AGO1 and HSP90 buffer different genetic variants in Arabidopsis thaliana

Abstract

Argonaute 1 (AGO1), the principal protein component of microRNA-mediated regulation, plays a key role in plant growth and development. AGO1 physically interacts with the chaperone HSP90, which buffers cryptic genetic variation in plants and animals. We sought to determine whether genetic perturbation of AGO1 in Arabidopsis thaliana would also reveal cryptic genetic variation, and if so, whether AGO1-dependent loci overlap with those dependent on HSP90. To address these questions, we introgressed a hypomorphic mutant allele of AGO1 into a set of mapping lines derived from the commonly used Arabidopsis strains Col-0 and Ler. Although we identified several cases in which AGO1 buffered genetic variation, none of the AGO1-dependent loci overlapped with those buffered by HSP90 for the traits assayed. We focused on 1 buffered locus where AGO1 perturbation uncoupled the traits days to flowering and rosette leaf number, which are otherwise closely correlated. Using a bulk segregant approach, we identified a nonfunctional Ler hua2 mutant allele as the causal AGO1-buffered polymorphism. Introduction of a nonfunctional hua2 allele into a Col-0 ago1 mutant background recapitulated the Ler-dependent ago1 phenotype, implying that coupling of these traits involves different molecular players in these closely related strains. Taken together, our findings demonstrate that even though AGO1 and HSP90 buffer genetic variation in the same traits, these robustness regulators interact epistatically with different genetic loci, suggesting that higher-order epistasis is uncommon. Plain Language Summary Argonaute 1 (AGO1), a key player in plant development, interacts with the chaperone HSP90, which buffers environmental and genetic variation. We found that AGO1 buffers environmental and genetic variation in the same traits; however, AGO1-dependent and HSP90-dependent loci do not overlap. Detailed analysis of a buffered locus found that a nonfunctional HUA2 allele decouples days to flowering and rosette leaf number in an AGO1-dependent manner, suggesting that the AGO1-dependent buffering acts at the network level.

Keywords: AGO1; HSP90; HUA2; Plant Genetics and Genomics; buffering; capacitors; epistasis.

Fig. 1.

Perturbation of AGO1 increases phenotypic variation among isogenic seedlings. a) Early seedling trait measures for wild-type (Col-0 WT), ago1-46, and ago1-27 seedlings. Ten-day-old seedlings were scored for 3 different morphological traits: lesions, asymmetrical rosettes, and organ defects. The data represent 2 biological replicates (2 replicates, n = 144 for ago1 mutants and n = 216 for Col-0 WT, *P < 0.05, ttest). b) Hypocotyl mean length and variance differ between wild-type and ago1-mutant seedlings. Hypocotyl length was measured for 7-day old, dark-grown seedlings. ago1-27 mutant seedlings showed greater variance than Col-0 wild-type seedling in hypocotyl length (Levene’s test, P < 1.0E−03; n = 475 for ago1-27, n = 486 for Col-0 WT). Inset: boxplots of hypocotyl length means. Y-axis represents hypocotyl length (mm), **P < 1.0E−15, Mann–Whitney Wilcoxon test.

Fig. 2.

Perturbation of AGO1 buffers genetic variation. a) Experimental design to examine the phenotypic consequences of genetic variation within the STAIRS in the context of the ago1-27 mutation. b) Summary of examined quantitative traits with evidence for AGO1-dependent or Ler-dependent variation in each tested STAIRS line. AGO1 perturbation reveals a cryptic genetic variant if this variant’s contribution to a quantitative trait can be detected only in an ago1-mutant background. AGO1 perturbation conceals a genetic variant if this variant’s contribution to a quantitative trait can no longer be detected in an ago1-mutant background. Genetic variation in the respective Ler segments can epistatically interact (i.e. mask) the phenotypic differences observed between Col-0 wild-type seedlings and ago1-27 mutant seedlings in the Col-0 background. For STAIRS line N9472, 78 seedlings were measured for hypocotyl length in the dark; for STAIRS lines N9448, N9459, and N9501, 100 seedlings were measured for this trait. At least 32 plants were measured for all other traits. See Supplementary Tables 2 and3 for trait values and assessment of significance. c) Three examples of AGO1-dependent and 1 example of Ler-dependent genetic variation are shown for 3 different traits. left, Col-0 WT; left-middle, STAIRS; right-middle, ago1-27 in the Col-0 background; right, ago1-27 in a STAIRS background.

Fig. 3.

