SNP-markers in Allium species to facilitate introgression breeding in onion

Plant materials

Transcriptome sequencing, data processing and assembly

For SNP discovery, RNA was isolated from leaves of 4–8 week old plants (one plant per cultivar, grown in the greenhouse) using the Trizol protocol (Invitrogen, Carlsbad, Ca, USA) and purified using the RNeasy MinElute kit (Qiagen, Hilden, Germany). cDNA synthesis and bar-coding per sample, followed by paired-end Illumina sequencing were carried out by BaseClear (www.baseclear.com). The Illumina Genome analyzer HiSeq2000 was used to sequence three cDNA samples in a first run (RF and cultivars Bravo and Jumbo) and in the second run two cDNA samples of eight cultivars were multiplexed per lane (Table 1). Paired end reads from fragments shorter than 2× the read length were merged to one single elongated read using the stitch software (github.com/audy/stitch). The resulting sequence fragments and the unjoined reads were all imported as single reads in the CLC Genomics Workbench (v 4.03). Before assembly the reads were processed to increase quality by removal of: a) the first 5′ base, b) sequences with more than one ambiguous nucleotide (N), c) low quality bases on both ends of the reads. De novo assembly with the software CLC was carried out using the following settings: a) 95 % sequence homology over minimal 80 % of the read length; b) mismatch/insertion/deletion costs were set to 2,3,3, respectively; c) global alignment; d) the consensus sequence per contig for variable positions is determined by the most abundant nucleotide; e) reads that match more than one contig (non-specific matches) are not incorporated into any contig, and; f) a minimum contig length of 180 bp is required. Contigs were constructed for each genotype separately and for the onion genotypes combined.

SNP discovery

All contigs were submitted to QualitySNP^ng for SNP discovery using the default settings of the programme [28, 29]. QualitySNP^ng is designed for filtering out SNPs from paralogous sequences without prior knowledge about the reference sequence. If the number of haplotypes exceeded two per SNP locus this is most likely the result of either paralogs assembled in one contig or sequencing artefacts. SNP regions were selected to have 75 bp flanking the SNP position at both sides without additional polymorphisms in these flanking regions. SNP regions were compared with all contigs using BlastN with Expectation value E lower than −20 to select unique SNP regions. Only SNP regions that mapped uniquely were kept [30]. These SNP markers were considered as usable.

SNP selection and genotyping

Two types of SNP markers were selected for genotyping and construction of molecular linkage maps: RF-SNP markers (names start with RF_ctg ) and onion-SNP markers (names start with al_ctg). Selection criterion used for RF-SNP markers were that SNPs were polymorphic in the RF-hybrid and monomorphic in all tested onion cultivars. For onion-SNP markers, the criterion was presence of the major allele in at least five cultivars and the minor allele in at least four cultivars. Genotyping was done with the use of KASP™ assay by VHL Genetics (www.vhlgenetics.com) for the RF-SNPs, and by LGC Genomics (www.lgcgroup.com) for the onion-SNPs.

For genotyping, DNA was extracted [31] from young leaves of the CCxRF population, the F2(CxR) population and parental plants of these populations. For the onion parent, plants of onion cultivars ‘Jumbo’ and ‘Bravo’ were used. Parents of the populations were genotyped at least in triplicate to test reproducibility of the KASP assay. SNP data were visualised using the software programme SNPViewer of LGC Genomics (www.lgcgroup.com/products/genotyping-software/snpviewer).

Genetic mapping

For the CCxRF population, only the SNPs polymorphic between the A. roylei parent and the A. fistulosum parent were used. These markers are expected to segregate in a 1:1 ratio as the selected SNP markers were polymorphic in the RF hybrid (heterozygous) and monomorphic in onion (selected for homozygosity of the onion allele). A genetic map based on SNP markers segregating in 154 progeny plants of CCxRF was constructed using JoinMap® 4.1 [32]. When markers showed an identical segregation pattern, only one was kept for the analysis. Linkage groups were made based on regression with a threshold LOD ≥ 8. Recombination frequencies were converted into map based distances in cM using Haldane’s mapping function.

Both onion-SNP markers and RF-SNPs that were polymorphic between onion and A. roylei were used for the production of a linkage map for the F2(CxR) population. These markers were expected to segregate in a 1:2:1 ratio. The F2 population consisted of 67 genotypes previously used for the identification and mapping of AFLP, SCAR, CAPS, and isozyme markers [20, 21] and 32 newly added F2 plants. From the original 67 genotypes, a limited set was still available (as DNA or plants) and used in the SNP analysis: 46 genotypes in the analysis with the RF-SNPs and 29 with the onion-SNPs. For mapping also marker data obtained previously were included. MapChart 2.2 [33] was used to visualize the genetic maps.

