SNP selection and genotyping results
Gerbera cDNA reads were clustered and assembled into 36,770 EST contigs within which a large number of specific and common SNPs were detected in the parents from the two populations (Fu et al. 2016). For genotyping in the population S, a set of 677 polymorphic SNPs markers, including 477 SNPs common to both populations and 200 specific SNPs, was selected. Similarly, there were 675 SNPs markers selected for population F, including 477 common markers and 198 specific SNPs.
A summary of segregation type for all SNP markers in both populations is shown in Table 1. Of all the selected SNP markers, 68% were successfully genotyped by KASP in population S and 72% were successful in population F (Fig. S1a–c illustrate the three visualised segregation results in SNPviewer). A number of markers showing a single group call are considered as non-polymorphic (Fig. S1d); these include 166 SNPs in population S and 147 in population F. Markers showing scattered segregation without clear grouping were noted as not-fitting segregation (Fig. S1e). The percentages of markers showing a not-fitting pattern were 7% in population S and 6% in population F.
In a number of cases in both the S and F populations (Table 1, null allele), parental genotype scores do not seem to fit the found offspring genotypes. These segregating SNP markers could be further analysed assuming the presence of null alleles. For example, in marker WGC19112, the genotype of two parents are A:G (P1) and G:G (P2), respectively, and the expected segregation in progeny should be [A:G]:[G:G] = 1:1, but the visualised genotyping result in SNPviewer (Fig. S1f) showed three genotype cluster plots [A:A]:[A:G]:[G:G] ≈ 1:1:2 (74:69:133). The possible explanation is the presence of a null allele in P2 (G:Ø) and the actual progeny segregation is [A:G]:[A:Ø]:[G:G]:[G:Ø] ≈ 1:1:1:1, because the genotyping technology cannot distinguish the genotype G:G and G:Ø (they are in the same cluster), also the P2 genotype G:Ø is recognised as G:G.
To use the marker information, we rescored these markers, like WGC19112, with the consideration that both parents are heterozygous. However, information content differed between the parents for such a marker. P1 is heterozygous and WGC19112 is used as a fully informative <lmxll> marker (a). Both A:G and A:Ø offspring clusters are rescored as “lm” and G:G (in fact containing [G:G] and [G:Ø]) as “ll”. P2 is also heterozygous and WGC19112 is here regarded as <nnxnp> marker (b). However, only offsprings within the groups A:A and A:G are informative for this parent (group A:A scored as “nn” and group A:G as “np”), the mixed group G:G (containing [G:G] and [G:Ø]) is discarded. To distinguish the two ways of scoring, we added a letter “a” or “b” at the end of the marker. Markers in which null alleles were demonstrated with an a and b at the end were mapped at almost the same position on the integrated maps, but in eight markers, sufficient linkage was only found in linkage groups of the most informative parent and not in the other parent.
Linkage map construction
Both maternal and paternal maps of the two populations were constructed, as well as integrated maps per population and a consensus map of the two populations. There were 30, 29, 27 and 28 linkage groups constructed in SP1, SP2, FP1 and FP2, respectively (Table S1). Total marker number ranged from 259 in parent FP2 to 350 in parent FP1. The observed parental map lengths varied from 1103 to 1498 cM and the average marker distance varied from 3.50 to 4.41 cM per parental map (Table S1).
Parental linkage maps could be aligned via the presence of bridge markers (<hkxhk> type markers) that are segregating from both parents (Table S2). Based on the position of the bridge markers, marker order on parental maps showed good consistency, but the distance between the markers on parental linkage maps varied as can be expected. For instance, on maternal linkage group FP1_08, the distance between the markers WGC9125 and WGC18021 is 7.6 cM, while the distance on the paternal linkage group (FP2_08) is 12.9 cM (Fig. S2a). By using these bridge markers, the two parental linkage maps could be combined into one integrated linkage map with the same linkage group number code (see Fig. S2a).