AGO1 perturbation uncouples the traits days to flowering and rosette leaf number. Plants for Col-0 WT, STAIRS9472, Col-0 ago1-27, and STAIRS9472; ago1-27 were grown in a random block design in LD, n = 30–36. Days to flowering were recorded and rosette leaf numbers at the onset of flowering were counted. Blue, Col-0 WT; yellow, STAIRS9472; red, ago1-27 in the Col-0 background; green, ago1-27 in the STAIRS9472 background. a) Days to flowering. The ago-1 mutant flowered ∼9 days later than Col-0 WT (P = 5.51E−12, Mann–Whitney Wilcoxon test); no significant difference was found between STAIRS9472 and the ago-1 mutant in the STAIRS9472 background (P = 0.4714, Mann–Whitney Wilcoxon test). The Ler introgression in STAIRS9472 was epistatic to ago1-27, *P < 0.0001, Mann–Whitney Wilcoxon test. b) Rosette leaf number. Col-0 WT plants showed fewer leaves than ago1-27 mutant plants, consistent with the mutant’s late flowering phenotype. In the STAIRS9472 background, ago1-27 mutant plants showed no change in the number of days to flowering; however, these plants developed significantly fewer leaves (P = 3.45E−12, Mann–Whitney Wilcoxon test). c) Scatter plot of the 2 measured traits days to flowering and rosette leaf number in the 4 tested genotypes. The traits were less well correlated in the ago-1 mutant in the Col-0 background (compare blue and green dots); however, the normally tight correlation was fully lost in the STAIRS background (compare red and yellow dots). d) Known flowering time genes are residing within the Ler chromosome 5 region of the STAIRS9472 line. e) Quantitative PCR measurements for candidate gene expression. TFL1 and SPL4 were significantly increased in expression in the STAIRS9472; ago1-27 background. Fourteen-day-old plant tissue was collected at ZT16 (Zeitgeber 16; 16 h after dawn). Mean expression data represent 2 biological replicates, each with 3 technical replicates. Standard error is indicated (*P < 0.05, **P < 0.005, T-test).

Fig. 4.

Bulk segregant analysis identifies the nonfunctional Ler hua2 allele as a candidate AGO1-dependent locus. a) F₂ plants from ago1-27 × STAIRS9472 cross were grown in LD, phenotypes were recorded, and plants were genotyped for the ago1-27 allele. For bulk segregant analysis, we selected plants that were homozygous for the ago1-27 mutation and flowered with 6 or fewer leaves (n = 100), resembling the AGO1-dependent STAIRS9472 phenotype. Representative F₂ago1-27 plants at flowering are shown. Scale bar = 1 cm. b) Bulk segregant analysis. Red dots represent Ler-allele frequencies on chromosome 5 (bp, x-axis). Allele frequencies (y-axis) were estimated as the fraction of reads supporting a Ler allele divided by the number of reads mapping to that locus. Dashed blue line represents sliding window-based allele frequencies as estimated by SHOREmap. Dashed black line represents window-based plot boost as estimated by SHOREmap. The Ler hua2-5 allele emerged as the candidate AGO1-dependent locus because Ler-allele frequencies were highest at this locus compared with other regions on chromosome 5. d) F₂ plants homozygous for the ago1-27 mutation with 6 or fewer leaves at flowering were PCR genotyped for alleles at FLC, HUA2, and MIR156f loci. The near perfect enrichment of Ler hua2-5 allele validates the result of our bulk segregant analysis.

Fig. 5.

The ago1-27; hua2-4 double mutant uncouples the traits days to flowering time and rosette leaf number in the Col-0 background. An F₂ population segregating for the ago1-27 and hua2-4 mutant alleles was grown in LD. Days to flowering were recorded and rosette leaf numbers at the onset of flowering were counted. Gray, Col-0 WT; white, ago1-27 parent; red, ago1-27; HUA2+/+ F2; blue, hua2-4 parent; green, ago1-27; hua2-4. See Supplementary Table 5 for further details. a) Plants carrying a homozygous ago1-27 allele flowered ∼8.6 days later than Col-0 WT with ∼16 leaves. Plants carrying a homozygous hua2-4 allele initiated flowering ∼4.5 days earlier than Col-0 WT. As observed for STAIRS9472; ago1-27, the hua2-4 mutant allele was epistatic to ago1-27. *P < 0.0283, **P < 1.0E−06, Mann–Whitney Wilcoxon test. The double mutant ago1-27; hua2-4 plants showed a similar mean value but greater trait variance. b) The rosette leaf number phenotype of the double mutant ago1-27; hua2-4 plants resembles that of the STAIRS9472; ago1-27 line. ago1-27; hua2-4 plants flower with 5 leaves on average (*P < 0.0283, **P < 1.0E−06, Mann–Whitney Wilcoxon test, for (a) and (b)]. c) Scatter plot with rosette leaf number on the x-axis and days to flowering on the y-axis. Data are shown for F₂ plants that are homozygous for the ago1-27 mutant allele and segregate for the hua2-4 mutant allele. d) Known flowering time genes with differential expression in the ago1-27; hua2-4 double mutant as determined by RNA-seq. e) Suggested intersection of miRNAs and HUA2 in flowering time pathways. Depicted in red is the proposed connection between HUA2 and SPL4 such that SPL4 expression is regulated by both mir156 and HUA2. In bold, genes that are overexpressed in the double mutant ago1-27, hua-2 relative to the single mutant ago1-27.