Phenotyping for resistance to Botrytis squamosa

Botrytis squamosa isolate MUCL 31421 (Belgian Coordinated Collection of Micro-organisms, Belgium) was multiplied on water agar covered with onion leaves at 15 °C. For spore production, plates were placed under near-UV (wave length 300–400 nm) or black light (wave length 315–400 nm). Resistance tests were carried out in June 2010 (Exp. 1), September 2010 (Exp. 2) and August 2011 (Exp. 3), comprising 49, 48 and 92 genotypes (in each of the respective experiments) of the CCxRF population and one genotype each of A. cepa cultivar Jumbo, A. roylei, A. fistulosum and the RF hybrid. Genotypes were multiplied in vitro and 10 replicates per genotype were used in each experiment. The F2(CxR) population was not tested, as most plants were not available anymore. Four week old tissue-culture plants were transplanted into the greenhouse in trays containing a mixture of steamed peat soil and sand (v:v = 2:1). Two weeks later, plants were transplanted to 1.5 l pots containing onion peat soil (1 plant per pot) and kept in the greenhouse. In Exp. 1, plants were transferred 6 weeks after transplanting into a plastic fog chamber with 100 % humidity and a temperature of 15 °C (day and night) and inoculated by spraying the plants with a spore concentration of 1.10⁵ spores ml⁻¹. Two days after inoculation, temperature was increased to 18 °C (day and night) and the fog chamber was removed while the humidity was kept at ~80 %. In Exp. 2 and Exp. 3, the same approach was followed but with plants of 15 and 13 weeks after transplanting, respectively. Plants were scored 3 and 4 days post inoculation (dpi) (Exp. 1), 3 and 6 dpi (Exp. 2) and 3 dpi (Exp. 3). Plants were scored in classes from 0 to 4, where 0 = no symptoms, healthy plant; 1 = a few spots on a single leaf; 2 = several small spots on more than one leaf; 3 = several larger spots on more than one leaf and 4 = more than half of the leaf covered with spots (Fig. 1).

As disease scores were obtained on an ordinal scale, the data could not be analysed under the assumption of normality. The data were modelled with reference to an underlying latent variable and threshold values associated with the ordinal scores (Proportional Odds Model) [34]. The parameters threshold values and means were estimated by the maximum likelihood method [35] employing Genstat 18th Ed. (Lawes Agricultural Trust, Rothamsted Exp. St. UK). Positions of the genotype distributions on the latent variable scale were used in subsequent QTL analyses. Broad sense heritability (H²) was calculated as the ratio of the genetic variation and the phenotypic variation.

QTL mapping

The analysis of quantitative trait loci (QTLs) was performed using MapQTL® 6 [36] through interval mapping. Co-dominant markers in these regions were used as co-factors in multiple-QTL mapping (MQM). Significant LOD thresholds were determined using a genome wide permutation test with 1000 iterations. The QTL graphs were prepared with MapChart 2.2 [33].

Transcriptome sequencing, SNP identification and development of markers

For genotyping of the CCxRF population SNPs were selected that were polymorphic in RF and monomorphic in the nine onion cultivars. Among the RF-SNPs only 2525 met this criterion (Additional file 1: Table S1). From these, 1100 (one per contig) were chosen for the development of a KASP assay (Additional file 2: Table S2).

A random selection of RF-SNPs polymorphic between onion and A. roylei was used for genotyping the F2(CxR) population. This set was complemented with onion-SNPs that were selected from the 243,879 onion cultivar contigs containing reliable SNPs. In 14,591 contigs 20,229 reliable onion-SNPs were identified using QualitySNP^ng (Additional file 3: Table S3). To maximize the chance that these SNPs would also be useful with onion cultivars, 1600 onion-SNPs were selected using the criterion that the major allele was present in at least five and the minor allele in at least four of the nine onion cultivars (Additional file 4: Table S4), and these were used for the development of KASP assays.

Molecular mapping

In total, 301 SNP markers (182 onion-SNPs, 119 RF-SNPs) were used to construct a linkage map for F2(CxR). These markers were complemented with 76 markers mapped previously [20, 21] allowing the assignment of chromosome numbers to the linkage groups. The map consisted of 805 cM divided over eight chromosomes (Additional file 5: Table S5).

