.

Gene expression divergence in evolved lineages containing CNVs. a) A clustered heatmap of the ratio of evolved over ancestral transformed data (methods) expression at multiple levels (RNA, RPF, and MS) of genes from all strains. The 218 genes shown are all significantly different in abundance (DESeq2 (RNA, RPF); Welch's t-test (MS); BH adj. P-value < 0.01) at multiple levels of gene expression, or are CNV amplified in at least 1 strain. b) Scatterplot of RNA abundance of genes with significantly different (DESeq2, BH adj. P-value < 0.05) expression between each of the 4 evolved strains and the ancestor. All CNV-amplified genes are significantly higher in the evolved strain. c) To assess if there were changes in transcription efficiency, we corrected RNA abundance for gene CN and found that a minority of genes in each strain (median of 36%) exhibit significantly different CN-corrected expression (DESeq2, BH adj. P-value < 0.05).

Fig. 3.

Divergence in translation efficiencies and potential regulatory mechanisms. a) 561 genes have significantly different translation efficiency relative to the ancestor in at least 2 evolved strains (FET, P-value < 0.05). Gene names in orange indicate CNV-amplified genes. b) Genes with significantly different translation efficiency and the presence of a high confidence uORF (circle), significant SSD1 motifs (square), or both (3.8-fold higher than expected at random, HGM, P-value = 1e−7). c) Ribosome binding within the TL (line) of UTH1 (box) in the ancestor and evolved strains. High confidence predicted uORFs are shown in gray, start codons and stop codons, low confidence uORFs are hollow boxes. Ssd1 binding motifs are shown in purple. The 5′ terminus of the transcript is denoted by the flat arrowhead. d). Example of uORFs and SSD1 binding motifs upstream of SSD1. The 5′ terminus of the transcript extends between the indicated region as indicated by the pointed arrowhead.

Fig. 4.

Divergence in protein expression efficiencies and potential regulatory mechanisms. Evolved and Ancestor strain protein expression efficiency ratios for a) genes with significantly different protein expression efficiency relative to the ancestor in at least 2 evolved strains) and for b) genes that have significantly different protein expression efficiency and are part of multimeric protein complexes are shown either above the diagonal (higher efficiency in Evolved strain) or below (lower efficiency in Evolved strain). Genes with significantly lower protein expression efficiency are 1.9-fold higher than random, HGM, P-value < 1e−7).