Following gene duplication, paralog interference constrains transcriptional circuit evolution

Abstract

Most models of gene duplication assume that the ancestral functions of the preduplication gene are independent and can therefore be neatly partitioned between descendant paralogs. However, many gene products, such as transcriptional regulators, are components within cooperative assemblies; here, we show that a natural consequence of duplication and divergence of such proteins can be competitive interference between the paralogs. Our example is based on the duplication of the essential MADS-box transcriptional regulator Mcm1, which is found in all fungi and regulates a large set of genes. We show that a set of historical amino acid sequence substitutions minimized paralog interference in contemporary species and, in doing so, increased the molecular complexity of this gene regulatory network. We propose that paralog interference is a common constraint on gene duplicate evolution, and its resolution, which can generate additional regulatory complexity, is needed to stabilize duplicated genes in the genome.

Fig. 1. Function and evolution of MADS-box proteins in hemiascomycete yeasts

(A) In K. lactis, an Mcm1 homodimer regulates the ARG genes by interacting with Arg81 and binding a specific DNA sequence. (B) In S. cerevisiae, an Mcm1-Arg80 heterodimer interacts with Arg81 to regulate ARG genes. (C) An Mcm1 homodimer interacts with Matα1 to regulate α-specific genes in K. lactis and (D) S. cerevisiae. (E) A maximum likelihood phylogeny of MADS-box domain proteins in hemiascomycete yeasts. A tandem gene duplication generated paralogs Mcm1 and Arg80 in the last shared common ancestor of Zygosaccharomyces rouxii and S. cerevisiae. Circles denote ancestral proteins reconstructed in this study. Asterisks on internal branches correspond to approximate-likelihood ratio support for the monophyly of the descendant clade: *** denotes support > 10.0; ** denotes support > 5.0. subs, substitutions.

Fig. 2. The preduplication ancestral gene complements both paralogs

(A) Growth of ancestral MADS-box gene strains using ornithine as a sole nitrogen source. Ornithine is converted into arginine and then modified to produce the other essential amino acids. In the absence of a functional ARG gene regulatory complex, strains cannot use ornithine as a nitrogen source. The preduplication AncMADS can supply the function of the modern Arg80 paralog (purple), but the postduplication AncMcm1 paralog cannot (blue). “Growth” on the y axis is the ratio of optical density at 600 nm (OD₆₀₀) at the indicated time point divided by OD₆₀₀ at time zero. (B to E) Gene expression profiling of ancestral MADS-box proteins in S. cerevisiae quantified with NanoString (www.nanostring.com). (B) MADS-box activated ARG genes. Row 1, CAR1; row 2, CAR2. (C) MADS-box repressed ARG genes. Row 1, ARG3; row 2, ARG5,6. (D) MADS-box activated mating genes (α-specific genes). Row 1, SAG1; row 2, MFa1; row 3, STE3. (E) MADS-box repressed mating genes (a-specific genes). Row 1, STE2; row 2, STE6. In each experiment, mean and standard error (indicated by error bars) were determined using three replicates.

Fig. 3. Divergence in cofactor and DNA-binding following gene duplication of ancestral MADS-box proteins

(A) Alignment of the N-terminal 63 amino acids of the MADS-box domain with residues that changed identity between AncMcm1, AncArg80, and AncMADS in color (24). α1 denotes a long α helix; β1 and β2 signify an antiparallel β sheet. (B) Gene expression profiling to determine the impact of mutants on the function of preduplication AncMADS protein in S. cerevisiae. Gene expression quantified using NanoString. Panel 1: MADS-box activated ARG genes; row 1, CAR1; row 2, CAR2. Panel 2: MADS-box repressed ARG genes; row 1, ARG3; row 2, ARG5,6. Panel 3: MADS-box activated mating genes (α-specific genes); row 1, SAG1; row 2, MFa1. Mean and standard error (indicated by error bars) were determined from three replicates. (C) After duplication, AncArg80 lost the ability to form a strong interaction with Matα1, and AncMcm1 lost the ability to form a strong interaction with Arg81. These losses destroyed the abilities of AncArg80 and AncMcm1 to regulate α-specific genes and ARG genes, respectively. (D) Half-lives of MADS-box ancestors on the S. cerevisiae CAR2 cis-regulatory sequence (see supplementary materials and methods). Saturating levels of unlabeled DNA were added at time point zero. (E) After the duplication of AncMADS, α-specific genes are regulated by a homodimer of AncMcm1, whereas ARG genes are regulated by a heterodimer of AncMcm1 and AncArg80 due to the reduced affinity of AncArg80 for DNA.

Fig. 4. Paralog interference between Arg80 and Mcm1

(A) The ratio of functional binding to inhibitory binding determines the extent of competitive interference between Mcm1 and Arg80. (B) Arg80, AncArg80, and AncArg80 mutant with DNA-binding surface changes reverted to ancestral state interfere with α-specific gene expression. α-specific gene expression was quantified by quantitative reverse-transcriptase fluorescence polymerase chain reaction using SAG1 transcript (normalized to URA6 transcript). Endogenous expression is driven by the native ARG80 promoter, and overexpression is driven by the TEF1 promoter. For gene expression experiments, mean and standard error (indicated by error bars) were determined from five replicates.