For the CCxRF population 627 SNP markers segregated as expected for an AA:AB type of marker. The other 140 markers segregated in AA:AB:B0 or in A0:B0 (0 indicating a null allele) fashion when CC was A0 or 00 respectively, and were also used for mapping. The RF-SNP markers that were mapped in both F2(CxR) and CCxRF were used to assign the linkage groups of RF to chromosomes. In total 667 markers were mapped on the CCxRF map, resulting in eight linkage groups (Additional file 6: Table S6 and Additional file 7: Figure S1) with at least 35 markers per linkage group. The length of the maps varied from 83 to 207 cM, with a total length of the CCxRF map of 1051 cM. For both populations, several chromosomal regions showed skewness (with a probability < 0.001). For CCxRF, skewed regions occurred on all chromosomes, except on Chromosomes seven and eight (Additional file 5: Table S5) and for F2(CxR) such regions were seen on five of the eight chromosomes (Chromosomes 1, 3, 4, 6 and 8).

Screening for resistance to Botrytis squamosa

Three independent evaluations were carried out for B. squamosa resistance in the CCxRF population. Leaf spots were observed in all tests already at 2 dpi. Partial resistance was observed in A. roylei, whereas A. fistulosum was almost as susceptible as onion. The level of infection of RF did not significantly differ from A. roylei, clearly indicating dominant inheritance of the resistance. Progeny plants of CCxRF showed a continuous variation in level of infection. Compared to the first two experiments, Exp. 3 showed the largest variation among genotypes, even though the level of infection was already high 3 dpi. In this third experiment, plants of A. roylei also showed infection symptoms. Heritability scores for resistance were 0.89 in Exp. 1 and Exp. 3, and in Exp. 2 0.71 (day 3) and and 0.77 (day 6).

QTL for B. squamosa resistance

The mean values of infection over the classes as well as the mean values of infection obtained by the proportional odds model were used as input in MapQTL. For resistance to B. squamosa, one QTL region originating from A. roylei was identified on Chromosome 6 in each of the three experiments (Fig. 2). The QTL region overlapped between the three experiments, but small differences for the 1 LOD QTL interval were seen. Over the experiments, 27 to 54 % of the total variance was explained (Table 4).

Exp	Position	Locus	LOD	Explained variance (%)
1 day3	13	RF_ctg_14098_355R	5.31	41.2
1 day4	11	RF_ctg_48161_346F	6.86	49.7
2 day3	42	RF_ctg_48477_420F	8.01	54.4
2 day6	42	RF_ctg_48477_420F	5.20	39.9
3 day3	15	RF_ctg_60255_530R	5.91	26.6

SNP discovery and marker development

SNP markers were designed to facilitate the introgression of traits from the onion related species A. roylei and A. fistulosum that possess, amongst others, disease resistances that are not present in onion cultivars. Transcriptome sequencing has proven to be an efficient approach to obtain SNP markers in diverse crops reviewed by [37, 38]. The advantage of transcriptome sequencing is that the sequencing is limited to parts of the genome that are more likely single copy, which is especially useful for crops with large, highly repetitive, genomes such as onion (16 Gb) [39–41]. We used transcriptome sequencing to obtain two sets of SNP markers that will be useful for Allium breeding: RF-SNP markers, which are polymorphic in the interspecific RF hybrid PRI 91021-8 and monomorphic among onion cultivars, and onion-SNP markers, which are polymorphic among onion cultivars. In total, 2525 RF-SNPs and 20,229 onion-SNPs were identified and regarded as good candidates for the development of KASP assays. Conversion of 1100 RF-SNPs and 1600 onion-SNPs in KASP genotyping assays and validation was successful for 70 % of the RF- and 77 % of the onion-SNP markers. These success rates are comparable to those reported for onion cultivars (74 %) [24] and for lily, another outcrossing species with a large genome (76 %) [30]. With the validation of 1237 new SNP markers for onion and 767 for the related species A. roylei and A. fistulosum, the amount of SNP markers that is currently publicly available has more than doubled: 43 SNP markers validated in 2005 [42], 93 in 2012 [41] and 930 in 2013 [24]. SNP markers are valuable tools for cultivar identification, determining genetic relatedness and diversity estimates among cultivars [43]. In a next step, SNP markers of various onion research groups could be combined to obtain a global consensus map of onion and other crossable Alliaceae [44]. Such a global consensus map is highly valuable for the onion research and breeding community, as it will allow the comparison of QTLs for specific traits detected in different populations. Mapped SNP markers are also valuable for improving the quality of the de novo genome assembly of onion, which is currently being carried out [45].

