Yeast-based evolutionary modeling of androgen receptor mutations and natural selection

Abstract

Cancer progression is associated with the evolutionary accumulation of genetic mutations that are biologically significant. Mutations of the androgen receptor (AR) are associated with the development of prostate cancer (PCa) by responding to non-androgenic hormones, and the lack of annotations in their responsiveness to hormone ligands remains a daunting challenge. Here, we have used a yeast reporter system to quickly evaluate the responsiveness of all fifty clinical AR mutations to a variety of steroidal ligands including dihydrotestosterone (DHT), 17β-estradiol (E2), progesterone (PROG), and cyproterone acetate (CPA). Based on an AR-driven reporter that synthesizes histidine, a basic amino acid required for yeast survival and propagation, the yeast reporter system enabling clonal selection was further empowered by combining with a random DNA mutagenesis library to simulate the natural evolution of AR gene under the selective pressures of steroidal ligands. In a time-frame of 1-2 weeks, 19 AR mutants were identified, in which 11 AR mutants were validated for activation by tested steroidal compounds. The high efficiency of our artificial evolution strategy was further evidenced by a sequential selection that enabled the discovery of multipoint AR mutations and evolution directions under the pressure of steroidal ligands. In summary, our designer yeast is a portable reporter module that can be readily adapted to streamline high-throughput AR-compound screening, used as a PCa clinical reference, and combined with additional bioassay systems to further extend its potential.

Fig 1. Correlation of AR-mediated reporter activities between designer yeast and mammalian systems.

(A) Schematic representation of our designer yeast platform compared with the mammalian luciferase reporter system. (B) The list of 56 PCa-associated AR-LBD mutations accumulated in ARDB during a 22-year period (from 1990 to 2012). Top ten frequent mutations were marked in blue. (C), The Correlation analyses were based on fold changes in activation of AR-LBD mutants as normalized to the wild-type AR between the yeast and mammalian reporter systems, using the top ten frequent mutations in (B) against a variety of concentrations of DHT, E2, PROG, and CPA. Pearson’s correlation coefficient (R) and P-values were listed in the figures. Hormone concentrations in the yeast assay were DHT 10⁻⁸ M, E2 10⁻⁵ M, PROG 10⁻⁵ M, and CPA 10⁻⁵ M. Hormone concentrations in the human Hep3B reporter assay were DHT 10⁻⁹ M, E2 10⁻⁷ M, PROG 10⁻⁹ M, and CPA 10⁻⁷ M.

Fig 2. The functional analysis of 50 clinical AR mutants recorded in ARDB in response to steroidal ligands.

The “+” mark indicates a full response of the AR mutants to indicated ligands was observed in plate yeast assays compared with LBD-WT, while the “–” mark indicates the opposite. The “Partial” mark indicated a partial ligand-induced response of the AR mutants. The concentration of all ligands in the yeast assay were 2 μM. Plates were photographed 60 h after incubation. Thirteen mutants (R630Q, K631T, S648N, E666D, Q671R, L702H, V731M, S783N, Q799E, R847G, M887I, K911R and Q920R) behaved similarly to the wild type protein showing full response to the physiological ligand DHT and partial response to a high dose (2 μM) of E2. Eight mutants (Q641X, K721E, W742X, M750I, W752X, V758A, R787X, Q868X) were identified as loss-of function mutations failed to respond to any tested steroids. Mutations ending in X like Q641X indicated that a stop-codon was generated in this position resulting in the expression of a truncated AR protein.

Fig 3. Simulation of AR evolution by combining designer yeast with random mutagenesis.

(A) Characterization of the AR-LBD EP-PCR library. (B) Yeast clonal selection against compounds based on solid culture in 96-well plates without the presence of 3-AT. Ligands were 10⁻⁸ M DHT, 10⁻⁵ M PROG, 10⁻⁵ M E2, and 10⁻⁵ M CPA. The heatmap represents OD₆₀₀ values of liquid cultures measured 21 h post-incubation. LBD-WT is indicated by pentacles; the × mark indicates blank wells. (C-D) Functional analysis of 19 identified AR mutations using both liquid (n = 6) and plate yeast assay in the presence of 3-AT against DHT 2 μM, E2 2 μM, PROG 2 μM, and CPA 2 μM. The incubation time was 60 h. Bars indicate mean ± s.d.

Fig 4. The evaluation of identified AR mutants in a dose-dependent yeast assay against steroidal ligands.

(A) Liquid dose-dependent yeast assay with 25 mM 3-AT added (n = 5). OD₆₀₀ values were measured 48 h post-incubation. Bars indicate mean ± s.d. (B) Plate dose-dependent yeast assay with 25 mM 3-AT added. Plates were photographed 48 h post-incubation.

Fig 5. Sequential simulation of AR natural evolution.

(A) Two representative mutations per type are shown in yeast assays (n = 6) and luciferase assays (n = 3). OD₆₀₀ values were measured 60 h post-incubation. The concentration of each ligand was 2 μM for yeast assays and 10 nM for luciferase assays. (B) The assessment of double-mutations identified in the second-round evolution, K778E/T878A and K778E/T878S, against tested steroidal ligands in liquid (n = 6) and plate yeast assays (with 25 mM 3-AT added). The concentration of each ligand was 2 μM. OD₆₀₀ values were measured 48 h post-incubation. Plates were photographed 48 h post-incubation. Bars indicate mean ± s.d.