SFF×HM2

Thesis figure 2.4 is a merged linkage map of carrot chromosomes that incorporates significant QTL for M. incognita nematode resistance from three populations (Br1091xHM1, SFFxHM2, HM3). The bars represent 1.5 LOD support intervals and the populations are coded with Br1091xHM1 as solid bars, SFFxHM2 as open bars, and HM3 as cross hashed bars. Numbers in parentheses indicate the largest LOD score followed by the percent phenotypic variation explained by that QTL.

Preliminary screening of several hundred cultivated and wild carrot populations led to the discovery of M. incognita resistance in “Brasilia” seed lot 1091 (Br1091) (Matthews et al. 1999), the Syrian cultivar “Homs” (HM), and in a population derived from an cross between the European cultivars “Scarlet Fancy” and “Favourite” (SFF). A Br1091 plant was crossed with a HM plant to generate the Br1091xHM1 F₂ population from a single F₁ plant. A SFF plant was intercrossed to a second HM plant to generate the SFFxHM2 F₂ population from a single F₁ plant. HM1 and HM2 were siblings derived from a self-pollinated HM selection. A third HM plant was self-pollinated to generate the HM3 population. This population had undergone five generations of selfing before the final self-pollination to produce the population HM3, and it was still segregating for resistance. All parent plants (Br1091, SFF, HM1, HM2, and HM3) had been previously evaluated in a greenhouse screen as described below (Br1091) or evaluated in a M. incognita infested field and identified as resistant to M. incognita with limited gall formation on the carrot root. F₃ families were derived from 95 Br1091xHM1 F₂ and 34 SFFxHM2 F₂ plants.

Linkage maps were constructed with JoinMap 3.0 software (Van Ooijen 2001). Markers and genotypes with more than 10% missing data and markers that significantly deviated from expected segregation ratios using a Chi-square test (P<0.01) were removed. For linkage groups with significant segregation distortion, all markers were used to generate the linkage map for that distorted linkage group. Linkage groups were obtained at a LOD threshold >3.0. The regression mapping algorithm was used with Haldane‟s mapping function to calculate distances between markers. Haldane‟s mapping function provides more accurate marker placement according to the carrot physical map than the Kosambi‟s mapping function (data not shown). Each marker was coded twice, once for each parental phase. The linkage groups were properly phased by using marker scores for individuals related to the parents (Gomez et al. 1996; Vivek and Simon 1999). The marker order was further examined using CheckMatrix (http://www.atgc.org/XLinkage, Truco et al. 2013) for inconsistencies and markers with more than one inconsistent score were removed. To remove redundant markers in the Br1091xHM1 population, a genetic bin map was developed. For each linkage group pair-wise recombination values among all markers were calculated. Adjacent markers with zero recombination among them were assigned to the same genetic bin. In addition, adjacent markers with “false” recombination due to missing data were considered to belong to the same genetic bin. The marker with the least missing data was chosen to represent each genetic bin. SNPs and SSRs with known chromosome locations were used to anchor the linkage groups. After being assigned to chromosomes, linkage groups were oriented and numbered following the chromosome orientation and classification of Iovene et al. (2011).

QTL analysis was performed in all three populations using R/qtl with the multiple imputations method (Broman and Sen 2009). QTL detection for each population included preliminary QTL identification using scanone followed by QTL modeling. The largest LOD peak from the analysis was added to the QTL model and if the QTL model was significant, it was retained. This process was then repeated using addqtl, instead of scanone, followed by QTL modeling and testing for interactions until adding additional QTL to the model was no longer significant. The final step used addpair to add a pair of interacting QTL or the interaction between a QTL in the model and a newly identified QTL. The support intervals were calculated using a 1.5 LOD drop (Broman and Sen 2009). QTL were named Mi-population-C_-Q_ where “population” is the population in which the QTL was identified, “C_” is the chromosome on which the QTL was identified and “Q_” is the QTL identifier from the QTL model. For example, Mi-BrHM1-C2-Q1 was mapped in the Br1091xHM1 population, is on chromosome 2, and is QTL that explains the most variation in the model.

Preliminary screening of several hundred cultivated and wild carrot populations led to the discovery of M. incognita resistance in “Brasilia” seed lot 1091 (Br1091) (Matthews et al. 1999), the Syrian cultivar “Homs” (HM), and in a population derived from an cross between the European cultivars “Scarlet Fancy” and “Favourite” (SFF). A Br1091 plant was crossed with a HM plant to generate the Br1091xHM1 F₂ population from a single F₁ plant. A SFF plant was intercrossed to a second HM plant to generate the SFFxHM2 F₂ population from a single F₁ plant. HM1 and HM2 were siblings derived from a self-pollinated HM selection. A third HM plant was self-pollinated to generate the HM3 population. This population had undergone five generations of selfing before the final self-pollination to produce the population HM3, and it was still segregating for resistance. All parent plants (Br1091, SFF, HM1, HM2, and HM3) had been previously evaluated in a greenhouse screen as described below (Br1091) or evaluated in a M. incognita infested field and identified as resistant to M. incognita with limited gall formation on the carrot root. F₃ families were derived from 95 Br1091xHM1 F₂ and 34 SFFxHM2 F₂ plants.

Linkage maps were constructed with JoinMap 3.0 software (Van Ooijen 2001). Markers and genotypes with more than 10% missing data and markers that significantly deviated from expected segregation ratios using a Chi-square test (P<0.01) were removed. For linkage groups with significant segregation distortion, all markers were used to generate the linkage map for that distorted linkage group. Linkage groups were obtained at a LOD threshold >3.0. The regression mapping algorithm was used with Haldane‟s mapping function to calculate distances between markers. Haldane‟s mapping function provides more accurate marker placement according to the carrot physical map than the Kosambi‟s mapping function (data not shown). Each marker was coded twice, once for each parental phase. The linkage groups were properly phased by using marker scores for individuals related to the parents (Gomez et al. 1996; Vivek and Simon 1999). The marker order was further examined using CheckMatrix (http://www.atgc.org/XLinkage, Truco et al. 2013) for inconsistencies and markers with more than one inconsistent score were removed. To remove redundant markers in the Br1091xHM1 population, a genetic bin map was developed. For each linkage group pair-wise recombination values among all markers were calculated. Adjacent markers with zero recombination among them were assigned to the same genetic bin. In addition, adjacent markers with “false” recombination due to missing data were considered to belong to the same genetic bin. The marker with the least missing data was chosen to represent each genetic bin. SNPs and SSRs with known chromosome locations were used to anchor the linkage groups. After being assigned to chromosomes, linkage groups were oriented and numbered following the chromosome orientation and classification of Iovene et al. (2011).

QTL analysis was performed in all three populations using R/qtl with the multiple imputations method (Broman and Sen 2009). QTL detection for each population included preliminary QTL identification using scanone followed by QTL modeling. The largest LOD peak from the analysis was added to the QTL model and if the QTL model was significant, it was retained. This process was then repeated using addqtl, instead of scanone, followed by QTL modeling and testing for interactions until adding additional QTL to the model was no longer significant. The final step used addpair to add a pair of interacting QTL or the interaction between a QTL in the model and a newly identified QTL. The support intervals were calculated using a 1.5 LOD drop (Broman and Sen 2009). QTL were named Mi-population-C_-Q_ where “population” is the population in which the QTL was identified, “C_” is the chromosome on which the QTL was identified and “Q_” is the QTL identifier from the QTL model. For example, Mi-BrHM1-C2-Q1 was mapped in the Br1091xHM1 population, is on chromosome 2, and is QTL that explains the most variation in the model.