Genetic maps and distortion of segregation

Selection of SNPs that are polymorphic in the RF hybrid but monomorphic in onion, turned out to be very useful for creating an inter species linkage map for A. roylei -A. fistulosum. The SNP linkage map spanned 1050 cM, which is longer than the AFLP maps of 661 cM and 886 cM earlier obtained [46, 47]. This increase in map length is likely due to the increase in the number of mapped markers at the telomeres. For the F2(CxR) population, discrepancies were encountered when mapping the newly obtained SNP markers together with 76 from the 526 previously mapped AFLP, CAPS and isozyme markers [21]. These are likely the result of the low number of identical genotypes in both studies (only 29). The intention of using the F2(CxR) population was to obtain sufficient information to assign linkage groups to chromosomes for the CCxRF population, which was successful. Distorted segregation of SNP markers was observed on six of the eight chromosomes of RF and five of the chromosomes of CR. In the past, the CCxRF population was used for the construction of a genetic linkage map based on AFLP markers [46], as well as for physical mapping [27]. Later, the population was extended with progeny plants from additional crosses between onion plants and the original RF genotype to identify QTLs for mycorrhizal responsiveness, plant dry weight and number of stem-borne roots per plant [47]. For the SNP genetic map presented in this study, progeny plants obtained from a single cross between an onion plant and the RF plant were used in order to prevent distorted segregation of markers due to unequal contributions of alleles from different mother plants. For CCxRF and F2(CxR) population, distorted segregation was reported earlier [20, 46]. Also in other crops, like tomato, distortions from the expected segregation ratio often occurs in crosses between cultivated material and wild species [48]. Distorted segregation may be the result of post- and pre-fertilization barriers between species. The phenomenon of disturbed pollen tube growth was observed in the style of interspecific F1 hybrids between A. cepa and A. fistulosum using A. cepa pollen [49]. Distorted segregation may also be the result of preferential genome or allele transmission, as was observed in F2 progeny plants of crosses between A. cepa and A. fistulosum as well as suppression of allelic expression and cyto-abnormalities in mitosis and meiosis [50].

BLB resistance and mapping of the Bs1 gene

In the current study, the focus was on the identification of the locus for resistance to B. squamosa from A. roylei. Previous studies indicated that resistance to B. squamosa in A. roylei is probably conditioned by a single dominant gene, for which Bs1 was proposed [7]. Both A. roylei and the interspecific RF hybrid showed comparable low levels of infection by B. squamosa and were clearly resistant, whereas onion plants and plants of A. fistulosum were highly infected and thus susceptible. Our results also demonstrated that resistance to B. squamosa in A. roylei is based on a high level of partial resistance, indicating that plants still can be infected by the pathogen. This is in line with other studies in which Botrytis leaf blight symptoms were also observed in onion plants homozygous for the Bs1 gene [10]. These plants were obtained after two generations of back-crossing to onion of an interspecific F1 hybrid between onion and A. roylei, followed by three to four generations of selfing. Under heavy disease pressure, heterozygous plants had levels of BLB symptoms between those of homozygous resistant inbred plants and susceptible controls [10]. After crossing the interspecific RF hybrid with onion, we found only one CCxRF plant with an infection level similar to A. roylei and RF, whereas all other plants, that we also considered resistant, had higher levels of infection. In case of one resistance gene, we would have expected a more defined segregation of resistance in this population, even if resistance was partial. Therefore, we hypothesise that one or more QTLs with minor effects originating from A. roylei or from A. fistulosum may play a role as well. Preliminary results obtained from a detached leaf assay indeed point in the direction of a minor QTL originating from A. roylei (Scholten and Burger, unpublished results).

QTL mapping for resistance to B. squamosa resulted in the identification of a QTL region on Chr. 6 of approximately 50 cM that explained 27 to 54 % of the variation over the experiments. Although this is a wide region, this is the first paper describing a locus originating from A. roylei conferring resistance to B. squamosa the causal agent of Botrytis leaf blight in onion. Results of earlier studies with CCxRF showed that recombination between the genomes of A. fistulosum and A. roylei and in back-crosses with onion between onion and this RF hybrid took place to a large extent [22, 23]. To narrow down the QTL region, repeated backcrossing with onion needs to be carried out followed by a recombination screening. In a recombination screening also markers that become homozygous are useful for selection.

Another approach to zoom in on the location of the Bs1 gene is genotyping the four available BLB-resistant onion lines containing the same source of resistance, CUBLB-R1, -R2, -R3 and -R4 (generation F1BC2S3 or S4) [10]. With the use of RF-SNP markers of Chr. 6 that segregate between onion and A. roylei we may identify the presence of A. roylei fragments still present in these homozygous resistant progeny plants obtained after five or six generations of selection. Indications for applicability are favourable, as heterozygous F1 hybrids obtained after crosses between the CUBLB-R1 to -R4 lines and parental onion lines had similar levels of resistance to B. squamosa in most experimental field trials as the homozygous lines, without observing any drawbacks in terms of reduced bulb size or yield [10].