Local adaptation is common in plants, yet characterization of its underlying genetic basis is rare in herbaceous perennials. Moreover, while many plant species exhibit intraspecific chemical defense polymorphisms, their importance for local adaptation remains poorly understood. We examined the genetic architecture of local adaptation in a perennial, obligately-outcrossing herbaceous legume, white clover (Trifolium repens). This widespread species displays a well-studied chemical defense polymorphism for cyanogenesis (HCN release following tissue damage) and has evolved climate-associated cyanogenesis clines throughout its range. Two biparental F₂ mapping populations, derived from three parents collected in environments spanning the U.S. latitudinal species range (Duluth, MN, St. Louis, MO and Gainesville, FL), were grown in triplicate for two years in reciprocal common garden experiments in the parental environments (6,012 total plants). Vegetative growth and reproductive fitness traits displayed trade-offs across reciprocal environments, indicating local adaptation. Genetic mapping of fitness traits revealed a genetic architecture characterized by allelic trade-offs between environments, with 100% and 80% of fitness QTL in the two mapping populations showing significant QTL X E interactions, consistent with antagonistic pleiotropy. Across the genome there were three hotspots of QTL co-localization. Unexpectedly, we found little evidence that the cyanogenesis polymorphism contributes to local adaptation. Instead, divergent life history strategies in reciprocal environments were major fitness determinants: selection favored early investment in flowering at the cost of multi-year survival in the southernmost site vs. delayed flowering and multi-year persistence in the northern environments. Our findings demonstrate that multi-locus genetic tradeoffs contribute to contrasting life history characteristics that allow for local adaptation in this outcrossing herbaceous perennial.

Dryad files contain the following information:

1) Raw phenotype data used for all quantitative trait and genetic mapping analyses (four.csv files for the two reciprocal comparisons: DMN, GFL-DG, STL, and GFL-SG).

2) Full results from QTL mapping analysis (R printouts from all models as.txt files)

3) R scripts for various analyses (corresponding figures and tables from the study are indicated in parentheses):

-λ calculations and correlations with Year 1 floral count (Figure S1)

-within-site trait variance and broad-sense heritability (Table S3)

-within-site trait correlations and PCAs (Figure S3,S6)

-across-site trait variance (Figure 1, Table S5, Table S6)

-rGE correlations (Figure 2, Table S4)

-QTL mapping analysis (Figure 3, Table 2, Table S7)

-QTL×E post-hoc analysis (Figure 4, Figure S7,S8, Table S8)

Abbreviations and terms used in files:
HCN, cyanogenic response (0=acyanogenic, 1 = cyanogenic)
Cyanotype, cyanogenic response defined with respect to the two required cyanogenic components: AcLi (both components present), Acli (cyanogenic glucosides present without linamarase), acLi (linamarase present), acli (neither cyanogenic component present)
Units for Flowering duration data: days
Units for Vegetative area data: square cm