A two-step adaptive walk rewires nutrient transport in a challenging edaphic environment

Most well-characterized cases of adaptation involve single genetic loci. Theory suggests that multilocus adaptive walks should be common, but these are challenging to identify in natural populations. Here, we combine trait mapping with population genetic modeling to show that a two-step process rewired nutrient homeostasis in a population of Arabidopsis as it colonized the base of an active stratovolcano characterized by extremely low soil manganese (Mn). First, a variant that disrupted the primary iron (Fe) uptake transporter gene (IRT1) swept quickly to fixation in a hard selective sweep, increasing Mn but limiting Fe in the leaves. Second, multiple independent tandem duplications occurred at NRAMP1 and together rose to near fixation in the island population, compensating the loss of IRT1 by improving Fe homeostasis. This study provides a clear case of a multilocus adaptive walk and reveals how genetic variants reshaped a phenotype and spread over space and time.


Fig. S1. Relative chlorophyll content correlates with chlorosis observed by image analysis in a representative subset of Fogo population.
Linear correlation of the relative chlorophyll content and the greenness and the segmented area in cm 2 in a representative set of 31 accessions from Fogo. R 2 = Pearson's R 2 , P = p-value and n = number of genotypes. Pictures were taken 24 days after sowing. Relative chlorophyll content was evaluated with the multispeQ tool 5 to 6 weeks after sowing. Each dot represents the median across 3 replicates per accession.

Fig. S2. Variation in photosynthetic activity in the Fogo population.
Variation in relative chlorophyll content, quantum yield in Photosystem II (Phi2) and nonphotochemical quenching (PhiNPQ) in two different experiments. The plants were grown in standard potting mix and greenhouse conditions (12 hours of light, 21 ºC at day, 14 ºC at night). The measurements were taken at bolting. Each dot represents one replicate per genotype.

Fig. S3. Variation in leaf elements accumulation in the Fogo population relative to Santo Antão and Morocco.
Distribution of 19 elements in the leaves of Fogo (orange), Santo Antão (blue) and Morocco (green) plants. The plants were grown on standard potting mix in controlled growth chamber conditions (12 hours of light, 21 ºC at day, 14 ºC at night, 70% humidity). The values on the x axis are indicated in µg•g -1 of dry weight (ppm) and correspond to the median across replicates. The tissue was harvested five to six weeks after sowing.  Distribution of 23 elements in topsoils from Fogo in orange and Santo Antão in blue. Values on the x axis are shown in ng•g -1 of dry weight (ppb) except for Na, Mg, Ca, S, K and P indicated in µg•g -1 of dry weight (ppm); and correspond to the median across extractions per field site. The extractions were done with water. P = p-value for Mann-Whitney-Wilcoxon test. Values are shown in µg•g -1 of dry weight (ppm). Each dot represents one replicate per genotype.   . The x and y axes correspond to the position at chromosome 1 for the TAIR10 reference and the de novo assembled genome of F13-8, respectively. (C) Difference in NRAMP1 expression in the roots of 12 days old seedlings grown on MS media (3 biological replicates, 3 technical replicates per sample). NRAMP1 mRNA levels were normalized to the PP2A gene and compared to Col-0. Statistical significance was conducted with a Student's t-test (p-value indicated).

Fig. S10. NRAMP1 protein sequences are identical within Cape Verdean accessions.
Multiple alignments of NRAMP1 protein sequences from Fogo and Santo Antão plants compared to Col-0. Predicted transmembrane domains are highlighted in red.

Fig. S17. NRAMP1 tandem duplication haplotypes in the Fogo population.
IGV browser view of the genomic region at NRAMP1 in F4-2 (CT-TD haplotype -2 copies), F9-2 (ATG-TD haplotype -5 copies) and F13-8 (AAGACTAA-TD haplotype -3 copies). The arrow indicates NRAMP1. The reads are shown as pairs and aligned to the TAIR10 reference genome.      NRAMP1 copy number estimates in the Fogo population based on the analysis described in Fig. 4B. The different NRAMP1 haplotypes are annotated by color (AAGACATAA-TD in beige, ATG-TD in burgundy, CT-TD in pink and the single NRAMP1 copy in blue).

Fig. S25. NRAMP1 expression analysis.
NRAMP1 expression in roots of 14-day-old seedlings grown on agar-solidified Hoagland media. mRNA levels were quantified with a probe-based digital PCR assay (2 to 4 biological replicates). NRAMP1 copy numbers are indicated within the circles. NRAMP1 mRNA levels were normalized to the PP2A gene and compared to Col-0 (+Mn). The statistical significance was conducted with a Kruskal-Wallis test. Significance level used 5% after Bonferroni correction. Groups sharing a letter are not significantly different.

Fig. S26. Correlation between NRAMP1 copy number and chlorosis variation in the Fogo population.
Correlation between NRAMP1 copy number and greenness, segmented area in cm 2 and relative chlorophyll content in two independent experiments. R = Spearman's rho, P = p-value and n = number of genotypes.

Fig. S27. Correlation between NRAMP1 copy number and 19 elements in the leaves of Fogo natural accessions.
The plants were grown on standard potting mix in controlled growth chamber conditions (12 hours of light, 21 ºC at day, 14 ºC at night, 70% humidity). The leaf elements concentrations on the y axis are indicated in µg•g -1 of dry weight (ppm) and correspond to the median across replicates per accession. The values on the x axis correspond to the NRAMP1 copy numbers. The tissue was harvested five to six weeks after sowing. R = Spearman's rho, P = p-value, n =103).      Table S14: Chip heritability estimated with GEMMA using a linear mixed model; PVE corresponds to the proportion of phenotypic variance explained by genotypes or chip heritability; PVENRAMP1 corresponds to the chip heritability using NRAMP1 copy number as covariate. Table S15: Variants at the genomic region from CLR overlapping with IRT1. Table S16: Alleles age estimates based on Relate trees. Table S17: Inferred selection coefficients for IRT1 130X using Clues. logLR corresponds to the log likelihood. s corresponds to the selection coefficient. Table S18: Inferred selection coefficients for NRAMP1 AAGACATAA-TD and ATG-TD using Clues. logLR corresponds to the log likelihood. s corresponds to the selection coefficient. Table S19: Primers used in this study.