Similarly, with the help of common markers, identical parental linkage maps of both crosses could be also be identified and aligned. For instance, there are around 35 markers in linkage group SP1_01, of which some markers are also found in two linkage groups in the parents of the F population (i.e. common markers) indicating that these linkage groups are homologues of SP1_01. So, these fragments are named as FP1_01.1 and FP1_01.2 and as FP2_01.1 and FP2_01.2. The same situation happens on SP1_03, SP2_03, SP1_12, SP2_12, etc. (Table S1). Few maternal and paternal linkage groups (e.g. SP1_21, SP2_23 and FP1_17, FP1_24, FP2_16, FP2_20) could not be aligned to a linkage group of another parent because just a single bridge marker was present or there was a lack of informative common markers (Tables S1 and S2).
These parental linkage maps, with a total of 285 common markers present, can be integrated into a consensus map (see Fig. S2b). In total, 24 consensus linkage groups were merged (Fig. 1, Table S3). As is described in Table S3, the consensus linkage map of 687 SNPs covered 1601 cM. The marker density on the consensus map varied from 1.32 cM on linkage group 09 (LG09) to 5.16 cM on LG17, with an average density of 2.57 cM. There were 14 gaps larger than 15 cM observed in the consensus linkage map.
Phenotypic traits evaluation for Botrytis resistance
Phenotypic data of resistance to B. cinerea were assessed in three tests (whole inflorescence, bottom and ray floret). Histograms of disease testing, resulting from these three traits in the mapping populations S and F and indicating transgressive segregations are shown in Fig. 2. The means of the phenotyping data in population S for whole inflorescence, bottom and ray floret were 2.42 ± 0.55, 2.96 ± 0.63 and 2.98 ± 0.79. Means in population F were 3.64 ± 0.40, 3.80 ± 0.40 and 3.14 ± 0.80, respectively. Based on the skewness and kurtosis scale of the distribution curves, all three tests in the two populations were considered as approximately normally distributed and no transformation of data was performed for QTL analysis.
Disease index of the three disease tests in both populations was analysed by Pearson correlation (Table S4). The coefficients of whole inflorescence and bottom tests in both populations showed moderately high correlations (R = 0.83 in population S, R = 0.67 in population F), but no significant correlation was found to the ray floret tests.
QTL analysis was first performed on the four parental linkage maps individually. The genome-wide (GW) LOD significance thresholds (P < 0.05) for whole inflorescence, bottom and ray floret were obtained using a permutation test (Table 2). Markers, with LOD scores above the GW threshold in every QTL after interval mapping (IM), were chosen as co-factors for multiple QTL models mapping (MQM mapping). Significant QTLs detected from the four parents are shown in Table 2 and Fig. S3a–c.
In the S population, seven significant QTLs for Botrytis resistance were detected by MQM mapping and 13 QTLs in the F population. The difference in numbers of QTLs found between the two populations is defined by the number of QTLs associated with Botrytis resistance in ray floret. There is only one ray floret QTL found in the S population but seven in the F population (Table 2). Phenotypic variance explained by single QTLs ranged between 5.7 and 11.4%, with three QTLs (RBQB4, RBQWI4 and RBQWI6) higher than 10%. Three QTLs, RBQWI2, RBQWI4 and RBQWI6 from SP1, FP1 and FP2, respectively, were found on LG23 at similar positions in the consensus map (see Fig. S4) indicating this may be a single QTL. Interestingly, a QTL for whole inflorescence and bottom (RBQB3 and RBQWI3) shared an identical position with marker WGC18733_346_S2F on LG11 of population S.
Several QTLs were detected on both parental linkage groups separately and showed overlapping positions on the integrated linkage group, like RBQWI1 from SP1 and RBQWI3 from SP2 on LG11 and RBQB5 from FP1 and RBQB6 from FP2 on LG9. In these cases, alleles from both parents contributed to the resistance in the progeny. We identified the favourable and unfavourable alleles from the parents of these QTLs. Progeny can then be divided into four groups: progeny with the presence of two favourable alleles (+/+), with one favourable allele from one of the parents (+/− or −/+) and no favourable allele present (−/−). The mean disease score of each progeny group for each QTL is shown in Table 3. The mean disease scores of individuals with two favourable alleles (+/+) were all significantly lower than those for individuals with no favourable allele present (−/−) and also show advantage over individuals with one favourable allele only.