L8708×Z020

Prior field screening led to the identification of an F₂ population, L8708 x Z020, which segregates for plant height, shoot biomass, root shape, and root color. This population was derived from a cross between L8708, an orange inbred line with a medium-long storage root and compact shoots, and Z020, a yellow, cultivated landrace from Uzbekistan with a short, blunt-tipped storage root with broad, prostrate leaves. A single F₁ plant was selected from this cross and selfed to produce the F₂ population used in this study.

Linkage maps were constructed using the JoinMap 4.1 software (Van Ooijen 2011). Markers and genotypes with more than 20% missing data and which deviated from expected segregation ratios based on a Chi-square test (P < 0.001) were excluded. All linkage groups were obtained at a LOD threshold greater than 10. The regression mapping algorithm was used with Haldane’s mapping function to calculate the distance between markers (Haldane 1919). Haldane’s function, which assumes no crossover interference, was chosen over Kosambi’s mapping function based on previous observations in carrot that Haldane’s function provides a more accurate marker placement relative to the physical map (Parsons et al. 2015). Marker names corresponded to physical locations in the genome, allowing assessment of accurate marker placement and identification of inconsistencies. After initial mapping, markers defined as having insufficient linkage were flipped to the opposite phase and remapped. Maps from round 2 were used for QTL analysis.

QTL analysis was conducted in R 3.3.2 (R Core Team 2016) using the R/qtl package (Broman and Sen 2009). Genotype probabilities were calculated using a step value of one for the entire linkage map and an assumed genotyping error rate of 0.001. Missing genotype data was replaced with the most probable values using the Viterbi algorithm (method = ‘argmax’) in the ‘fill.geno’ function.

Multiple QTL mapping (MQM) (Jansen and Stam 1994) was performed in R/qtl using the ‘mqmscan’ function with an additive model and cofactor significance set to 0.001 (Arends et al. 2010). Cofactors were set at a fixed marker interval of 5 cM. Genome-wide LOD significance thresholds (α = 0.01) were determined separately for each phenotype after running 1,000 random permutations (Churchill and Doerge 1994) with the assumed genotyping error rate set to 0.001. For each QTL, confidence intervals were determined using the 1.5 LOD drop off flanking the highest peak of the QTL. QTL were named using an abbreviation for the trait (e.g. ht, height) suffixed with the linkage group (1-9), and finally the serial number of QTLs in the linkage group (e.g. ht-2.1, ht-2.2).